Author: Robert Hyatt
Date: 19:29:17 01/08/00
Go up one level in this thread
On January 08, 2000 at 10:45:32, Graham Laight wrote: >On January 07, 2000 at 23:02:17, Stephen A. Boak wrote: > >>>On January 07, 2000 at 10:57:55, Graham Laight wrote: >> >><snip> >>>I just think >>>that those who advocate using the Century V GM series as a guide to top computer >>>strength would have been happier to read that Century had won the game 12-8, >>>rather than losing it by that margin. >>> >>>-g >> >>Hi Graham, > >Hi Stephen! > >>There you go again. Maligning an entire group of people by imputing--without a >>basis--that they would rather have their personal opinions validated by >>empirical evidence than to seek or discover the truth by objective and unbiased >>gathering of empirical evidence even if it differs from their prior opinion. >>Sound like familiar behavior to you? It should. You do this all the time. >> >>In psychology, this behavior is called 'projection'. Attributing undesirable >>characteristics to other people that one fears most in himself. > >This is true. People who criticise others are usually telling you more about >themselves then they are about the other people, Stephen. > >>The people who say 'look to the Rebel matches for evidence' are countering the >>likes of you who happily promote opinions without much in the way of evidence or >>logic. They say that only comp-human testing, under rigorous conditions, will >>shed light on the dispute. >> >>It is you, not they, who would love to be vindicated by comp-comp match results >>that support your opinion that top programs are GM strength versus humans. >>Absent actual evidence by comp-human testing results, under rigorous conditions, >>to support your own opinion (and you have come up blank there so far), you would >>love to twist and turn each and every other event (even if illogically connected >>to the issue) to assert it supports your own unfounded opinion. >> >>So being the unbiased and fair fellow that you are, ahem, you fall back to the >>same old and deflated argument that a comp-comp result somehow bears value on > >If there's a major group of chess players for whom the Elo rating system does >not work, this is bad for the Elo rating system. > >i.e. if, in one group, player A scores a significantly higher rating than player >B, but when they play in another group, player B scores a significantly higher >rating than player A, the Elo rating system is damaged. No it isn't... this is about statistics and sampling theory. The Elo rating system predicts game outcomes between players in the _same_ pool of players. Because as they play each other, their ratings are adjusted to reflect the outcome, and since humans don't magically become stronger overnight, that history of performance is a very good predictor of future performance. But if you take two groups, A and B, and the people within A play each other to establish ratings, and the people within B play each other to establish ratings, there is _nothing_ that can be used to figure out how a player from A will do vs a player from B. For a simple example, imagine that somehow group A is nothing but GM-level players. They will play each other and their initial rating will hit around 1500 since you have to start them somewhere. Some will hit 1700, some 1400, but remember, they are _really_ GM players. Group B is a group of beginners. Their ratings start off at roughly 1500 also, and they will spread out from about 1400 to 1700, because of how the system works. But now what happens when a 1700 from group B plays a 1400 from group A? He gets smashed. Because the two ratings are meaningless outside the rating pool. That is what is wrong with the SSDF ratings. They are fine for comparing the different programs and predicting match outcomes. They are worthless for comparing to FIDE ratings to predict game or match outcomes... > >Likewise, if, in SSDF, Tiger were to score more highly than Century, but in FIDE >(if they were allowed to play), Century scored more highly than Tiger, one would >have to say that one had been misguided by the ratings. > Rebel probably should not score higher in one than in the other, assuming that Rebel and Tiger play enough to keep their ratings in line. But it will definitely happen. Because the two pools of players are different. >>the comp-human strength debate. No proof, mind you, but it sounds good, doesn't >>it--to your own ears. Even better, you allude that a particular comp-comp >>result will somehow make your detractors feel bad, if comp A beats comp B, or >>the reverse. >> >>Wrong! It is not they who argue that comp-comp results have meaning in the >>comp-human debate. It is you who have fallen prey to that fallacy. You are >>merely trying to smear them with the muck with which you have smeared yourself. > >Why is the suggestion that comp-comp ratings have a bearing on comp-human >ratings "muck"? I wouldn't say "muck" myself, just "wrong". Based on a whole lot of data points from watching on ICC. > >>Sadly for you, the muck doesn't stick to those you fling it at--they don't care >>a whit about comp-comp results in their quest for measuring the true strength of >>comps versus humans. You think you are pulling the wool over sheep's eyes. >>However, these sheep have seen the mud and know how to avoid it, unlike more >>ignorant creatures that often are befouled by the paths they choose. >> >>But you don't stop there, do you. You can't resist trying to twist the needle a >>little more, to prick those who disagree with you. You have to impute emotional >>attachment to 'vindication' in your gainsayers (ahhh, projection at its >>finest!), to demean their character in some manner and establish the hoped for >>bias that makes you feel superior. >> >>As pointed out before, you demean *yourself* with your antics. Again you hoist >>yourself with your own petard (that petard must be getting pretty worn out--you >>might wish to order a few new ones). You try to smear your opponents by having >>the pot call the kettle black. You fabricate the presumption that your >>opponents actually care about a comp-comp result (when it is really you who >>believes it matters), such that a particular comp-comp result (A beats B, or B >>beats A) would make them feel bad and lead them to doubt their claims about >>computers not being GM strength against strong human opponents. >> >>So your circular reasoning, devoid of logic in the first place, seeks to place >>your own inadequacies on other decent human beings. Once again, the circle >>closes and you are left facing yourself. Once again, a feeble attempt >>boomerangs onto the initiator--you. You created the stink, it sticks to you, >>and you can't rid yourself of it that easily by complaining about the odors of >>others. It is yourself you are smelling. >> >>Your constant tirade of besmirching remarks reminds me of a snotty child trying >>to start a fight. You go as far as you possibly can to bait others (we call >>this trolling on ICC and other websites) and then you smirk and smile and blame >>them for any and all misunderstandings. Good job, by the way--you are much more >>subtle and congenial than other trolls, although just as repugnant. >> >>You are very good a taunting. Ever think of being an attorney? You would be >>ecstatic at cross-examining a hostile witness when the judge gives you a rather >>free hand. > >May I return the favour and offer you some career advice as well, Stephen? You'd >make a good comedy script writer. > >-g > >>What will it be next? Oh, I know--you will again rely on Selective Search >>figures, again without documenting the basis for those hybrid ratings (comp-comp >>mixed with comp-human results), and surely you will again blithely mix up >>comp-human results at fast time controls to support your opinions about comp vs >>human strength at slower tournament controls (40/2). Will you repeat your >>slanders about people's motivations, solely to keep your fires fueled? I >>believe so. >> >>By all means, get those new petards ordered now! I recommend the striped ones, >>the black and white ones. You don't want to be hoisted by the same old ones, do >>you? And you'd stink a whole lot better in new clothing, wouldn't you? >> >>Value-added postings. The New Year's resolution for 2000. A worthy goal for us >>all. >> >>Laighten up, will you! >> >>--Steve Boak :)
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.