D.AI.SY: DECENTRALIZED AI SYSTEM
AI-POWERED INVESTING UNLOCKING
HIGH-RISK HIGH-RETURN ALPHA
Whitepaper • Version 1.0 • Feb, 2021
Table of Contents
Legal Disclaimer
4
D.AI.SY Project Overview
5
Executive Summary
6
Background
8
Market Analysis
10
Enormous, Global Investor Opportunity
Investment Vehicles for High-Risk High-Return Investment
Systematic Trading
14
Technical Analysis
Endotech’s Technical Methodology
Fundamental Analysis
D.AI.SY Deep Reasoning Methodology
D.AI.SY Methodology
19
D.AI.SY Basics
D.AI.SY Raw Data Pools
D.AI.SY Data Collection and Verification Methodology
D.AI.SY Mapping Movements and Money Flows
D.AI.SY AI Deep Reasoning
D.AI.SY AI Volatility Models
D.AI.SY Algorithmic Trading
Endotech’s Trading Framework
30
Endotech Trading Methodology
Trading Capacity
Whitepaper • Version 1.0 • Feb, 2021
2
/ 57
Technology And Architecture
35
Architectural Concept
Ingest & Store
Analyze & Prepare
Learn
Roadmap
38
D.AI.SY Project Team
39
References
40
Appendix
43
Risk Return Defined
Global Market Analysis
Algorithmic Hedge Funds
Global Money Map
Whitepaper • Version 1.0 • Feb, 2021
3
/ 57
Legal Disclaimer
Please read this notice carefully before reading this whitepaper. This legal disclaimer refers to every
person and organization that reads this document. This notice is also subject to changes in the future
without notice.
This whitepaper does not constitute a legal document. The purpose of whitepaper is to provide back-
ground information about the technical aspects underpinning the D.AI.SY project and the market
opportunities. Significant efforts have been made to ensure this document contains true and current
information.
This document does not constitute professional investment advice nor is it to be considered invest-
ment recommendation for purchasing any asset including cryptocurrency.
Potential investors and buyers should consult professional legal, investment, tax, accounting advice for
determining the suitability of any investment. Readers alone are responsible for evaluating the merits
and risks, the potential benefits and possible consequences of actions connected with participation
in the D.AI.SY project.
Whitepapers are not legal documents and have no legal standing in any possible direct or indirect
claim for compensation for any kind of damage, including possible damage caused by loss of profit
or investment.
Whitepaper • Version 1.0 • Feb, 2021
4
/ 57
D.AI.SY Project Overview
Endotech Ltd. is an Israeli company working in the field of Artificial Intelligence and finance for more
than a decade. It has created some of the most enduring and successful algorithmic breakthroughs
in the field. Dr. Anna Becker has worked with banks, hedge funds, trading brokers, family offices and
individual traders across the globe to develop proprietary algorithms.
In 2018 Dr. Becker launched the renowned ‘eponymous’ project at Endotech.io to provide high-risk
high-return high-probability strategies with their fully automated execution for both retail and insti-
tutional traders.
In January 2021, EndoTech announced the Equity Crowd Funding agreement with D.AI.SY Global for
the development of superintelligence under the name of D.AI.SY project. D.AI.SY crowdfunding model
operates through a smart contract on the blockchain, and allows retail investors to participate in the
future of EndoTech. Note – Endotech is not responsible for D.AI.SY marketing operations.
D.AI.SY’s mission is to improve the probabilistic returns by using superintelligence in fundamental
analysis. It will become an integral part of EndoTech’s automated investment solution for Endotech
network members.
Whitepaper • Version 1.0 • Feb, 2021
5
/ 57
Executive Summary
D.AI.SY was established by AI algorithm experts with vast investment experience in the financial space.
Their goal is to revolutionize the scope of data and understanding that underpins financial markets
and capture the alpha therein. D.AI.SY harnesses proven algorithmic tools. to explore, map and har-
vest repeatable high-risk high-return investment opportunities using super-intelligence AI techniques.
To address this challenge, the joint scientific-investment team is building a robust infrastructure that
consists of a full, validated investment datapool, super-intelligent relational understanding, a series
of AI tools for identifying and extracting opportunity and automated execution engine to capitalize
on alpha.
THE RETAIL INVESTMENT INDUSTRY REQUIRES MORE SCIENTIFIC RIGOR TO IDENTIFY AND
EXTRACT ALPHA OPPORTUNITIES.
Current data-base investing tends to harvest small pockets of limited data and is bounded by AI insights
rooted in only technical analysis. Unfortunately, those AI pockets often concentrated amongst several
hedge funds, unavailable for retail investors seeking high-risk high-return opportunities. Furthermore,
that AI alpha has proven sporadic and unpredictable as it is limited both in terms of breadth of data
and depth of scientific understanding. Even today’s AI-powered hunters lack super-intelligence for
prolonged high-risk high-return results. There has yet to be a serious scientific expansion of invest-
ment data learning, and it is not accessible to retail investors to participate.
Today’s scientific investor is at best descriptive – not predictive. Today’s scientific investor is institu-
tional – not retail. Today’s scientific investor only surfaces temporal alpha – not sustainable opportu-
nities.
D.AI.SY IS THE FIRST COMPREHENSIVE, SCIENTIFIC APPROACH TO INVESTMENT UNDERSTANDING
USING RELATIONAL SUPER-INTELLIGENCE.
D.AI.SY’s data ecosystem gathers and validates terabytes of investment data from sources beyond tech-
nical data. It integrates, validates and conducts relational causality understanding between elements.
In this process, super-intelligence is created to determine the predictive causes of various investments
behaviors and outcomes. With refinement and accuracy, this creates repeatable, predictable High-risk
high-return trading opportunities.
The D.AI.SY data ecosystem ensures transparent, decentralized and vibrant investment informa-
tion for super-intelligence mining. With the creation of D.AI.SY’s Data Pools platform it is possible
to broaden and deepen the sources of scientific investment study and apply new informatics programs
to their understanding.
D.AI.SY IS FOCUSING ITS INITIAL RESEARCH ON APPLICATION IN HIGHLY LIQUID, VOLATILE ASSETS.
The work is difficult. The approach is known. The challenge is bold. The returns are great. To focus the
scientific work, the D.AI.SY team will be concentrating the initial efforts on a narrow investment set,
Whitepaper • Version 1.0 • Feb, 2021
6
/ 57
with a super-intelligence approach to prove the approach. The team understands that by limiting the
scope to highly liquid, volatile assets, it will be easier to identify opportunities for member activation.
Furthermore, through the established Endotech algorithmic investing, it will be easier to capitalize
on identified, sustained Alpha opportunities.
NEW SCIENTIFIC APPROACH TO GENERATE NEW INVESTMENT RESULTS.
Through Endotech, D.AI.SY will expose new financial investment data, use super-intelligence to under-
stand relationships between stimulus and leverage proven AI tools to capture market alpha. It’s first
target is tradeable, predictable alpha rooted in volatility.
Whitepaper • Version 1.0 • Feb, 2021
7
/ 57
Background
While historic investment strategies centered around conservative savings plans, the latest genera-
tions have shown that investing is no longer savings-focused. A growing number of retail investors are
no longer content with traditional, long term, slow growth strategies. Simply put, they want aggressive
returns.
As individuals press on with their traditional work days, they have developed a penchant for more
aggressive investments. Often, having secured their savings targets, they then look to other vehicles
for higher risk, higher return investments. We have seen a generation of investors move from passive
savers to active investors that are “leaning in” to market opportunities to seek out higher return oppor-
tunities. The flourishing foreign exchange, ETF and Crypto markets are just some examples of this
new drive towards aggressive upside.
While the investment industry creates new high-risk high-return vehicles, it is important to highlight
that despite the incredible progress in financial knowledge, fundamental analysis, scientific rigour,
the probability of netting a high-risk high-return is, at best, unpredictable.
While data abounds in technical analysis, these high-risk high-return algorithms are threatened
by time-horizon. Endotech’s decades of experience in algorithmic investing has proven the alpha.
However, there is a reality that even advanced technical analytical algorithms cannot account for fun-
damental shifts. Inevitably, their results will be time-bound. It’s with this lens that Endotech embarks
on D.AI.SY – to extend its scientific rigour to retail algorithmic investing rooted in fundamental anal-
ysis.
Hedge Funds have explored the use of deep data on fundamental analysis. While this appears
to be oxymoronic, it isn’t. There are specific data signals that can be derived from fundamental anal-
ysis that can be systematically identified and activated. And, their super-natural, enduring results are
dramatic. However, the data work and systematic execution that has been applied to date is limited
and restricted: Limited by their modeling approach of deep learning (instead of deep reasoning) and
restricted to high-net worth individuals or algorithmic developers.
The early and dramatic financial opportunity in systematical fundamental analysis has been shown:
There are proven examples of use of data science for fundamental-derived signals to inform enduring
systematic trading.
D.AI.SY is set to capture fundamental analysis signals, cleanse them and map them to a definitive data
pool for systematic modelling. Harnessing the proven methodology from other scientific fields, this
data pool will provide the training ground to develop models well beyond today’s Sharpe ratio of 2.
With super-intelligence, D.AI.SY will surface, and then execute these high-risk high-return opportuni-
ties for retail investors to auto-trade.
Whitepaper • Version 1.0 • Feb, 2021
8
/ 57
This scientific undertaking will revolutionize the scope of data and understanding that underpins
financial markets and capturing the alpha therein through proven algorithmic tools. D.AI.SY is set
to explore, map and harvest repeatable high-risk high-return investment opportunities using super-in-
telligence AI techniques.
Whitepaper • Version 1.0 • Feb, 2021
9
/ 57
Market Analysis
ENORMOUS, GLOBAL INVESTOR OPPORTUNITY
Investing has become so prevalent globally, that investors now ‘swipe’ to trade. While there is no defin-
itive number of retail investment accounts, the growth is obvious worldwide. Even as global coronavirus
infections continue to mount and more companies issue profit warnings, stockbroker switchboards
from Sydney to Singapore have lit up with calls from masses wanting to invest.
Furthermore, the breadth and scale of the COVID-19 lockdowns and business closures across the globe
are also unprecedented, prompting governments to pump in trillions of dollars in support and central
banks to slash interest rates, fuelling the supply of cheap cash. Many small investors seem to feel this
is an unrivalled opportunity – particularly as markets have rallied in 2020. Almost invariably across all
key global markets, we can see enormous upswings of activity.
