Ok, based on the lack of *public* response, I don't feel a burning desire to follow up much on the above, especially since following up in any exhaustive way (which was my initial intention) would reveal much 'secret sauce', and be a flat-out handout of information that I've worked mighty hard to attain. . but I will provide this much for anyone that might be interested:
1. Every standalone statistical data point that one might expect to be predictive (profit factor and sharpe for example) proved to be at least slightly predictive, when viewed most broadly. . meaning that if you're faced with a handful of backtest or optimization results, chances are at least slightly better than the one with the higher profit factor (for example) will end up performing better than the one with the lower profit factor. This shouldnt surprise anyone.
2. There are significant differences in the standalone-value of each standalone statistical data point.
3. There are even more significant differences between instruments,
4. . . and even more significant differences between hour-of-day of chosen entry. . . indicating that, based upon historical data, and that there are very significant differences in 'predictability' based upon the hour-of-day of your entries. Many blocks of time showed that getting a predictive edge was all but impossible to find (just 'noise'), whereas other blocks of time showed significant possibilities in terms of exploiting market inefficiencies.
5. These time-blocks mentioned in point 4 correlated very heavily intra-sector. . meaning, crops have a similar 'hot' block of time, as do energy instruments, as do currencies, etc.
6. One of the most exciting finds, there *are* certain combinations of statistical variables (the 'calculations' I mentioned in my first post) that do yield products/sums that are very predictive indeed, in a linear manner. . . meaning, the higher the product/sum attained by the calculation after a backtest/optimization, the more likely the strategy is to perform well in live walk.
7. There is one specific combination (a calculation created by using a few different statistical outputs to create a singular linear output which can then be used as a 'score' or 'measure' of a strategies likely success in the future) that proved enormously more accurate/predictive than any other, at least 6x as predictive overall (meaning, when this singular value is HIGHER, the strategy tends to perform well. . when this value is lower, it tends to perform poorly) relative to anything else I've dreamt up and tested. . however its not only too complex to mention here, but also extremely secret-saucey. . its extremely strong/robust, proving itself over 2million+ tests over different strategies/instruments/time frames. . I mention this only to state that such predictive data points *do* exist, if one is willing to do the work to uncover them.
Those who have seen my posts scattered about know that I'm always very interested in mutually beneficial collaboration, so if you think such a thing might be possible, please send me a message . . .