What do you think - is this game for real?

[Jonathan Rogers]

I am afraid this is another thread about inferring cheating from games played online. I wondered whether using an example might help, and even if not, I have become so distracted by this recent experience of mine that I would just welcome other people's views!

This is the game, not played in any competition, (and I was the loser, as you will have guessed) and the game was played at 10 minutes each with a three second increment. My opponent played at a reasonably consistent speed throughout, using maybe 12-15 secs per move, though some moves were played more quickly.

1 e4 e5 2 d4 Nc6 3 d5 Nce7 4 Nc3 Ng6 5 Nf3 Bc5 6 Na4 Be7 7 a3 Nf6 8 Bd3 O-O 9 h4 d6 10 h5 Nf4 11 Bxf4 exf4 12 h6 g6 13 Qd2 c6 14 dxc6 bxc6 15 Qc3 c5 16 b3 Ne8 17 Rd1 Bf6 18 Qd2 Bd7 19 Bc4 Be6 20 Bxe6 fxe6 21 e5 dxe5 22 Qxd8 Rxd8 23 Nxc5 Rxd1+ 24 Kxd1 Nc7 25 c4 Be7 26 Ne4 Bxa3 27 Ke2 Rb8 28 Rg1 Rxb3 29 Nf6+ Kf7 30 Nxh7 Be7 31 Rd1 Rb8 32 Rd7 Rc8 33 Nxe5+ Ke8 34 Nf8 Kxf8 35 Nxg6+ Kf7 36 Nxe7 Rh8 37 Rxc7 Kf6 38 Rxa7 1-0

Now, at the time I even sent a "well played" to my opponent; but later I began to think more about the striking contrast between his rating and quality of play, and then wondered more about the style of his play. There was a certain remorselessness about it, how many humans just leave that knight on a4, not trying to regroup it, and how many humans play 34 Nf8 as perhaps the cleanest win? And don't get me started on drawing inferences from 28 Rg1, which is certainly not a computer move and I suppose could have been a waiting move (realising that I was going to play ...Rxb3 and make my position yet worse). But again it seems a highly, highly unusual choice, even as a "human" trap, with this being a quick game and there being so many other tempting moves available - so could it also be a non computer move played as an attempt to divert suspicion?

I am not going to reveal his username here and I don't myself know his real identity - and whoever it is, he seems to be staying at the same low rating, so maybe he did just have a really good game against me (or maybe I had a very bad one and cannot bring myself to see it!). In which case - well, I already said "well played" to him.

But there you are, I do think it is a rather odd game and would be interested to hear the views of other forumites, including messrs Conlon, Wu and Turner.

[Ian Thompson]

I'd be wondering whether 28. Rg1 was a mouse slip. 28. Ra1 is the best move according to the computer and 28. Rd1 and 28. Rb1 are also plausible. Why would you move the rook to any other square on the first rank?

[Jonathan Rogers]

maybe though when I have mouse slips with long rook or queen moves, I still fall only one square short of my intentions.

[David Robertson]

Extremely dodgy - for the reasons you give. But also, intuitively, for the patternless, unfathomable, quasi-randomness of the moves (his, not yours ). And then, once Qs come off, W's pieces suddenly harmonise with uncanny fluency. Nf8 is symptomatic of a pure engine move (the earlier Qc3 looked dodgy too). I've been on the receiving end of hundreds of games like this; playing through this one felt very familiar. Hard/impossible to be sure though. The engine doesn't need to be the latest or the best to be good enough. But I say: engine

[Jonathan Rogers]

Thanks, yes I wondered about Qc3 as well. My own moves are obviously what they seem - a human trying and failing to find a good plan!

[Joseph Conlon]

I (personally) would be suspicious of this game - but not more than that without other evidence - unless there was objective reason to think the white player was very strong.

Running it through lichess's analysis gives 0/0/1 with the one blunder being Rg1 (which looks like a mouse slip on dragging the rook, being neither a human or computer move). If that move is taken away, then you have a 0/0/0 game with an acpl of 15 or so - not impossible for a strong player having a good game.

https://lichess.org/GDEwoiZS

To me 23. Nxc5 and 34. Nf8 both look unusual for humans, the former as exchanging the rooks first seems more natural, and the latter as its a hard move to see. What would make me very suspicious is if 35. Nxg6 took a similar or longer length of time than 34. Nf8.

Was there a very large rating differential? Sometimes I feel people see a scalp and decide to cheat for that game (e.g. a 1400 on lichess who berserked against me in bullet and then ran out 35 strong moves in 14 seconds).

[Jonathan Rogers]

Thanks. You ask about the rating differential - the answer is, not much at the time. We were randomly paired by the server, having similar ratings since I had only started to play at this speed. All he saw was a provisional rating of mine with a question mark attached to it, since I had only played and won two or three games. Perhaps he guessed from the question mark that I might be much stronger (a couple of days later, I was already some 400 points higher than him, despite losing a hatful in the above game) but that seems speculative.

So one's suspicions might be somewhat allayed by his staying at around the same rating; cheating is presumably not his habit, and it is not so obvious why, if he did cheat, he would just make an exception for me. It still remains that if he is a career 1800, he had a remarkably good, quasi-engine like game (though of course these things can happen).

[Li Wu]

Yup agree with game looking sus mostly from a human perspective, with 19.Bc4 and 22.e5 also looking strong to me.

From a pure move comparison perspective though a lot of the game was when white was already winning- so can't really be used for "simple" cheat detection techniques such as comparing with other human games.

As Joseph says- some timings would be good to know.

[Alan Walton]

Personally looking at the position the knight on e5 cannot move the rook is best placed on d7, therefore what to do with the other knight, after some relatively simple calculations Nf8 becomes obvious

[Matthew Turner]

Personally looking at the position the knight on e5 cannot move the rook is best placed on d7, therefore what to do with the other knight, after some relatively simple calculations Nf8 becomes obvious

Yes, but Black is just screwed, so why not just play Kf3. Would you really get on to thinking about Nf8

[Alan Walton]

Personally looking at the position the knight on e5 cannot move the rook is best placed on d7, therefore what to do with the other knight, after some relatively simple calculations Nf8 becomes obvious

Yes, but Black is just screwed, so why not just play Kf3. Would you really get on to thinking about Nf8

Good point, was only trying to say Nf8 isn’t that un-human move to make in this position, just some logical thinking

[Richard Bates]

You were playing a neural net.

[David Sedgwick]

maybe though when I have mouse slips with long rook or queen moves, I still fall only one square short of my intentions.

My mouse slips almost invariably result in my inadvertently coming to rest on the square next to the one I have vacated. Rg1 here would be typical of them.

I agree with those who have said that the game is suspicious. But I would give the winner the benefit of the doubt in the absence of other evidence.

[MartinCarpenter]

You were playing a neural net.

Not LC0 as set up on my machine at least.

The game actually looks pretty believable to me. Some club players simply do play relentlessly (and well!) for short term tactics/traps - hence Qc3, Nf8 etc - with only minimal positional sense and when the position suits them it can be very annoying indeed to play against.

[Jonathan Rogers]

Thanks all.

What everyone has said makes some sense, so I think the answer is that I am probably right in "not knowing what to think", and we shall never know for sure. Given that (judging by his rather stable rating) whoever it is at any rate is evidently not winning games like this on any serial basis, I don't think I will bother the site with the matter of looking at his other games.

It has still been interesting to hear other views, so thanks again.