Cheaters plague

I'm really sursprised that cheating isn't discussed on LG and there is nothing happening against it. I am a FIDE Master, and yet I can't get past rating ~1750 because of tons of cheaters. On other chess sites, cheaters at least try to hide, don't admit, don't play all top engine moves etc. And LG had reached a point where such things happen:

Me: Are you using an engine?

Opponent: Yes

Me: Do you realise it is cheating and you should hget banned for it?

Opponent: Yes

This is sickening. I have yet never seen cheaters behave like that. So my question is: will anything ever be done against cheating? Or are we just supposed to sit waiting until there are more cheaters that fair players? Or is Richard trying to do something, but it is hard and time-consuming?

I have also noticed that the number of people cheating in chess is probably the bigest of all LG games. I don't know if I should continue playing chess here...

One more question, is it allowed to post names of cheaters in forums? On some sites it isn't.

Continuing the conversation:

Me: Is you = xxxx?

Opponent: sure

Cheating is a big problem and I wish Richard would at least ban obvious cheaters before they got their ratings high enough to give them satisfaction.

That being said, My rating is in the 1700s these days and used to be in the 1800s and my USCF OTB strength is also no higher than the 1700s. Maybe I'm making better use than you of a correspondence environment, analyzing position by myself. But I'm certainly not using an engine and these days I'm not even working hard at analysis. If you're really 600 points stronger than me OTB and no better here, you must be playing a remarkably high proportion of your games against rising (and likely but not necessarily cheating) opponents.

Chess is boring, anyway - play other games instead!

If you have firm evidence that multiple accounts are either the same person ("sockpuppets") or are at least collaborating, then please do present your evidence (which will necessitate naming them). Similarly, if you identify moves that humans would be very unlikely to consider but the best engines suggest, again, present your evidence.

I'm not saying anything will be done, but at least others will have the opportunity to evaluate the evidence, and perhaps blacklist those players if it is convincing. Similarly, the Tour of Little Golem has a strict policy, with the ability to kick out known cheaters, so please make these things known.

As you know from my mistakes, not all evidence is 100% bulletproof, so be moderate in any accusations.

three points:

1. it is quite easy to slip in cheating mode - I guess for most 'cheaters' it started that way, just using a program to see where they went wrong, and then using it to make sure you don't miss an obvious blunder, and then let's see if the program has another idea, and then... oops.

2. is there free software out there that automatically analyses a game versus program answers? Like, I load a complete PGN from the site here, feed it in the program, and (after an hour running) get a result like: '99% best three moves used, so probably cheater'? I would be willing to spend some time analyzing like the first league.

3. ever thought about an amnesty? everybody who declares that he cheated in the past, and wants to became clean again, is considered a 'refound son' and excused? Or would this be too much? Maybe that removes a bunch of them?

P.S.: Today is coming-out day, I heard!

I suggest, at the risk of repeating myself, playing a game where the human players still beat the best programs. That protects you from engines. Hex is such a game. And I don't think there are collaborating accounts ("sockpuppets") in Hex. Havannah is another such game.

I know of no other 2-person game with perfect information and no chance (hence theoretically deterministic) that has both chess's scope for creativity in a single game and chess's wide variety of very different-feeling paths that a game may develop (thanks mostly to the different combination of pieces that may be left when other pieces are traded off). So even though cheating is a problem, I will continue to play the game that first brought me to Little Golem and is still the one I enjoy the most.

If you don't like chess, you don't speak for everyone. There's no reason to say that cheating doesn't matter just because it's not your preferred game.

"There are many cyclists using steroids."

"Then do archery or sailing."

Does this make any sense?

At a different site I caught a player cheating at reversi. It turns out he's a famous player in a different game, has won on-the-board tournaments in that particular game, has invented several variants of that game, and he's the founder of the American Federation for that particular game.

Yet he feels the need to use a software program to win reversi games online. WTF?????

Cheating has been a big problem on LG - actually, these days the problem is significantly smaller than what it used to be! And somar96, if you think chess has the biggest problem, you probably don't plaay Reversi! :-)

I suggest, as FatPhil does, that you name the cheaters you find - but preferrably after confronting them with the accusation and being pretty sure about their cheating. Such information is valuable to others who want to avoid playing against cheaters, and also it enables us to have tournaments like the Tour where we ban cheaters. Additionally, some cheaters might read the post, become embaressed and stop playing here. There's also a small but probably non-zero chance that Richard finds the evidence so strong that he bans the cheater from LG - but I wouldn't bet on it...

Dimo admitted to me he was a cheater. He didn't say as much, but he is a Mahmoud.

That's 17383 (currently Dimo games) and of course (current name) MM, who's 16932 and (current name) Time Lost, who's 16931. 17386, whose current name is j, and 21128 (Abo_Mostafa) are also Mahmouds.

