|
Doctor Malaver posted:Disappointing game from Carlsen. He went directly from prep to draw. Not something you'd expect from the world champion, and one of the best players ever, playing with white pieces. The structure of the tournament doesn't really incentivize a player who thinks they're better at rapid to try to win the classical games
|
#
¿
Nov 30, 2021 22:20
|
|
|
|
| # ¿ May 16, 2024 03:06 |
|
|
I know that, since chess at the 3600 level is just a guaranteed draw, computer chess tournaments all always start from positions at the edge of theory, playing each position twice (so each program gets to be white). I wonder if that could be adapted to human play for competitions between super GMs? To force them into positions that are more like +0.5 than +0.05. I already thought of two or three issues with this even as I was typing it, but I don't think that the core idea is wrong that classical chess at the 2800 level is too drawish to really determine the stronger player reliably is far off.
|
|
#
¿
Dec 1, 2021 02:55
|
|
|
I feel like you could build a machine learning model that estimates how many centipawns the average player of a certain rating will lose over the next few turns. It wouldn't be too different from the bots that attempt to mimic human play, I don't think. How accurate they'd be is another question, though.
|
|
#
¿
Dec 14, 2021 00:38
|
|
|
dhamster posted:Computer chess tournaments have many games that end in mate, even though each side is doing the top computer move Don't computer chess tournaments also start from the middle game (using an opening database) specifically to force imbalanced situations
|
|
#
¿
Apr 9, 2022 19:24
|
|
|
Honestly at lower elos you're probably better off keeping your queen and hoping they don't find mate and then blunder away their advantage
|
|
#
¿
Apr 17, 2022 04:45
|
|
|
It seems to me that the ability to eventually find the best move is a different (but related) skill to finding a good move quickly, and it's not obvious to me that the former is a better measure of a player's chess skill than the latter
|
|
#
¿
Jul 22, 2022 23:22
|
|
|
The weird corner cases make a lot more sense if you consider an alternate version of chess where the goal is just to capture your opponent's king. Your king would die first so you would lose. Checkmate is just a formalization of the idea that your king being captured is inevitable
|
|
#
¿
Sep 15, 2022 19:59
|
|
|
Redmark posted:While it logically makes sense stalemate doesn't fit into this model at all. yeah stalemate should be a loss let's overthrow fide
|
|
#
¿
Sep 15, 2022 20:30
|
|
|
neither magnus nor hans has founded a rival chess organization so it's not even up there for recent scandals
|
|
#
¿
Sep 15, 2022 22:35
|
|
|
Lol
|
|
#
¿
Sep 19, 2022 18:51
|
|
|
magnus' prep for the game had a big "if hans plays c4 I got nothing". This is just more confirmation that his prep was stolen
|
|
#
¿
Sep 19, 2022 20:08
|
|
|
fart simpson posted:care to explain for the rest of the class? no cheat detection is flawless and they tend to be easily beaten if cheaters know what the algorithm is looking for, because the algorithms don't actually detect cheating, they detect more easily measurable stuff that's correlated with cheating. this isn't a chess only problem, it's common to many games
|
|
#
¿
Sep 20, 2022 09:12
|
|
|
I feel like the magnus interview was a lot of words that amounted to "my lawyer has advised me not to comment"
|
|
#
¿
Sep 21, 2022 21:32
|
|
|
the clause about electronic devices allowed by the arbiter is a kind of funny loophole because a while back a blind player cheated by claiming his blutooth audio device was just reading back the moves of the game, when in fact it was connected to the internet and he was being fed moves by a collaborator
|
|
#
¿
Sep 22, 2022 02:42
|
|
|
Simply make 1 minute chess the main prestige format so nobody has time to cheat
|
|
#
¿
Sep 22, 2022 05:01
|
|
|
I don't remember the details but the methodology in that image is suspect, apparently
|
|
#
¿
Sep 22, 2022 05:10
|
|
|
I recall people saying that the chart does not accurately identify which events had live streaming, and is thus a bad data source before you even do any statistics
|
|
#
¿
Sep 22, 2022 20:24
|
|
|
I mean, you have to confirm that your analysis method doesn't just declare everyone a cheater. That's just basic responsibility, not hand-wringing.
