4NCL Online

Venues, fixtures, teams and related matters.
User avatar
JustinHorton
Posts: 10364
Joined: Mon Aug 04, 2008 10:06 am
Location: Somewhere you're not

Re: 4NCL Online

Post by JustinHorton » Tue Jul 14, 2020 4:37 pm

OK, I've read Alex's article (I would have done so before, but my laptop was in computer hospital) and although most of it is just good straightforward and helpful explanation. There are some small aspects of it that I'm not sure about, or want to take issue with:
  • I'm not sure what's meant by "a lead of 4-pawns or a lead of 9-pawns"
  • Nor am i sure what "training data" is
  • Where it says "using a method published in academic papers with Guy Haworth in 2011" I wouldn't mind seeing their titles and other details, in case anybody wants to look at them.
Perhaps unsurprisingly, this annoyed me
A common thing I have heard over the past few months is that websites rush into flawed judgements with regard to issuing bans. My experience is the opposite. The burden of proof required by a website to automatically flag someone is high
since my experience is very much the opposite. Still, people have different experiences, mine therefore being no more valid than his, but I wouldn't mind having seen some scepticism in the piece, which is very much no problem, nothing to see here where flawed judgements and so on are concerned. You would think reading this paper that all the relevant experience was on one side.

But if you're going to write
I can’t comment on how any platform’s systems work, because I have not seen them
then don't! Don't comment!

Also:
so far as I am aware, chess websites actually work to “comfortable satisfaction”
How are you aware of this? Do they say so? Where do they say so?

Incidentally, this:
You can never be 100% certain, if a court applied this test then it would never convict anyone if the defendant didn’t admit it
isn't true, is it?
"Do you play chess?"
"Yes, but I prefer a game with a better chance of cheating."

lostontime.blogspot.com

Ian Thompson
Posts: 3551
Joined: Wed Jul 02, 2008 4:31 pm
Location: Awbridge, Hampshire

Re: 4NCL Online

Post by Ian Thompson » Tue Jul 14, 2020 5:33 pm

JustinHorton wrote:
Tue Jul 14, 2020 4:37 pm
  • Nor am i sure what "training data" is
There are several explanations of it here - https://www.chessprofessionals.org/cont ... ting-chess
JustinHorton wrote:
Tue Jul 14, 2020 4:37 pm
  • Where it says "using a method published in academic papers with Guy Haworth in 2011" I wouldn't mind seeing their titles and other details, in case anybody wants to look at them.
Lots of papers referred to here - http://talkchess.com/forum3/viewtopic.php?t=65772

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

Re: 4NCL Online

Post by JustinHorton » Tue Jul 14, 2020 6:38 pm

Mmm, they really need to be referred to in the piece
"Do you play chess?"
"Yes, but I prefer a game with a better chance of cheating."

lostontime.blogspot.com

Hok Yin Stephen Chiu
Posts: 148
Joined: Sun Dec 31, 2017 7:52 pm

Re: 4NCL Online

Post by Hok Yin Stephen Chiu » Sun Jul 19, 2020 8:52 am

Conscious I'm a bit late to the thread, but this is quite interesting line of discussion.
Matthew Turner wrote:
Wed Jun 17, 2020 12:42 pm
The probability that a 2400 plays at 2800 and the probability that a 1400 plays at 1800 are exactly the same, but you (and I) think it isn't. There is a good reason for thinking this and a bad reason for thinking this.

Bad reason, we can think of examples of a 1400 playing at 1800 - this isn't because it is more likely, it is simply that there are more 1400's than 2400's
Good reason, how accurate/stable a rating is is inversely correlated to how high the rating is.

Ken Regan tries to counter-balance this by effectively artificially boosting the rating of lower rated players. I don't know how well this works, but I do think you should be extra careful with bans for low rated players.
One wonders, whether the 1400-1800 example is particularly similar to the 2400-2800 example?

I, for example am effectively a retired chess player, I aim to play about one-two games a year to stay on various League Committees, and have managed an awe-inspiring grade of 77 (haha!), but during my undergraduate days, I was around 120, and even reached 128 after a particular good showing at the london chess classic. It is not particularly a stretch to say I could with some hard work, effort, and the right help revert back, and could plausible exceed my historic average by maybe 20-25 points, in maybe a space of 3-6 months. The probability of a 1400 playing at a 1800 level is much higher, owing to the fact that a 2400 has had to be much more consistent to even be a 2400 player - whereas any drunk can be a 1400 player!