RETAIL MARKET OPPORTUNITY
The United States Federal Reserve Bulletin suggests that 51% of American families own stocks and
the majority of those are direct holdings. More than that, millions more American families are active
in other asset categories like foreign exchange and crypto currencies.
Of course, this thirst for investment is not limited to any region or investment class. Retail investors
seek high-return opportunities across different markets and assets in droves with hundreds of mil-
lions retail foreign exchange traders and crypto currencies.
With gross savings plateauing in the last decade, investors look for new investment vehicles. It is now
clear that investing no longer means savings. And Investing is no longer limited to conservative, long
term investment.
Whitepaper • Version 1.0 • Feb, 2021
1
0 / 57
GROSS SAVINGS (% OF GDP)
A fuller global analysis can be found in the Appendix
Whitepaper • Version 1.0 • Feb, 2021
1
1 / 57
INVESTMENT VEHICLES
FOR HIGH-RISK HIGH-RETURN INVESTMENT
1.
SUSTAINED HIGH-RISK HIGH-RETURN INVESTMENTS
While sustained high-risk high-return investment is a new retail ambition, it has long been the
goal of hedge fund managers. Despite the fact that Hedge funds actually under-performed the
market in 2020, there are good reasons to believe that certain hedge fund algorithmic approaches
to investment represent a good model for high risk High-Return retail investors.
For example, Renaissance Technologies, based in the United states, currently has $84 billion
under management and has been delivering superior and consistent high-risk high-return for
three decades. Founded in 1982 by James Harris Simons – a famous American mathematician
who deciphered Soviet codes during the Cold War. Until 2009. Renaissance Technologies
is a quantitative trading fund whose strategies are based on statistical and mathematical analysis.
And their Medallion investment funds have the best history of profitability on Wall Street –
60% per annum for 30 years.
2.
RETAIL INVESTORS HAVE NOT HAD HIGH-RISK HIGH-RETURN VEHICLES
While certain hedge funds have been able to master financial wizardry through combinations
of fundamental and technical analysis, combined with excellence in execution, these vehicles are
not accessible to the millions of retail investors identified above.
As such, these hedge funds are for only extremely high net individuals, foundations or institutions.
As of today, there are no financially robust investment vehicles available to retail investors
to capitalize on sustained high-risk high-return investment approaches. Are there opportunities
to create similarly successful mechanisms for the millions of retail investors?
3.
HEDGE FUNDS HAVE PROVEN ALGORITHMIC OPPORTUNITY AND BOUNDARIES
As to 2020, hedge funds have approximately $ 9.6 trillion under general management. Despite the
strongest market volatility, over the past year, the 20 best performing hedge funds in the world
earned $63.5 billion for their clients, breaking a decade’s record, writes Reuters.
While large sectors of the economy were forced to close, and millions of people lost their jobs,
the top 20 hedge funds in the world were able not only to earn decent money, but also to exceed
the profit indicators of 2019 of $ 59.3 billion. At the same time, 2020 was not as profitable as the
previous one for hedge funds in general when they made $ 178 billion.
According to Hedge Fund Research, the average return of a hedge fund in 2020 was 11.6%, which
is 16% less than the growth of the S&P 500 index.
There is however a huge disparity in returns. Some hedge fund managers, such as Saba
Capital’s Boaz Weinstein and Pershing Square’s Bill Ackman, have taken advantage of these
Whitepaper • Version 1.0 • Feb, 2021
1
2 / 57
extraordinary conditions, correctly calling some of the big market moves. Others have been
caught out.
In macro trading — betting on global bonds, currencies and other assets — years of famine have
been replaced by a time of plenty. Some of the moves have been spectacular, such as the fall in the
US 10-year yield from 1.9 per cent to 0.7 per cent, as its price surged. Funds such as Brevan Howard,
Caxton Associates and Rokos Capital Management are enjoying some of their best gains in years.
Some of the highest-profile computer-driven fund firms, on the other hand, have found their
models unable to cope well with wild market swings. In some cases the switch between sectors
of the market such as faster-growing stocks and cheap stocks has caught them off guard.
Jim Simons’ Renaissance Technologies, David Harding’s Winton Group and California-based
artificial intelligence pioneer Voleon Group are among those who have nursed losses from the
first-quarter rout that were not made up in the market rebound. US quant giant AQR’s Equity
Market Neutral fund — which is designed to hedge out the risk of equity market moves — is down
nearly 15 per cent this year.
In equity investing, Lansdowne Partners and Larry Robbins’s Glenview Capital are among those
down sharply this year, even as funds such as Ross Turner’s Pelham Capital are making double-
digit gains, according to numbers sent to investors.
Turbulence in credit markets, meanwhile, has left many funds in the sector in the red for 2020.
However, LCH Chairman Rick Sofer is confident that 2020 can be considered one of the best
in terms of profit for the most famous and large hedge funds, as the net profit received by 20 top
managers for their clients in the amount of $ 63.5 billion was the highest in decades.
See Appendix for more about Algorithmic Hedge Fund and Crypto Hedge Funds.
Whitepaper • Version 1.0 • Feb, 2021
1
3 / 57
Systematic Trading
There are two widely known schools of thought with regards to trading analysis: technical analysis
and fundamental analysis. With respect to systematic, or ‘algorithmic’ investing, most of the efforts
(and successes) to date have been in technical analysis, while the greatest alpha opportunity seems
to be in finding systematic opportunities in combination of technical and fundamental analysis.
TECHNICAL ANALYSIS
Technical analysis has been used by traders, analysts, and investors for centuries and has achieved
broad acceptance among regulators and the academic community – particularly with regard to its
behavioral finance aspects.
Systematic Trading rooted in Technical Analysis has proven to be fertile ground for out-sized results.
With automated synthesis of existing technical patterns, investors have been able to identify, model
and extract value from short-term opportunities. Common systems identify movement patterns
in shorter, high frequency, high volatility trades and take advantage of arbitrage opportunities. Herein
lies the successes to date, particularly for breakthrough retail investors.
Endotech’s successful track record in technical analysis is rooted in surfacing volatility.
ENDOTECH’S TECHNICAL METHODOLOGY
STEP 1: STATISTICAL PATTERN RECOGNITION
Endotech creates a comprehensive statistical analysis of the markets behaviour/patterns for multiple
time frames. For example in ETH/USDT, Endotech identifies the following trends that become predict-
able once captured:
Whitepaper • Version 1.0 • Feb, 2021
1
4 / 57
STATISTICS ON ETH TRENDS 2018-2020
60%
45%
30%
15%
0%
Trends of 10% and up from Jan 2018 show – consistent high-volatility present.
STEP 2: GOAL SETTING
Once trends are identified – Endotech specifies model precision and accuracy:
•
Precision – depicts the number of patterns we achieved to penetrate,
•
Accuracy – depicts how accurately each pattern was penetrated.
For example, if a pattern is a trend of 10%+ (meaning that price moves 10% and up), then:
•
precision equals the number of such bullish trends captured in long positions plus the number
of bearish trends the system captured in short positions divided by the total number of trends.
•
accuracy is combination of entry and exit position accuracy, where entry accuracy is a percent
of move
STEP 3: SETTING PERFORMANCE MARKERS
When launching strategies, Endotech system defines exact parameters for preventing losses, by stop-
ping the system or reducing portfolio allocation when performance markers become different from
expected ones, e.g. success rate or accuracy rates drop below the threshold, max number of losses
or max drawdown are reached, slippage, number of traders, etc.
STEP 4: DEFINING TECHNICAL TOOLS
Once patterns are identified, Endotech starts by running individual proprietary indicators with very
few parameters, and then increases details as unified patterns and variations are identified. Then
Endotech cross-checks each indicator and parameters to verify that it belongs to a unified model that
works on each type of asset / and individual asset.
STEP 5: PARAMETERS OPTIMISATION AND SYSTEM ADAPTATION
From the outset, systems have a pre-configured optimization that applies machine learning with
proprietary score, resolving overfitting and underfitting, that use predefined constraints. We run
Whitepaper • Version 1.0 • Feb, 2021
1
5 / 57
dynamic Cuda-based optimization, using combinations of gradient descent, proprietary goals, and
filter functions to find a range of parameters values for the algorithms.
And then we run re-optimisation once a day, to ensure that we adapt the system to market conditions,
either reduce allocation, or stop it completely.
But even this dynamic re-optimisation is not always sufficient and identification and analysis
of changes in patterns needed – usually when there is a new – extreme volatility.
Volatile markets need robust technical analysis systems to generate opportunities. However, over long
periods of time and varied circumstances, these models have proven to have Fading Performance, due
to limited efficacy as they are unable to learn and adapt to new market conditions. When modelled
‘signals’ are no longer present, or there is a fundamental shift in extra-ordinary conditions, models
are not able to identify, let alone act. They simply cannot recognize, understand or internalize the
externalities that render their systematic approach irrelevant and unprofitable.
To address this eventual Fading, Endotech is looking at D.AI.SY to find volatility in systematic funda-
mental analysis.
FUNDAMENTAL ANALYSIS
Fundamental Analysis seeks the true value of assets based on their fundamental worth through a lens
of micro and macro economic understanding.
Applying a systematic approach to uncovering patterns in fundamental analysis is rife with challenges.
The underlying data is both endless and often meaningless. On the one hand, there are endless inputs
of varying quality that must be considered. On the other, there are very few significant movements
in asset values to study. The relationship between the inputs and outputs is inaccurate and not mean-
ingful.
Systematic Trading anchored in Fundamental Analysis is the new holy grail for algorithmic investors.
Today, it is also thought to be the secret sauce that propels successful hedge funds like TwoSigma and
the like. But this is an enormous scientific endeavor to try to understand sustained data and opportu-
nities from an unending pool of information.
Herein lies the challenge of fundamental analysis. The breadth of underlying factors is innumera-
ble – social trends, demand, business leadership, political movements, regulation… So the modelling
is equally complicated.
Hedge fund success in tapping fundamental-derived, systematic returns is rooted in a narrow view
of the data. Each fund will find their angle on fundamental analysis and derive their corresponding
signal to systematically trade. And the results, though limited to elite hedge funds, is dramatic. Sus-
tained, systematic returns.
Whitepaper • Version 1.0 • Feb, 2021
1
6 / 57
To make this real for readers less versed in systematic signals derived from fundamental analysis,
consider these examples.
Corporate earning reports are rife with data and words. Systematically
picking up the frequency and position of various key words can
be a powerful input towards algorithmically understanding volatility.
Or, consider the understanding necessary to decipher the impact on fundamentals
of a high-profile investor like Elon Musk moving into Bitcoin, as compared to other
investors. How is this fundamental move to be understood and read systematically?