|
|
#
¿
Sep 27, 2022 05:52
|
|
|
if you have enough data, you can massage statistics to say almost anything you want the trick is to have a specific hypothesis and a corresponding null hypothesis, and a specific test for whether or not it's acceptable to discard the null hypothesis, backed by actual math and not intuition you cannot just toss out a number, say "this looks sus", and have it mean anything in a situation like this
|
|
#
¿
Sep 28, 2022 01:17
|
|
|
it's worth noting that while I think the statistics we've seen so far are extremely shoddy, that doesn't necessarily mean that the conclusions drawn are wrong. You can come to the correct conclusion shoddily. You should just be wary. I give more weight to non-quantifiable things like "it's very unusual for a GM to not be able to explain their moves" than poorly-quantified things like the tweet up above--though I'd give even more weight to a properly done statistical analysis
|
|
#
¿
Sep 28, 2022 06:11
|
|
|
Categorically excluding unquantifiable evidence (as opposed to quantifiable but as-yet-unquantified evidence) is an extremely narrow epistemological view that, if actually followed rigorously, leaves you basically unable to functionFangz posted:The difference between the distributions for Magnus and Hans above is easily statistically significant. The question is whether you can come up with a plausible alternative explanation. A statistical difference between Hams and Magnus just says that Hans and Magnus play differently, which is expected and obvious. You'd want to do a comparison between Hans and a wide sample of comparable-elo players and then make the assertion "Hans does not play like a GM." That graph is bizarre enough that I think the proper analysis would very likely turn up the expected result but I'm just frustrated at the lack of rigor in everything I've seen purporting to use numbers to prove something. And yeah, while the n is high enough it's probably statistically significant, I'd still like to see some notion of a proper test being done to prove it. The advantage of numerical evidence is that math doesn't care about vibes and can't be tricked by reading preconceived notions into charts, which is why I think its important to do the math right It sounds like the chess.com cheating team is working on a detailed report which will probably have the statistics I crave so I will look forward to that. I'm not sure twitter will ever have the level of rigor I really want because of the nature of the medium
|
|
#
¿
Sep 28, 2022 18:35
|
|
|
the actual way to test that hypothesis would be to compare Hans to all similar-elo players and see if his standard deviation is an outlier, not compare him to a specific player who you think he ought to resemble
|
|
#
¿
Sep 28, 2022 19:41
|
|
|
Nissin Cup Nudist posted:How is is that every world champion from 1948-2000 (except Fischer) was Soviet/Russian? Were the soviets running chess gulag camps and injecting kids with experimental big brain drugs? Paying off FIDE? Did the rest of the world not give a single poo poo about chess? there were definitely accusations of soviet players and arbiters conspiring to get their best players all the way to the top. Not knowledgeable enough about decades-old drama to have an opinion on whether it's true but people definitely claimed it
|
|
#
¿
Sep 28, 2022 22:06
|
|
|
having the government pay you to play chess definitely would have a positive effect on how many potential GMs you convert into actual GMs
|
|
#
¿
Sep 28, 2022 22:07
|
|
|
Anal beads
|
|
#
¿
Sep 29, 2022 04:19
|
|
|
uPen posted:It's not a great sign for them that their methods for catching cheaters are so flimsy that knowing how they do it invalidates the method. That's extremely common in many online games, because you can't actually detect cheating, you can only detect behaviors correlated with cheating.
