Iyama Yuta losing it?

[gowan]

Analysing performances of go players in a match or matches is difficult using ordinary statistical or probabilistic techniques. In particular in a series of games between the same two players the individual games are not independent. I'm sure we all know two players who are equal in strength according to ratings but one, for some reason, wins most of the time. This does not mean that the ratings are not correct because ratings are based over many games with a large variety of opponents. Thus for any particular two players the ratings might not correspond to the actual chance of one winning. And, in a series of games between the same two players, there are strong psychological effects on the results.

[topazg]

I was not implying that go performance was random. In terms of the testing of theories, random is for the "null hypothesis" (that the effect being observed, in this case that a top player losing a string of games was random).

Look, if player A is a certain amount better than player B we expect A to win a higher precentage of games than player B. But each game is an independent event. In other words, if because of the difference in strength we expect player A to win 60% of the games against player B this does not mean that out of ten games played between them player A will win six of them and player B four of them. It means that in each of the games the probability of player A winning that game 0.6 and player B 0.4 The equivalent of a somewhat unfair coin being flipped.

But there is a probability that of the ten games, player B would win all ten of them, a probability of winning nine of them, of eight of them, etc. What I was saying is that people have a tendency to seek meaning when "just random" a perfectly acceptable alternative. That is what the coin example was all about. Somebody watching one of those coins come up five heads in a row too quick to decide "that must be an unfair coin". With enough coins being flipped, some of them will exhibit "runs" even though they are fair coins.

Look, this matter has been researched. For the ball player in a slump, no help at all. But analysis of the number and duration of slumps is actually around what would be expected if random. The same with "hot streaks".

There are enough top go professionals that some of them can be expected to win more or win less than evaluations of their strength indicate might be expected. It would tke a huge advantage over the other players (very high probabity of winnning each game) for us not to expect Iyama to lose some titles.

Yes, all true, not disputing any of it, and also irrelevant. The point uberdude made was that "Iyama is losing it (his dominance) of Japanese Go". The number of simultaneous titles he held before compared to the number he holds now is indeed basically a proof of his point. He never claimed that "losing his titles" was unexpected, it was just a factual observation.

[Uberdude]

Seeing as I find Go more interesting than being retaught basic statistics I learnt years ago, here is the final game.

http://igokisen.web.fc2.com/jp/sgf/39kiseit7.sgf

A feisty performance from both players with a lot of sharp tesujis and interesting shapes (e.g. move 32 wedging a one point jump shimari, just as in the recent European Pro qualifiers). And what would EdLee say about n9?!