| |
BGonline.org Forums
"Easy" improvement to PR calculations
Posted By: Henrik Bukkjaer In Response To: "Easy" improvement to PR calculations (Bob Koca)
Date: Friday, 12 April 2013, at 11:32 p.m.
>> If this ever gets implemented and used in a PR tournament where you win by playing with a lower PR then there will problably be several more little tricks to use and of course I still think variance reduced results would be better.
Now, I'm trying to improve the PR measurement, which is a normalized average error-rate. I'm not trying to convince you that it should be the preferred measurement of choice in what you call a "PR tournament" (should probably be called "performance tournament" or "error tournament" if it is the concept you refer to).
But IF you were to use PR in such a tournament, I think my enhancements would be useful, and would limit the number of "tricks" you could use to your advantage.
Think about playing a blitz. Player A blitzes player B. Typically what happens, is that player A has a high number of "fully live" decisions to make, throughout such a game. Going into the blitz, executing it, etc. Also player A will typically have a good number of cube decisions - especially if he ends up playing "too good". Player B on the other hand, will have a limited number of decisions, dancing and being closed out on the bar. Many of his rolls wil be forced moves entering maybe just one checker from the bar. The non-forced moves will typically be easier for player B also, since on of his two checkerplays will be forced from the bar. Also, he only get one cube-decision (take decision). Now, at some point, player A opens his board, and player B will enter, trying to same a gammon.
When the dust has settled, even though the forced moves doesn't count in current PR calculation, player B ends up with the easy part here. A lot of semi-forced moves (half decisions) and the number of non-contact moves being relatively higher, due to the fewer non-forced moves!
In an "error tournament", with the current PR calculation, you bet I'd rather be the person who gets blitzed!
>> If one is consciously adjusting to weaker play then the entire concept of using PR to compare different players who were not playing each other is inherently unsound.
True, but still:
If you make an error of size X, consciously to adjust to a weaker opponent, then you would expect that to affect you PR by adding X to the sum(err) part of the PR equation. While that is true, it is also true that by making such an adjustment, you risk getting fewer non-contact decisions, which are easier decisions to be honest. Thus, your adjustment affects, indirectly, not just the numerator but also the denominator, making your PR suffer more than it should.
Thus, it might be unsound to compare different players not playing each other (or rather, not both playing equally strong opponents), but you would get a much better comparison by making the adjustments I propose.
So, if a player plays online to minimize his PR, and gets as low as 2.3, he might play live to maximize his winning chances, and end up with a PR of over 4! Using the proposed enhancement, he could end up getting 2.5 online and 3.7 live! Maybe not the golden answer we dream about, but an improvement nevertheless.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.