The people are talked about yesterday are exactly Dimo games, Mahmod and MM. The one with whom the dialogue I posted in 1st post is Mahmod aka MM. The re surely the same cheaters, and should be banned straightaway.

I asked Dimo as well, and he said "only in the opening". Which is cheating as well. You are just allowed to use a book/game explorer, which is different to an engine. And then opening has different meaning :-) But I am not 100% sure it's the same person.

Another dialogue:

Me: Do you use an engine?

Opponent: If I do use engine I need more time per move.

How would he know how much time is needed if he didn't use an engine? :)

Aganju, the software is here:

http://www.chess.com/download/view/chessanalyse-26

I haven't tried it though, I don't install much stuff on my computer.

And I heard about the day as well :)

To anyone who doesn't like chess, I am hopin and planning to make a living from chess in the future, and no way you can do that in other games I think :) An the cheating problem doesn't appear OTB (at least now).

Marius Halsor, I'm like a 1200 player in reversi and there is no way I can tell if someone is c ehater or not, but I can in chess in most cases.

For the sake of completeness, many people in the world make a living out of Go :-)

Yes. Every time you pass GO you get $200.

What does it mean to "pass GO"?

That's a phrase from the game "Monopoly", somar. The start area is dubbed "Go", and then you move round and round the board, and everytime you pass "Go", you get $200. It's just a joke :-)

I agree that suggesting you try another game is not a solution to your problem. The ideal solution would be that Richard banned known cheaters, but that's unlikely to happen. You COULD of course play only against players you trust, but that means no MCs or Championships. You can create user tournaments with passwords, and invite people you trust to join you. If you invite the top non-cheaters, that may even be considered a "shadow championship" or something. Then there are always other sites, of course, but who wants to leave LG?

Unfortunately you're almost right about only chess being suitable to play for a living among the games on LG. But only almost, as pointed out above - there are also professional GO-players. To my knowledge we have no professional Hex-players yet, but who knows?

Ok, I don't remember having played english monopoly, even though I like the game.

I would volunteer to help Richard with the bannings, but I don't think that will happen. Now I'm at least hoping for Richard's answer to this thread...

And it would be hard to make sure there are no cheaters in tournaments, there are also occasioanl cheaters which are almost undetectable...

You can't avoid all cheaters, but as you play a lot, you eventually come across many people whom you trust. Invite THEM to tournaments. I can not prove this, but I am very confident that in the Monstership (chech the forum for details) there were no cheaters, only honest, very good players.

I think absolutely differently. Cheaters who behave and pretend to be nice, honest people are as bad cheaters as the blatant ones. And I have no membership to make tournaments. Actually I will time out all of my games probably because I'm leaving for World Youth Championships in 5 days.

The main form of cheating Mahmoud does in man-in-the-middle, though maybe he's now doing a lot of relying on engines, too. When you play any of his aliases, you are very likely really playing an opponent of one of his other aliases, or perhaps someone he's supposedly playing on another site. He just relays the moves back and forth.

@somar: The link you gave me cost me a lot of time already, and all for naught. First, I needed to install it as 'admin', which forced me to close all open browsers and programs (not that big a deal, but I di that once a month normally). Then it claimed I need .NET 3.5 first. That cost half an hour to download and install. Then it claimed I need the SP1 for that .NET stuff. Another hour (why is the SP 1 double the size of the original??). Then, it crashed whenever started.

After some reboot and fiddling, I could run the program, but it cannot read LG PGN files. And it is a 'free' demo version only and wants 40 $ before doing anything useful.

So all for naught...

Does anyone know how to massage a LG-PGN in a Chess-Analyze-2.6-readable PGN?

oh, and in addition, it 'works with any engine', but doesn't bring one. So even after paying 40 $, and being able to convert PGNs to PGNs, it is no use, as I would still need to get a chess engine.

crafty is a good free engine. At least it's good enough for my purposes, which is to use a few times a year to go over a finished game where during play I felt just slightly out of my depth. You don't need any particular front end for it. You can run it by itself, but a front end may make it more pleasing.

somar: I agree that "Cheaters who behave and pretend to be nice, honest people are as bad cheaters as the blatant ones". However, after playing here for several years, I have learnt to know several of the players. And I trust them. Just like I trust my friends in real life when playing otb. When you've been around for a while, I'm sure you'll understand what I mean.

Dvd, if you haven't looked at Stockfish, please make it a point to do so. Stockfish is also a free engine, also with open source code, and runs circles around Crafty.

The latest 40/40 rating list (http://computerchess.org.uk/ccrl/4040/) lists the 1.7 version of Stockfish as rated 281 points higher than Crafty.

The latest 40/4 rating list (http://computerchess.org.uk/ccrl/404/) indicates Stockfish 1.7 as 371 points higher than Crafty!

