Farewell

Most of his higher rated opponents also berserked, including DJ Haubi on those two games. So effectively these were all blitz games with blitz opponents.

I can see this will not be resolved any time now. Online cheating proof is done by stats. "You played too well so you must have cheated". And he's passed that threshold on this tournament. Q.E.D by modern methods. I dare say that in the past, criminal convictions were done on the best evidence they had at the time, and there must have been miscarriages of justice.

If he did cheat, we don't know how. Like they didn't know how Borislav Ivanov was cheating until they found something in his shoes. Even that must have been a very clever technical thing to do.

I don't think this is really about one person - it's about how online chess is going to progress and be effectively policed. It's a major issue, much discussed.

The percentages above are comparing Magnus Carlsen in one tournament (where he played stronger opposition) to an online player here in one tournament. Maybe pepellou played better in this tournament than Magnus did in that one. It doesn't mean he's a better player as consistently, over a longer period of games, Magnus would have better stats, and we're not comparing like with like, as they weren't up against the same opposition.

It's like comparing tennis players and their unforced errors. The errors may appear to be "unforced" but I guess you'd make a lot more of them if the guy over the other side of the net is going to punish you badly if you don't play your best stroke, i.e. the stronger the opponent, the more likely you are to make an unforced error against him as you're playing under a lot more pressure.

Many of us were asking "on what evidence was he banned" but lichess wouldn't provide it, either to us or to him. That left us speculating. Maybe why this topic has gone on so long.

I mean, how many of us have played a good game of chess, totally fairly, but then been worried we maybe played too well that nobody would believe us? So then, maybe, we run it through the engine and are "relieved" that actually we didn't play as well as we thought we had. But this is the fear we face - if we're not believed, we can be booted out with no way to return.

Sadge

@cashcow8 said in #101:
> [...]
> If he did cheat, we don't know how. Like they didn't know how Borislav Ivanov was cheating until they found something in his shoes. Even that must have been a very clever technical thing to do.
> [...]
This is, from the best evidence I've read about that case, a long-running myth, often repeated. It is reasonable to assume that Ivanov was indeed using something in his shoes because of his refusal to remove his shoes when asked. But exactly how he cheated has never been conclusively established.

But that he did cheat (over the board in his case) is beyond all reasonable doubt. And that situation, I suggest, is typical of everyone busted for online cheating too.

By the way, there are methods involved in detection and proof beyond reasonable doubt other than just statistics and correlation with engine moves, though those statistics do indeed form the dominant method when it comes to online cheating detection.

This is intriguing. Just sharing my two cents on this. Let's get the facts straight first. What we know so far and agree on.

1. So from the web archive, we can see that Pepe has played tens of thousands of games spanning several years here - http://web.archive.org/web/20221126163648/https://lichess.org/@/Pepellou

2. Looks like he was banned after participating in this tournament - lichess.org/tournament/buowiSL6

3. He did beat a grandmaster (2550) there, but I don't find anything suspicious in that game - lichess.org/Np1ZQ3F1. In the game, he had 0 inaccuracies, 2 mistakes, and 1 blunder made with a 92% accuracy and 31 average centipawn loss. It is worth noting that the only blunder he made seems to be a mouse slip. But the moves he found, there isn't any good evidence to suggest he used an engine to find them. There aren't any spectacular moves or lines that only engines are likely to find. Only a nice tactic here - lichess.org/Np1ZQ3F1/white#51, but considering his rating (Almost 2400), I find it reasonable for him to play like this. The grandmaster congratulated him in the chat and it felt sincere too, suggesting he didn't find anything suspicious in the game as well and perhaps wasn't the one who reported him.

4. Then he goes on to play this game with a 2000-rated player - lichess.org/3ca65f9R. Only 3 inaccuracies, 0 mistakes, and 0 blunders, with a 95% accuracy and 20 average centipawn loss. Which is fine, and again, nothing suspicious. The moves he played seem natural enough for a player with his rating. And he slowly converted his advantage.

5. What I find strange is that after this game, there is a comment from his opponent in the chat room saying "byebye pepelou", as can be seen here - lichess.org/tournament/buowiSL6. And that seems to be his second last game in Lichess, after which he was banned.

