Framework | Softwares | AI Research Processes

TagoTrader was invented and developed by Amir Arazi in the period of 2007 to a stable version in Aug 2018.
All development process documents arranged in a very orderly manner, as a research project for a higher purpose.

Jan 2017 is where Amir came up with the AI idea which eventually (in Aug 2018) led him to the answer for the most important question in the entire trading system, then in this year his research project for the algorithm took a completely new turn, and he found himself running and managing couple of massive IT projects, all in rapid development processes for the AI algorithm to evolve as fast as possible.

Amir had to feed the system with a lot of development hours and real-time hybrid data worth hundreds of thousands of dollars in each of those two years (2017-2018).

The TagoTrader Framework and Softwares knowledge base, real-time recorded data & analysis contains more than 30TB of powerful and valuable information, which contains, among other things, a huge number of market situations, based on real positions.

The entire system was developed using real money positions, the system was not developed by back-testing nor by any traditional optimization methods, nothing in the system is based on demo trading nor on virtual results of any kind.

This means that each question, idea, method and any other algorithm in the system was developed based on real money positions in real markets, Everything was tested with real money and only after a real genuine confirmation, the research process progressed.

At no point did Amir agree to install a version of the software on a co-located server in a data-center anywhere, even though Amir have a lot of practical experience in the field - The attitude has always been that a powerful system is not the one that depends on the latest shout in computing, a smart algorithm and database design is much more important for a better overall performance than any tweaking on machine equipment or internet ping and bandwidth.
The TagoTrader Framework and Softwares is based on a trade cycles of millions of (mostly) USD dollars, meaning that even 100 milliseconds will not make it or break it.

The system can connect and work on any financial instrument which traded on an electronic exchange anywhere in the world and is available to the system through an API, including: Futures, Stocks, ETF's, Bonds, Forex and Crypto - providing huge financial management abilities to leverage and increase the managed capital value.

The system on real live capital runtime - only receives information through a very fast and secure TC (Thin-Channel) which is used as the first conduit to submit the real-time data throughout the entire TagoTrader system, while the system itself is not connected to the Internet at all, which will be clarified later on.

The main local server machine is running the TagoTrader Framework which enables the synchronized work of all the softwares together including all computers on the internal network or server farm, managed mainly by the TagoRT code syntax and less by user interface tools, and could run both on Linux and/or Windows based systems across the internal network.

The choice between running the software through a small local cross platforms computer network OR a bigger higher cost server farm, comes mainly for two reasons, either the capital and various financial markets and instruments amounts managed by the system, OR whether or not the system is set on "Action mode" or "Research mode".

In both modes (Action/Research), the system must obtain data by real capital market positions so that it can operate and advance its processes, otherwise it will not work - the fundamental premise of the entire system unites is that it cannot draw conclusions without introducing some capital to the market.




Technologies | Softwares

The TagoTrader softwares uses 6 in house technologies, each software can work completely independently (except TagoCB) to provide conclusions, answers and solutions in real time, plus all softwares can communicate with each other according to the Tago One-Step methodology that will be clarified further below.

6 In-House Technologies:

- Tago DB®
- Tago Random®
- Tago Position Matrix®
- Tago Virtual Exchange Simulator & Analyzer®
- Tago Encrypt®
- Tago Framework®

Technologies constitute exclusive intellectual property of Tago Trading & Investments Strategies LTD, an IT private company founded in 2009 by Amir Arazi.

7 Softwares:

- Tago RT Calculator.
- Tago CR (Calculator & Recorder).
- Tago RT Trader (TagoTrader).
- Tago RM (Risk Manager).
- Tago PM (Portfolio Manager).
- Tago CB (Cycle Builder).
- Tago LA (Liquidity Analyzer).




TagoDB®

TagoDB handle large amounts of data in excess of terabytes with a very fast access rates, supporting data from multiple hybrid sources, designed to provide availability and high scalability at the same time without compromising system performance, software and hardware alike, ensures a database that can be counted on for high volume of mission-critical real-time data, with a very low latency regional outages - perfect to deal with massive data feeds from multiple exchanges across the globe, at the same time, in real-time.

TagoDB contains: Object oriented, Hierarchical, Relational & Flat file based database capabilities all together, and capable of running on Linux, Windows, and some UNIX based workstations.
TagoDB is using the Tago Real-Time Archive® technology.




Tago Random®

An algorithm based technology that provides a solution for scenario-calculations that require true random data.
A more effective random calculation solution to replace the pseudo-randomness computers usually generate.




Tago Position Matrix®

A technology that shows the state of the main interest between potential liquidity providers in relation to an open position in real time, including a unique visual display that can perform complex and effective calculations regarding the open position management, according to the risk manager and other parts of the software which signals an event of conclusion or action.




