False Positives on Chess Cheating

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
noctiferus
Posts: 364
Joined: Sun Oct 04, 2009 11:27 am
Location: Italy

Re: False Positives on Chess Cheating

Post by noctiferus » Sat Jan 11, 2014 4:36 pm

Everything becomes clearer using the ROC curve:

http://en.wikipedia.org/wiki/Receiver_o ... acteristic

Uri Blass
Posts: 8435
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: False Positives on Chess Cheating

Post by Uri Blass » Sat Jan 11, 2014 6:34 pm

CRoberson wrote:It comes down to which of two systems is best:
1) You convict 110% of the cheaters.
2) You convict 90% of the cheaters.

Both systems are off by 10%, but the first one falsely convicts some innocents while the second case doesn't quite catch all the cheaters.

I say system #2 is best. Don't convict the innocent.
If you use system #1, eventually over time you will convict all people of cheating.
If you use system #2, eventually over time you will catch all cheaters without convicting the innocent.
I think that the options are different.
Option 1:you convict 20% of the cheaters (10% is false positive)
Option 2:you convict 10% of the cheaters (less than 0.0001% is false positive that is practically 0%)

I do not believe that you can convinct most of the cheaters.

Note that people can cheat in online internet games not only by chess programs and they can simply get an advice from their friend who watch the game.
I see no way to catch them and you have to accept that cheating is common in games that are not OTB games.

leavenfish
Posts: 254
Joined: Mon Sep 02, 2013 6:23 am

Re: False Positives on Chess Cheating

Post by leavenfish » Sat Jan 11, 2014 11:48 pm

I wonder about Chess.com's ability to detect cheaters on 'turned based'...essentially email games. It strikes me that it would be considerably harder to prove statistically someone is cheating IF they are a pretty good OTB player.

I ask because I found myself in a tournament multi-round tournament, basically I wanted a chance to try out some relatively new openings and generally passive openings.

First round - 1 decent player and he blundered right when I was about to offer a draw in one game and got his queen trapped in another. No problem in the other games...some even simply quit.

Round 2
- slightly better competition. One guy I slightly suspected of cheating...but figured maybe he was just a good player.

Round 3 - opponents relatively weak...1 good player but he too blundered and now it looks like I am going to advance to the final round. I really don't want to play any more, but I find myself getting wrapped up in the games and that keeps me away from my OTB training.

I haven't lost any games at all...about 4 draws I think. My old ICCF was 2398 - quit over a decade ago, it part because I really had trouble winning games....drawing wasn't a problem usually and only top flight players made me look foolish. USCF OTB somewhere under 2200 for about 20 yrs.

In these Chess.com 'turn based' games, when positions get complicated...you can move the pieces around all you want so you don't miss so much tactically. OTB...well, I miss more than I would like.

My Chess.com rating is something like 2350 right now so I know there isn't any real world comparison. Can't be...except that again, if you are only playing a few games at a time, it doesn't require one be named Carlsen to not make many tactical mistakes and get a high rating...just that you be a fairly accomplished player at the game. So, anyway, I'm just wondering how they (or anyone...ICCF lets you 'cheat') can even fairly reliably detect cheaters after a certain rating level.

bob
Posts: 20373
Joined: Mon Feb 27, 2006 6:30 pm
Location: Birmingham, AL

Re: False Positives on Chess Cheating

Post by bob » Sun Jan 12, 2014 1:51 am

leavenfish wrote:I wonder about Chess.com's ability to detect cheaters on 'turned based'...essentially email games. It strikes me that it would be considerably harder to prove statistically someone is cheating IF they are a pretty good OTB player.

I ask because I found myself in a tournament multi-round tournament, basically I wanted a chance to try out some relatively new openings and generally passive openings.

First round - 1 decent player and he blundered right when I was about to offer a draw in one game and got his queen trapped in another. No problem in the other games...some even simply quit.

Round 2
- slightly better competition. One guy I slightly suspected of cheating...but figured maybe he was just a good player.

Round 3 - opponents relatively weak...1 good player but he too blundered and now it looks like I am going to advance to the final round. I really don't want to play any more, but I find myself getting wrapped up in the games and that keeps me away from my OTB training.

I haven't lost any games at all...about 4 draws I think. My old ICCF was 2398 - quit over a decade ago, it part because I really had trouble winning games....drawing wasn't a problem usually and only top flight players made me look foolish. USCF OTB somewhere under 2200 for about 20 yrs.

In these Chess.com 'turn based' games, when positions get complicated...you can move the pieces around all you want so you don't miss so much tactically. OTB...well, I miss more than I would like.

My Chess.com rating is something like 2350 right now so I know there isn't any real world comparison. Can't be...except that again, if you are only playing a few games at a time, it doesn't require one be named Carlsen to not make many tactical mistakes and get a high rating...just that you be a fairly accomplished player at the game. So, anyway, I'm just wondering how they (or anyone...ICCF lets you 'cheat') can even fairly reliably detect cheaters after a certain rating level.
I don't believe reliable cheating detection is possible, except in OTB events. And as we have seen, more and more, it is not THAT easy in OTB games, there are lots of ways to get outside assistance that are very difficult to detect. It is possible to hold a cheating-proof event, but the costs are going to be really outrageously high. Entire tournament hall has to be a Faraday cage to keep all electromagnetic signals out. Power has to be filtered to prevent external signal introduction through that pathway. No windows to prevent any sort of optical signaling (infrared, etc). It's a problem, to be sure. In non-OTB events I think it might be almost impossible, if not impossible. Hence the ICCF allowing computers. Why make a rule (no computers) where there is no real hope of enforcing it?

Post Reply