What are the best predictors of real time model success in quantitative trading, i.e. # of transactions, length of backtest, StdDev, Skewness, Excess Kurtosis etc and what are the numerical values of the best predictors?

  • $\begingroup$ I don't know. But when you have 30 years of data and thousands of transactions like you do, that is probably enough data. I would be more concerned about omissions or biases in your backtest and would want to have the system development reviewed by someone I trust. $\endgroup$ – Alex C Apr 25 '18 at 14:42
  • 1
    $\begingroup$ Out-of-sample testing through a variety of different trading environments may be one useful indicator. Though this could give false results depending on how you train your model before those test periods. I refuse to call anything a "best predictor" because...well, you'll see :) $\endgroup$ – amdopt Apr 25 '18 at 19:28