3 ms·
>The funnest part of writing a bot is using all the historical data to observe how the bot would have done during that period. I ended up checking thousands of
by notpc 10y ago
>The funnest part of writing a bot is using all the historical data to observe how the bot would have done during that period. I ended up checking thousands of permutations of parameters by brute force to determine the best ones, but in the end transaction fees killed me.
That's a sure way to find a bad strategy. You end up doing the equivalent of overfitting. If you look at enough variables, the chance of there not being spurious correlations is basically zero, so you're guaranteed to land on sets of parameters that look good with historical data but fail going forward.
- BickNowstrom 10y agoThe writing does not exclude proper backtesting. A more favorable reading would have him/her use a holdout or test set that is in the future of the historical data used for training. Then there would be no problem with that approach in regards to overfit/poor generalization.
- drawnwren 10y agoThere is still the problem of the market not reacting to his trades. It would be very easy for me to show a trade of $1m making money, but very hard to execute such a trade.