Use of chess engine

@Toscani are you even aware that we're talking about chesscom ratings, which have a similar range as Lichess ratings?

Flip a coin and the value is still the same. Yet we have a winner. The one with the face up. A loser with the face down.
We need to change this perspective, so that we can appreciate more the game than the rating gain. Chess ratings have increased since engines started to existing. The more we train with engines, the more the ratings will increase.

@Cedur216 said in #26:

Yes, this needs an actual source.
Well what more do you want? I told you where to go see it. Tell me you don't know who Levitov is? Go to his channel, Fabi has only been on a few times. That episode is in english (as Fabi doesn't speak Russian to my recollection). Go watch. And the last C^2 podcast just had Danny on with Fabi and Christian. If you can stand to watch that pathetic snake for more than 5 minutes, watch that too. Fabi is real polite and doesn't get into a pissing match about the levels of cheating there, but just listen to what he says and doesn't say. He's not buying the 2% thing. Nor do I.

HOWEVER, with all that said, this isn't cc, and here they actually DO care about trying to stop it. I'm not saying the job isn't a huge one and they stop 100%, but they at least do not turn a blind eye to it. IF the people here are cheating enmasse, then they're a damn site better at hiding it than the ones on cc. To me it sure looks like the graphs of my OTB games. Again, I may be wrong, but I see few to 0 examples of error free chess here. And the games in the past that made me think I was cheated, NEVER gave me winning advantages, not even for a moment. Ymmv.

@Toscani said in #27:

The only real way to make a true rating grow is by the ratings in the pool. If you are not exposed to ratings in the master level you will never get a master rating. So 1800 is just an earned number and in this case it represents GM level, knowing the player was a GM. This is a prime example why ratings cannot be compared or assumed they mean something more than what they really are, a standings number in the group of players that caused that effect.

So, ...
Ratings are not universal indicators of skill;
They only measure rank relative to a pool;
Without some form of cross-pollination between strong and weak pools, rating inflation or deflation will occur; and
Always consider the context of a rating, not just the number !!

This is not really a valid argument, or at least not one that was completely thought out.

The trouble here is that the players he was supposedly playing were not 1800 level players, they were 3600 stockfish playing on 1800 accounts. With that in mind, it's not that they're not 1800, it's that 1800 isn't 1800. I'm sure it gets into a mess considering who you're paired with, if they cook-the-books on the pairings by having different pools of players (which I think they admit to having, as they have here).

So yes, you're correct about 1800 not meaning anything concrete, as in skill level. But I'd say you're coming at it from the wrong direction. Then too there is a lot of nuance in how we should interpret the "numbers". Imho, a rating number is only relevant to the person it is attached to, and only in a context of its delta over time. And even then, that is irrelevant if that person goes from playing cheaters to humans or the reverse. Then again, rating change is meaningless because your 1800 rating is going down because you're getting beat by a 1600 who's really a 3600... at times...

@Woland52 here is Grok 4’s response;

I totally get your frustration—online chess can feel like a minefield when suspicion creeps in, especially in longer formats like classical games where there’s more opportunity to consult an engine. It’s disheartening to pour time and skill into a game only to wonder if your opponent is getting a silicon boost. Let me break down my thoughts on this, based on what I’ve seen and know about the chess world.
First off, your estimate of 50% cheating might be on the high side, but cheating is definitely a real and persistent issue. From what data is out there, platforms like Chess.com report detecting and banning thousands of accounts monthly for engine use— for instance, they’ve closed over 1 million accounts since 2020 for fair play violations, with a small but notable percentage being titled players. Lichess, being open-source, also flags suspicious games, though their stats are less publicized. Studies and reports suggest the actual cheating rate in rated games hovers around 1-5% overall, but it spikes in higher-rated or prize-money events, and yes, classical games do see more of it because of the time factor. That said, your observation about erratic play (blunders early, then superhuman accuracy) is a classic red flag that anti-cheat algorithms look for—things like sudden spikes in move accuracy correlating with engine evaluations.
The anti-cheating systems aren’t perfect, though they’re getting smarter. They use a mix of statistical analysis (comparing moves to top engines like Stockfish), behavioral patterns (like unusual mouse movements or tab-switching if detectable), and even machine learning to spot anomalies. But cheaters do evolve—some use subtle methods like checking engines only in critical positions or employing “sandbox” setups to evade detection. It’s not that the systems miss it entirely; it’s more that false positives are a risk (no one wants innocent players banned), so thresholds are set conservatively. Platforms are investing heavily here—Chess.com’s Fair Play team, for example, reviews reports manually, and there’s ongoing talk in the community about AI-driven improvements or even requiring webcams for high-stakes games.
I agree it’s a problem that needs better solutions, and limiting it is crucial for the integrity of online chess. Some ideas floating around include stricter verification for accounts (like phone/ID linking), more transparent reporting tools, or even hybrid modes where engines are allowed but flagged (though that defeats the purpose for purists). In the meantime, if you’re fed up with classical online, you might consider switching to over-the-board (OTB) tournaments through local clubs—they’re harder to cheat in and often more socially rewarding. Or stick to platforms with strong reputations like Lichess for casual play, or even try variants like Chess960 to throw off engine-preppers.
What about you—have you reported suspicious games, and if so, what was the outcome? I’d love to hear more about your experiences to refine my take. Hang in there; the chess community is vocal about this, so pressure for change is building.