So while just anecdotal, these examples show the potential of identifying, cleaning fundamental data
pools, translating them into systematic signals for modelling.
D.AI.SY purely systematic approach to trading (with no manual interference), is rooted in an exact
element of fundamental analysis. In this case, volatility is the familiar and enduring input that will
guide algorithmic development, modeling and execution.
While there are too many specific signals that can be modelled and systematically derived from fun-
damental analysis, volatility is a purpose-driven choice for D.AI.SY.
D.AI.SY is a high-risk high-return approach that seeks to systematically create, recognize and act upon
trading opportunities for 100% + returns. From technical analysis alone, volatility has proven to be one
of the most integral ingredients for successful, profitable – though temporal – modelling.
Volatility will be a key enabler to this scientific undertaking. While traditional investment models are
celebrated for achieving Sharpe ratios of 2, the depths of data that can be surfaced from fundamental
analysis is much deeper. Using super-intelligence, and a restricted view as to what needs to be mined
from fundamental signals, D.AI.SY sets itself a more audacious accuracy goal.
Using a scientific approach familiar in other realms, it is possible to achieve a Sharpe ratio of 5. This
goal – the scientific standard in Physics – will enable sustained, systematic trading. With a stated goal
of discovering volatility from fundamental analysis, D.AI.SY can bring the scientific rigor to achieve
the sustained high-risk high-return trading.
AI IN FUNDAMENTAL ANALYSIS
Successful scientific approaches require validated data to be normalized and mapped in order
to understand their relationships. To date, the majority of approaches have attempted to “solve” for
asset movements and not understand relationships. These efforts have floundered as there has been
a weak relationship between single inputs and outputs. Deep learning AI has not been able to identify
the significant data from macro/micro to point to systemic results.
Whitepaper • Version 1.0 • Feb, 2021
1
7 / 57
DEEP LEARNING
DEEP REASONING
Deep learning has yet to mine the myriad
Deep reasoning, or super-intelligence, is the
of fundamental inputs to find single,
scientific approach that maps the outputs
significant, inputs that consistently account
and seeks to understand the underlying
for asset movements. But there are other
interplay of elements that created that.
approaches to scientific analysis beyond
It is a far more bounded approach to scien-
deep learning. Enter deep reasoning.
tific study and can shed light on the narrower
set of factors (not any individual factor) that
underlie a movement.
D.AI.SY DEEP REASONING METHODOLOGY
The combination of lessons learned from deep learning short-coming, and the breakthroughs in fun-
damental analysis adopted from parallel scientific communities create a clear path for modelling
fundamental analysis.
With a known result as its goal – volatility – D.AI.SY, will apply deep-reasoning to a bounded financial
fundamental dataset to map data and reasoning around volatility.
Using techniques proven in other scientific domains, D.AI.SY will create a definitive source of vali-
dated data, to model relationships between significant inputs that underpin volatility. The technical
details are expanded upon below but fall under various categories.
Whitepaper • Version 1.0 • Feb, 2021
1
8 / 57
D.AI.SY Methodology
D.AI.SY BASICS
D.AI.SY’s modelling of fundamental market movements follows a very deterministic approach. The
methodology will enable accelerated progress towards learning and application of its goal of system-
atically surfacing fundamental-rooted volatility.
The methodology has proven itself in both other algorithmic financial arenas and other scientific
sectors. While this undertaking is audacious, the methodology has proven robust enough to make
breakthroughs in understanding.
D.AI.SY’s proprietary methodology follows these steps:
1.
Establishing a detailed, validated view of important data pools. This includes different forms
of money and assets as they stand across critical assets, geographies and more. This is the
fundamental layer and acts as ‘baseline’ for understanding future movements.
2.
Data is validated and integrated. The challenge of normalizing and quantifying fundamental
signals is significant. The process of quantification, categorization and validation is a significant
effort.
3.
Mapping key movements and flows for money and capturing fundamental changes. This
includes noting change in types of money, assets as well as changes that have occurred
in parallel like geopolitical upheaval, trade, education and more. This is capturing the changes
in type and direction of assets and capturing the suspected contributing factors in a data-based
fashion.
4.
Crafting and testing various AI deep-reasoning hypotheses to understand the relations between
various fundamental shifts and money movements. Modelling and testing the interplay between
flows, pools and contributing factors. This machine-learning sets to act methodologically upon
terabytes of data to establish the veracity of any factor on market movements.
5.
Building an ensemble/authoritative model from validated hypotheses to systematically identify,
capture and understand fundamental stimulus and compute the relationship with anticipated
market volatility.
6.
Empowering algorithmic trading to systematically harness these fundamental signals and
enable technical analysis algorithms to identify and act on volatility. Leveraging the model
towards building and capturing market volatility indicators with algorithmic trading for retail
investors.
Whitepaper • Version 1.0 • Feb, 2021
1
9 / 57
D.AI.SY’S PROPRIETARY METHODOLOGY
AI Technical Analysis
6. Algorithmic Trading
Volatility
5. AI Volatility Models
MODEL
MODEL
MODEL
MODEL
4. AI Deep Reasoning
Money Flow
Money Flow Asset
Money Flow Types
3. Mapping Movements and Flows
Country
Money Map
Geo-Political
Behavioural
2. Verified & Normalised Data Pools
3rd Party Sources
Social Media
Geo-Political
Sentiments
1. Raw Data Pools
Side-effect of this approach: In 1981 Robert James Shiller published a famous article in which he chal-
lenged the efficient-market hypothesis driven by emotion instead of rational calculation. He argued
that a huge set of data is required for the market to operate efficiently. He added that the use of mod-
ern technology can benefit economists to accrue data of broader asset classes that will make the
market more information-based and the prices more efficient.
D.AI.SY RAW DATA POOLS
D.AI.SY project is building a comprehensive model of the liquid markets volatility across the world.
As a first step, D.AI.SY needs to identify key fundamental money pools for monitoring and inges-
tion. The first critical component that needs to be understood is the current snapshot of the Global
Money Map (see Appendix). D.AI.SY views the money map from multiple dimensions of the money
metrics – what money type, how money is used, who owns, where located.
Key spheres of activity that impact the market include economics, politics and society. In order to fur-
ther map the market, their preliminary classification and gradation is necessary. Furthermore, the
classification of money in accordance with their origin and material values represented by them.
Whitepaper • Version 1.0 • Feb, 2021
2
0 / 57
Several categories of money/value can be distinguished including:
1.
Money equivalent to gold and other absolute material values, such as natural resources, land,
exclusive works of art.
2.
Money of global industry, as energy, basic materials, machinery, tools, heavy vehicles, etc.
3.
Consumer money. Manufacturing of widely consumed goods, as textiles, food, agricultural
sector, etc.
4.
Young money. This is money earned by investing in financial assets and transactions with them.
5.
Unsecured money. Money earned from non-resource or non-financial industries such
as tourism, show business, the Internet, information and science-intensive technologies.
ECONOMY
Consideration of economic processes is based on well-known classifications of economic sectors, such
as the Global Industry Classification Standard (GICS) developed by MSCI and The Refinitiv Business
Classification (TRBC) developed by the Reuters Group. The classification of industries is indexed and
brought into line with categories of money.
AN EXAMPLE OF A METHODOLOGY FOR INDEXING GLOBAL INDUSTRIES.
Type of Production
Universal
Specialized
Special
Direct Production
Maintenance
Communications
Role in the Marketing Chain
Whitepaper • Version 1.0 • Feb, 2021
2
1 / 57
AN EXAMPLE OF A METHODOLOGY FOR INDEXING NON-RESOURCE INDUSTRIES.
Scope of Application
Economics
Quality Of Life
Society
Direct Production
Maintenance
Communications
Role In The Marketing Chain
POLITICS
The study of political processes is based on geopolitical concepts of control over the territory and
related regularities of distribution and redistribution of the influence spheres for various states and
associations. Spheres of geopolitical influences and interests have always been closely associated with
economics. In the modern world of globalization and high informatization of space, one can observe
a direct flow of all geopolitical trends into the economic sphere. And, as a consequence, the direct
impact of these processes to the market. For the analysis, the classical ideas, principles and concepts
of geopolitical science are used (see Appendix).
SOCIETY
For the structuring and further analysis of information about social processes, a historically formed
stable triplex of basic institutions that regulate the functioning of the main social subsystems is used.
These subsystems are economics, politics and ideology. In macrosociology they are called institu-
tional matrices. The term was first defined by K. Polanyi, then it was used by D. North. The theory was
further developed in the works of S. Kirdina. Based on the analysis of extensive empirical material,
it was shown that, as a rule, one of two institutional matrices is steadily dominating in the structure
of society: either the X – or the Y-matrix, which are qualitatively different from each other in the con-
tent of the underlying institutions that form them:
Whitepaper • Version 1.0 • Feb, 2021
2
2 / 57
Institutional spheres
X-matrix
Y-matrix
Economy
Redistribution economy
Market economy
Politics
Unitary political structure
Federated political structure
Ideology
Communitarian ideology
Individualist ideology
Europe, North America,
Domination
Russia, Asia, Latin America
Australia, New Zealand
Redistributive economy is a term by K. Polanyi, characterized by X-inefficiency of H. Leibenstein. For
complete description see Appendix.
MULTIDIMENSIONAL MATRIX
All statistical information on the behavior of economic sectors, political and social processes is indexed,
analyzed and approximated using technologies for analyzing the truth of information and multidi-
mensional analysis and synthesis tables of the Futurum Foundation http://www.fundfuturum.com.
Decision making is carried out by building a four-dimensional map of the financial market, identifying
the most significant links in the sectors of interest. The map coordinates are:
•
time-approximated indicators of economic sectors;
•
indices of geopolitical weight and stability, built on the above principles, the fluctuations
of which are recorded and processed according to monitored press reports and statistical
indicators;
•
the state of society, identified by the share distribution in society of dominant and
complementary social institutions;
•
military-political events.
Whitepaper • Version 1.0 • Feb, 2021
2
3 / 57
MARKET INFLUENCE MAP.
Geopolitical processes
USA
England
England
Russia
Military-political processes
Whitepaper • Version 1.0 • Feb, 2021
2
4 / 57
The given layout of the table is used, on the one hand, to study connections and qualitative logic of the
market at the initial stages of analysis and on the other hand it represents in the schematic form
a four-dimensional vector field, the behavior and connections of which in time and space of the above
coordinates are modeled by vector analysis methods with finding divergence and gradients of the field,
revealing stable trends and critical locations of the market.
