4NCL Online

Venues, fixtures, teams and related matters.
Hok Yin Stephen Chiu
Posts: 148
Joined: Sun Dec 31, 2017 7:52 pm

Re: 4NCL Online

Post by Hok Yin Stephen Chiu » Thu Jun 11, 2020 2:16 am

Roger de Coverly wrote:
Wed Jun 10, 2020 10:25 pm
Peter Hornsby earlier made a valid point. He suggested that players need to be in a position to defend themselves against false accusations and they could do this by setting up a video and audio record of themselves whilst in play. That's less intrusive and probably easier to set up than giving an external arbiter access to home webcams.
it was once claimed in the 80s that nobody will ever need more than 640KB of RAM; lucky, nobody has yet suggested that we will ever need more than 640 TB of hard drive or cloud space! If so, we may find that even 640 exabytes of hard drive might be insufficient for holding all the video and audio files of each player playing two hours of chess each week at 4NCL!

I'm sure the engineers of the 50s who developed the first hard drive would have marvelled at the suggestion that their technology would one day go on to holding thousands of videos of chess players playing chess in their rooms, during a pandemic, to ensure that we were absolutely sure no cheating was ever taking place!
G. Secretary, https://WarwickChessAlumni.blogspot.com/
Delegate - Leamington
FIDE Arbiter

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: 4NCL Online

Post by Roger de Coverly » Thu Jun 11, 2020 7:13 am

Hok Yin Stephen Chiu wrote:
Thu Jun 11, 2020 2:16 am
I'm sure the engineers of the 50s who developed the first hard drive would have marvelled at the suggestion that their technology would one day go on to holding thousands of videos of chess players playing chess in their rooms, during a pandemic, to ensure that we were absolutely sure no cheating was ever taking place!
How then does dashcam and security camera footage work?

Paul Habershon
Posts: 550
Joined: Sat Aug 07, 2010 5:51 pm

Re: 4NCL Online

Post by Paul Habershon » Thu Jun 11, 2020 9:07 am

Just sharing some recent posts on the Quota Digest website.

Does lichess have less cheating compared to chess.com or not?
On Chess.com if you cheat, your userid will remain there for all to see the cheating stigmata next to it. Some of your friends will even know your real identity. On LiChess if you cheat, you can delete your account immediately, and it cannot be viewed again. So ask yourself: Where would a cheater rather roam?


André Cesarino
June 5
By “cheating”, it would mean “computer-assisted” moves, right? How can they know if you’re pulling this off?



Cecil Ohlson
Original Author · June 5 · 24 upvotes
T1 is the computer program’s (engine’s) best move. T2 is how often you played one of the top 2 moves. T3 means one of the top 3 moves. These are the stats for Magnus Carlsen: T1=40.85%; T2=57%; T3=70.43%

Hikaru Nakamura: T1=40.10%; T2=57.41%; T3=69.53%

When you consistently produce better stats than these then you are cheating, especially with large volumes.

So the sites simply look if you have a chess program running in the background. Now fools have tried all sorts of combinations to avoid detection, but the bottom line is that eventually every cheater gets caught.

Cecil Ohlson
Original Author · 21h ago · 1 upvote
Cheaters typically play one of the computer’s top 3 moves. When you do this more than 70% of the time, then you are cheating. Obviously the more games you play, the more certain they are of their conclusion. So to combat this, they cheat for a short while only. Some cheat at the beginning only, some at the end. Some cheat until they are up then play honestly. Some wait until they are down, then start cheating. Some do it for one or 2 critical positions only. The joke is on them because just as a human cannot beat a top computer program, so too, a cheater cannot remain undetected for long by a top anti-cheat program no matter what strategy they employ. In addition, certain moves are identified as engine-only moves. If you blindly play those moves because your secondary computer advises it, you are highly suspect but it could still be a coincidence. However, play too many of them and your coincidence defence will no longer be valid.

