Another way to absolve Stick!
Posted By: Henrik Bukkjaer In Response To: Another way to absolve Stick! (Joe Russell)
Date: Friday, 31 August 2012, at 10:49 p.m.
In Response To: Another way to absolve Stick! (Joe Russell)
I think I see your point Joe.
So say you log into Othello's new BG match db here:
Search for Sticks matches, and find some live and some online ones... open them and analyze them.
If it turns out, that
...in the highly prestigious IIBGF Grand Master online matches, he plays error rates of say 0.8 and 1.2, 13 point matches without a single blunder, and
...in the live 11 point matches found there he averages over 4, and blunders are there in "normal" concentration,
then that would be suspicious patterns you were looking for?
Question is then, how do you get enough of a sample-size to perform such analysis, without cherry-picked files or the opposite bad apples?
And there would still be a lot of other factors left - some matches may contain more adjustments than others, some easy, some tough, IF someone is cheating, then that person doesn't necessarily cheat all the time, etc.
Plus, there's a limit to how many blunders you'll find when a player plays PR sub 2 anyway - we are at the end of the scales here, making things a bit unusual. But of course, if the patterns from the above "examples" are found throughout a larger dataset, then it could indicate foul play.
Just a few observations and comments to various tidbits in these threads - in no particular order:
* It is cheating if; You check just a single position in some of your matches. You get outside help apart from what's available through the client (pip-counts, etc.), that is, looking up in tables with important %-points, getting help from another person, etc. You don't have to cheat systematically on a big scale, to be "cheating".
* Maybe a discrepancy of 2.3 PR points between online and live play doesn't seem like much, in particular if we are looking at only a few sample matches. But I think one should look at it relatively, not in absolute numbers, because there's a big difference between 2.0 and 4+. If you truly hover around 2.0, then I would expect that you are so strong, that you don't accidental play 5 or 6 (two and a half to three times the PR) in a lot of matches. The sample size from online play is big, the live one maybe no so (I don't know how many matches exists and the conditions for them, selection, etc.), but one of Bill Phipps' aims, is to get a set of solid data to compare to and analyze upon.
* If it is true, that one set of matches are played to achieve lowest PR possible, and the other set of matches are played to maximize win-chances against an imperfect population of players, no holds barred, then a PR difference of about 2 points certainly could be partly explained. I for one truly believes that reasonable adjustments and optimal play against a weaker opponent, costs you error rate wise more than just the errors you introduce! Conservative play in this regard would be late doubling and more passing - both adjustments leads to more games not being played to the end, thus fewer of the "easy" non-contact decisions. I've tested and verified this hypothesis.
HOWEVER, should that be the case here, then:
1) It can easily be seen in the error rate statistics (more cube errors than checkerplay errors, and cube errors heavily skewed towards missed doubles and wrong takes).
2) If a proctored or live test takes place, such as Bill Phipps suggests, it will be easy for the player in question, in this case Stick, to play without these adjustments.
* Just because you are one of the better players in the world, doesn't mean you have "no reason" to cheat. I guess even the second best player in the world, have at least one reason to cheat :-) But in a game where low PRs have become the new black, everyone with an ego has the incentive to polish their PR a bit. Want to go up the giant list? Honestly, what single better mean than to chop of .5 of your average PR on Iancho's list??
* Error rate statistics, to a very large degree, cannot serve as "proof" of diddly squat, it can only indicate certain things or make foul or fair play probable. We must always remember this, no matter who we are talking about, or whether we are clearing or condemning people.
* Apart from error statistics, other things could be taken into account - such as but not limited to: checkerplay playing tempo and variance in that, time to consider doubles, kind of decisions chosen to ponder about, effect of thinking time (eg. do you more often than not, come our with the right decision when you think deep) - the latter would be a strong indication of cheating, if long delays yield exceptional high percentage of correct decisions.
Messages In This Thread
BGonline.org Forums is maintained by Stick with WebBBS 5.12.