9

To start off - due to my anxiety, I do not play any form of chess against other people (except some friends of mine, but that obviously doesn't matter for the context of this post).

I recently got really interested in how cheating in chess operates and how it's detected. Specifically, this post gave me a lot of ideas how cheating is detected in general.

However, I have been wondering about the following scenario: let's imagine somebody is playing online chess, and they can see the evaluation bar in real time, with no additional information. For example, after blundering a piece, the player would immediately see that his opponent is better off. The same would be true the other way around - the player could see that his opponent made a mistake immediately, but without knowing what mistake that was exactly, without seeing any engine moves or possible threats.

So now my questions are:

  1. How would something like this even be detected? Especially at lower rating, since the player would probably not be able to interpret the results correctly (most of the time anyway).
  2. How much of an advantage would this really give? I assume for lower-rated players this would barely do anything, but for more skilled players, would this really be a lot?
  3. How much would the time control affect the impact of this? Would fast chess benefit less than classical?

Do you know / have you heard about this method being used by players? Any insights are welcome, as this is a question asked out of pure curiousity.

1
  • You can try this for yourself. Select a chess.com bot and a challenge level where you get hints or takebacks but don't use them. Jan 2 at 12:23

5 Answers 5

11

For 1):

  1. The player that uses an eval bar will successfully capitalize on opponents' mistakes more often than the player that does not, playing like a higher-rated player in this regard. However, they will not make fewer losing mistakes than a player at their true strength.

  2. The player will almost always spend more time thinking about a move following an opponent's blunder, no matter how complicated the tactic they blundered is.

A chess platform could perform some statistical test that reflects either effect (for example, 1 may affect the distribution of the player's "game shapes" on chess.com), and the player would be found as an outlier and examined further.

7
  • 2
    What does "For 1): 1. 2." mean here?
    – user24703
    Dec 31, 2021 at 12:14
  • 2
    @Quasímodo Answering "Question 1" with two different detection approaches Dec 31, 2021 at 13:39
  • These sound good in theory, but my head hurts trying to think about actually implementing a fair algorithm that would identify a cheater correctly without false positives constantly. The variance in people is so great, I don't think this could be done well. Plenty of players actually are better at capitalizing on an opponents mistakes than they are at avoiding their own (more focused on attack, than taking the time to picture the board from their opponents side) - I don't have stats but I'm pretty sure I'm around ~1500(at least 1400 on chess.com) and do this all the time still.
    – TCooper
    Jan 4 at 0:02
  • (cont) while I've played plenty of people who positionally teach me A TON, but miss an easy skewer or something when I muck up. On the second one... I don't even know where to start. People vary the amount of time in-between moves constantly for any number of reasons. We're talking online chess - I leave my app to answer texts then come back all the time. You can't avoid banning real players by these rules. I know you mentioned further review, but I think the number of flags with any software like this would overwhelm human staff quickly.
    – TCooper
    Jan 4 at 0:04
  • @TCooper How many times do you need to be distracted right after an opponent's blunder to be in the 1st percentile for this statistic? 0.1st? 0.01st? The people doing this analysis would know, and they can temper what percentile corresponds to "almost surely cheating" based on this information. And further review could very well be checking standings on the other statistical test(s).
    – Edward
    Jan 4 at 0:27
5

I assume for lower-rated players this would barely do anything, but for more skilled players, would this really be a lot?

I would expect it to be the opposite. Very low Elo players often hang a piece for several turns, with neither players noticing it. If a player sees the eval bar suddenly move several points, they can look for a hanging piece, available fork, etc. For higher ranked players, larger swings in eval are rarer, and at really high levels, if an engine says someone made a blunder, there's a significant chance it's wrong. Now, if a high ranked player could, for each move, make one tentative move and see the resulting eval before committing to the move, that could be very valuable, as they could identify a safe move and a risky move, and fall back on the safe move if the risky move turns out to not be sound.

3

Further to Edward's answre, suppose that a weakness of the player is that they are poor at spotting sound combinations that start with sacrifices. They might see BxP+ but reject it because it loses bishop for pawn, without seeing the longer-term advantage of a sound attack over the next few moves. Then their knowledge of the eval bar might encourage them to make such sacrifices -- if the eval bar goes down, they know to cut their losses; otherwise, that encourages them to go ahead and continue the attack so long as the eval doesn't go down. The eval bar has given them a dishonest proxy for long-range vision.

1
  • 2
    Eh, but if you've already played BxP you're probably better off continuing the attack even if it was unsound., hoping the opponent makes a mistake.
    – D M
    Jan 1 at 15:04
2

How much of an advantage would this really give? I assume for lower-rated players this would barely do anything, but for more skilled players, would this really be a lot?

Knowing the score is a huge advantage in an online chess game. You can adjust your strategy based on the score. This is like inside stock trading.

How would something like this even be detected? Especially at lower rating, since the player would probably not be able to interpret the results correctly (most of the time anyway).

As I mentioned in the other post, as soon as you use the information for your advantage, you will statistically perform better than everybody else in your rating group.

4
  • +1 on the first part, I don't think anyone else has made the massive advantage so obviously clear. Regarding the second, how do you account for the people who are actually improving in their own time? i.e. I play countless games on lichess anonymous for variety/casual gaming - then focus and try hard on chess.com - I'm sure my rating increase there is anomalous, but it's not dishonest. Not saying this won't work (the ambiguity in yours makes it impossible to disprove at least) but how do you avoid false positives? aka outside theory, how does that work?
    – TCooper
    Jan 4 at 0:12
  • Real improvements will be slow. Nobody can improve for hundreds of ELO in just weeks or days.
    – SmallChess
    Jan 4 at 0:13
  • 1
    Professor Ken Regan has a paper for the methods.
    – SmallChess
    Jan 4 at 0:14
  • Good call, not that many games(mostly win streak, actually), but it did take me a year+ to go from ~800-1500 - will be reading the paper tonight, not because I need to... but sounds really interesting. Thanks!
    – TCooper
    Jan 4 at 0:14
1

This kind of cheating provides a big (150-200, maybe more) elo advantage, according to a GM I asked.

Check out this position:

[FEN "5k2/8/5pK1/3B1P1P/3n4/8/3b4/8 w - - 0 1"]

(Black to play)

This position is from Carlsen-Caruana, World Chess Championship 2018, game 6. Before consulting an engine, how would you evaluate the position?

I suspect most humans would feel the game is drawn. Black can try, but it's hard to dislodge White from the corner. However, turn on an engine and you'll see that Black has a forced mate in 30. The line is deep enough that Kasparov wrote no human could've found it.

Or could they? Caruana's second, Rustam Kasimdzhanov, makes a telling observation:

"Maybe the move was indeed difficult to see. But it was not impossible. However, to find such a move you have to believe that you can win this endgame. And even if it is objectively drawn you can still try some tricks"

Based on the highlighted text, if Caruana did know Black has a mate in 30, my money is on Caruana finding the win. Here is another "missed" win of this kind; if Ding Liren had access to the engine eval I would also bet on him finding the win.

In the same way, if someone shows you a position (puzzle) and tells you "White to move and win", it becomes easier to solve. If they just asked you for what you think of the position, it's much harder. If you know there is a win (or if you know there is a defense), you look until you find it.

As for detecting such cheating, I have not seen any sources on how it might be done. I imagine most of it would come from time management: the cheater thinks hard and finds obscure wins or defenses that most other humans would not have found.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.