New Ranking System Optimizer Spreadsheet

Research
21

Yuval,

I hope I sent you this in an email attachment. I had a little trouble (did not go twice but seemed to this last time) so let me know if it did not go through. Is your email yuvaltaylor@portfolio123.com?

Thanks.

-Jim

22

I am new to all this, but where can i find more information on how to optimize a ranking system?

Further, how do i know that my optimized ranking system will be the optimal performer in the future?

thanks for helping me understand!

Best,

23

Click on the gear box in the ranking system and then click ā€œOptimize.ā€ Unfortunately the documentation for this feature, which is ancient, seems to have been lost, but you might be able to figure it out. Some of the other posts in this thread might be helpful. Personally, I havenā€™t tried it in years. Itā€™s a bit of an artifact.

There are lots of other ways to optimize a ranking system. One could, for instance, find weights that work pretty well through trial and error, then add 3% or 4% to one while subtracting 3% or 4% from another, see if that improves it, and repeat ad infinitum. One could randomly sample a half-dozen to ten ranking factors out of your set of ranking factors and equal weight them, test it, repeat the process hundreds of times, and average the best-performing systems. One could test ten or twelve sets of random weight assignments, take the best performing one, and test ten or twelve variations of those weights, and go from there.

Itā€™s very unlikely that a system that is optimized for one specific period will be the optimal performer in a future period. But itā€™s also very unlikely that a system that is not optimized will be the optimal performer in a future period. We basically have no way of knowing what will be the optimal performer in a future period. We can, however, deduce (though we canā€™t know for certain) that itā€™s more probable that a system that would have worked in the past will work in the future than a system that would have utterly failed in the past. There will always be, however, exceptions.

24

Does anyone still have this spreadsheet? and is someone using it?

1 Like
25

Would be interested in this spreadsheet too. Thanks in advance!

2 Likes