Machine Learning in NT8

[ashmanz80]

Hello,

I would like to know if anyone has worked on any machine learning (ML) algorithms using C# to trade on NT8. It seems using the old technical indicators isn't enough these days and incorporating ML increases the odds.

Any comments/suggestions would be helpful.

[NinjaTrader_PaulH]

Hello ashmanz80,

Thanks for your post.

This should be an interesting thread and look forward to participation by the community.

[ashmanz80]

Hello ashmanz80,

Thanks for your post.

This should be an interesting thread and look forward to participation by the community.

Thanks Paul!

I honestly think anyone really interested in automated strategies should consider this. I've found this on GitHub, https://github.com/accord-net/framew...Classification

It shows all the different classifiers for ML that can be used for trading!

[koganam]

Thanks Paul!

I honestly think anyone really interested in automated strategies should consider this. I've found this on GitHub, https://github.com/accord-net/framew...Classification

It shows all the different classifiers for ML that can be used for trading!

You mean implementing rules like:

"If the smallest MFE of the last 3 trades was 20 ticks, make the next trades target 20 ticks."

or:

"If the historical probability of making 8 ticks is much greater than retreating 16 ticks, and the values calculate to a positive statistical expectation, make the next trade, even though the Stop is larger than the Target?"

Yes, there are some of us using such approaches. I know of one person who keeps statistics of what time on each day is the most profitable for a reversal of price from one standard deviation off of it, to the VWAP. He issues orders based on those times. That person is active on these fora but I cannot give you his identity without his permission. He will probably see this thread though, as it is on essentially his approach.

[ashmanz80]

Actually, more towards using

- Support Vector Machines or simply SVM's
- Random Forest

Using different classifiers and regression analysis to basically compute the future prices. It can take a simple strategy such as cross and push it to new limits. I'm not talking about just using historical data sets to 'sort' of predict the next move.. I'm talking about strategies that are smart enough to be predictive and dynamic rather than constant and linear.

an example

https://www.inovancetech.com/how-to-trade-rsi.html