Cheating in chess

[Alistair Campbell]

Well yes, can't see how a mere 51% probability would stand up to any legal scrutiny for a start.

What does the term "on the balance of probabilities" mean?

[Tim Harding]

Well yes, can't see how a mere 51% probability would stand up to any legal scrutiny for a start.

What does the term "on the balance of probabilities" mean?

It's the test in civil cases for a prosecution or lawsuit to succeed, as opposed to "beyond reasonable doubt" in criminal cases, so it's a much lower bar.

I'm not a lawyer but I don't think it is tied to a specific percentage, but rather deliberately left vague? Maybe some expert can clarify.

[Leonard Barden]

Leonard, I enjoyed your Guardian piece as always, but one correction - the decision was not made by the Fair Play Commission, which is a general body within FIDE one of whose functions, as I said above, is to recommend bans for players whom they consider to their comfortable satisfaction to have been guilty of fair play violations. The decision was made by the Fair Play Panel appointed for this tournament specifically. Only two of the members of the FPP were also members of the FPC.

disqualified by a minimum vote (2-1 with 1 abstention) of Fide’s Fair Play Panel and placed last.

Fixed!

[Matt Bridgeman]

It just means more likely than not. So would the jurisdiction for a civil case be Australia as Tornelo is an Australian chess platform? You’d want to name both FIDE and Tornelo jointly in any counter-actions, as David Cordover’s evidence would torpedo FIDE’s position.

Having been in front of plenty of Australian magistrates, I wouldn’t want to be on the prosecution end when the defence start tendering medical evidence regarding eye sight and outline there was no way for the player to be part of the internal processes.

Certainly in the policing world ‘reasonable suspicion’ is never a judgement, but usually an avenue to gather more information through tools such as questioning and searches. The flip-side of all this is perhaps the authorities have been pushed into taking a bad stand, by the sheer numbers of normally rule abiding over the board players, descending into regular online cheating. Trying to draw some lines in the sand has backfired this time.

[Paolo Casaschi]

>The comment in brackets seems to me extremely poorly worded.

I agree; I should not have said 'of course'.

However, the investigating committee will be operating in English because that is the lingua franca of the FPC. It is true that in theory a non-English speaking player could be interviewed live through an interpreter. I am not aware of that having been done or of the resource being available to do it. In practice I think non-English-speaking defendants are likely to interact with the committee in writing.

In my experience in courtrooms, having to operate with an interpreter significantly reduces the merit of hearing live, as opposed to written, evidence.

When FIDE organizes an event pretending to be a "World Championship" I find it difficult to argue for discriminating non English speakers simply because it's more convenient for the people appointed to make a decision.
If discriminating non-English speaking players is a new FIDE policy, I'd strongly suggest to translate FIDE's motto into English

The bigger concern however, from your wording of how such decisions are made, is for looking mainly at arbitrary computer statistics to the point that talking with the accused becomes irrelevant and you should not even bother if they don't speak the language convenient to the accuser.

[Kevin Thurlow]

"Having been in front of plenty of Australian magistrates, I wouldn’t want to be on the prosecution end when the defence start tendering medical evidence regarding eye sight and outline there was no way for the player to be part of the internal processes."

Quite. When I play online chess, I'm looking all over the place.

(I'm assuming the first part of your sentence is not as defendant!)

"If discriminating non-English speaking players is a new FIDE policy, I'd strongly suggest to translate FIDE's motto into English "

Or conduct hearings in Latin.

[Matt Bridgeman]

"Having been in front of plenty of Australian magistrates, I wouldn’t want to be on the prosecution end when the defence start tendering medical evidence regarding eye sight and outline there was no way for the player to be part of the internal processes."

Quite. When I play online chess, I'm looking all over the place.

(I'm assuming the first part of your sentence is not as defendant!)

"If discriminating non-English speaking players is a new FIDE policy, I'd strongly suggest to translate FIDE's motto into English "

Or conduct hearings in Latin.

Police Prosecutor for a 5 year stint in sunny Oz.

[MartinCarpenter]

As to what is 'a reasonable suspicion', I believe that Mr Martynov has said that if the FPP considers the probability of cheating to have occurred to be 51% or higher then it should disqualify the player.

I had heard that referenced before, but not from what I would call a reputable source. Are we really saying that a player can be excluded from an event with everything that potentially entails when there is a 49% chance that they are innocent? Surely, any right-minded individual would wish to distance themselves from that approach?