6. There is also this game he played against a 2450 - lichess.org/npLB3hqH. Here, on move 20, he thinks for a good 30 seconds and blunders from -7.8 (Depth 37) to -2 (Depth 37) by playing axb3 instead of the engine-recommended Qa5. This makes a lot of sense as a human because the queen sacrifice is cool and opens a bunch of tactical and mating opportunities (which was what actually happened later on). On the contrary, the move suggested by the engine is not so obvious as it seemingly hangs the bishop

7. He seems to be banned from chess.com as well for violating the fair play terms - www.chess.com/member/pepellou. At the same time, lots of users are vouching for him saying he is the last person who would cheat.

Now that we have our facts straight, let's try to fit all of these together.

1. It is entirely possible that Pepe cheated on this tournament. Although from what I have examined so far, I do not see any conclusive evidence for this claim.

2. It is also possible that Pepe didn't cheat, and this could be a witch hunt. I am not familiar with how cheating detection and banning work in Lichess, but could it be that some users reported Pepe after his strong performance in that tournament? It could be also automatically picked up by the system. And then a moderator investigated and concluded that this is a case of cheating and banned his account.

3. It is also possible that Pepe didn't cheat in this tournament, but cheated on some of his earlier games. He got flagged for his performance in this tournament, and after a thorough investigation, the moderators found strong evidence for this and closed his account. Combine this with the fact that his chess.com account is also banned for fair play violation, then this makes sense. But this also begs the question, that why was it undetected for so long in Lichess, unless he only cheated in the more recent events.

4. What I find strange though is that comment I mentioned in the tournament earlier, "byeybye pepelou". I couldn't find a timestamp for that comment. It could take place after Pepe's account was banned, but if it was shortly right after the game BEFORE his account was banned, then this seems somewhat fishy to me. Because the message is quite clear, concise, and confident. "goodbye pepelou", as if the user who wrote that knows for certain Pepe will be gone. Assuming that is indeed the case, I am not sure how that could be unless the user in question is a moderator.

No matter what happened, banning someone's account is a big step. It doesn't matter if it's automatic or manually done, with great power comes great responsibility and we should bear in mind that someone should be innocent until proven guilty with a strong conviction. Without visibility and accountability, power is often misused. Even when it's not, it gives the perception that it is being misused.

Pepe should not be given any special privilege for being a renowned member and contributor in Lichess, he should be treated the same way as any other. But every user deserves this. A message or an explanation if their account was banned, because this is a big step. I don't think it should be a blanket term only, that they violated the terms and conditions. It should be more than that. If not initially, then definitely after the appeal. If someone has already investigated, then it doesn't take much effort to write a few words explaining the reasons behind the decision, does it? It gives clarity and visibility and ensures a fair process for everyone. As I said, without clarity and visibility, it often gives the perception of unfair treatment even if a fair process is ensured in the background.

Apologies for the long post. I do realize I am not very familiar with how moderators and cheating detection work in Lichess, and I appreciate all of you for the work that you do and keep this site running for free with all your hard work and dedication.

@sgtlaugh said in #104:
> 4. What I find strange though is that comment I mentioned in the tournament earlier, "byeybye pepelou" [...] as if the user who wrote that knows for certain Pepe will be gone.

the tournament was over. of course the user knew that pepe would be gone, as well as all the other participants of the tournament. that's what happens when a tournament ends, everyone is leaving the tournament page & tournament chat to do other things. saying goodbye was only being polite. but of course, we can also act like that was super strange and try to construct a conspiracy theory, why not.

@glbert said in #105:
> saying goodbye was only being polite. but of course, we can also act like that was super strange and try to construct a conspiracy theory, why not.

It seems like the definition of "polite" in this case really hinges on how calibrated you are to see a remark as passive-aggressive.

While I think @sgtlaugh presented a fairly neutral and valid summary of this thread, it's unlikely the comment in the tourney was from a mod, if anything it's more likely it was from the player who first reported Pepellou for cheating.

To me, it sounded like something a sore loser would say. Or someone gloating right before filing a cheat report.

Framing the conversation as either completely null—or full of wild conspiracies—is not a good-faith argument toward the larger issues being discussed.

@cashcow8 said it best in #101: "...how many of us have played a good game of chess, totally fairly, but then been worried we maybe played too well that nobody would believe us? So then, maybe, we run it through the engine and are "relieved" that actually we didn't play as well as we thought we had. But this is the fear we face - if we're not believed, we can be booted out with no way to return."

Whether Pepellou was actually a cheater or not doesn't change the fact that cheat detection as it currently stands, at the very least, creates the perception of a ceiling for fair play.

Regardless of what happens with his specific case, acknowledging, or at least addressing, in a good-faith discussion the issue of human bias within the human-machine cheat detection system would be a step in the right direction towards resolving this in a positive way for the community.

