All of us had the bad feeling when we faced cheaters. Very often we may report, very rarely lichess may ban the cheater. That makes a task very difficult for decent fair players. What should they do? One natural response to lichess administration ignorance is that a player reports more and more, even if he/she is not sure that the opponent was a cheater.
Let us first see why lichess.org very often ignores our reports.
1. They might just don't conider reports except that one player has received many complains. If this is the case, it puts pressure on shoulder of fair players.
2. Weak engine detector: Not every cheater plays all moves based on engine suggestion, they could simply check their moves with engine and if it is not bad, then they will do it. In critical moments (e.g. when engine says +-4) they follow engine precisely. Or another approach is to do 2nd or 3rd best move of engine if they aren't bad. It is unlikely engine detectors take into account these factors.
What should be the strategy of the site against cheaters?
In my humble opinion, the strategy against cheaters should be zero tolerant. They should lose all their credibility immediately, no matter how many times they have cheated. The rationale behind this is that, we make life hard for cheaters and please civil players. Otherwise if they know site accepts certain level of cheating, they cheat occasionally.
How to help to improve cheater detectors:
Of course chess detectors (if there are automatic detectors) will consider certain factors like: are all move played with constant speed? Are there too much similarity with engine suggestions? Is there a lengthy tactical sequence (e.g. 4-5 non-trivial moves) that has been decided in matter of a second (if you watch lives of Hikaro in such cases thinks quite a bit). I think these factors have been considered by chess engine detectors, if not, then we are doing very poorly.
However, I believe there are other important factors should be considered:
1. The rating vs Performance: I, with a 2200 rating, lost against a 1200 player. I had a feeling that, that player wasn't an engine cheater, but another human played with that account. That player had 2400 performance in that tournament and nothing happened. This is also cheating. If someone is above 2000, in the first 5 games it goes above 2000, not that after hundered games remaining in 1200.
2. Tactical rating of suspicious player: if their tactic is not quite a bit higher than their rating in chess, then it might mean that they are quite vulnerable. Getting higher in sole tactics is much easier than in chess itself, therefore if someone has rating X in chess and its tactic is not at least 100 more, then they shouldn't be good at detecting tactics in the games as well, while this is not the case with cheaters. I have seen players with tactic ratings of 1800 and chess rating of 2200. Note that here I'm talking about well stablished rating, not a rating that is in its initial format.
You might say a cheater will also cheat tactics. But, it is more difficult to cheat on tactics, think about it, setting up the whole position, just to get some point and do this 100 times to reach the expected level? It is hard.
3. What is their performance in last 20-30 seconds of the game: since an engine cheater should simulate the game, they have difficulty when they fall into time scramble. It is not possible to properly use engine anymore, therefore they lose on time pressure very often, since their rating doesn't reflect their capabilities and most likely they play with a similar rating player, which of course has a better capabilities.
Why top players don't face cheaters too often?
They also face cheaters, but they win most likely and such cheaters will be banned earlier, since a report of a GM is more valuable than ordinary person (this is my feeling about lichess). The reason they lose is that, a cheater has to look at the engine and then move, therefore loses time. At the end they fall in time scramble and lose easily. Here is an examples that a grandmaster lost to a cheater: www.youtube.com/watch?v=IY-8VE1GoqE (unfortunately it is only in German).
I believe after FIDE Arena, lichess has the least credibility in dealing with cheaters (compare to its competitors chess.com and chess24). Therefor I want to raise this concern further here, and in live streams and discuss with you about the ways we can catch the cheater, or help lichess to improve its cheator detectors. When I want to discuss this huge issue, I will announce it a day in advance, so that we can see each other. Please share your opinion about other methods to detect cheaters, either here or in live stream. We may create a group for that: players against cheaters, to analyse possible cheating games together.
Let us first see why lichess.org very often ignores our reports.
1. They might just don't conider reports except that one player has received many complains. If this is the case, it puts pressure on shoulder of fair players.
2. Weak engine detector: Not every cheater plays all moves based on engine suggestion, they could simply check their moves with engine and if it is not bad, then they will do it. In critical moments (e.g. when engine says +-4) they follow engine precisely. Or another approach is to do 2nd or 3rd best move of engine if they aren't bad. It is unlikely engine detectors take into account these factors.
What should be the strategy of the site against cheaters?
In my humble opinion, the strategy against cheaters should be zero tolerant. They should lose all their credibility immediately, no matter how many times they have cheated. The rationale behind this is that, we make life hard for cheaters and please civil players. Otherwise if they know site accepts certain level of cheating, they cheat occasionally.
How to help to improve cheater detectors:
Of course chess detectors (if there are automatic detectors) will consider certain factors like: are all move played with constant speed? Are there too much similarity with engine suggestions? Is there a lengthy tactical sequence (e.g. 4-5 non-trivial moves) that has been decided in matter of a second (if you watch lives of Hikaro in such cases thinks quite a bit). I think these factors have been considered by chess engine detectors, if not, then we are doing very poorly.
However, I believe there are other important factors should be considered:
1. The rating vs Performance: I, with a 2200 rating, lost against a 1200 player. I had a feeling that, that player wasn't an engine cheater, but another human played with that account. That player had 2400 performance in that tournament and nothing happened. This is also cheating. If someone is above 2000, in the first 5 games it goes above 2000, not that after hundered games remaining in 1200.
2. Tactical rating of suspicious player: if their tactic is not quite a bit higher than their rating in chess, then it might mean that they are quite vulnerable. Getting higher in sole tactics is much easier than in chess itself, therefore if someone has rating X in chess and its tactic is not at least 100 more, then they shouldn't be good at detecting tactics in the games as well, while this is not the case with cheaters. I have seen players with tactic ratings of 1800 and chess rating of 2200. Note that here I'm talking about well stablished rating, not a rating that is in its initial format.
You might say a cheater will also cheat tactics. But, it is more difficult to cheat on tactics, think about it, setting up the whole position, just to get some point and do this 100 times to reach the expected level? It is hard.
3. What is their performance in last 20-30 seconds of the game: since an engine cheater should simulate the game, they have difficulty when they fall into time scramble. It is not possible to properly use engine anymore, therefore they lose on time pressure very often, since their rating doesn't reflect their capabilities and most likely they play with a similar rating player, which of course has a better capabilities.
Why top players don't face cheaters too often?
They also face cheaters, but they win most likely and such cheaters will be banned earlier, since a report of a GM is more valuable than ordinary person (this is my feeling about lichess). The reason they lose is that, a cheater has to look at the engine and then move, therefore loses time. At the end they fall in time scramble and lose easily. Here is an examples that a grandmaster lost to a cheater: www.youtube.com/watch?v=IY-8VE1GoqE (unfortunately it is only in German).
I believe after FIDE Arena, lichess has the least credibility in dealing with cheaters (compare to its competitors chess.com and chess24). Therefor I want to raise this concern further here, and in live streams and discuss with you about the ways we can catch the cheater, or help lichess to improve its cheator detectors. When I want to discuss this huge issue, I will announce it a day in advance, so that we can see each other. Please share your opinion about other methods to detect cheaters, either here or in live stream. We may create a group for that: players against cheaters, to analyse possible cheating games together.