0
$\begingroup$

Generally, a trading strategy is effective for a certain duration which is a function of the environment, news etc. I am currently learning about basic strategies that are (or once were) used. Some include simple mean reversion or standard deviation approaches others use more complex methods like machine learning for example. I tried a simple mean reversion strategy on some live data I queried and it turns out that that the end of day results were positive. My question is, why haven’t these strategies (or this one specifically tried) lost all the “alpha”? From what I’ve learned in classes, strategies like this that are well known should no longer be effective.

Excuse me if there is something obvious I’m overlooking before asking this question here. Thanks!

$\endgroup$
5
  • $\begingroup$ "if there is something obvious I’m overlooking" - you need tradable prices (you mention "end of day"), to factor in transaction costs, slippage/market impact, financing costs, etc. If it looks too good to be true, it probably is. $\endgroup$
    – user42108
    Feb 17 at 0:16
  • $\begingroup$ Thank you for your comment. I apologize for not specifying that this was done on live intraday data from yfinance. Would what you said still apply if there are no transaction costs from, say, using a commission free brokerage? Thanks $\endgroup$ Feb 17 at 0:19
  • $\begingroup$ The best way to locate inefficiencies in your strategy is to perform a backtest on historical data for several years which needs to be of the same type. Factor in slippage as @user42108 suggested which means that the strategy generally won't be able to transact at the same price as in historical data, the price will be slightly worse, depending on the spread. $\endgroup$ Feb 17 at 6:42
  • $\begingroup$ You wrote "I tried a simple mean reversion strategy on some live data". Did you try only one? Most people try many many strategies or variants of strategies. Of course if you run 20 tests 1 will appear significant at 5% level but will in fact be worthless and will not produce a profit going forward. This is a big problem with backtesting in general: the problem of multiple comparisons or p-hacking. $\endgroup$
    – noob2
    Feb 17 at 14:31
  • $\begingroup$ "Would what you said still apply if there are no transaction costs from, say, using a commission free brokerage?" - you still have to account for slippage (even if you think your market impact is negligible) which will typically be much larger than your TC. See @SergeiRodionov comment on backtesting your results. $\endgroup$
    – user42108
    Feb 17 at 15:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.