Build your own Trading System with Machine Learning and AI

Have you always wondered, how to best combine different technical indicators into a single strategy? What are the best parameters, and to which pairs should the strategy be applied to?

Algominr

Algominr can figure this out with machine learning and statistical techniques.

I've released the commercial version of Algominr.
- Includes a strategy wizard
- Use a visual scripting tool to create, change, and explore your strategies
- Various machine learning techniques such as: Artificial Neural Networks, Support Vector Machines, Decision Trees, Random Forests, AdaBoost, Logistic&Linear regression.
- Lightning fast backtests

It includes support for MT4&MT5 as well as a direct connection to the Oanda v20 API. I will support IB TWS in the near future.

https://algominr.com/

It's not unreasonable to assume that people are doing this, e.g. Renaissance and their Medallion Fund, one of the best performing funds worldwide, have similar techniques in their toolbox.
So why can this still work?:

1. You still have to define the structure of the strategy, usually through a hypothesis, e.g. you think that the gold price influences the EURJPY through some mechanism
2. The search space is really, really big. There are a huge number of possible strategies, more than atoms in the universe. Nobody can search all of that, so still a lot of undiscovered potential
3. To justify the large computational power, a large sum needs to be invested. However, many of the effects that you can find in the markets, only work for strategies that allocate less than 100'000 USD. This is not interesting for large funds.
4. ...

Well, the same is true for discretionary trading funds.
Also, as I pointed out, you are not necessarily in direct competition with hedge funds which will only trade strategies where they can move a certain volume.

Personally, I much prefer the computational-based approach, as it allows for rigorous testing of strategies before deployment. I also believe that the combination of a hypothesis with computational search (for parameters and currencies etc.) is the best approach at the moment.

I doubt I would be able to create discretionary strategies with a sharpe-ratio larger than 3. Maybe you can.

Overview video


Algominr