-4
$\begingroup$

I created a strategy using a regression on a price series. I tested it with many walk-forward analyses and it has passed. I am currently live trading it with real capital (the ultimate test). My question is how many live trades have to occur to have a high confidence that I have not overfitted or am subject to short term variance but actually have a statistical edge?

$\endgroup$

closed as unclear what you're asking by LocalVolatility, Attack68, vonjd, Helin, byouness Jun 18 '18 at 13:47

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 3
    $\begingroup$ It is unclear what you are asking for. What kind of strategy do you consider? The "long run" is a concept about time and not about the number of trades you perform in your strategy. $\endgroup$ – Bernd Jun 17 '18 at 7:05
  • $\begingroup$ The strategy is a swing trading strategy. How many trades are necessary for statistical significance? $\endgroup$ – Ryan Schmitt Jun 17 '18 at 17:28
  • 1
    $\begingroup$ What are the practical curcumstances that gave rise to such a question? Maybe that helps us understand .... $\endgroup$ – Bernd Jun 18 '18 at 6:03
  • $\begingroup$ What else needs to be done to reopen this question? $\endgroup$ – Ryan Schmitt Jun 20 '18 at 19:58
  • $\begingroup$ People need to cast a reopen-vote to open it or they can ask for more clarification. $\endgroup$ – Bob Jansen Jun 20 '18 at 20:28
1
$\begingroup$

I think your question is somehow missleading.

When one is looking for a good trading strategy that shall be based on statistica/mathematical/econometrical methods he typically proceeds as follows:

  1. Estimates the process that drives the prices, e.g. of stocks. This could be done with a lot of models. For example RW (random walk), CAPM (capital asset pricing model), ARMA-GARCH (Autoregressive moving average with generelized autoregressive conditional heteroscedasticity), ...

  2. Then he performes "tests" of several trading strategies for the estimated proccess to find the best trading strategy for his pourpuse.

Your question "Exactly how many trades are necessary for statistical significance?" doesn't refer to one of the above steps clearly. Statistical significance has to be obtained in step one. Because this is the step where you are estimating something. But the number of you trades are at best showing up in step two (if you simulate the trades numerically). Or they will show up in the end, when you strategy is implemented. Additionally, the question for an 'exact number' seams to be missleading. Typically, answers would be of the 'the more, the etter' type.

$\endgroup$
  • $\begingroup$ You say the more the better but a sample of 1000000 has a huge probability to contain heteroscedascicity more so than a small sample. $\endgroup$ – Ryan Schmitt Jun 18 '18 at 11:59
  • $\begingroup$ Heteroscedasticity doesn't depend on the size of the sample. It does also not come with more or less probability. Either it is part of the model structure or it isn't. $\endgroup$ – Bernd Jun 23 '18 at 4:52

Not the answer you're looking for? Browse other questions tagged or ask your own question.