Nicely articulated @spidersneedlovetoo, and I couldn't agree more.

You're right when you say framing the conversation is not a good-faith argument (it is not a logical argument at all). It could be anything, so let's not deep dive there since we don't know for sure what happened. But I did find it strange, which is why I mentioned it, and it also felt to me like something a person would say after reporting.

I agree with @cashcow8 and you, and perhaps a lot of other users. After seeing more incidents like this, I would be genuinely concerned and somewhat scared if I played in a tournament or any serious event and performed well, better than I expected, better than my level. I am sure many other users can relate too.

What seems even more dangerous to me is that many users here, including high-rated users, seem very convinced that he cheated. These users are not moderators, not part of the Lichess team, but are coming to a conviction after analyzing his games by themselves and from their knowledge and experience. Yet, these users are failing to provide good insights or reasoning behind their convictions. I am sorry, but just because you took his games and crunched some numbers on a tool, what does that prove? Humans are not machines, and statistics, while they can be very useful, are often a double-edged sword if misused. How well a player is playing varies a lot, for most of us. It depends on our skillset surely, but also our mood, emotional state, concentration, settings, momentum, and even our opponents.

Granted, even then, if a lower-rated player was finding absurd engine lines consistently, it'd pretty much leave no plausible doubt that they were cheating. But I honestly did not find anything of that sort from Pepe's recent rapid games. He often made blunders and had several missed opportunities. And it was not just with lower-rated opponents. I had a conversation in DM with someone who said there is a discrepancy in how he plays with lower-rated and higher-rated players because with lower-rated players he doesn't need to use an engine but with higher-rated players he does. And the mistakes and blunders he made with higher-rated players are intentional to bypass the cheating mechanism, or probably he wasn't using engines on those games.

Now, I do understand that it may very well be the case. What bothers me is their strong conviction. Because it seems to me like it has more to do with personal biases and faulty analysis and not thinking through. There could be other plausible explanations too, ones I find more likely actually. If you say he made mistakes intentionally to avoid the cheating system, then couldn't that be said for everyone? How do you differentiate a cheater? The only reliable metric from their games alone would be how consistently they are performing or finding good/engine moves compared to users of a similar rating. And if it's a high statistical deviation with a very high confidence level, then we could argue that they most likely cheated.

And regarding having more accuracy with higher-rated players, this just demonstrates my point. That this is a complex scenario with so many different factors, we can't always cherry-pick or think from the surface level. If you think about it initially, someone having higher accuracy with very strong opponents, but lower accuracy with low-rated opponents, why may that be the case? Surely if they can play better with high-rated opponents, they should at least be able to play on the same level if not better against lower-rated opponents, right? And if they don't, if they consistently have a lower accuracy with weaker players, then it possibly implies that they are using external assistance when they are playing higher-rated players.

Guess what. On the surface, it makes so much sense. But if you think about it, the argument falls apart when you try to find some other explanations for this behavior. Although it seems counter-intuitive, it makes sense that someone can have higher accuracy against higher-rated players. In fact, I believe it will be true for most players. If you know how accuracy is calculated, then it makes sense. Lower-rated players will play worse and will provide you with more opportunities to exploit. Granted, you will exploit most or some of them and will probably win the game. But the ones you won't exploit, guess what, they count as inaccuracies/mistakes/blunders and reduces your accuracy. Most of them will be mistakes and blunders. On shorter time controls, you will move fast and won't even expect sometimes that you could exploit these mistakes. Because it could be the case that you are not used to playing weaker opponents, most of your games happen between opponents of similar ratings. And they won't make mistakes like that, especially in the opening/middle game.

Compare that with playing against higher-rated players. Where they will have a more solid approach. They will provide fewer opportunities to you, which means chances are you are going to lose, or that it's going to be a grueling fight and they will slowly convert the game to their advantage, strategically and positionally in the long run. That'd actually mean more accuracy for you, and if you manage to do the same and win the game, even more so.

The funny thing is, I didn't find anything of that sort in Pepe's games. The last two of his games are against 2000-rated opponents, both of them having 90+ accuracy. Did he cheat on those too? What's more ironic is that the user who mentioned such an argument, I quickly examined his last few rapid games. His last two rapid games against 2050-rated opponents seem to have an accuracy of 87% and 88%. On the other hand, his very last game against a1700 had an accuracy of 58% and another one down the list against a 1500 had an accuracy of 79%.