But the penalties for cheating are too lenient. It should be a lifetime ban for first offenders. Your parents, children, school teachers, friends should be contacted. Regular posts should be made on your FaceBook page. It should be made available for your job interviews. When you get married, it should be mentioned during the speeches. A wiki page should be opened in your honour, permanently detailing your cheating exploits…

Paul Habershon
Posts: 550
Joined: Sat Aug 07, 2010 5:51 pm

Re: 4NCL Online

Post by Paul Habershon » Thu Jun 11, 2020 9:10 am

Typo: Quora not Quota

User avatar
JustinHorton
Posts: 10364
Joined: Mon Aug 04, 2008 10:06 am
Location: Somewhere you're not

Re: 4NCL Online

Post by JustinHorton » Thu Jun 11, 2020 9:15 am

Paul Habershon wrote:
Thu Jun 11, 2020 9:07 am
The joke is on them because just as a human cannot beat a top computer program, so too, a cheater cannot remain undetected for long by a top anti-cheat program no matter what strategy they employ. In addition, certain moves are identified as engine-only moves.
What does this mean?
"Do you play chess?"
"Yes, but I prefer a game with a better chance of cheating."

lostontime.blogspot.com

Paul Habershon
Posts: 550
Joined: Sat Aug 07, 2010 5:51 pm

Re: 4NCL Online

Post by Paul Habershon » Thu Jun 11, 2020 9:26 am

JustinHorton wrote:
Thu Jun 11, 2020 9:15 am
Paul Habershon wrote:
Thu Jun 11, 2020 9:07 am
The joke is on them because just as a human cannot beat a top computer program, so too, a cheater cannot remain undetected for long by a top anti-cheat program no matter what strategy they employ. In addition, certain moves are identified as engine-only moves.
What does this mean?
Just to stress that those are not my words. They are copied from posts on Quora Digest. I probably should have italicised them in my original post.

Li Wu
Posts: 79
Joined: Tue Aug 09, 2011 3:01 pm

Re: 4NCL Online

Post by Li Wu » Thu Jun 11, 2020 9:41 am

MartinCarpenter wrote:
Wed Jun 10, 2020 9:33 pm
Its about what you feed in/how you use the numbers coming out of it in practice. That isn’t the behaviour of someone who’s given up on the whole idea, is it?
This is reasonable for someone to say if they wanted to have a proper discussion, but also reasonable for someone arguing in bad faith to say as well. Like I said, if we play a game of "name the largest number", the second person to speak always wins (and you haven't been a "first speaker" ever as far as I can tell in this thread).

Since you have failed to address my (especially the bold) points, can I confirm if this specific thing is addressed then you are happy with everything? Have a good think- if you can truly be satisfied. I don't mean to sound condescending, but be honest with yourself if your concerns can ever be fully alleviated (helps by writing out all of them in order). There is no problem if the answer is false, but it means that all of these arguments are pointless.

MartinCarpenter
Posts: 3044
Joined: Tue May 24, 2011 10:58 am

Re: 4NCL Online

Post by MartinCarpenter » Thu Jun 11, 2020 11:03 am

Li Wu wrote:
Thu Jun 11, 2020 9:41 am
Since you have failed to address my (especially the bold) points, can I confirm if this specific thing is addressed then you are happy with everything?
Your bold points were mostly avoided for politeness - as I said you'd seriously misunderstood what I was saying.

As for what would make me happy? That there was a clear cut policy, based on a false positive rate agreed acceptable, with correspondingly calibrated clear cut statistical cut offs and clear specifications of what grounds might be considered for appeal.

That would hopefully avoid all the problems with teams taking huge offence and pulling out, which has definitely been a huge problem this season.

From the discussions we've got an upper bound on the false positive rate on current (4NCL) policy of ~1/1000. There is quite a bit of uncertainty that will act to make it less good than that. Say 1/300 or something.

I'd quite happily convict online accounts on that, but personally I'd want a couple more zeros before making accusations that affected real people. They feel like quite a 'big' thing to do, and its very clear that people and teams are reacting that way in practice.

This might well make it very difficult for the stats to police based on just 1 season of 10(+) 4NCL online games.

For me that's how it is. They're not magic. The very real chances of post hoc bans changing who won each year are clearly a big problem for some people though.

We might have to somehow get people to agree that the bans are not to be taken as indicating guilt in any specific case.

These sorts of trade offs are really what we should be discussing.


There's several things that LiChess are doing that either have error rates that can't be understood/quantified (the NN), or seem much better designed to catch innocents than actual cheats. Those I really don't think should have a place in a properly constructed system.

Matthew Turner
Posts: 3600
Joined: Fri May 16, 2008 11:54 am

Re: 4NCL Online

Post by Matthew Turner » Thu Jun 11, 2020 11:17 am

"From the discussions we've got an upper bound on the false positive rate on current (4NCL) policy of ~1/1000. There is quite a bit of uncertainty that will act to make it less good than that. Say 1/300 or something."