|
|
#
¿
Sep 30, 2022 20:19
|
|
|
it consdiers it brilliant because it's a good piece sacrifice: you're hanging your other knight, but it's still the best move (apparently)
|
|
#
¿
Oct 2, 2022 23:30
|
|
|
chess.com also hands out brilliant moves more often the lower your rating
|
|
#
¿
Oct 2, 2022 23:30
|
|
|
mortal posted:Was that a brilliant move, or a great move? A great move is the only good move in the position, so if your opponent hangs a piece, taking it is often a great move. A brilliant move is a move the engine (at whatever depth) did not consider to be the best move, but after it's made, it reevaluates it as the best move. unless they changed it again, a few years back chess.com changed it so brilliant just means "good sacrifice"
|
|
#
¿
Oct 3, 2022 06:11
|
|
|
Can anyone copy the text out of the paywall?
|
|
#
¿
Oct 4, 2022 20:37
|
|
|
Nvm found it in a reddit comment: https://www.reddit.com/r/chess/comments/xvo7u4/wsj_chess_investigation_finds_that_us_grandmaster/ir22f7u/
|
|
#
¿
Oct 4, 2022 20:38
|
|
|
It also notes that his otb elo pattern doesn't really resemble that of other young grandmasters, though it stops short of making an accusation
|
|
#
¿
Oct 4, 2022 20:50
|
|
|
FIDE's investigation I possible otb cheating (as well as a simultaneous investigation into Carlsen's behavior) is ongoing so the ride isn't over yet
|
|
#
¿
Oct 4, 2022 20:52
|
|
|
Being good at chess does not make you good at tech, or have any idea what anti-cheat detection might be capable of
|
|
#
¿
Oct 4, 2022 21:03
|
|
|
Fortunately, they seem not to have made any otb accusations. They flagged a few tournaments that seemed suspicious to them but admitted their lack of expertise and deferred the verdict on those tournaments to the fide investigation. That seems like pretty responsible behavior to me The public airing of their evidence of online e cheating is pretty inevitable when Hans directly contradicts what they know on the subject While the full 72 page report doesn't seem to be public yet, the WSJ summarizes it as basically "Hans lied about the degree, extent, and time frame of his cheating on our website. You should not take his statements on the matter at face value, and here's why. Here's a few things we hope the otb cheating experts investigate." It seems perfectly in line with what an investigation by chess.com can or should be able to say, to me.
|
|
#
¿
Oct 4, 2022 21:15
|
|
|
Morrow posted:I'm not sure this is a joke post, but uh the thing that's different here is that it's a very big, public, messy scandal in which hans directly contradicted information that they had access to. There is a strong public interest in chess.com putting up the evidence for "Hans is lying" in a way that isn't true of other online cheating cases because they didn't become matters of public interest
|
|
#
¿
Oct 4, 2022 21:16
|
|
|
Chromatics posted:Do we know if it ever going to be public? if it doesn't become public then chess.com is startlingly bad at this and I'll jump on the conspiracy train
|
|
#
¿
Oct 4, 2022 21:18
|
|
|
I wouldn't be surprised if part of the reason chess.com doesn't reveal their methods (and lichess leaves the details buried in the source code of a github, rather than having a big page summarizing it) is so that the people who might otherwise go to a lot of effort to cheat in an undetectable way instead cheat in ways that they think are undetectable, but are not, actually
|
|
#
¿
Oct 4, 2022 23:33
|
|
|
qsvui posted:I feel pretty good after playing what I think is a clean game. But Stockfish says I made like 6 blunders and then I feel bad. as a general rule, if stockfish says something is a blunder, and then you look at the board for a bit and can't figure out why, then it's not a blunder at your level--your opponent won't see jow to capitalize on the supposed mistake either (probably)
|
|
#
¿
Oct 5, 2022 07:02
|
|
|
|
| # ¿ May 16, 2024 03:06 |
|
|
because the question of whether Hans cheated was of public interest and they had relevant evidence of far more cheating than Hans admitted to, op
|
|
#
¿
Oct 5, 2022 07:37
|
|