Note that Stockfish version 1.9 was recently released, which is almost certainly a tiny bit stronger than 1.7.

I use WinBoard as my front end, most of the time. (I also have the Fritz12 program, which I can use with Rybka and Stockfish and all the others, but I usually prefer WinBoard.)

I haven't used that ChessAnalyse, so sorry about that Aganju :( I know it works by some other people, and you can still do quite a lot for free. I have no idea why it doesn't work, too many possibilities...

And yes, Stockfish is much better than Crafty.

And mahmoud is certainly using an engine, I can see that.

Ed, I'm not sure why I need a stronger engine. I suspect for what I use it for, anything over (at most) 2500 is superfluous. But if I do need one, why Sotckfish over the free version of Rybka? Is the free version crippled enough that Stockfish is better? Or is it just a preference for open source programs?

@Dvd - here's RoRoRo's latest Sons-of-Mahmoud list:

SonOfMahmod members #16929 mahmod

#16931 Time_Lost

#16932 MM

#16933 WHY

#16935 Back Again

#16944 Alexander

#16945 Vin

#17383 Dimo games

#17384 Black Horse

#17385 July

#17386 j

#17954 No Problems

I've not investigated 21128 yet, if the evidence is compelling, I'll add it to the list (and it's worth checking ids +/-1 too, as you can see from the above.)

How does RoRo do it? I agree with July, j, no problems, they are all cheaters no matter if they are all the same person. The rest I haven't played against.

Banning this one person shouldn't be much of a problem...

and Vin too

And how about http://www.littlegolem.net/jsp/info/player.jsp?plid=21926 ?

Regarding choice of engine - the best engine to use in order to detect cheaters is the same engine as the cheater! So that probably means the most popular program on the most popular platform. Strength is mostly irrelevant, and an engine that's too strong may even deviate significantly from what the cheaters' engines are doing, and thus not flag a match.

@somar - rororo simply maintains the database, he relies on me to tell him what to store.

Yes, but does it see the it by ip address?

I've thought about getting RoRoRo to sniff out IP addresses, but that requires him to play against the cheaters. He therefore becomes susceptible to feeding them ratings points. There are other techniques that I could use, but they would be considered hacking, and I'd rather not go down that route.

This is the kind of thing that should be addressed at the server level, I'd like to assist Richard in coming up with some kind of solution.

I'd like to see how a completely above-board bot, using the best available engine for long time controls, would do in chess here. I think its rating would help us humans understand what our own ratings mean in the context of this server. There are some problems, though. It would be unfair for the bot to compete in the championships or monthly cups and MiM cheaters would engage it as much as possible in random rated games.

Banning cheaters is the best solution :)

ok, I spent another three hours, and got a bit of that analysis program running.

Main issues (and I'm looking for solutions):

- the PGN files generated by LG do not contain '+' for check, but the analysis program considers this an error. I have been going through all the games one by one and adding the + but that is tedious at best.

- only 100 moves of a player can be analyzed (or pay 40 $ for full version)

- only the best move is compared, not second best etc.

- only 60 games can be analyzed (or pay 40 $ for full version)

Anyway, I got it running. I started on the top league of the running chess championship (15.1.1).

For example, I looked at one game each of Henrik and YHW (just a random pick from me); they come in at 50%/41% matches. I also run a full analysis on KARPOV (also just a random pick from me, no implications!), and found him to be in the 50-55% range of matches; and 'janowol' (also full analysis) 73.1% of a total 308 moves. In addition, for janowol, I looked at the difference distibution, and found several moves up to 0.5 worse than the best move, even a bad engine would never be so far off.

All done with 5 secs/10 plies limit per move, using the Rybka 2.3.2a engine.

Does it tell us anything yet? I don't think so. I should analyse a 'known cheater' to get a baseline, so give me known bad one. And then I need to learn to read the numbers. It will not be black/white, I'm afraid.

I analyzed a game Spasski/Fischer, they had 38-41% both.

I also analyzed a game I knew was played by an engine, and it came to 61% (same settings as before). Just as some input data.

@somar96: remember our game 1198868? Your value there is 53%.

Ug, I briefly considered trying to bolt stockfish into RoRoRo, but then I downloaded stockfish-191-linux.zip only to find this:

phil@geespaz:tmp$ unzip stockfish-191-linux.zip<br/>Archive:  stockfish-191-linux.zip<br/>   creating: stockfish-191-linux/<br/>  inflating: stockfish-191-linux/changes.txt <br/>   creating: __MACOSX/<br/>   creating: __MACOSX/stockfish-191-linux/<br/>  inflating: __MACOSX/stockfish-191-linux/._changes.txt <br/>...

and this:

phil@geespaz:stockfish-191-linux$ more src/mersenne.cpp<br/>/*<br/>   A C-program for MT19937, with initialization improved 2002/1/26.

and decided that I really didn't want to touch their code. (99.9% of people who use the mersenne twister know nothing about random numbers, and I taint the other .1% with the same brush in the hope that the stigma causes them to rethink its usage.)