Martin,
Could you explain where these number have come from or what they mean, I have to confess I haven't the foggiest

Li Wu
Posts: 79
Joined: Tue Aug 09, 2011 3:01 pm

Re: 4NCL Online

Post by Li Wu » Thu Jun 11, 2020 12:34 pm

Fair enough-all sound like good points.

Do you at least see the problem in attacking a fixed target? You state numbers without context which may not even have real meaning. If you wanted to seriously argue a point, go download and do some analysis into the effects you are describing as "might obscure the data". Maybe some potential issues are negligible? For example, it was stated that elo today is inflated compared to previous years, but this has been proven false. Not everything you state as a potential problem has been researched however, and unfortunately you can always argue with a new idea (and note ppl like Ken Regan have been doing research and analysing anything of interest for years). You are also not satisfied with the fact that these things were taken into account as a point of discussion.

The policy atm seems quite clear cut to me, and gives a good appeal process. As some others have posted- unfortunately team captains and players have to deal with a new reality- someone be banned with evidence that they might not see, or even if they do see, they are unable to understand.

I would argue that teams taking offence is something that can improve over time. Most chess players (especially those not technology or scientifically savvy) haven't been exposed to what cheating in chess or catching cheats mean, and there is a huge lack of understanding which broods mistrust. All will improve over time.

Keith Arkell
Posts: 928
Joined: Tue Dec 30, 2008 2:10 am

Re: 4NCL Online

Post by Keith Arkell » Fri Jun 12, 2020 4:12 am

As many of you don't appear to be on facebook I'm going to copy and paste a valuable piece Kenneth Regan wrote on my wall last night:

'The one thing I'll fix is that there may not be any context in which it is right to say my confidence levels are higher than those of Lichess. What is true is that the statistical model begins with a statement of odds based on considering a datum in isolation, which I call "face-value odds." It is really the natural frequency of the datum assuming no cheating. The "face-value odds" then get whittled down based on how many players there are (presuming they all would be fully tested if they had high results because people notice them), on the prior unlikelihood of a given person cheating, and supplementary information. I describe the process in my recent article https://rjlipton.wordpress.com/.../the- ... y-argument.../ but see also my long-ago article https://rjlipton.wordpress.com/2013/09/ ... woods-law/ and others like it on the prominent Maths/CS blog I partner. In particular, Lichess may have---and need to begin with---a much higher "face-value odds" threshold because they have vastly higher game volume.'

MartinCarpenter
Posts: 3044
Joined: Tue May 24, 2011 10:58 am

Re: 4NCL Online

Post by MartinCarpenter » Fri Jun 12, 2020 9:20 am

Matthew Turner wrote:
Thu Jun 11, 2020 11:17 am
"From the discussions we've got an upper bound on the false positive rate on current (4NCL) policy of ~1/1000. There is quite a bit of uncertainty that will act to make it less good than that. Say 1/300 or something."

Martin,
Could you explain where these number have come from or what they mean, I have to confess I haven't the foggiest
The number of people falsely convicted/The number of actual cheats caught. Very crudely its the 3-5% cheating rate and something like 1/30,000 being taken as enough statistical evidence.

There's then a bunch of factors of uncertain magnitude but which probably increase the rate a bit, and none I can really think of that might help. Someone did the calculation a bunch of posts back then David pointed out it probably was worse than that in practice.

Matthew Turner
Posts: 3600
Joined: Fri May 16, 2008 11:54 am

Re: 4NCL Online

Post by Matthew Turner » Fri Jun 12, 2020 9:52 am

Martin,
So you are saying you could run some numbers, but then you'll artificially reduce them to be prudent. That leaves you with with a situation where for every 301 people you ban for cheating one will be innocent. Or to put it another way, if the 4NCL ban 30 people this year there is a 1 in 10 chance that someone is innocent. If that is true then I think most people would be happy to accept that.

User avatar
JustinHorton
Posts: 10364
Joined: Mon Aug 04, 2008 10:06 am
Location: Somewhere you're not

Re: 4NCL Online

Post by JustinHorton » Fri Jun 12, 2020 9:59 am

One in ten? I wouldn't be.
"Do you play chess?"
"Yes, but I prefer a game with a better chance of cheating."

lostontime.blogspot.com

Matthew Turner
Posts: 3600
Joined: Fri May 16, 2008 11:54 am

Re: 4NCL Online

Post by Matthew Turner » Fri Jun 12, 2020 10:24 am

So what would you be prepared to accept?