SPECIFIC DATA SOURCES
D.AI.SY will harness large datapools that exist and integrate others on its own. For example, The
Research Division of the Federal Reserve Bank of St. Louis and its IDEAS database, provides links
to over 1,200,000 full-text articles. Most contributions are freely downloadable, but copyright remains
with the author or copyright holder. It is among the largest internet repositories of academic material
in the world. This resource is constantly growing as materials to RePEc can be added through a depart-
ment or institutional archive or, if no institutional archive is available, through the Munich Personal
RePEc Archive. Institutions are welcome to join and contribute their materials by establishing and
maintaining their own RePEc archive.
Additionally, leading publishers, such as Elsevier and Springer, have their economics material listed
in RePEc. RePEc collaborates with the American Economic Association’s EconLit database to provide
content from leading universities’ working papers or preprint series to EconLit. Over 1500 journals
and over 3300 working paper series have registered, for a total of over 1.2 million articles, the majority
of which are online.
D.AI.SY DATA COLLECTION
AND VERIFICATION METHODOLOGY
While D.AI.SY uses third party data, we have a strict data verification process to verify, clean, normal-
ize and complete missing data using AI.
Certainly no data set can be perfectly complete. There will be data challenges: there will be missing
data, there will be incorrect data, redundant data, non quantifiable data. We will use the internet
as a main resource to get additional data. Crawlers in social media for sentiment analysis, companies
reports, news, etc all will be pulled into buckets to verify and complete the data. Initially, data can
be structured, non-structured, statistical, or streaming.
In practice, the quality of the final models depends much more on the quality of the prepared data
than on the choice of the model itself and its optimization. Therefore, after collecting data, it is nec-
essary to perform data preprocessing. This process includes data normalization (if necessary) as well
as feature extraction and feature selection.
Normalizing numeric data involves bringing parameter values to a specific numeric range, for exam-
ple, taking a logarithm or a trigonometric function. Bollean parameters can be converted using one-
hot encoding. In this case, each feature is represented by a binary vector. Missing or invalid data can
be removed from the dataset or replaced with a row averaged and closest top or bottom. But of course,
there are cases that we need to take care with extra-caution not over generalising.
Whitepaper • Version 1.0 • Feb, 2021
2
5 / 57
It is also important to make a careful selection of significant exogenous parameters before modeling.
This will significantly improve the accuracy of the forecast. Insignificant variables introduce only
noise into the model, having almost no effect on the result. And when there are enough of them, which
is assumed in our system, it is necessary to separate the wheat from the chaff.
In practice, this task arises due to the fact that at the time of data collection, experts do not yet know
which variables will be most significant in the analysis. For the development of D.AI.SY, the collection
of economic, political, social, and natural data will be carried out and the selection of relevant informa-
tion is important. Significance analysis can be performed using correlation analysis, multicollinearity
tests using the t-test, or ridge regression. Feature extraction and feature selection techniques can also
be used in the modeling process if the model does not meet the required quality level.
D.AI.SY MAPPING MOVEMENTS AND MONEY FLOWS
Then, the second level dynamics are assessed: how this Money Flows across the assets, owners and
the world. To model such flows, one needs again to face multiple dimensions of the information.
On one side there are different models for each type of the flow including it’s own time periods and
connections, from other all of the supporting macro and micro economic data that affects typical and
abnormal money movements.
When selecting key factors affecting economic processes and the movement of money, we focus
on: historical economical development of countries and regions, level of political stability, ecological
situation, epidemiological situation, features of internal consumption, mentality and traditions.
While these flows and changes are enormous, their impact on the markets can be quantified and
modeled to surface what is truly impactful. To do so, it’s necessary to capture the changes in type and
direction of assets and capture the suspected contributing factors in a data-based fashion – as outlined
in D.AI.SY’s Deep Reasoning
D.AI.SY AI DEEP REASONING
As we described it before, D.AI.SY data pools will be collected, verified and stored in multidimensional
structures, to be further modeled using AI networks. To craft and test various deep-reasoning hypoth-
eses to understand the relations between various fundamental shifts and money movements we will
use motifs to model interplay between flows, pools and contributing factors, hopefully following
Pareto rules.
For example, when we model money flow in certain countries, probably the main 80% movements
will be based on the same model (motifs) as in other countries, and the remaining 20% will be guided
by specific models.
Whitepaper • Version 1.0 • Feb, 2021
2
6 / 57
Of course, there are a huge number of external factors that comprehensively affect the money flows
and determine its behavior. For a complete analysis of these impacts, a special method of multidimen-
sional classification and structuring of factors influences to be used.
The state of the market in various spatial and temporal locations is predicted as a probabilistic super-
position of these influences using multidimensional logical truth-falsity complexes.
Important to mention, that we rely on a dialectical approach, one of the expressions of which in math-
ematical formulations is the well-known Gödel incompleteness theorem. In its most general form, the
idea is that it is impossible to obtain a reliable all-encompassing representation about the system from
itself; a systematic dialectical leap is necessary.
D.AI.SY AI VOLATILITY MODELS
1.
GOALS OF VOLATILITY MODEL
The goal of D.AI.SY’s AI modelling is to uncover volatility. Machine Learning will focus on the money
flows within/into/and out of exchanges – both centralized and decentralized.
While volatility will be the primary modelling interest, there will be additional learnings that will
serve to sharpen modelling. For example, by-products such as increase or decrease in open interest,
increased sentiment of interest. funds flowing into emerging assets/exchanges, or renewed interest
in existing markets, will all serve to further refine model dynamics and improve probability in our
technical analysis models.
2.
BUILD – CREATE ASSEMBLY OF NETWORK COMPONENTS
Given the complexity of modelling both money maps and flows, and then create multidimensional
structures and models to extract the information about volatility – it is important to build ensemble
models. Ensemble models are machine learning paradigms where multiple models (often called
“weak learners”) are trained to solve the same problem and combine to produce better results.
The primary hypothesis for D.AI.SY’s modelling is that with the right mix of weak models, it is
possible to create more accurate, reliable models.
In practice, such ensemble approaches have already demonstrated high prediction accuracy thanks
to three strategies:
1.
Bagging.
Where weak homogeneous models are trained in parallel and independently of each other, and
then they are combined, following some deterministic averaging process.
2.
Boosting.
Where weak homogeneous models are trained in a sequential adaptive way (the model depends
on the previous ones) and then combined following a deterministic strategy.
Whitepaper • Version 1.0 • Feb, 2021
2
7 / 57
3.
Staking.
Where weak homogeneous models learn and combine them, training the metamodel to produce
a prediction based on the predictions of various weak models.
D.AI.SY also leverages cluster analysis which studies the data structure to find the hidden relation-
ships and patterns within and between elements. In D.AI.SY’s case, clustering will be used in every
element of modelling.
3.
UNDERSTAND – INTRODUCE DEEP REASONING FOR VOLATILITY
While modeling is the foundational step, it is not sufficient. To achieve the machine learning goal,
models must understand the interplay between elements that relate to volatility.
Deep reasoning must perform automated reasoning between a myriad of factors. This is much more
complex than deep learning – that more simply seeks patterns. Deep reason models provide an effec-
tive integration of learning and reasoning mechanisms is a long-standing research problem at the
intersection of many different areas.
D.AI.SY’s deep reasoning approach will focus on understanding the natural intelligence between
elements and the adoption of suitable mathematical means for rigorous modeling.
Deep reasoning is relational and evolutionary as if it’s a ‘cognitive process’ of knowledge and behavior
acquisition. D.AI.SY’s deep reasoning will fall into five categories:
•
object identification,
•
cluster classification,
•
functional regression,
•
behavior generation,
•
knowledge acquisition.
4.
NARROW – REDUCTION OF DIMENSIONS (FACTORS)
Through the preceding stages, data and variables increase logarithmically. In order to bring the data
towards models that can provide relevant modelling, it is critical to understand what are the dimen-
sions and their importance. In this process, the number of dimensions are reduced to enable deeper
reasoning to thrive on more informative data sets.
Through cluster analysis methods, D.AI.SY will actively reduce the dimension of the problem. D.AI.
SY intelligence is based on different approaches to learning. Since there is a lot of accumulated
experience in the world, it makes sense to use supervised machine learning algorithms to extract
the rules. Some positive results of using this approach for predicting market volatility, crisis
and abnormal situations have been confirmed by a number of studies and presented in scientific
publications.
Also, to solve clustering problems and in the absence of labeled data, it is advisable to form
knowledge on unsupervised learning algorithms.
Analysis of scientific publications in recent years shows that machine learning models such
as support vector machines, random forest, convolutional neural networks and recurrent neural
Whitepaper • Version 1.0 • Feb, 2021
2
8 / 57
networks are used to solve problems similar to D.AI.SY problems, as well as their sometime
joint execution. To a greater extent, recurrent networks of the LSTM type are used to analyze
time series data, and their success is due to the fact that they are architecturally able to relate
previous information to the current task and are capable of learning long-term additions.
5.
CREATE – ARCHITECTURAL FRAMEWORK (DONE)
To ensure effective modeling and functioning of D.AI.SY, it is necessary to use computers with
parallel architecture CUDA, which has been successfully implemented in graphics accelerators
from NVIDA. Since the volumes of analyzed information are large enough, the multiprocessor
power of the graphics processors is justified. Thus, researches in volatility forecasting and crisis
situations are being actively conducted using various methods and techniques. All of them are
largely oriented toward specific markets and are limited in time and space, which is a result of
insufficient accuracy and limited range of application.
6.
REFINE – MODEL CHALLENGES AND SOLUTIONS
Through modelling, there are known obstacles that will be encountered. Below is a short outline of the
challenges and D.AI.SY’s scientific team’s approach to solving them.
Overfitting While the main goal of the project is to incorporate every available piece of financial
information to make a final decision, it inherently leads us to “overfitting” the problem. We cannot
put all the information into the Network and hope that it will use Machine Learning to learn
what explains various volatility, because there is significantly LESS specific output events than
inputs that can explain it. Simply put, there might be one or two completely non-related inputs that
perfectly correlate with volatility, and the model will end up resulting in a useless model.
Curse of Dimensionality In the process of modeling, the problem of “Curse of Dimensionality”
may arise. This problem is associated with an exponential increase in the amount of data due
to an increase in the number of dimensions of the investigated space. Given D.AI.SY incorporates
data from all over the world – this issue is anticipated. Complex dependencies and volumes of
computation may not give the required productivity. To address this, D.AI.SY will move on to
solving the dimensionality reduction through clustering. To do this, the researcher can use the
method of principal components, discriminant analysis, matrix factorization or their modifications.
