I am developing a Trader Expert that captures and tracks a Traders expert knowledge as well as managing price/transaction/performance data. The expert engine will aim to deliver a decision support tool and a structured approach for any trading strategy - no matter how convoluted the strategy or process may be and would not require any programing – reducing risk and helping to increase profitability. Just started the project on my own but am interested in getting any ideas on this and gauging interest if others would find it useful for their own trading?
This is what happens when technique grabs a solution and goes looking for a problem. To have an expert system work, you must have stable conditions → stable data. In finance, you will have a different outcome: https://xkcd.com/1570/
Brokerage firms do this. Not to develop trading strategy but instead to manage risk on an aggregated level. This won't work unless you have a massive amount of data to evaluate.
Agreed the description was’nt very clear and don’t worry if I have to close down my trading accounts then it will be due to my trading not any AI expert machine! The aims of the software are far more limited and mundane. Every trader has “expert knowledge” on which they depend. The software will let you capture this expert knowledge and have it available to your team/customers (ie anyone who uses your trading strategy) or used as a personal trading decision tool/knowledge store. It also attempts to be as flexible as possible for any trader to implement their own bespoke strategies and calculations. A graphical flowcharting tool will be used by the trader to express the step by step methods (and any convluted branching!) of the trading algorithms they use (any number of these strategies would be supported). The algorithm may also include calculations/formulas (eg risk, money management, entries/exits …etc) as well as market data, fundamental, macro-economic etc data processing. General mathematical calculations/formulas will be expressed to the system using a formula editor and these will be either automatically or manually processed by the system. When a trader wants to enter, manage or close a trade the software will respond with a series of questions and depending on the responses given the system will advise accordingly. A log is kept of every trading session so that a trader can check back on what their responses were and the software will explain how it arrived at its decision. A discretionary or system trader does not have to follow the advice given and I have no plans to build a trade execution platform either. Other features of the software include storage and processing of multiple sources of large scale securities data sets via enabling rapid corporate actions price adjustments, etc. A transaction store, P&L reporting, trader configured key performance indicators, trader dashboard and charting will all be available. I see a number of ways the software could be used and others no doubt others could add to the list: As a stand alone application providing and supporting trading decisions helping to reduce risk. As a application helping traders to capture their expertise and trading plans in an easily accessible knowledge base of their own. As a way of storing existing trading strategies, general trading processes (eg order failure procedures) and support information in a user-friendly system. This could then be made available to others within an organisation/trading team or for customers using a particular trading strategy. As a trading journal and audit trail. This audit trail of questions and responses will also be very useful when building the knowledge base. As a way of introducing some structure into trading decisions to avoid impulsive trades As a bespoke trading risk/money management calculator P&L reporting and Performance (KPI's, statistics etc) monitoring tool Cleaning and applying corporate actions (eg dividend formulas) to multiple price databases. The database could also include meta-data storage. Including exchange, vendor, broker and symbol matching tables etc. The inference engine is kept separate from the knowledge base so traders have the ultimate flexibility in the creation of their own trading “expert”. Having started this work fairly recently I have perhaps 50% of the above coded already and would hope that an initial prototype could be completed relatively quickly. Not sure if this is a much better description or whether it would be applicable to others but it has nothing to do with AI/deep learning in that sense.