I've seeing lately a lot of players that doesn't get detected by the current system as cheaters, but looking at their history i would say that you can guess they cheated several times. For example, they would have several loses against 1100-1300 players, but several wins against 2000+ players in the same period of time. Maybe there are multiple people using the account of course, altough i think that that it's not allowed here?
What do you think about this?
I can lose against 1600 while beating 2800 in the same session. Not sure if that's a clear sign of cheating.
CheatsOnlyNL
Depends on the time controls, right? In bullet chess, I would imagine that there's a lot of variance in results.
If someone is losing to 11-1300s, but beating 2000+, they might actually be a 2000+ who purposefully loses against lower rated players so the higher rated players lose more points than they should; this is called artificially increasing/decreasing their rating, and I've seen that this always leads to a ban. Stealth cheaters do exist though; I have caught some people who might cheat 1 out of 20 games or so; this gets really suspicious when they play perfect games against someone 500+ points better than them, but then perform with their actual strength against players at their level. When I see a 2600+ rated player lose to a 1600s, I run an analysis, and if the 1600 played a perfect game, this is super suspicious in my opinion. If I suspect someone of cheating, but I'm not sure, I look at other games. If they are a cheater, there will probably be other games that are really suspicious. I hope this explains something on how to catch cheaters. :D
@TricksOnlyNL is that on classical, blitz or bullet?. I was refering to classical games, where i think that is a lot more improbable, but of course possible.
@ijh yes, perfect games are obviously suspicious, but by "stealth" i was refering to those players that don't cheat in that obvious way, but by looking at their history with hints like those i was pointing out, you can say that there is a high probability that they cheat.
After reading your opinions, i think then, it's difficult to tell if that's is a sign of cheating.
There are a number of reasons for this phenomenon.
1) The very low rated player's rating is not accurate. For instance the other day I played against a 1200 player the other day who played like a 2200.
2) Suppose a boyfriend who is 2100 rated has a girlfriend that is 1400 rated. She plays on the same account as him because they share a computer, and he never bothered to log out.
3) Sometimes versus lower rated players you let you either let your guard down, or you play too aggressively, and make a mistake because of this.
4) The low rated player they lost to is using an engine for help.
5) They could just be having a bad game. If you botch an opening vs a lower rated player you can get in trouble pretty quick.
That is not cheating. I remember losing to a 1300 and winning against a 2000+ in the same Hourly Bullet Arena.
The problem with judging cheater by playing perfect games (~0 centipawn loss) is, that one has to check the opponents play manually. Some games have a small centipawn loss because the opponent played obvious blunders and they were just realized (a machine would do the same) ... I mean if the opening was a no-brainer, the opponent just hung pawns. In the middle game your opponent forcibly starts to lose pieces and you just take em. No brainer again and still perfect game, ~0 centipawn loss. Playing perfect endgames then is a no-brainer as well if the machine always calculates till the mate, you dont get additional centipawn loss (eg. mate in #4,#3,#6,#5#4#1,checkmate leads to no additional centipawn losses for both sides and even lowers their average cp-loss. so this can also be continued the whole time to lower the average centipawn loss artificially but ... who does this?). Also 19-moves-won-games without centipawn loss are not rare and are therefore "perfect games" alone without taking a look at the opponents play are a poor indicator. By the way this is also a reason why I dont value the average centipawn loss that much as an indicator for a good game myself very much. When the games on the other hand get complicated enough, even gms play at relatively high cp-loss rates.
Additionally creative play (playing opening traps, deviating from optimal lines, sacrificing inaccurately to draw the opponent into a complicated game) also leads to higher cp-losses (referring to those players who say that playing for openingtraps is not optimal usually). Who knows if the cheater on the other side just wants to exploit that fact and plays that stuff on purpose the first 10 moves? There are a lot more reasons which come to my mind. The report function is necessary. But looking at the cp-loss alone is not enogh to be sure. As well as a high cp-loss is not a criterion to be sure that someone does not cheat. An arbiter, probably a good tactician should check the game for "humanity". This is therefore a necessary subjective criterion imho.