D.AI.SY ALGORITHMIC TRADING
To bring research to reality, and to empower algorithmic trading to systematically harness these fun-
damental signals and enable retail investors to act on volatility, we incorporate through signals into
existing EndoTech’s end-to-end solution elaborated in next chapter.
Whitepaper • Version 1.0 • Feb, 2021
2
9 / 57
Endotech’s Trading
Framework
Investment decisions clearly rely on a tremendous interplay between science and art. What is often not
as evident, are the other aspects that enable those trades and investments to be executed profitably
and effectively.
Without advanced execution advances, even the best-informed trading ideas are impotent. To capital-
ize on competitive advantages in decision making, it is critical to have a well-informed, technologically
robust trading methodology and trading capacity.
ENDOTECH TRADING METHODOLOGY
Advanced trading methodology must consider numerous competing interests simultaneously. Even
novice investors come to understand the competitive advantage of automated execution as it system-
atically manages various parameters in ways well-beyond human capacity.
The burden of managing multiple assets, volatility, geographies, human emotions all while juggling
evolving realities of portfolio management, money management, order execution, risk control and
more is impossible.
To address these challenges there are important executional capabilities that must be adopted:
1.
PORTFOLIO MANAGEMENT
•
Active rebalancing is key. Buy and Hold is not a valid approach in such a young market,
like Crypto markets, when the “power struggle” is still very high, technologies are still
immature, any crypto-currency can quickly lose its value to another player. Monitoring
and keeping currencies in a manner that allows you to properly rebalance your
portfolio and minimize risk/maximize return (this means keeping them on a reputable
exchange, or in a wallet that allows you to quickly transact currencies – more on security
in a future post).
•
Rely on trading strategies to add assets to the portfolio to increase or decrease the
share of an individual holding. These strategies are carefully chosen to provide adaptive
technology for quickly-changing markets.
Whitepaper • Version 1.0 • Feb, 2021
3
0 / 57
2.
CHOOSING ASSETS
D.AI.SY volatility plays the main role in choosing currently (or predicted) volatile assets to enable
technical analysis strategies. Further assessment of investor’s preferences is a function of the
choice of assets to be included in the portfolio. The decision is made after evaluating how much
each asset contributes to the opportunity, how liquid it is, and how risky it is.
3.
CHOOSING APPROPRIATE ET SIGNALS
Given the high volatility of the crypto market, the buy-and-hold strategy is rarely the winning
one; for now, at least. Using AI/ML algorithms to identify patterns, we can spot which assets are
breaking out and when, their time frames, and their correlation. Everything to determine what
the optimal return targets are, whilst simultaneously minimizing losses.
4.
CHOOSING ET ALLOCATION
In choosing how much capital to allocate to each portfolio asset, in order to get the risk/return
combination in line with the investor preferences, the main instrument available in finance is the
Capital Asset Pricing Model (CAPM).
This model, used in traditional markets for more than fifty years, simply expresses the expected
return on a security or portfolio as a function of the market risk premium, reduced by a factor
(β) that represents the risk/return ratio.
Despite its age, the CAPM has proved itself a valuable tool in the optimization of portfolios
on the crypto space too. In this context, we plan to leverage our know-how and to create dynamic
portfolio weights in order to optimize investment strategies and maximize expected outcomes.
5.
EXECUTION AND CAPACITY
Being a young market, Crypto involves a lot of uncertainty at the execution stage. Where to execute
orders and how to do it? In ENDOTECH we put emphasize on fully automated execution both
from technological and monitoring perspective. Being connected to multiple exchanges, enabling
execution on tenth of thousands of accounts, managing significant amounts by the strategies
requires an massive infrastructure that ENDOTECH built for algorithmic trading.
Fortunately, the situation for the investor is less complicated than for the trader. Nevertheless,
it remains a delicate topic, because if the investor cannot execute the order or the exchange
is hacked, or files for bankruptcy, no matter how well the portfolio is managed, the investor
carries 100% of the execution risk.
We also caution investors about spreads and slippage — on some exchanges bots benefit from
market transactions, and while a few percentage points on the spread doesn’t seem too high,
during choppy markets when transactions may take place weekly, it can kill a portfolio’s profits
(see Capacity chapter).
6.
CHOOSING ET MONEY MANAGEMENT
Money management is the mathematical process of increasing and decreasing the number
of contracts/shares/options. The purpose of utilizing money management should be to increase
the profitability during positive runs and protect those profits during losses.
Whitepaper • Version 1.0 • Feb, 2021
3
1 / 57
Money management is represented by the ensemble of decisions and strategies regarding how
to reinvest the profits or to handle losses. A common opinion is that money management really
has the capability of moving the needle for investment return.
There are many aspects and methods — compound vs fixed, active capital %: fixed dollar amount,
percent at risk, trading with optimal F, etc. For example, we believe that in a market like Crypto,
where results can be as high as those available in leveraged markets, most of the models can
be considered as too extreme. Yet, since investor appetite is high, we use 90% of the capital
as active capital, and a compound reinvestment strategy to keep up with Bitcoin returns.
COMPOUND MONEY MANAGEMENT
In triple-H investing (high-risk high-return high-probability), the compounding effect of money
management is crucial.
While there are multiple tools that allow trades and investors to increase their Returns (like
leverage, increasing value of trading capital, different compound strategies), it also increases the
value of Risk. And since Risk is also High – 20 to 40%, by setting leverage to 2.5 one might increase
risk to a value of 100% (=40%*2.5) which is a risk of total loss…
In D.AI.SY it is critical to consider all available tools to increase returns, without significantly
affecting risks (percentage wise, not dollar wise) and without changing probability. However,
it is noteworthy that the simplest and main tool for increasing returns is compound effect of the
investment.
7.
CONCLUSION
While decision making has received much consideration, there are enormous technological
advances in the execution of trades. To efficiently capitalize on trading ideas, the fundamental
principles must be applied scientifically to surface the trading potential. We, at ENDOTECH have
developed an extensive proprietary solution that we use since 2017.
From executional excellence, we now turn our attention to a often-forgotten consideration:
Trading Capacity.
TRADING CAPACITY
Trading capacity reflects the potential market for any investment or strategy. It can serve as the limiter
to the otherwise boundless opportunities identified in trading decisioning. But, trading capacity can
evaporate even the best trades and their execution if there is no trading capacity.
The main concern is that certain algorithms work in certain trading niche areas, and inherently will
experience a bottleneck that when reached becomes less profitable.
Whitepaper • Version 1.0 • Feb, 2021
3
2 / 57
As an example, within high-frequency or arbitrage style trading, the bottleneck is the number of short-
term opportunities within both the market and the available volume of level 2 books. Once participants
close the spreads and differences between exchanges, the profitability goes down significantly.
Capacity is an inevitable part of the process for any scaled trading solution. Our methodology is to know
up front where and when such capacity issues can hit. To have solutions prepared in advance, to avoid
surprises.
With growing AUM – it is important to understand and navigate the trading capacities for each asset
class. Practically speaking, ongoing trading success relies on actively managing the capacity and
diversifying trading across asset classes including Crypto, Forex, Commodities, ETFs.
Specifically, D.AI.SY’s trading capacity on Crypto market is $480M and delivers a profit of 237%. It is
marginally less than its current 250% on a $50M AUM.
Whitepaper • Version 1.0 • Feb, 2021
3
3 / 57
ENDOTECH’S ALGORITHMIC TRADING INFRASTRUCTURE
Mobile/PC
UI Gateway
Account
Portfolio
Management
Management
Wallet
Crypto
Fiat
Drill Down Portfolio Module
Portfolio Management
Money
Asset Selector
Portfolio Allocation
Signal/Strategies
Management
Portfolio Trading
Order Execution
Smart Order
Portfolio Orders
Quote Engine
Prediction
Execution
Exchange Connectivity
Exchange1
Exchange 2
Exchange N
Whitepaper • Version 1.0 • Feb, 2021
3
4 / 57
Technology
and Architecture
The Data Analysis and Machine Learning Pipeline architecture enables effectively collecting and
storing different formats of data, received from various sources. Additionally, it supports a unified
data analysis and machine learning workflow.
ARCHITECTURAL CONCEPT
The architecture solution is based on AWS technological stack and contains three fully integrated
modules:
•
Ingest & Store – Enables real-time data ingestion from various data sources (including data
uploaded by the user) in real-time data storage on a data lake. This functionality is specifically
tailored for situations where there is a need for storing and organizing large amounts of real-
time data on a data lake.
•
Analyze & Prepare – Provides scheduling and orchestrating ETL and data analysis workflows
with data that is stored on a data lake.
•
Learn – Provides scheduling and orchestrating Machine Learning workflows with datasets
prepared on the previous stage, as well as training and deployment of different ML models.
Eventually enables getting real-time predictions from the trained and deployed model.
INGEST & STORE
In this module, data is ingested from various sources or sample data uploaded into an S3 bucket.
Whitepaper • Version 1.0 • Feb, 2021
3
5 / 57
1.
The following options are supported for data ingestion:
•
Custom proxy application which runs on Amazon Elastic Cluster Service REST API using
Amazon API Gateway. API Gateway is a fully managed service that makes it easy to create,
publish, maintain, monitor, and secure APIs at any scale. It also provides tools for creating and
documenting web APIs that route HTTP requests to Lambda functions.
•
Data upload – mainly used to test the streaming capability of the architecture. In this case,
a user uploads a sample CSV data into Amazon S3 bucket. Uploading the data triggers an AWS
Lambda function. When the Lambda function is triggered, it reads the data.
2.
The data sent in streams to Amazon Kinesis Data Streams. Kinesis Data Streams is a massively
scalable and durable real-time data streaming service. Alternatively, the data can then
be streamed via Kinesis Data Streams.
3.
The Kinesis streaming data is then automatically consumed by Amazon Kinesis Data Firehose.
Kinesis Data Firehose loads streaming data into data lakes, data stores, and analytics services.
It’s a fully managed service that automatically scales to match the throughput of your data
and requires no ongoing administration. Data captured by this service can optionally
be transformed and stored into an S3 bucket as an intermediate process.
4.
The stream of data in the S3 bucket is loaded into an Amazon Redshift cluster and stored
in a database. Amazon Redshift is a fully managed, petabyte-scale data warehouse service. The
data warehouse is a collection of computing resources called nodes, which are organized into
a group called a cluster. Each cluster runs an Amazon Redshift engine and contains one or more
databases.
ANALYZE & PREPARE
In this module, data analysis events are scheduled using EventBridge. It is a serverless event bus, which
allows it to regularly run the ETL data analysis & normalizing pipeline.
