| |
BGonline.org Forums
PR against XG vs. live PR
Posted By: David Presser In Response To: PR against XG vs. live PR (Timothy Chow)
Date: Friday, 11 July 2014, at 10:34 p.m.
My response is not just to Timothy's post. It is a reply to this discussion.
We didn't really define "Unforced Errors" in bg context, so for the purpose of this discussion I will define it as errors that a player could avoid given his current skill-set. In that case they can happen in online and live play.
In my first post it may have been incorrectly implied that I am able to avoid unforced errors when playing online. This is not the case and such errors do happen. However, I am more likely to have unforced errors when playing live. My initial point was that by transcribing my matches I was able to observe this problem, which in my opinion is the first and most important step in dealing with a weakness. I worked on this problem in the same way that Timothy did and detected patterns and ways to improve. Today, I am still suffering from unforced errors both online and live, and am still more likely to make unforced errors live but not as much as I used to.
As for John's comments - in general I agree that it is harder to play when leading and needing to balance your game plan with the score while when playing from behind you "can't do too bad". Nonetheless, it should be noted that by "online PR" I referred to playing against XG and/or humans and not just XG. I will not search for data to support that, but intuitively I can say that when trailing many positions that are clear ND are considered as a cube decision according to XG (for having a difference not greater than 0.2 between ND and D if I recall correctly) and increasing the denominator which lowers the PR, whereas the leader has games that he just jumps from over 0.2 ND to over 0.2 TG such that technically XG considers it as no cube decisions during the game (unless you blunder that one positions which increases the denominator by adding 1 while the numerator increases by adding the magnitude of the blunder which is greater than 0.2).
While we are on the subject, over time and thanks to other players inputs I was thinking of a simple model that links live PR and online PR. I am aware that I am the least experienced and least knowledgeable among this group, and I am sure that there are other sources that I am not familiar with or didn't read, so your feedback would be appreciated:
I think that on average there is a difference between a players live PR and his online PR and this difference can be quantified. The difference isn't a constant, but more of a function of his online PR (It can also be a function of his live PR) such that if a player has an online PR of X his average live PR would be A*X. Intuitively I would say that A ranges between 1.1 and 1.2 (possibly even more, or maybe it is exponential). That is If we randomly pick a player from a pool of players with a 3 PR, their average live PR would be between 3.3 and 3.6 and for a pool of a 5 PR players, their average live PR would be between 5.5 and 6. The range in A could be partly explained by those unforced errors that were discussed and happen in live matches but not online. I think that the lower range of A should still be greater than 1 for natural reasons that make most players play better online (for example, you play online at your convenience, as opposed to playing live which sometimes is at an unusual hours so it won't hold the tournament). We are aware of players that play better live than online, and/or play drastically better online than they play live. I consider those players as outliars. This simple model addresses some of the issues that were raised in this discussion thread while supporting consistency of the better player in the same way that the the better player is expected to have lower variance in his PR.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.