This leads onto another consideration, a traditional ECF/FIDE grade is based on all your games played in a competitive environment. Whereas, analogously one's online grade is effectively calculated as if, all your casual over-the-board games in a pub several pints in, are combined with your serious league/congress matches, to calculate your grade. So, any system that has an element of relying on cheat detection by looking at ALL previous online games runs to an ever compounding problem.

Whatever our thoughts are - effectively, anti-cheating software of any kind is an application of computational intelligence, drawing a mix of ideas from neural networks and cluster analysis (cheat or not cheat). We will not have a perfect system by any stretch of our imagination, the best we can hope for is what Michael said earlier, we just want something where justice is not just *done*, but that,
Michael Farthing wrote:
Wed Jun 17, 2020 8:49 am
justice is seen to be done.
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 » Sun Jul 19, 2020 10:14 am

Hok Yin Stephen Chiu wrote:
Sun Jul 19, 2020 8:52 am
One wonders, whether the 1400-1800 example is particularly similar to the 2400-2800 example?
According to Elo's theory, rating is an estimate of strength which is unknown but can be estimated using results. At a 1400 level, plausibly the estimate is likely to be less reliable, because it's comparatively easier for work at chess to be rewarded by playing at a higher level.

Neil Graham
Posts: 1938
Joined: Thu Apr 12, 2007 8:36 pm

Re: 4NCL Online

Post by Neil Graham » Tue Jul 21, 2020 10:45 pm

There have been around 2,000 contributions to this thread and the "Cheating in Chess" thread in General Chat (including the other odds and sods ones that have cropped up). Frankly I lost track of it all around posting 763.

Can I therefore ask some questions which may (or may not) have been answered?

Firstly at the half way stage there was a FairPlay update on 11th May 2020 which detailed 19 cases of players on the barred list. Seven of the players had already been debarred by Lichess and eight of the remaining 12 had been identified as using computers plus four others. In Alex's article on Fair Play he mentions a figure of 30 players.

a) Is there going to be a further FairPlay update on the On-Line League?
b) Lichess ban players for using computer assistance "we have the right to ban or close an account for any reason without warning". They do not reveal their methodology and the appeal rights are extremely limited. How many players have actually appealed to Lichess and have any been successful? 4NCL states it has nothing to do with banning by the host platform (Lichess).
c) On the other hand 4NCL have a clear Fairplay policy. How many players have failed the Regan cheating test in the 4NCL On-Line? How many appeals have there been and were any successful?
d) Of the 30 players mentioned in para 3 only two owned up to cheating. As far as I understand, the other 28 said "no - not me guv" or words to that effect. Could this be confirmed?
e) Several teams withdrew in protest about alleged cheating - is there any comment on this?
f) Two semi-finals didn't take place when teams withdrew - was this because of cheating?
g) Where cheating is established should there be a retrospective amending of results? Promotions/Relegations were clearly affected this time round.
h) Was cheating established at the recent 4NCL On-Line Congress? If so are we likely to see a report on this?
i) Various players have been named on this site. Is it right and proper that the sole penalty for cheating is a ban? Should a list of banned players be shown together with reasons so other players/organisers are aware.

Enough to be going on with?

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

Re: 4NCL Online

Post by Roger de Coverly » Wed Jul 22, 2020 12:22 am

Neil Graham wrote:
Tue Jul 21, 2020 10:45 pm
h) Was cheating established at the recent 4NCL On-Line Congress? If so are we likely to see a report on this?
A player made 4 from 5 in the Open, but was flagged by lichess almost as soon as the results were available. Subsequently the lichess account was deleted. Other circumstantial evidence was that the player had been eligible for Senior events for a number of years and had not previously demonstrated a track record of being able to score well against a typical field in an Open. That's despite quite a number of attempts over the board.

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

Re: 4NCL Online

Post by MartinCarpenter » Wed Jul 22, 2020 9:55 am

Neil Graham wrote:
Tue Jul 21, 2020 10:45 pm
How many players have failed the Regan cheating test in the 4NCL On-Line?
....
g) Where cheating is established should there be a retrospective amending of results? Promotions/Relegations were clearly affected this time round.
i) Various players have been named on this site. Is it right and proper that the sole penalty for cheating is a ban? Should a list of banned players be shown together with reasons so other players/organisers are aware.
You're thinking in slightly overly black and white terms here. We're basically in a bit of a twilight zone where we can state with some confidence that there was some cheating and that some people were playing surprisingly well enough, to a degree making a temporary ban for the remainder of a season a reasonable measure.

What no one is remotely confident in doing is outright saying that a specific individual X was cheating. Hence your questions (g)/(i) are a bit moot. No one has been established as cheating.

