Just to interfere little bit. I do not think you need that kind of software any at all. It did not help me on my overall trading experience. Or just to get enlightened that there is no good strategy working on realtime tracking. And also beware of too much OOS looking, then you get just a big in-sample, until you find THAT strategy that has good OOS results. You will find of course if you search enough, but that does not mean it works then in reality on realtime tracking. This OOS thingy is just another backtesting fallacy.
Build alpha or GSB
- Thread starter Maynuts
- Start date
Testing on out-of-sample is part of forward testing, but as @TrAndy2022 noted. the out-of-sample data becomes in-sample once it's used to change things. Build Alpha has a good article about this and more.
Of course, forward testing on future data with simulated or actual trades can show what it's like to run a newly-created system.
I've written software like this for my own use to help create rules for trading with genetic optimization and genetic programming. For example, here, here, and here.
And, as with everything, it works -- sometimes.
Yes this is the right place and with Xmas approaching, we have some time for entertainment. No trading for me during the last two weeks of the year. We closed the books already. It was a good year, it could have been better but let us not get too greedy.
Stay away from any software that "develops trading strategies", do yourself a huge favor. Do not waste any time and avoid the frustration and the losses.
These software programs will find profitable strategies even in random data. Even noise will pass all robustness checks and out of sample tests if you try many times. Eran Ravin has demonstrated this with code you can run yourself. Do not be lazy, run the code. If you have never heard of him you are probably a newbie in the business.
The Quantopian story is a good example of how data mining will slap you on the face. They put thousands of quants to develop strategies and selected all the good ones that passed all robustness and out of sample tests. They did not make any money and closed down.
Just take a look at the recent drawdown of the other but more competent and interesting outsourced data mining fund, Numerai. They took a massive hit after March as the stock market rebounded. Numerai is the most advanced quant trading operation afaik. They know the business of data mining and machine learning. If you think you will buy software for X amount of $ and do better than Richard Craib and these geniuses you are mistaken, I am sorry. The probability of loss is close to 100%. No market will pay you any profits for data mining. You are wasting your time, and possibly money. Thank me later.
It is ultra hard to find something that has value and produce good results. This can take many years to find. So ask yourself if nothing works (really good) what would you do ? How would you trade/speculate/invest then if you know there is no strategy at all which is better than random entries and exits ? I am not talking about the many mediocre strategies where you can make up to 20% p.a. with at least also 20% risk as drawdown in any given year.