Hacker News Reader

Comment by swyx | Hacker News Reader

yes but against magnus, who is supposedly two levels above Hans, this is not just a one move cheat, he'd have to cheat + have a continuous absence of mistakes, which is an awful lot of information to transmit.

i dont have a horse in this race i just like thinking about things in terms of information theory since this is a remarkable applied case

another way to decide this - have them play blitz (where the moves are way too fast for info transmission to happen), and see if the skill level scales accordingly?

View on HN

have them play blitz (where the moves are way too fast for info transmission to happen), and see if the skill level scales accordingly?

Not a fair contest. There are plenty of top classical chess players who are weaker in blitz and vice versa. It’s a different skillset. Classical is all about preparation for the opening followed by some deep thinking in the midgame. Blitz is all about pattern recognition and the ability to simplify down to an ending where you can blitz out the exact solution from a database.

Well, even endgames with as few as 5 pieces on the board are beyond what a human can solve with memorization. I don't disagree that blitz is mostly pattern recognition and rapid tactical thinking, but that applies all the way down to the end.

It's true that some people are worse at blitz, but if Niemann's OTB blitz rating is as good as his OTB rating at slower time controls, that's evidence against him cheating ... at least in ways that materially affect his rating. I guess it would be possible to cheat with an engine tuned to your actual rating just to make it less stressful.

Two levels if he isn't cheating. If he is, perhaps 3 or more. Recently magnus played hans on the beach: https://www.youtube.com/watch?v=T1OBEY99inw

magnus completely destroyed hans in two games, as black. I think the ease with which magnus took hans apart in these beach games, presumably added to his suspicion when hans played so much better in the Sinquefield cup.

Eh casual play has so many factors. I wouldn't put much weight on how badly a gm loses on the beach. For example, drugs could have been involved.

I do think Hans is cheating, but I think the proof will lie in statistical analysis of his games and demonstrating that he has an unusual (>3200 rated) propensity to clutch out specific moves. I think everyone suspects at this point that if Hans is cheating, its only a handful of moves per game.

>this is not just a one move cheat, he'd have to cheat + have a continuous absence of mistakes

Blunders are exactly what a device like the one described would seriously help with. If the buzz means both "there is an only move here and it's not immediately obvious" and "at least one of the natural moves here is a blunder or very inaccurate" then you need to just send a buzz and you've probably cut inaccuracies significantly. That said, a very simple communication device like this is probably badly hurt by a 15 minute delay.

> at least one of the natural moves here is a blunder

Interesting, I'm not sure if a computer has the ability to recognize something as a "natural move but also a blunder." It would require a very human-like way of thinking about moves, which computers don't generally have.

In any case, you would need a human operator who is a very strong chessplayer himself.

Magnus Carlsen has commented that Hans' mentor Maxim Dlugy "must be doing a great job"

https://youtu.be/c50PJmOj2-U?t=83

Anyone 1500+ can recognize the natural move--that's what makes it natural.

Probably the easiest case of "natural move but blunder" is anything that is a top 3 engine move when looking 3-5 moves deep, but losing significantly on deeper evaluation.

Also, this sort of categorization is at the heart of how chess puzzle collections are automatically assembled. A good chess puzzle contains an unnatural move that wins--the exact opposite of the natural but blunder. Chess sites scan their online play databases for these all the time, and serve them up as puzzles.

> Anyone 1500+ can recognize the natural move--that's what makes it natural.

Any human 1500+ can recognize the natural human move. The way the computer thinks about moves is different.

I really don't believe that Stockfish can tell you "this move is tempting, but wrong."

I'm sure you could build something in that might kind of work, but until I see it I'm skeptical.

Carlsen has broken 3000 in blitz before... Unfortunately he's too good for it to mean anything I think.

> he'd have to cheat + have a continuous absence of mistakes, which is an awful lot of information to transmit.

Perhaps a top-level player can jump to a higher level if they can stop worrying about coming up with brilliances in the macro strategy, and instead focus entirely on making their micro-level play spotless.

This is the opposite of what a computer-assisted player would do. Computer chess engines (generally speaking, somewhat less true of the latest generation) are not great at high level strategy but will never miss a tactic (micro level play).

That's outdated, now that we have deep learning the machines are great at high level strategy too.

> two levels above Hans

What does that even mean?

ELO gives you a statistical evaluation of how likely victory is for one of the player. Hans rating means he has non insignificant chance of winning against Magnus.

Hans can win without cheating as this last game proves. There is not a shred of evidence against him after all.

Reply on news.ycombinator.com