Step Functions and Lambda functions used to orchestrate the data querying workflows. The archi-
tecture triggers and controls an AWS Batch job to run SQL queries on the data lake using Amazon
Redshift. The results of the queries are stored in the dedicated S3 bucket. Those results are normalized
datasets which used to keep the Machine Learning model up to date.
Additionally Apache Spark might be used for data analysis. It is an open source, Big Data framework
that is part of the Hadoop ecosystem. Many computational tasks are implemented in Apache Spark
much faster due to recurrent in-memory processing. The ability to repeatedly access data loaded into
memory allows us to efficiently work with machine learning algorithms. It is supported by the Python
programming language, which has long been effectively used to solve machine learning problems.
Python might be used to model the behavior of financial markets.
Whitepaper • Version 1.0 • Feb, 2021
3
6 / 57
LEARN
This module fully orchestrates different Machine Learning workflows.
1.
Live stream data preprocessing using Amazon SageMaker Processing service. It processes
the raw extract, transform, and load (ETL) data and makes it ingestible by Machine Learning
models. It launches a processing container, pulls the query results from the S3 bucket, and runs
a custom preprocessing script to perform tasks such as feature engineering, data validation,
train/test split, and more. The output is then stored in the dedicated S3 bucket.
2.
Training and deploying Machine Learning models using Amazon SageMaker solution.
It launches a Machine Learning training job to train on the preprocessed and transformed data,
and then store the model artifacts in Amazon S3. It then deploys the best model trained via
an automatic ML approach on an Amazon SageMaker endpoint.
3.
REST API using Amazon API Gateway with Lambda integration. It allows for real-time inference
on the deployed model. The Lambda function accepts user input via the REST API and API
Gateway, converts the input, and communicates with the Amazon SageMaker endpoint
to obtain predictions from the trained model.
Whitepaper • Version 1.0 • Feb, 2021
3
7 / 57
Roadmap
The project is planned for six quarters starting with the launch of the crowdfunding process on Janu-
ary 10th, 2021. The preparatory phase is due mid-February.
2021
PQ1
JAN
FEB
Crowdfunding
Whitepaper, Team
system
PQ4
PQ3
PQ2
OCT
JUL
APR
Commodities,
Forex market
Core App
Ultra ETF testing
testing
Development
2022
PQ5
PQ6
PQ7
JAN
APR
JUL
Alpha D.AI.SY
Beta D.AI.SY
D.AI.SY product
Whitepaper • Version 1.0 • Feb, 2021
3
8 / 57
D.AI.SY Project Team
Dr. Anna Becker
Dmitry Gooshchin
CEO and Co-Founder
COO and Co-Founder
PhD in Artificial Intelligence from the Technion
MSc in Astrophysics from Tel Aviv University.
Institute. Has founded and successfully sold the
Over 15 years of servicing fintech companies,
fintech software company Strategy Runner.
owns a business optimization agency. Chess
Author of the textbook on Bayesian networks
Grandmaster, patents in wireless technologies.
(ASIN: B005X5AYC6).
R&D Team
QA Team
Led by Taras Dorozhovets
Led by Lana Steshenko
Financial Analysts
Trading Desk
Led by Dr. Greta Tovarovski
Led by Ilya Zolotykh
Operation Management
Product Management
Led by Adam Rubin
Led by Margo Zolotykh
Whitepaper • Version 1.0 • Feb, 2021
3
9 / 57
References
1.
Vidmant O.S. Forecasting the volatility of Financial Time Series by Tree Ensembles. The world
of new economy. 2018;12(3):82-89. https://doi.org/10.26794/2220-6469-2018-12-3-82-89
2.
Rongjun Yang, Lin Yu, Yuanjun Zhao, Hongxin Yu, Guiping Xu, Yiting Wu, Zhengkai Liu,
Big data analytics for financial Market volatility forecast based on support vector machine,
International Journal of Information Management, Volume 50, 2020, Pages 452-462, ISSN 0268-
3.
4.
5.
Aristeidis Samitas, Elias Kampouris, Dimitris Kenourgios, Machine learning as an early warning
system to predict financial crisis, International Review of Financial Analysis, Volume 71, 2020,
101507, ISSN 1057-5219, https://doi.org/10.1016/j.irfa.2020.101507.
6.
Peiwan Wang, Lu Zong, and Yurun Yang. 2020. Predicting Chinese Bond Market Turbulences:
Attention-BiLSTM Based Early Warning System. In Proceedings of the 2020 2nd International
Conference on Big Data Engineering (BDE 2020). Association for Computing Machinery, New
York, NY, USA, 91-104. DOI:https://doi.org/10.1145/3404512.3404521
7.
Škare, M. and Porada-Rochoń, M. 2020. Forecasting financial cycles: can big data help?.
Technological and Economic Development of Economy. 26, 5 (Aug. 2020), 974-988. DOI:https://
doi.org/10.3846/tede.2020.12702.
8.
9.
10.
as-retail-investors-dive-in-bet-on-post-virus-bounce-idUKL4N2BK1LQ
11.
irish_fund_industry_association.pdf
12.
vozmozhnostey/
13.
14.
15.
16.
17.
report-may-2020.pdf
18.
Whitepaper • Version 1.0 • Feb, 2021
4
0 / 57
19.
20.
21.
fondov
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
energy-transition-potential-impacts-from-decarbonization-in-the-european-power-sector
34.
35.
FAE3FDF1D69C6B0704EEC81B617B706A
36.
companies.html
37.
38.
39.
40.
41.
solutions/Ready_for_AI-Solution_Overview.pdf
42.
methodology-january-30-2018_4_0.pdf
43.
44.
html
45.
46.
47.
Whitepaper • Version 1.0 • Feb, 2021
4
1 / 57
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
ChosenLang=en
69.
70.
71.
72.
73.
74.
75.
pdf/588470PUB0Weal101public10BOX353816B.pdf
Whitepaper • Version 1.0 • Feb, 2021
4
2 / 57
Appendix
RISK RETURN DEFINED
D.AI.SY’s high-risk high-return approach to algorithmic investments requires an understanding
of some key concepts defined:
•
Return – is an expected outcome. The focus of this discussion is about the expected VALUE
OF PROFIT,
•
Risk – is a VALUE OF LOSS. This is often misunderstood. Here, we look at the value, not
probability of the loss. You might lose a certain percentage of your capital, you may lose
it, or you most probably lose this risk-part of your capital,
•
Probability – this is a third concept that is often unexamined. In every model, every physical
event, we know what is the probability for it to happen. It seems easy for us to nail three
values – 0, 100, and 50. We say: it NEVER happens, it ALWAYS happens and it’s 50/50 meaning
maybe yes, or maybe not. In high-risk high-return investments we are targeting 80% probability,
meaning that
Algorithmic Investment
Initial investment
$1
High-Return
$1-5
High-Loss
$0.4
Profitability of Profit
Probability of 80%
What is your probability of hitting green
So, there is potential for high-return of 100%-500% annual return. There is still a systematic risk
of losing 40% of your capital, and when we talk about risk, there is a market/exchange/leverage and
other risks that combined might result in 100% loss of your capital.
Whitepaper • Version 1.0 • Feb, 2021
4
3 / 57
GLOBAL MARKET ANALYSIS
Further to the global market analysis, you can see specific retail investment trends described below.
ASIA
After the 2002-2003 SARS outbreak, which was mostly confined to Asia and hit Asia’s markets hard-
est – the rebound was sharp. Hong Kong’s Hang Seng index. HSI dropped 18% in about four months,
then surged by almost a third in the second half of 2003. The decline in the 2008/2009 financial
crisis and the recovery both took much longer. The MSCI All-Country index .MIWD00000PUS lost
56% over 10 months, with plenty of false dawns along the way. It was nearly five-and-a-half years
before it regained 2008 levels.
Where Asian indexes have fallen the furthest, interest from retail investors has jumped the most
as they deploy long-held savings or draw down loans to buy shares.
In India, brokerage Zerodha has opened a record 140,000 new accounts in mid-March 2020 – double the
average. The S&P Sensex index. BSESN 30% drop since January makes it one of the world’s worst-per-
forming markets.
CommSec, Australia’s largest retail broker, said account openings had increased fourfold in March
2020. Brokers in Manila, Hong Kong, Bangkok, Tokyo, Kuala Lumpur and Jakarta reported surges
as well.
In Korea, where the benchmark Kospi .KS11 is down by a quarter since January 2020, retail investors
lifted their broker deposits 53% to a record 41 trillion won ($34 billion).
In China, where the number of stock trading accounts rose nearly 12% in February 2020 only, inflows
have swamped funds that facilitate international investment.
In Japan, around 40% of the population are active retail investors. It is especially significant for a coun-
try known for its higher living standards.
EUROPE
Based on the report made by European Securities and Markets Authority (ESMA) in October 2020,
in recent years bank deposits have offered near-zero returns on household savings. The picture
of a dominating banking sector that limits capital markets activity is, however, not homogenous across
individual EU Member States.
While across the EU, the share of households’ financial assets held in investment funds is around
10%, at national level recent figures range from under 1% in Ireland, together with Estonia and Latvia,
to around 16% in Belgium, for example.
The proportion of households that own listed shares goes from around 1% in Estonia, Hungary and
Portugal to 20% in Cyprus. Ireland is around the EU average, with households’ holdings of listed
shares at around 4% .
Whitepaper • Version 1.0 • Feb, 2021
4
4 / 57
At the same time, patterns of participation may also vary in relation to indirect holding of financial
assets by households, through pensions and insurance policies. For example, the rate of the indirect
participation in investment funds varies from under 1% in Greece to over 30% in the Netherlands,
Sweden and Denmark. In Ireland and Germany this rate is above 20%.
LATIN AMERICA
In the past decade and a half, the enrollment rate in higher education in Latin America has increased
to 45%. Even in less developed regions with a more impoverished population, the higher education
enrollment still increased to 25% in 13 years. As a result, people have become increasingly familiar
with economic and financial notions.
Latin America has access to the Internet, and ambitious investors can also educate themselves, which
has also led to an increase in the number of independent stocks and Forex traders. Compared to the
trading regulations of the US market, Latin America is a bit different in terms of trading leverage and
barriers to entry, but still, in the past years, more and more people have been using online resources
to learn how to grow their wealth.
Latin America is made up of more countries, so there are regional economic fluctuations. However,
it appears that all countries have something in common: the median age of the population. According
to the latest data, the median age of Latin American people is 31 years old. This means a higher con-
centration of Millennials and Gen Z, who, statistically, are more inclined towards innovative ideas,
experimentation, and appetite for investment.
