4NCL Online

[MartinCarpenter]

I have indeed told my teammates that were they to be accused of cheating I would throw them under the bus at the first possible opportunity!

Depends on the team mate a bit. For quite a few my a priori probability of them cheating online would be enormously low, so I'd be very sceptical of anything short of overwhelming statistical evidence.

The double check with Regan's stuff that the 4NCL are doing isn't really set to quite strict enough parameters to do that. Not unreasonably from their perspective given that they'd have to actively have a fight with LiChess to overturn a ban.

[NickFaulks]

Would that actually solve (or even really address) any of the potential problems arising?

I had the impression that all the disputed bans were those imposed by lichess. Has the 4NCL taken any action on its own initiative using the Regan software?

I wouldn't mind seeing Roger's question answered, if anybody knows.

Yes, it was my understanding that this was fairly clear: all the games were put through the software and it was upheld.

That is not an answer to Roger's question.

[Rhys Cumming]

Would that actually solve (or even really address) any of the potential problems arising?


I wouldn't mind seeing Roger's question answered, if anybody knows.

Yes, it was my understanding that this was fairly clear: all the games were put through the software and it was upheld.

That is not an answer to Roger's question.

Apologies, I misread. I don't know the answer to that one: I have only seen bans initiated by lichess first but that's a very small sample.

[Kevin Thurlow]

"1. Not playing after being at work all day and being tired as a result
2. Different time limits for games
3. Not wanting to finish the game quickly to get home
4. Putting more effort into FIDE rated games"

I agree - I have frequently done better at game-a-day events than the evening leagues. And it is unfair basing an argument on one sample, but if there's a pattern, it would be worth looking at. Some people are just inconsistent of course.

[Rhys Cumming]

I have indeed told my teammates that were they to be accused of cheating I would throw them under the bus at the first possible opportunity!

Depends on the team mate a bit. For quite a few my a priori probability of them cheating online would be enormously low, so I'd be very sceptical of anything short of overwhelming statistical evidence.

The double check with Regan's stuff that the 4NCL are doing isn't really set to quite strict enough parameters to do that. Not unreasonably from their perspective given that they'd have to actively have a fight with LiChess to overturn a ban.

I think herein lies the issue: whether we do indeed have overwhelming statistical evidence. I obviously don' t have the specific report produced in this case, but these bans are only being upheld when there is a high probability of cheating. You have a teammate being accused of cheating and are given a report saying that statistically we think there is a high level of probability that the player cheated. If there are specific reasons why you might believe your player to have been falsely flagged then that is one thing, but 'I know my player and they wouldn't do something like this' does not, in my opinion, justify the sort of 'my player is innocent and is being unfairly persecuted' style reaction we have seen in certain cases.

[Thomas Rendle]

If these players are later shown to have been cheating do you think we're likely to get apologies from teams that then chose to withdraw?

[NickFaulks]

OK Matthew, I've no clue what players we are talking about, but I think what you are saying is that we can search for and find their games, and draw our own conclusions?

But, what we can't do is see the statistical evidence, unless the players asks for this, and chooses to put it in the public domain?

Has anyone actually seen what a player gets when they ask to see the statistical evidence? Maybe I should get myself banned for cheating, just to find out how the process works.

[Matthew Turner]

So a z score of 4 equates to a circa 1 in 30,000 event. There are about 750 players in the 4NCL, so we would expect to naturally see a z score of 4 every 40 years. You can adjust your figures a bit, but even if it is a once in twenty year event then if Lichess bans are upheld by a Regan test with a z score of 4 or above then I think it is clear that there are a very limited number of false positives.

[MartinCarpenter]

So a z score of 4 equates to a circa 1 in 30,000 event. There are about 750 players in the 4NCL, so we would expect to naturally see a z score of 4 every 40 years. You can adjust your figures a bit, but even if it is a once in twenty year event then if Lichess bans are upheld by a Regan test with a z score of 4 or above then I think it is clear that there are a very limited number of false positives.

1/40 years clearly isn't right because they're not running the test at the level of an entire 4NCL season comprising a single data point. If its down to the level of single games then its more like 1/4 years.

I'd also be distinctly worried about how good the normal approximation is for small sample sets, harder to quantify but does cut into the confidence.

You've also had, what?, 2 people banned this year. So even at 1 in 40 years, you're getting 1/80 accusations wrong. That's not especially comforting.

[Matthew Turner]

Martin,
They are not running the software on single games, they are running it on all the available games.

[Rhys Cumming]

So a z score of 4 equates to a circa 1 in 30,000 event. There are about 750 players in the 4NCL, so we would expect to naturally see a z score of 4 every 40 years. You can adjust your figures a bit, but even if it is a once in twenty year event then if Lichess bans are upheld by a Regan test with a z score of 4 or above then I think it is clear that there are a very limited number of false positives.

1/40 years clearly isn't right because they're not running the test at the level of an entire 4NCL season comprising a single data point. If its down to the level of single games then its more like 1/4 years.

I'd also be distinctly worried about how good the normal approximation is for small sample sets, harder to quantify but does cut into the confidence.

You've also had, what?, 2 people banned this year. So even at 1 in 40 years, you're getting 1/80 accusations wrong. That's not especially comforting.

Significantly more than 2 (those are just the teams that pulled out in protest). I don't know the exact figure (those will presumably be released after this final round) but this was the update after 5 rounds:

"Players who were using computer assistance during their games: 12, of which
- 8 were using an engine during their games
- 4 were attempting to gain assistance using other methods during competition games"

[Thomas Rendle]

You've also had, what?, 2 people banned this year. So even at 1 in 40 years, you're getting 1/80 accusations wrong. That's not especially comforting.

I think it's more like 4-5 bans from the main (rather than junior) 4NCL. We got a 'Fair play update' on 11th May at the half-way point (round 5). I suspect we'll be hearing more after round 10 is complete (tomorrow night).

Some of the players caught have certainly played to a very, very high standard and don't forget computer correlation is only part of evidence in some cases.

Also I think it should be made clear that it's better to admit guilt where a case is reasonably proven. Players and team captains need to think carefully before going for a blanket denial, which as we've seen undermines the integrity of an online event completely.

[MartinCarpenter]

Martin,
They are not running the software on single games, they are running it on all the available games.

They're obviously not waiting until the end of the season either though, so its definitely <1/40 years.

What level of false accusations are people happy with? Even assuming the stats are working perfectly - very slightly generous - you're still looking at something like 200 or 300 to 1 right <-> wrong.

I'd be much happier if it they waited until the end of the season then grouped it all up but I presume LiChess effectively don't let them do this.

[Rhys Cumming]

You've also had, what?, 2 people banned this year. So even at 1 in 40 years, you're getting 1/80 accusations wrong. That's not especially comforting.

Also I think it should be made clear that it's better to admit guilt where a case is reasonably proven. Players and team captains need to think carefully before going for a blanket denial, which as we've seen undermines the integrity of an online event completely.

Agreed. Also the team captains withdrawing and making a fuss on a player's behalf seem primarily to be drawing attention to their incident. The 4NCL is doing its best to not draw attention to which player is involved, which withdrawals in protest highlight. Serious competitions online are obviously a new venture and I personally think the chess world will be relatively forgiving for initial lapses in judgement if those are appropriately acknowledged. I do wonder whether these player's will come to (or already) regret the grandstanding done on their behalf.

[Thomas Rendle]

It's a difficult balancing act. It can be a demoralising experience to play against a cheater, especially in longer games (rather than blitz) where it starts to feel rather hopeless. You don't want that to continue any longer than necessary if you have the evidence you need.