Usually I prefer to praise good blog posts, this time I have to be more critical, sorry.
A: "One chief distinction [...] Computers, by contrast, reductionistically calculate any and all possibilities undiluted by prejudicial notions of what a “good” or “bad” or “promising” position might be."
-> computers calculate (a lot) more variations than humans per second, but the sentence as stated is simply not true.
- there is engine selectivity / pruning (they do not calculate "any and all")
- at the end of a variation engines evaluate the position (using quite different and fascinating ideas), it rarely happens but this assessment at the end of a horizon can be wrong, quite similar to humans, and precisely a "prejudicial notion of good/bad/promising"
B: "instead favors Euwe (who was likely not even in the top ten after WW2)"
-> Euwe was the chess world champion 35-37. claiming he was not top10 in 45/46/47 requires citing some source
C: "All this makes me skeptical that “accuracy” alone correlates particularly well with what sets great players apart from good"
- accuracy does correlate particularly well with player skill
- in fact this correlation works so well that - at least across several games - it is used as a reliable cheat detection algorithm online (in addition to repetitive time usage patterns)
- the word "alone": in case the point is there exist other metrics that correlate similarly well and might explain player skill better if they are used in combination, these metrics should be introduced including how to calculate them (this has be done in the past by some authors and is usually interesting), since it's not been done I assume that wasn't even the point
-> of course a slightly higher accuracy does not always imply a slightly higher rating,
first of all a "historical player accuracy" can be slightly confounded by opening preferences,
second even at highest levels human chess can involve intentional (slightly) unsound moves to get an advantage on the clock or get positions that are difficult to play for the opponent e.g.,
but noone disputes those things, otherwise the "correlation" that you mentioned would be even closer to 1 than it is
D: "I cannot help but feel a sort of despair when I see so many players doubt a Grandmaster’s [Hans Niemann] abilities based on how a computer evaluates, not their moves, but merely their impromptu thoughts about which moves they are considering."
I admit I didn't fully get the connection to Hans Niemann in this blog post.
let's just assume there is a good connection. the statement still appears wrong.
--> I have seen quite a few players publicly stating their opinion on Hans Niemann on different platforms.
- most if not all players or "influencers" stated they do not accuse him of cheating since there is no hard evidence
- "doubts" are sometimes presented but not only based on moves or "impromptu thoughts":
instead confirmed online cheating behavior in the past, observed weird behavior otb, and one of the fastest high rating rises in chess history are mentioned to name a few
A: "One chief distinction [...] Computers, by contrast, reductionistically calculate any and all possibilities undiluted by prejudicial notions of what a “good” or “bad” or “promising” position might be."
-> computers calculate (a lot) more variations than humans per second, but the sentence as stated is simply not true.
- there is engine selectivity / pruning (they do not calculate "any and all")
- at the end of a variation engines evaluate the position (using quite different and fascinating ideas), it rarely happens but this assessment at the end of a horizon can be wrong, quite similar to humans, and precisely a "prejudicial notion of good/bad/promising"
B: "instead favors Euwe (who was likely not even in the top ten after WW2)"
-> Euwe was the chess world champion 35-37. claiming he was not top10 in 45/46/47 requires citing some source
C: "All this makes me skeptical that “accuracy” alone correlates particularly well with what sets great players apart from good"
- accuracy does correlate particularly well with player skill
- in fact this correlation works so well that - at least across several games - it is used as a reliable cheat detection algorithm online (in addition to repetitive time usage patterns)
- the word "alone": in case the point is there exist other metrics that correlate similarly well and might explain player skill better if they are used in combination, these metrics should be introduced including how to calculate them (this has be done in the past by some authors and is usually interesting), since it's not been done I assume that wasn't even the point
-> of course a slightly higher accuracy does not always imply a slightly higher rating,
first of all a "historical player accuracy" can be slightly confounded by opening preferences,
second even at highest levels human chess can involve intentional (slightly) unsound moves to get an advantage on the clock or get positions that are difficult to play for the opponent e.g.,
but noone disputes those things, otherwise the "correlation" that you mentioned would be even closer to 1 than it is
D: "I cannot help but feel a sort of despair when I see so many players doubt a Grandmaster’s [Hans Niemann] abilities based on how a computer evaluates, not their moves, but merely their impromptu thoughts about which moves they are considering."
I admit I didn't fully get the connection to Hans Niemann in this blog post.
let's just assume there is a good connection. the statement still appears wrong.
--> I have seen quite a few players publicly stating their opinion on Hans Niemann on different platforms.
- most if not all players or "influencers" stated they do not accuse him of cheating since there is no hard evidence
- "doubts" are sometimes presented but not only based on moves or "impromptu thoughts":
instead confirmed online cheating behavior in the past, observed weird behavior otb, and one of the fastest high rating rises in chess history are mentioned to name a few
A) is a general observation about the differences between the two approaches. Computers examine more actual positions and humans use various less quantifiable methods to compensate. That computer programs must at some point employ some positional criteria in their evaluations isn't in question. That said, I took the blog post down after about 30 minutes in part because I felt the larger points could easily get bogged down in the minutiae of this subject.
B) Euwe's results in major tournaments after the WW2 is not particularly remarkable and includes last place by a large margin in the '48 World Championship and similar scores in the Candidates tournaments of the 50's. There's no reason to think of him as the strongest player in the world for any point in time after 1937 or so.
C) Accuracy of course correlates with skill to some degree, but to claim it is the only metric is tantamount to, say, ranking concert pianists by how many right notes they play. It's unquestionably significant but not the only possible consideration, as for humans chess is still something of an art as well as science and cannot be reduced entirely to statistics.