@V1g1yy said in #24:

In a Levitov interview, Fabiano Caruana said he made an anonymous account and couldn't make it past 1800 rapid on the other site.

I think I found this interview, and you seem to have mixed up a few things. Fabi reached 3100 as expected, but encountered the most cheaters at the 1800 level, which I find rather surprising. Fabi did not mention the time control at which he played, unless I overlooked something.

https://www.youtube.com/watch?v=ADK1o1sZ0s0&t=1725s

^^^^ Correct me if I'm wrong, but, (and I didn't listen to everything from there on) he did also say, and I quote "So yea, that was like 1 in 4 that got banned for sure, uh, and then eventually I managed to raise my rating and I stopped playing on that account".

Then too, I may be recalling some stuff that was said in follow-up interviews, and I know for a fact he mentioned he estimated 50% and got taken to task on that, and still does. He tiptoes around stuff here (as he should) in order to not cause a bigger stink than is necessary, but when I read between the lines, be mindful he said ONE FOURTH of the people were BANNED. Not 1/4th of the people were suspected. He suspected a LOT more. And listening to the sequence he told those stories, it wasn't until he got rating refunds that he managed to eclipse that rating range. IOW, he stalled there and needed points refunded to get past 1800. May not be specifically in that interview, but he did add that, (though I think it's in that video as well).

(edit to add) "Fabi did not mention the time control at which he played, unless I overlooked something."
Rapid

I agree that cheating of some form might be more common at longer time controls (I assume that's what OP meant by "classical").

But I don't think the number of cheaters comes close to 50 percent. Indeed, I cheerfully believe that cheating here is pretty freakin' rare.

Do I know that for sure? No, of course not. But there are two ways to get a feel for the answer:

1. If somebody's rating is DRASTICALLY higher at slower time controls I'd be more inquisitive (although not at all certain) --- that might be a subtle clue or .... it might just imply the player is older. As we age, our blitz prowess seems to slump a bit. Or maybe it's just me. And some people, even when young, just play much better slower. So, the test doesn't yield certainty, to say the least.

But if the blitz and classical ratings are similar, I wouldn't even be inquisitive. In my inexpert opinion, that would eliminate the possibility that somebody might be cheating.

2. Have the Lichess engine "evaluate" the game and see if the last half is played remarkably --not just better, but like Fischer's ghost.

But even that would not leave me certain that cheating is present. Even a patzer like me has, every once in a while, played a game at very, very high accuracy -- even 100 percent a few times. Or at least that's what I recall. Although I also remember being taller.

At blitz, I don't feel like I ever encounter cheating here. Indeed, I just don't feel like I encounter cheating here at all. Maybe I've been tricked! But I'm not going to worry about it unless I put money on the game -- and I never do.

.

The segment at 49:43 is another interesting one. "Here's your options... " Lol.

People that cheat do not necessarily cheat to rise through the ranks but, rather, to maintain the rating they believe they are entitled to have. Most of the "I'm realistically this good and need to stay here at all costs" cheaters are in the 1700-2100 range. They cheat not because they can't play at that level, but precisely because they can play at that level, thus they do not suffer too much cognitive dissonance (i.e., "I would have figured it out anyway, so why risk goofing up? Besides, it's late in the evening, and I'm tired.")

And the cheating is subtle; a few moves here and there, or help in guiding them through the opening ... "and then I'm on my own."

I think this is the most common form of cheating and it is impossible to differentiate from legitimate play. The problem is that so many people have been cheating in this manner for so long, that it's baked into the detection models -- that is, the standard reference for "normal" play already has millions of such cheated games baked into it.

My opinion is that if all engines disappeared tomorrow, ratings across the board would drop 300 points overnight.