I've copied the relevant graph for the first 6 rounds of the 4NCL from the rather good end of year report that got posted again here. Blue is the OTB 4NCL, Red is the online.
4NCL.png
4NCL.png (108.64 KiB) Viewed 1179 times
From this you can see that there was a distinctly unusually large cluster of people on the right hand side of the graph - ie playing much better moves than they'd have been expected to. It seems basically certain that some of this was caused by cheating.

Individual cases? Well. For example look at the 3-3.4 sigma results. You would expect maybe 5(?) of those if it followed the distribution/under performance, instead you have 10. Very good reasons to be suspicious at a population level, also nowhere near enough to firmly convict an individual.

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

Re: 4NCL Online

Post by Roger de Coverly » Wed Jul 22, 2020 10:14 am

MartinCarpenter wrote:
Wed Jul 22, 2020 9:55 am
What no one is remotely confident in doing is outright saying that a specific individual X was cheating. Hence your questions (g)/(i) are a bit moot. No one has been established as cheating.
Although they word it differently, lichess are confident enough to flag players based on a limited number of games. I believe it was lichess exclusions rather than 4NCL ones which triggered the squad withdrawals. The secretive lichess approach means that it's not actually stated which "terms and conditions" a player violated.

Thomas Rendle
Posts: 469
Joined: Tue Aug 10, 2010 8:31 am

Re: 4NCL Online

Post by Thomas Rendle » Wed Jul 22, 2020 10:35 am

MartinCarpenter wrote:
Wed Jul 22, 2020 9:55 am
What no one is remotely confident in doing is outright saying that a specific individual X was cheating. Hence your questions (g)/(i) are a bit moot. No one has been established as cheating.
That more because it's just not worth posting 'X' is cheating on a forum on the internet. There are individual cases I'm completely satisfied are the result of cheating and I'm still not going to name them here.

Mick Norris
Posts: 10328
Joined: Tue Apr 17, 2007 10:12 am
Location: Bolton, Greater Manchester

Re: 4NCL Online

Post by Mick Norris » Wed Jul 22, 2020 10:40 am

Thomas Rendle wrote:
Wed Jul 22, 2020 10:35 am
That more because it's just not worth posting 'X' is cheating on a forum on the internet. There are individual cases I'm completely satisfied are the result of cheating and I'm still not going to name them here.
Carl has asked us not to name individuals, so that's right
Any postings on here represent my personal views

NickFaulks
Posts: 8453
Joined: Sat Jan 02, 2010 1:28 pm

Re: 4NCL Online

Post by NickFaulks » Wed Jul 22, 2020 3:56 pm

Mick Norris wrote:
Wed Jul 22, 2020 10:40 am
Carl has asked us not to name individuals, so that's right
Particularly when there are so many other ways to make sure everyone knows who is being accused.
If you want a picture of the future, imagine a QR code stamped on a human face — forever.

Richard Bates
Posts: 3338
Joined: Fri Nov 14, 2008 8:27 pm

Re: 4NCL Online

Post by Richard Bates » Wed Jul 22, 2020 9:43 pm

Out of interest, does the graph for the OTB 4NCL use ratings at the start of the season, or official 4NCL ratings which change over the course of the season? The former being in principle more likely to produce outliers.

Alan Walton
Posts: 1393
Joined: Mon Nov 02, 2009 8:33 pm
Location: Oldham

Re: 4NCL Online

Post by Alan Walton » Thu Jul 23, 2020 3:25 am

Name them all, otherwise we end up going round in circles; these are players suddenly performing better than over the board; admittedly some players play better online, my rating is higher at blitz than my Fide rating, but some of the accuracy level go through the roof and there can only be one conclusion

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

Re: 4NCL Online

Post by MartinCarpenter » Thu Jul 23, 2020 9:26 am

Alan Walton wrote:
Thu Jul 23, 2020 3:25 am
Name them all, otherwise we end up going round in circles; these are players suddenly performing better than over the board; admittedly some players play better online, my rating is higher at blitz than my Fide rating, but some of the accuracy level go through the roof and there can only be one conclusion
The problem here is that its a pretty small sample (6 games). There's ~5 people playing 3-3.5 sigma worse than OTB and you'd definitely expect a few innocent people playing about that much better as well, even 3.5-4.

So you've got a group of people of whom ~80-90% appear to be cheating, but one or two not.

You can get more data, either by waiting to get more games or some other way, then properly ban people when you're sure.

Or you can ban everyone in the suspicious group, knowing one or two will be innocent. If you do that then you obviously need to try and keep it low key and low consequence for each individual.

Post Reply