Tago Virtual Exchange Simulator & Analyzer®

A technology to simulate an exchange behavior by referring directly to an existing market situation (not an open position), provides the system with clear instructions regarding the direction of the market at any given moment, AI processes included.




Tago Encrypt®

The general encryption system that protects all components of the software in real time, including the real-time data that flows between the various programs in the entire TagoTrader software suite, using cryptographic algorithms, without slowing down the system, providing a very high level of security, available by in-house internal encryption not known to any hacker.

In addition to that, the whole system is highly secured, backed by Tago's Cyber Security model where the system is always offline to the outside world, running from a local server machine, in an OGRTI mode (only getting real-time information) built-in the receiving pipeline of the TagoTrader Framework.

There is no threat and there is no hacker who could break into such a system simply because even at runtime it is never connected to anywhere by definition.




Tago Framework®

Tago Framework consists of a large number of parts, elaboration on four of them is clarified in a nutshell below:
- TagoRT Syntax
- Tago One-Step
- Tago RDM
- Tago NENI

The TagoRT Syntax

TagoRT is TagoTrader's proprietary codding syntax that works on the TagoTrader Framework only, where the TagoTrader Softwares communicate with each other on the internal network, meaning that instructions can be given and performed across the entire computer network, all in accordance with predetermined instructions set into the entire system using the TagoRT codding syntax.

The TagoRT code syntax is applied on all TagoTrader softwares, and is also the language used to manipulate features, customization options and editing or adding new algorithms in the AI components of the entire softwares suite.
Using TagoRT codding commands, the system accepts and implements Big Data volume changes in real time (employing on-the-fly compilation technics) to make progress and provide conclusions for the next Nano second (internally), or for tomorrow (tomorrow in a certain sense) gaining BI (Business Intelligence) for the next execution request received mostly by Tago RT Trader and Tago Risk Manager softwares.

Tago One-Step

One-Step methodology was developed for the TagoTrader framework mechanism which basically says "one step only" or "one step than explain" or "100% true or get help", this means that the entire system (framework) allows all softwares to come back with 2 kind of answers: 1 - that is 100% true, OR 2 - the whole system switched to RDM (Receiving Data Mode) which is like "receiving data by force", the system is now evolving (learning) and it will output the new algorithm that is missing or need adjustment by itself - this is where the fundamental premise of the entire system is reflected the most (that it cannot draw conclusions without introducing some capital to the market).

The longest time to receive a final answer throughout the entire life of the system was 14 hours and twenty-two minutes, which eventually opened a huge liquidity window of over X10 on one electronically traded financial instrument.


Tago RDM

Tago RDN (Receiving Data Mode) is the main situation in which the system goes automatically into a research mode.
When all components outputs stops call for an RDM mode in an internal research project, is in simple words, where the AI RDM's events in the system are only signaling for Enlargement-Tasks (actions to evolve and find new conclusions), rather than answering unresolved Risk-Manager questions for example.

It's basically generates the algorithm and say when it relies on it (temporary certificate) by automated research processes, once the algorithm has received a temporary certificate, it is automatically injected into all necessary system components that are using that algorithm - the old version of the algorithm is already saved by the database of the TagoCR software.

Part of the system's intelligence is based on Dependent-Data or Independent-Data, according to definitions and goals set by TagoRT syntax commands, a good example to explain this would be - like when a child asks, is 1 hour a lot of time? How much is 1 hour? The system is giving the answer a child should receive = that it depends, it may be a lot, and maybe 1 hour would be considered a bit.

This is applied when the system is asking a question of 2 channels, "One-Step", and the answer will be one of two, either 100% true or false (calling for RDM mode).

So, does the current position risk is 100% a true answer? if the answer by any of the system's components is: "it depends on: 1,2,3,X,Y,Z,etc." than RDM mode is activated and so on.

Tago_NENI

Tago_NENI refers to "new_Exchange_new_Instrument" and used for information that has entered the system and has not yet pass several days and algorithmic calculation rounds in the required AI components of the entire frameworks and softwares, so that it will evolve and learn enough to perform execution with real capital.

Tago_NENI in TagoRT syntax includes on-the-fly implementation without stopping any real time process in the system.
When all components outputs stops calling the RDM mode for a specific NENI property, the NENI mode is deactivated automatically, for the reason that events are no longer calling an unresolved issue that needs to go into AI processes, NENI can't accept enlargement algorithmic jobs, only evolving jobs.




Tago RT Calculator Software

The Tago RT Calculator runs a metaphysical algorithmic process, along with AI research processes - in order to keep the main CALC algorithm in-line with the markets in real-time.




Tago Calculator & Recorder Software

TagoCR is a robust real-time data analyzer, event and data recorder, with an artificial intelligence based event decoder processes.
all the other software units communicate with TagoCR mainly to receive answers, avoidance of discrepancies of any kind, verify and fill in information gaps, if there is a signal of data overload in the system, ensuring the flow of clean raw and approved information and dynamic data to the overall system in real-time.