RUSSIA
According to the Bank of Russia records, in the third quarter of 2020, brokers acquired another 1.6 mil-
lion individual clients, having set a quarterly growth record. As before, banks providing brokerage
services attracted by far a majority of clients. The number of individual investment accounts (IIAs)
topped 2.9 million, of which 525 thousand were opened in the third quarter alone.
Trust managers may view the third quarter as the year’s best so far. The value of trust assets rose
9.1% in the period, rising upwards of 1.4 trillion rubles. Trust managers were able to acquire some
79.3 thousand clients, all of them individuals. The number of their clients totalled almost 480 thousand.
Algorithmic Hedge Funds
The following represent leading hedge funds that leverage algorithmic solutions to maximize alpha.
Whitepaper • Version 1.0 • Feb, 2021
4
5 / 57
ALGORITHMIC HEDGE FUNDS
The following represent leading hedge funds that leverage algorithmic solutions to maximize alpha.
TOP 5 HEDGE FUNDS
Under
management:
D.E. SHAW & CO
D.E. Shaw & Co are among the pioneers of quantitative
$ 50 billion
trading and investing. They also use discretionary
Country: USA
approaches, but to a lesser extent.The fund ranked 3rd
Founder: David Shaw
in overall profitability among other hedge funds in 2018.
Established: 1988
They use algorithmic strategies in the stock and futures
markets, discretionary trading in the insurance market, the
credit market and the energy market.
TWO SIGMA
Here is what hedge fund founders David Siegel and John
$ 60 billion
Overdeck write about themselves:
Country: USA
Over the past
“We are not classic investment managers. We are
10 years, their
Founders: David
committed to technological innovation in the world
average annual
Siegel, John Overdeck
of finance. Factors such as machine learning and capacity
return has been
Established: 2001
allocation help us. More than 1,700 people believe
30%.
scientific approach is the best investment approach. This
is Two Sigma. “
David Siegel graduated from Princeton and later received
his PhD (Advanced Degree in Science) in Computer
Science from MIT (Massachusetts Institute of Technology).
John Overdeck worked for the D.E. Shaw and Amazon
before joining Two Sigma with David Siegel.
MAN GROUP
James Maine started trading sugar and rum back in the
$ 61 billion
18th century, unaware that in two centuries his company
Country: UK
would grow into one of the largest hedge funds in the
Founder: James Maine
world.
Date of foundation:
Today Man Group uses absolutely all markets, maximally
178
distributing risks across various instruments and
investment strategies.
RENAISSANCE
James Simons is a famous American mathematician who
$ 84 billion
TECHNOLOGIES
deciphered Soviet codes during the Cold War. Until 2009,
Medallion – 60% per
it was he who headed the company.
Country: USA
annum for 30 years
Renaissance Technologies is a quantitative trading
Founder: James Harris
fund whose strategies are based on statistical and
Simons
mathematical analysis. The company has 3 investment
Established: 1982
funds. One of them –
Whitepaper • Version 1.0 • Feb, 2021
4
6 / 57
BRIDGEWATER
Bridgewater’s clients are institutional organizations such
$ 160 billion
ASSOCIATES
as insurance companies, banks, and pension funds. The
The average return
fund uses macro strategies applying fundamental analysis
Country: USA
over the past
and quantitative trading.
10 years is 19%.
Founder: Ray Dalio
In 1997, Winton Group started with $ 1.6 million, attracting
Established: 1975
more and more investments annually.
Basically, the fund uses diversified macro strategies
and trend strategies.In 2020, Bridgewater Pure Alpha
II remained in the red: 18.6%. This is the worst result in the
last decade.
CRYPTO HEDGE FUNDS
The growth in the number of cryptocurrencies set the foundation for the rise of professional invest-
ment management firms. The goal of crypto hedge funds is to serve investors who are looking beyond
the traditional Bitcoin investment. Cryptocurrency hedge funds practice active management where
fund managers curate the portfolio and make capital allocations according to a trading style. Some
funds are more aggressive, while others are more conservative. Investing through crypto hedge funds
is done by investors who don’t have the time or the skills to invest on their own.
The entire cryptocurrency market is only worth a few hundred billion dollars, which is very little com-
pared to the gold market, for example, which is worth more than $ 7 trillion. Or compared to the real
estate market, the largest asset market in the world, which is worth over $ 200 trillion.
What attracts most investors in this market is the potential for huge profits. During 2017 and 2018,
the value of some tokens has skyrocketed in a short time frame. Some cryptocurrency hedge funds
(Pantera Capital) have reported returns of 10,000% or more. By comparison, the average annual return
on the S&P 500 over the past 90 years has been around 10%. The main investors in crypto hedge funds
are “family offices” and wealthy individuals. At the same time, Bitcoin remains the most popular asset
among such structures.
The total volume of assets under management of cryptocurrency hedge funds in 2019 increased from
$ 1 billion to more than $ 2 billion. At the same time, the share of funds that manage assets worth more
than $ 20 million increased from 19% to 35%.
In the Crypto Hedge Fund Report 2020, the specialists used the data they collected during the first
quarter of 2020. According to this information, the average cryptocurrency hedge fund manages assets
worth $ 44 million, which is almost twice as much as a year earlier. The median value also doubled,
from $ 4.3 million to $ 8.2 million.
PwC counted 150 active hedge funds on the market. At the same time, almost two-thirds of them
Whitepaper • Version 1.0 • Feb, 2021
4
7 / 57
were launched either in 2018 or in 2013. Experts noted the correlation between the number of newly
created funds and the value of Bitcoin: the more the largest cryptocurrency is, the higher the activity
of management companies.
TOP 5 CRYPTO HEDGE FUNDS
1.
GALAXY DIGITAL ASSETS FUND
Founded by Michael Novogratz, New York-based Galaxy Digital Assets Fund debuted on the
crypto landscape in 2018. The hybrid hedge fund invests in digital currencies, ICOs, and related
companies. According to CryptoFundList, it launched with $500 million in assets. It had also
planned to raise at least $200 million. In 2018, the fund suffered a loss of $273 million as crypto
prices crashed. The company also lost $68 million in the third quarter of 2019.
2.
ALPHABIT FUND
Domiciled in the Cayman Islands, Alphabit Fund is a hybrid between a hedge fund and an open-
ended mutual fund. It was launched with just $1 million of seed capital. The hedge fund raised
an estimated $300 million from investors in 2017. According to estimates by Crypto Fund Research,
Alphabit has more than $500 million in assets. The fund aims to outperform the price of Bitcoin
while generating lower volatility.
3.
POLYCHAIN CAPITAL
Olaf Carlson-Wee, the former Head of Risk at Coinbase, founded Polychain Capital in 2016. This
crypto hedge fund has raised funds from Andreessen Horowitz and Sequoia Capital. The San
Francisco-based fund had $967 million in assets at the end of June 2019, according to Crypto Fund
Research. The multi-strategy hedge fund invests only in cryptocurrencies, not in companies.
4.
PANTERA CAPITAL
Pantera Capital was launched in 2003 as a traditional investment fund. It shifted its focus
to cryptocurrencies in 2013 with the launch of Pantera Bitcoin Fund. The multi-strategy hedge
fund has an estimated $810 million in assets, according to CryptoFundList. Pantera invests
in cryptocurrencies, ICOs, and blockchain startups. Pantera has the backing of Fortress
Investment Group, Benchmark Capital, and Ribbit Capital.
Whitepaper • Version 1.0 • Feb, 2021
4
8 / 57
5.
GRAYSCALE INVESTMENTS
Grayscale Investments is a subsidiary of New York-based Digital Currency Group. According
to CoinTelegraph, it had a staggering $2.7 billion in assets in 2019. The crypto hedge fund invests
in a wide range of digital assets including Bitcoin, Ethereum, Litecoin, Ripple, and blockchain
companies. Grayscale has invested in several crypto companies including eToro, Circle, Ledger,
Shapeshift, and more.
Analysts noted that cryptocurrency hedge funds are most often registered in the same jurisdictions
as traditional ones: most of the organizations are registered in the Cayman Islands (42%), in the USA
(38%) and in the British Virgin Islands (8%). At the same time, more than half of the management
companies operate from the United States (52%), and the UK is in second most popular place (15%).
On average, each hedge fund serves 28 investors. The main contributors are either “family offices”
(48%) or individuals with a high level of income (42%). The average investment is $ 3.1 million, and the
median is $ 0.3 million.
GLOBAL MONEY MAP
Example of a journey: American astrophysicist Gregory P. Laughlin (Santa Cruz, California) announced
that planet Earth is estimated at $5 quadrillions (December 2020, Daily Mail). Fortunately, there
is no one to whom he can sell the planet. But he still can evaluate it. So we start our drill down into
approximate cost of Earth as an economical system – from natural resources to intellectual property
products.
Land and water resources are a unique type of resources that exist on its own, not as a result of human
activity.
Whitepaper • Version 1.0 • Feb, 2021
4
9 / 57
LAND RESOURCES
Square,
Significance for agriculture,
Category
million
%
timber and paper industries
km2
High
Forests
40,3
27
High
Meadows and pastures
28,5
19
High
Agricultural area
19
13
Neutral
Industry and human settlements
3
2
Neutral
Freshwater reservoirs
3,2
2,1
Low
Unusable lands
4,5
3
Low
Glaciers
16,3
11
Low
Polar and high-altitude deserts
5
3,3
Low
Tundra
7
4,7
Low
Swamps
4
2,7
Low
Deserts, sands, rocks
18,2
12,2
Land, total:
149
100
Green: categories with important significance for agriculture, timber and paper industries.
Red: categories without important economical significance.
FOR SEPTEMBER 2018, WORLDWIDE:
Lands and real estate
Estimation, $trln
Agricultural land, forestry sector lands
27,1
Residential real estate
220,6
Commercial real estate
33,3
Total
281 (+6% for 2018)
Whitepaper • Version 1.0 • Feb, 2021
5
0 / 57
MARKETS WITH MAXIMUM PRICES OF COMMERCIAL REAL ESTATE, SEPTEMBER 2018:
Rank
Country
Estimation, $trln
1
USA
8,1
2
China
3,6
3
Japan
2,8
4
Germany
1,7
5
UK
1,7
Total:
17,9
UP TO THE COVID-19 EPIDEMIC, LAND IN MAJOR CITIES OF THE WORLD, INTENDED FOR ELITE
BUILDINGS AND TOURIST BUSINESS FACILITIES, REACHED THE MAXIMUM VALUE:
Country
City
Price, 000$/m2
1
Monaco
Monte Carlo
47,578
2
Moscow
Russian Federation
20,853
3
London
UK
20,756
4
Tokio
Japan
17,998
5
Hong Kong
16,125
6
New York
USA
14,898
7
Paris
France
12,122
8
Singapore
9, 701
9
Roma
Italy
9, 166
10
Mumbai
India
9, 163
MINERAL RESOURCES
The most significant groups of mineral resources are two following ones:
•
energy feedstock (oil, natural gas, coal, uranium),
•
precious metals (at first gold).