All in all, to the users who are claiming Pepe cheated just by analyzing his recent games, I am very curious and interested to hear your arguments. Please don't suggest reading a particular book. Please don't provide shallow arguments or just some numbers you crunched with a tool. But actual logical arguments and reasonings which led you to your conviction. Surely if you spent hours reading a book or using a tool to analyze his games, it won't take much to present your reasoning to the community.

I like to believe that Lichess found very strong evidence that led them to ban a renowned community member like Pepe. But from what I've noticed from the community, there is often a witch hunt when it comes to cheating. And this is neither healthy nor fair for everyone.

> All in all, to the users who are claiming Pepe cheated just by analyzing his recent games, I am very curious and interested to hear your arguments. Please don't suggest reading a particular book. Please don't provide shallow arguments or just some numbers you crunched with a tool.

No significant mistakes in balanced positions despite the very short time control. "Mistake" meaning either giving opportunities or missing some.

Humans make mistakes, all the time, especially at short time controls.

@earlpurple said in #95:
> I just had a look at the closed chess.com pepellou account. Seemed it was active between around 2013 and 2015. It seems strange they would allow someone onto their staff who has had an account in the past closed for cheating, even if that was in 2015, but he is a developer on Chessable which was PlayMagnus and they took it over.
>
> The essence is that pretty much all online cheat detection is done by statistics. "I don't believe they can play that well so they probably used assistance". You could call it "circumstantial evidence".
>
> I'd quite like to take a bunch of these players and gather them to play either OTB or hybrid, mixed with a few lower titled players and see how they all get on.
>
> As we know, there are probably 2 types of cheat: those who, without assistance, would be totally hopeless, and those who would hold their own in the contest but with it, even a couple of key situations per game, are genuine contenders for winning.

But chess com claims when in doubt they have titled human players who check the games manually, not by statistics alone.

@hicetnunc said in #108:
> No significant mistakes in balanced positions despite the very short time control. "Mistake" meaning either giving opportunities or missing some.
>
> Human make mistakes, all the time, especially at short time controls.

I am sorry, but that's not a good argument. First of all, he did make mistakes and blunders. A blunder is by definition a significant mistake isn't it? Are you saying he didn't make as many mistakes as he was supposed to make considering his rating? If yes, then how are you arriving at that conclusion? Did you compare his performance in this tournament against his performance at previous tournaments? Against similarly rated players in this tournament and Lichess? And did you also take a look at those other players' past tournament results to eliminate the possibility that he didn't cheat, and he perhaps was in a good form and he just played well, combined with some luck from his opponents? Because you would only find this out if you compare similar-rated players AND their overall track record, not just from one tournament. That would give you good and reliable statistical confidence about the claim that he cheated.

You could also argue from another viewpoint. That how likely it is for his moves to come from an engine. Again, that'd require thorough statistical analysis. Unless something is really off, let's say he finds crazy ideas that seem very unnatural for someone of his caliber to find by himself. I took a look at all his games, I didn't find any such cases. You are welcome to share if you have any different findings.

You seem to focus a lot on the phrase "shorter time controls". I do agree, that in general, the more time you have, the fewer blunders you will make. However, this could also be somewhat subjective. Not entirely though. What is short-time control for you could be sufficient time for someone else. Wouldn't you agree? When I take a look at your profile, I see that the difference between your Rapid/Classical with your Bullet/Blitz is around 200 rating points. Perhaps you are more comfortable in longer time controls, and you feel that people will make significantly more mistakes in shorter time controls. This is not necessarily true. Take it from someone who plays Blitz/Bullet significantly more than Rapid/Classical. 3.5 minutes is a lot of time. Plus you get to think during your opponent's clock too, and it was 7+5 initially and he berzerked, so he had 3.5 minutes, plus 7+ minutes from his opponent's clock. I wouldn't exactly call it a short time.

If I take a look at his archived profile here - http://web.archive.org/web/20221126163648/https://lichess.org/@/Pepellou, I can also see that he played Bullet/Blitz a lot more than he played Rapid/Classical. Perhaps he felt more comfortable in shorter time formats.

Again, I have to emphasize that I didn't find any moves or engine lines from Pepe's games in that tournament that can lead me to question how on earth did he find that move given his rating and the time he had. But for playing naturally developing moves and punishing his opponents' mistakes, I don't think it is fair to conclude he cheated based on the presumption that people make mistakes all the time, especially in shorter time controls. It can be a very slippery slope.