Tago RT Trader Software

The Tago RT Trader is the exclusive decisions maker software for the Tago RT Calculator software on the main local server machine, and contains trading wisdom that has been evolved over many years.

Tago RT Trader software is the only software in the entire system that is able to not comply with instructions from the risk manager in certain cases.




Tago Risk Manager Software

TagoRM is a real-time risk management software that transforms the current open risk of a trade or an entire trade cycle into clear actions and decisions according to the current global settings and status in real time.
Trade Cycle means: a trade which the system sees from the perspective of one single transaction, though it contain an unlimited number of instrument and trading strategies combined and sync all together.

TagoRM is acting as the RM of each _NENI or non NENI properties (projects) in the system, simultaneously, provides risk management for each instrument and strategy running on real-time, separately and personally, and also for the big picture of all that is running together (along with TagoPM when necessary).

TagoRM analyze the current risk exposures, not only from RM point of view, but also from the perspective of each software component which intervenes in decisions that concern the RM decisions - and thus systematically reduces risks over time for and in all components.

TagoRM is responsible of many evaluation and prioritization processes including prioritize remediation actions based on dynamic risk rules rather than inactive rules/data, while a piece of dynamic information will always have to be re-signed by the RM software in teal-time (excluding Tago RT Trader Software in certain cases).




Tago Portfolio Manager Software

TagoPM make consistent plans and solutions for distributing capital practices and monitor progress against these plans on an ongoing basis.

TagoPM includes the ability to define an asset source from the available instruments on the system, based on key processes and functions from a bigger picture perspective.

TagoPM responsible for normalizing factors of all kinds, from multiple sources, including multi-currencies calculations for the entire system (via an exchange rates API).

With a process for identifying imbalances or efficiency in the various markets, TagoPM provides arbitrage real-time data (sent mainly to TagoRM) - to facilitate the evaluation of various portfolio management options available.

In addition to that, TagoPM distributing data quotes for all components in the entire system, and it does so through a very secure connection to a custom multiple data-feed client developed by Tago Trading & investment Strategies LTD.




Tago Cycle Builder Software

TagoCB is responsible for all aspects of a trade cycle, and gives instructions (without the need for internal signature in most cases) to the main Tago RT Trader Software.

TagoCB must communicate with all other softwares in order to reach a state of decision-making and conclusions, and is the only software that cannot operate completely independently.




Tago Liquidity Analyzer Software

TagoLA is an AI decision-making system with reference to risk-reward financial calculations rules and scenarios, responsible of sizing positions, prior to execution or open positions.

TagoLA is using the Tago Position Matrix® technology.




In Conclusion

The entire system was designed to discover, optimize and improve a certain algorithm, in August 2018, the research project completed successfully, providing the exact algorithmic solution the system was looking for.

Today, the system can be activated in a variety of different ways, to complete various goals and tasks related to trading and capital management in a broader perspective.

One example would be:



In the above example, the system is running using 3 machines in a network, to perform one specific action which can be a constant real-time search for an answer to one question, transaction execution, etc.

Machine 1: The only machine that is connected to the internet, running a custom multiple data feed client, providing real-time data for all the system components, using one route which passes through the a TC in OGRTI mode.
Therefore, the only attack that can be made to the system (worse case) will be to reach machine number 1 which does not contains any sensitive information.

Machine 2: runs TagoPM, Tago RT, Tago RT Trader & Tago CR.
TagoPM is among other things first of all is normalizing and distributing data quotes for all components, Tago RT, Tago RT Trader is also running, and Tago CR records everything.

Machine 3: runs Tago RM on first channel, then the Tago RM feeds the other two for running artificial intelligence processes.
In addition to that, another process of Tago RT Trader (in action mode) runs at the end of the chain to provide final answers or final action decisions in real time.
This is allowed by the system only if the first instance of the Tago RT Trader on Machine 2 is running on another mode.

There are lots of reasons why to run the system one way or another, according to the above example, the processes of intelligence receive data through the risk manager to take care of solutions related to risk management only or through the risk manager.
It is to illustrate some of the capabilities of the system.

Another example of running the system will be shown in the following example, when all the components communicate with each other, without initial barriers or any other system narrowing definitions.
This method of running the TagoTrader Framework is intended to provide answers and solutions to the most complex and critical issues, it requires many definitions and preparation in advance but it provides endless like possibilities to exploit the full power of the entire system.


In the above example, the system runs on a minimum of two machines in a network, each software can be run in a separate machine, and also the same software several times in different machines or in the same machine, provided it is in a different mode/state, all according to the project requirements for which the entire system ran for.