Dvd,

I should probably take this conversation out of this forum and start another, since it's starting to get off topic.

When I analyze my tournament games, or games of GMs, I like to use the strongest engine available, even though many of my weaker engines, of course, are still also much stronger than I am!

And why NOT use the strongest engine available? For many positions, a 2500 engine will do what you need it to, just as a 3000 engine would - it will show you the tactics you missed. There are, however, many quiet positions the stronger engines "understand" better, and it's those type of positions that I like to use them for. Also, some of the stronger engines see more deeply "faster", meaning the best line of play is revealed more quickly and you don't have to spend as much time analyzing.

Of course, using SEVERAL engines, and comparing the moves and lines they ALL suggest, is almost certainly ideal. I've seen many positions where the fastest win is found by other engines much faster than Rybka finds it, believe it or not.

Crafty is fine... I have a copy of the latest version and I've been using Crafty for many years. But if I was going to use just one engine only, I'd pick Stockfish over Crafty, simply because it is noticeably stronger. I want the best and I would always feel I might be missing out on something otherwise.

I don't know how Stockfish 1.9 compares to the free version of Rybka. That's a good question and maybe for fun, I'll set up an automated tournament in the near future and have a match between these two. If so, I'll report the results here. I'm pretty sure though, the latest Stockfish is probably stronger. The latest version of SF ranks right up there in strength with Deep Rybka 4, and I believe Deep Rybka 4 is noticeably stronger than the free version of Rybka. And no, the free version of Rybka, to my knowledge, is NOT crippled at all. It's the very same free version that existed back when Rybka was free, several years ago. If SF is stronger, is because SF is much more recent. As you know, with each year the engines are all slowly getting stronger, thanks to both hardware and software advances.

I had a lot of fun earlier this spring and summer downloading engines, comparing them, setting up tournaments against each other, analyzing games and positions, with all of these engines, running automated test suites on them and comparing the results, etc. While I was away at work my home computer was often hosting matches between my chess engines. I had it set up where the results of the games would then be posted to my website, automatically. At work I could then peek in and check to see how the tournament was progressing! Fun stuff.

Note: I have zero losses in my last 8 over-the-board tournament games! (6 wins and 2 draws.) I KNOW all this fun I've had playing with my chess engines has made me a stronger player.

Fischer Spassky didn't have any engine to help prepare, more apropos would be the very recent longplay games of carlsen/anand/topalov/kramnik/etc as almost every serious player uses them to improve.

Stockfish 1.9.1 is clearly stronger than the free Rybka!

If you read the rybka forum you might find some ready to disagree as to whether rybka is crippled and that is the current version they consider crippled. Rybka is selling a rental of the latest version and it is necessary for in for it to be a lot stronger than Deep Rybka 4 if people are going to be willing to pay for it as a rental!

Other free strong engines Protector 1.3.6, critter0.8, thinker (WB), Spark0.5c/spark0.4!

@Aganju: 5 seconds per move seems low. Wouldn't an engine user be letting his engine run longer than that? I guess it's a question of how often the move found after 5 seconds is different from the move found after 1-3 minutes.

@FatPhil: Mersenne twister? What does a chess-playing program want with the Mersenne twister? I can see why this would give you pause.

@MarleysGhost: Right; I was assuming (and that assumption might well be wrong) that the result will be indicative with a short time already.

I made it quick mainly because I wanted a result quickly - consider that analyzing ~45 games with ~80 moves for about ~3 min each makes nearly 8 days of continuously running.

I would consider running it longer once I have established a process and defined the other parameters (min/max # of plies, do we consider the opening?, how do we remove obviously 'pre-draw' games, where the two players make 15 or 20 useless moves and then agree on a draw).

No problem with 8 days of net computing time, even though the process is semi-manual; I would need about five mins of setup and result readout per participating player, and the computer can run through the night.

The standard settings of analysis and results are:

Engine: Rybka 3 or 4, Houdini or Stockfish (newest version)

30 or 40 seconds per move (5 is too little)

A clear cheater: 60% top choice, 75% top 2, 85% top 3 (+/- 5%)

If it is included in this +/- 5%, it is probably a cheater, but not 100% sure.

Spassky/Fischer would get 50-55% top choice

ok, I run again with 30 secs per move, no depth limit. Takes a lot longer though.

Still for the top league of the current championship, 15.1.1.

I started with Johubber this time; result: 82% (231 of 291) match to the engines best move (again, the demo version doesn't compare 2nd and 3rd best).

More will follow in good time.