Whitepaper • Version 1.0 • Feb, 2021
5
1 / 57
OIL
Proved oil reserves
by
regions
and the
most significant countries, bill.barr,
$trln:
In prices
% of total
Region/Country
12.1999
12.2009
12.2018
12.2019
02.2021,
reserves
$trln
N.America
232.8
217.8
245.5
244.4
14.1
14.16
Canada
181.6
175.0
170.8
169.7
9.8
9.83
S.&C.America
95.9
233.3
324.7
324.1
18.7
18.78
Venezuela
76.8
211.2
303.8
303.8
17.5
17.6
Europe
20.7
14.0
14.6
14.4
0.8
0.83
Norway
10.9
7.1
8.6
8.5
0.5
0.49
CIS
120.1
144.0
145.7
145.7
8.4
8.44
Russian Federation
112.1
105.6
107.2
107.2
6.2
6.21
Kazakhstan
5.4
30.0
30.0
30.0
1.7
1.74
M.East
685.8
753.1
833.9
833.8
48.1
48.3
Saudi Arabia
262.8
264.6
297.7
297.6
17.2
17.24
Iran
93.1
137.0
155.6
155.6
9.0
9.01
Iraq
112.5
115.0
145.0
145.0
8.4
8.4
Kuwait
96.5
101.5
101.5
101.5
5.9
5.88
UAE
97.8
97.8
97.8
97.8
5.6
5.67
Africa
84.7
123.0
125.7
125.7
7.2
7.28
(on 03.02.2021, barrel of oil = $57.93)
An example of the influence of political instability factor on the development of the industry: Vene-
zuela – maximum crude oil reserves in the world, political instability, US sanctions.
DYNAMICS OF OIL AND PETROLEUM PRODUCTS PRODUCTION IN VENEZUELA, MILL.T:
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
155.9
145.8
141.5
139.3
137.8
138.5
135.4
121.0
107.6
75.6
46.6
Whitepaper • Version 1.0 • Feb, 2021
5
2 / 57
An example of the influence of internal consumption factor on the development of the industry: India –
actively develops oil mining and processing, but because of highest internal consumption it provides
itself with oil products by less than 25%, not to mention turning oil into a source of income.
DYNAMICS OF OIL AND PETROLEUM PRODUCTS PRODUCTION AND CONSUMPTION IN INDIA, THOUS.
OF BARRELS PER DAY:
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
Prod
838
901
937
926
905
893
874
885
869
869
826
Cons
3298
3378
3542
3740
3781
3906
4230
4632
4860
5112
5271
PRECIOUS METALS
Gold is not the most expensive precious metal but traditionally always in great demand. Gold reserves
worldwide:
mt
$trln
Above the ground (mined for the entire
187-195 thousands (estimations
10,7-11,2
history)
of Australian Government and of World
Gold Council (WGC))
Under the ground (estimation of United
appr. 50 thousands*
2,87
States Geological Survey (USGS))
*When we focus on modern times of gold mining (appr. 3500 mt/year), the under-ground gold reserves
will last for about 15 years. If no new gold fields are discovered and no new technologies are developed
in industries using gold, a precedent situation will be created.
TOP-10 COUNTRIES WITH MAXIMUM GOLD RESERVES, 2020:
Whitepaper • Version 1.0 • Feb, 2021
5
3 / 57
Share in total volume of national
Country
Gold reserve, mt
reserves, %
USA
8133,5
79,9
Germany
3362,4
77,1
Italy
2451,8
72,9
France
2436,1
67,8
Russian Federation
2299,4
24,5
China
1948,3
27,9
Switzerland
1040,0
6,8
Japan
765,2
3,5
India
664,2
7,8
Netherlands
612,5
70,2
Gold production consists of mining (67-72%) and recycling (28-33%).
MINING VOLUME BY COUNTRY, DATA FOR YEAR 2019, PRICE 03.02.2021 = 1832.16 USD/OZ =
58.9 MILL USD/T:
Country
Mining volume, mt
Estimation, billion USD
China
383,2
22,57
Russian Federation
329,5
19,4
Australia
325,1
19,15
USA
200,2
11,79
Canada
182,9
10,77
Peru
143,3
8,44
Ghana
142,4
8,39
South Africa (RSA)
118,2
6,96
Mexico
111,4
6,56
Brazil
106,9
6,3
We have to consider that rich mining fields in third world countries usually belong to foreign mining
Whitepaper • Version 1.0 • Feb, 2021
5
4 / 57
companies (in total or in part, in different forms). That’s why profitability of these fields in their loca-
tion countries noticeably reduces.
EXAMPLE: GOLD FIELDS IN GHANA:
Gold Field
Mining company
Country-owner
Tarkwa, Damang
Gold Fields
South Africa (RSA)
Wassa
Golden Star Resources
Canada
Obuasi
AngloGold Ashanti
South Africa (RSA)
MONEY (FINANCE ACTIVES):
Agregat
Components
Estimation, $trln
bank’s cash, cash in circulation
М0
5
(157 currencies)
maximum liquid assets (funds
М1 – М0
on credit cards, demand deposits,
23,6
traveller’s checks)
М2 – М1 – М0
short-term deposits
31,5
long-term deposits (including
М3 – М2 – М1 – М0
20,8
pension), government loan bonds
Total (М3)
80,9
The total amount of global debt is $215 trillion, 35 % of the amount was formed after worldwide crisis
at 2008. Global sovereign debt is $60 trillion, 24 % is debt of EU countries, 30 % is debt of USA.
POLITICS
For geopolitical analysis the following classical ideas, principles and concepts of geopolitical science
were used.
T. Mahan concept that “Control over the seas is the central link through which countries accumulate
wealth.” The main work – “The Influence of Sea Power Upon History, 1660-1783” (1890).
The theory of “Heartland” H.J. Mackinder. Mackinder called the Heartland the central part of Eurasia,
Whitepaper • Version 1.0 • Feb, 2021
5
5 / 57
around which the inner arc (Europe – Arabia – Indochina) and the peripheral arc (America – Africa –
Oceania) are located. He summarised his theory thus: “Who rules East Europe commands the Heart-
land; who rules the Heartland commands the World-Island; who rules the World-Island commands
the world”. Mackinder, H.J. Democratic Ideals and Reality. New York: Holt, 1919.
Analytical method of N.J. Speakman. In his works “America’s Strategy in World Politics” (1942) and “
The Geography of the Peace “ (1944), he puts forward the following criteria for the geopolitical power
of the state: the surface of the territory, the nature of borders, the volume of population, the presence
or absence of minerals, economic and technological development, financial power, ethnic homogene-
ity, level of social integration, political stability, national spirit.
Works by Z. Brzezinski “The Grand Chessboard: American Primacy and Its Geostrategic Imperatives”
(1997) and “The Choice: Global Domination or Global Leadership” (2004), containing a description
of the alignment of the main world forces and interests, as well as geopolitical forecasts for future.
In order to further search for relevant information, a preliminary classification of the most important
geopolitical centers of the world and the most significant and economically stable countries in them
has been carried out.
•
Countries of the Transatlantic bloc: USA, England, Germany, France
•
Countries of Eastern Europe: Russia
•
Asia-Pacific block: China, South Korea, Japan
•
Indian Ocean Region: India
•
Central Eurasia: Israel, Turkey, United Arab Emirates
•
The most important countries of the Southern Hemisphere: Brazil, South Africa, Australia.
The classification will be refined and indexed according to the methods given above for the economy.
SOCIETY
The dominance of the basic institutions of either the X – or the Y-matrix ensures the integrity, survival
and development of the corresponding type of society. At the same time, complementary institutions
from the matrix of the opposite type play an supporting role, only complementing the institutional
social structure.
In the stable societies, the share of complementary institutions is about a third (30-35%). If this share
is significantly less, the total dominance of basic institutions leads society to crises or stagnation, and
the excessive introduction of complementary institutions leads to social upheavals and revolutions.
In order to analyze the social stability of the region, we have identified eight states of each of the matri-
ces, depending on the activation of certain groups of social institutions. For example, the activation
of the institutions of the market economy in the conditions of the X-matrix leads to a significant revival
Whitepaper • Version 1.0 • Feb, 2021
5
6 / 57
of all spheres of society. This is the situation observed now in China. And attempts to replace a unitary
political system with a federal one in combination with a market economy for X-type matrices lead
to a sharp change in course and total instability, which was observed during the collapse of the USSR
in the 90s of the 20th century.
It is also proposed to use a slightly more subtle classification of institutional matrices, which is being
refined.
1.
Polanyi, K. (2001). The Great Transformation: The Political and Economic Origins of Our
Time, 2nd ed. Foreword by Joseph E. Stiglitz; introduction by Fred Block. Boston: Beacon
Press. ISBN 9780807056431
2.
North, Douglass (1991). “Institutions”. Journal of Economic Perspectives. 5 (1): 97-112. doi:10.1257/
jep.5.1.97
3.
North, Douglass C. (1989). “Institutions and economic growth: An historical
introduction”. World Development. 17 (9): 1319-1332. doi:10.1016/0305-750X(89)90075-2.
4.
Empirical Studies in Institutional Change, Cambridge University Press, 1996 (edited with Lee
Alston & Thrainn Eggertsson) ISBN 0521557437
5.
S Kirdina. From Marxian school of economic thought to system paradigm in economic studies:
The institutional matrices theory, Montenegrin Journal of Economics 8 (2) (2012), 53-71.
6.
S. Kirdina. The Transformation Process in Russia and East European Countries: Institutional
Matrices’ Theory Standpoint. In Institutional and Organizational Dynamics in the Post-Socialist
Transformation, International Conference, January 24-25, 2002, Amiens (France) CRIISEA,
University of Picardie and OEP, University of Marne-la-Vallee. CD 6B 3 0037578/0902.021135
7.
Harvey Leibenstein (1978). General X-efficiency Theory and Economic Development. Oxford
University Press. ISBN 978-0-19-502380-0.
Whitepaper • Version 1.0 • Feb, 2021
5
7 / 57