D) I'm not sure what's being misunderstood here. Many people found Niemann's postgame analysis suspicious because it was at times at odds with computer evaluations. I'm saying that it is not only particularly poor evidence but also symptomatic of a fundamental misconception of how many strong human players think about chess.
Finally, I've no comment on Niemanns "weird" OTB behavior or his rating changes, as they do not strike me as compelling evidence of anything. His online cheating is another matter but irrelevant to the general points of what I've written above.
- Zug
B) Euwe's results in major tournaments after the WW2 is not particularly remarkable and includes last place by a large margin in the '48 World Championship and similar scores in the Candidates tournaments of the 50's. There's no reason to think of him as the strongest player in the world for any point in time after 1937 or so.
C) Accuracy of course correlates with skill to some degree, but to claim it is the only metric is tantamount to, say, ranking concert pianists by how many right notes they play. It's unquestionably significant but not the only possible consideration, as for humans chess is still something of an art as well as science and cannot be reduced entirely to statistics.
D) I'm not sure what's being misunderstood here. Many people found Niemann's postgame analysis suspicious because it was at times at odds with computer evaluations. I'm saying that it is not only particularly poor evidence but also symptomatic of a fundamental misconception of how many strong human players think about chess.
Finally, I've no comment on Niemanns "weird" OTB behavior or his rating changes, as they do not strike me as compelling evidence of anything. His online cheating is another matter but irrelevant to the general points of what I've written above.
- Zug
GM Susan Polgar tweeted some interesting stats who speak by itself ... twitter.com/SusanPolgar/status/1568835715437305859?s=20&t=z7v4n8MXfJu04lPQomLhKQ
@kasparovbab said in #4:
> GM Susan Polgar tweeted some interesting stats who speak by itself ...
> twitter.com/SusanPolgar/status/1568835715437305859
I do find these things suspicious and I'd like to reiterate that my concern isn't whether Neimann is cheating in some way or not (I've no idea one way or another) as much as what is considered legitimate evidence.
That said, if Susan Polgar's reputation seems to have somehow escaped untarnished from serious allegations of what strikes me as significantly more legally questionable acts (www.chess.com/article/view/lawsuits-chess-and-susan-polgar), I'm not sure Niemann couldn't either.
> GM Susan Polgar tweeted some interesting stats who speak by itself ...
> twitter.com/SusanPolgar/status/1568835715437305859
I do find these things suspicious and I'd like to reiterate that my concern isn't whether Neimann is cheating in some way or not (I've no idea one way or another) as much as what is considered legitimate evidence.
That said, if Susan Polgar's reputation seems to have somehow escaped untarnished from serious allegations of what strikes me as significantly more legally questionable acts (www.chess.com/article/view/lawsuits-chess-and-susan-polgar), I'm not sure Niemann couldn't either.
You criticize accuracy not being the only metric correlating with skill level, yet you dont provide alternative well defined metrics.
That chess can be understood as an art and should not be reduced to statistics is correct but not related to competitive skill level or cheating accusations.
That chess can be understood as an art and should not be reduced to statistics is correct but not related to competitive skill level or cheating accusations.
<Comment deleted by user>
@cmplx said in #6:
> You criticize accuracy not being the only metric correlating with skill level, yet you dont provide alternative well defined metrics.
>
I mentioned several, such as winning or tying for the World Championship (Bronstein), winning a record number of consecutive international tournaments (Karpov), and so forth. But this isn't really the point of the article and I suspect we're discussing past each other's points in any event. Suffice it to say that I'm not disputing the obvious correlation between move "accuracy" and chess skill, I'm simply suggesting there exist other factors.
> That chess can be understood as an art and should not be reduced to statistics is correct but not related to competitive skill level or cheating accusations.
This again seems to miss the main gist of the article which I summarized above:
> Many people found Niemann's postgame analysis suspicious because it was at times at odds with computer evaluations. I'm saying that it is not only particularly poor evidence but also symptomatic of a fundamental misconception of how many strong human players think about chess.
I honestly don't know if I'm somehow being abstruse in some way about this - if I am, I apologize, for I don't presently know how to make this clearer.
Sincerely,
- Zug
> You criticize accuracy not being the only metric correlating with skill level, yet you dont provide alternative well defined metrics.
>
I mentioned several, such as winning or tying for the World Championship (Bronstein), winning a record number of consecutive international tournaments (Karpov), and so forth. But this isn't really the point of the article and I suspect we're discussing past each other's points in any event. Suffice it to say that I'm not disputing the obvious correlation between move "accuracy" and chess skill, I'm simply suggesting there exist other factors.
> That chess can be understood as an art and should not be reduced to statistics is correct but not related to competitive skill level or cheating accusations.
This again seems to miss the main gist of the article which I summarized above:
> Many people found Niemann's postgame analysis suspicious because it was at times at odds with computer evaluations. I'm saying that it is not only particularly poor evidence but also symptomatic of a fundamental misconception of how many strong human players think about chess.
I honestly don't know if I'm somehow being abstruse in some way about this - if I am, I apologize, for I don't presently know how to make this clearer.
Sincerely,
- Zug
twitter.com/thestrongchess/status/1568810062399021062. This guy links a google doc in his tweet where he spent a fair bit of time on Niemann's tournament results 2019-2020, and lo and behold, the evidence susan polgar tweeted is nonsense.
Thanks @ZugAddict for the info ! I didn't know that story ... Anyway No human is perfect ... Who am I to make a final judgement... Time will tell the true ! Also the whole situation remind me this famous quote " perfection is just the disguise of insecurity " by Gary Vaynerchuk ;)