No new ideas, no optimisation of parameters, just tedious data cleaning and careful implementation. This is a reminder that paying attention to the details matters!
Show this thread
Spent most of the day paying serious attention to details of one trading strategy — cleaning noise out of the signal, carefully legging into trades, minimising unnecessary turnover etc — and improved Sharpe of the backtest from 1.5 to 2.8
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
i would simply make sure my orders got execution at exactly the price i wanted
-
*emily scratching her head wondering why the sharpe is still so low*
- Show replies
New conversation -
-
-
I, too, know how to overfit a model,
-
yes this is fair, in this case I’m pretty sure that the true improvement is < 1.3 but > 0
- Show replies
New conversation -
-
-
regarding "minimising unnecessary turnover": how do you know turnover is unnecessary without introducing lookahead bias?
-
Simple example would be, A can be traded as a spread vs B or C. If you currently have an A/B spread on with weak signal and the signal for an A/C spread is strong but not enough to put a trade on yet, rather than trading out of the A/B spread it may be better to
- Show replies
New conversation -
-
-
@pdfmakerapp grab this -
Salam, we have captured the page for posterity: https://threadreaderapp.com/pdfs/piUxfggdT . Share it only with your closest friends ;)
End of conversation
New conversation -
-
-
What kind of holding period / product? 2.8 is pretty nutty for non-HFT.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.