Updated logic only applied in the current month
If it's only on March then the update isn't about trading logic. Unfortunately the only one who knows why that happen is the creator of the EA.
Thank you for your comment. Literally, what this modification has done was like; rewriting the backtest results to make it look better for potential customers for the EA without improving any other months than Mar. 2024. I hope the creator gets back to me with reasonable explanations.
I got a shocking fact, and I would like to share this fact with others.
The author got back to me and said "I made the adjustment precisely at the last period of time. That is training the neural network for current market conditions.” This explanations were quite shocking to me because I now know that it is possible to improve backtesting results afterward, and such improvement can be targeted on a certain period to make bad periods to look good without touching any good periods at all. Naturally thinking, he perhaps didn't touch good periods as it makes no sense to touch good periods.
When I made my purchase decision, I was so excited about the unbelievable backtest results. But that was the outcome of such "training." It is possible to apply a certain trading mechanism on a limited period, and other mechanism on other periods. I am neither saying the EA I bought is bad nor the author is bad. But I just want people to be aware of it.
I got a shocking fact, and I would like to share this fact with others.
The author got back to me and said "I made the adjustment precisely at the last period of time. That is training the neural network for current market conditions.” This explanations were quite shocking to me because I now know that it is possible to improve backtesting results afterward, and such improvement can be targeted on a certain period to make bad periods to look good without touching any good periods at all. Naturally thinking, he perhaps didn't touch good periods as it makes no sense to touch good periods.
When I made my purchase decision, I was so excited about the unbelievable backtest results. But that was the outcome of such "training." It is possible to apply a certain trading mechanism on a limited period, and other mechanism on other periods. I am neither saying the EA I bought is bad nor the author is bad. But I just want people to be aware of it.
Neural networks are complex and will not always behave the way they should, like any other EA, market conditions change and because of that, no one can train a model accordingly because of the unknown, that's why self trained Ea's exist, but they have problems to, because the markets are never the same, so the training period gets confused and the strategy misbehaves. I have developed and written many. In my personal opinion, you can create a very good Ea with either python or MQL that works really well, if you have the market knowledge. NN's was a big thing, but is now failing many. Its good to know programming languages and how to work with them including tensorflow and such, but for trading, its not worth it. If there is any EA in existence that makes a ton of money, no one will share it
Neural networks are complex and will not always behave the way they should, like any other EA, market conditions change and because of that, no one can train a model accordingly because of the unknown, that's why self trained Ea's exist, but they have problems to, because the markets are never the same, so the training period gets confused and the strategy misbehaves. I have developed and written many. In my personal opinion, you can create a very good Ea with either python or MQL that works really well, if you have the market knowledge. NN's was a big thing, but is now failing many. Its good to know programming languages and how to work with them including tensorflow and such, but for trading, its not worth it. If there is any EA in existence that makes a ton of money, no one will share it
Thank you very much for your detailed input. I was totally wrong to have assumed backtesting would be a representation of future. You are right, if such EA exists, nobody shares it.
I had further communication with the author and knew he didn't try to cheat users with beautiful backtest. It was just a nature of neural network training. And your explanations further confirmed the complexity of it. Thank you!
Forum on trading, automated trading systems and testing trading strategies
Vladislav Boyko, 2024.03.18 23:03
Don't trust backtests as there are many ways to fake them. At a minimum, there should be at least monitoring of trading on a real account. Although, even good monitoring does not guarantee success in the future.
[Everything I said is my subjective opinion. I have no experience buying or selling robots.]
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Is it possible to modify the trading results during a certain period?
Recently, the EA I’ve been using got consecutive losses. Then the developer of the EA updated the EA after that. I have compared the backtest results between the previous version and the new version. The new version improved the backtest results in March 2024, but the improvement was seen in March 2024 only. Any older trades before March 2024 remained exactly the same, no improvement, no difference of positions.
I believe that if the trading logic is modified, then the trading results should be different throughout all the backtesting periods. But it made the difference only in March 2024.
I have compared each trade on the charts in Mar. 2024. It’s not like the updated version eliminated losing trades, but it’s more like the updated version is equipped with a kind of trend filter on the higher timeframe so that it takes trades on the right side.
So I do not think it’s a manipulation to make the backtest results look good. Because if it’s a manipulation, the author/developer would rather modify the trading results throughout the whole backtesting period.
But I simply don’t understand why the updated logic is applied in the current month (March 2024) only. Is it possible to do so? If so, why the update is applied in the current month only? Is there anyway to apply it on other months?
Could anybody give me suggestions on this issue? I’ve sent the same question to the author, and I am waiting for a reply. But I would like to hear opinions from other traders/programmers.