If it is made absolutely explicit that there is no judgement of guilt, and cheating was a really big problem then just maybe?!

The main obvious disadvantage would be getting people to accept the randomness inherent in the approach.

Of course if they're setting the statistics to generate a 51% chance, I'd imagine the false positive rate is much higher than 50/50.

[John Cox]

Paolo, you have to be realistic. Chess is a small sport with limited resources. Most players investigated by the FPP deal with the investigating committee in writing. I don't see, in practical terms, what you're proposing.

The major factor in most allegations of use of computers is what you term as 'arbitrary' computer statistics.

As to what 'reasonable grounds for suspicion' means; it is not a defined term. It has only come to be used in tournament regulations fairly recently, because FIDE has not previously involved itself in running online tournaments. I dare say Mr Martynov's formulation is as good an attempt to define it as any other.

In general, let me make some more trite observations.

Chess is, as I said, not a rich sport with limitless resources.

It is generally thought that online cheating is rife.

Given the pandemic and the limited resource available, it is more or less impossible to stop or detect determined and technically-savvy online cheats with certainty.

However, FIDE still wants to run online tournaments which provide a reasonably level playing field and afford a fair chance for innocent players to win.

Suppose there is a situation where upon investigation FIDE considers that there is a chance greater than 50% that the points winner has cheated, whereas there is a negligible chance that the second-placed player has cheated.

The present regulation would have the result that the highest points scorer was disqualified, and the second-placed finisher declared the winner. Some of you seem to think that is an outrage. All I say is that, in assessing that result, you do need to take into account the observations I made above, to consider FIDE's goal as outlined above, and, if you continue to think the present regulations are not good, say which ones would have been better.

[Matthew Turner]

John's example is perhaps quite understandably geared towards a situation where we have a small number of players as we had at the universities championships finals. However, I think this gives a somewhat misleading impression.

Let's imagine we have 200 players for The World U12 online Championships where we are applying a balance of probabilities to banning cheats

How many cheats do you think there will be?
How many cheats will be caught?
How many innocent players will be caught?

The innocent players arousing suspicion are obviously, in general, going to be those doing well, so I think the resulting medals table will end up being a pretty random affair.

Returning to the Universities Championships Finals. Had you asked the players whether they wanted a balance of probabilities, a comfortable satisfaction, or a beyond reasonable doubt approach to banning cheats I would suggest that zero would have gone for the first of these options.

I agree that there are no easy options, but this balance of probabilities approach cannot be right. You should set parameters where innocent players are going to be caught very infrequently, take step to limit cheating, but accept that some cheats will get away with it.

[Jacques Parry]

Had you asked the players whether they wanted a balance of probabilities, a comfortable satisfaction, or a beyond reasonable doubt approach to banning cheats I would suggest that zero would have gone for the first of these options.

But weren't they asked, in effect, whether they were willing to risk being disqualified if it was suspected that they were cheating, because that's what the regulations said? Do strong players really sign up for important online events without checking that they are content with the anti-cheating measures? The regulations don't refer to "balance of probabilities", but if mere suspicion is sufficient then balance of probabilities must also be sufficient, a fortiori.

[Matthew Turner]

Of course if they're setting the statistics to generate a 51% chance, I'd imagine the false positive rate is much higher than 50/50.

I don't believe this is true (in events with sufficient players at least). We have statistical analysis which is very reliable, so you could take a distribution of Regan Z scores and overlay the predicted distribution. Simply select a point where there are slightly more than twice the number of predicted players and ban them all.
So to my 200 player example. Our best guess is that is 'serious' events the level of cheating is 3 - 5% of players. So let's try to make the FIDE approach look as good as possible and take the lower figure, we then have 6 cheats. No cheat detection system is perfect so let say we catch 5 out of the 6. Using a balance of probabilities approach, we know that we are going to catch almost as many innocent players as cheats, but again let try to make FIDE's approach look as good as possible and say we only catch 3 innocent players. It stands to reason that these 3 are going to finish high up in the event, in fact it is very likely that the best innocent player will receive a ban (assuming the players have similar ratings and there isn't just a big outlier in terms of rating and ability who wins effortlessly).
So, our reasonable best case scenario might be that medals go to the 2nd, 4th and 5th highest finishing innocent players. However, in reality this scenario will never happen because the medal positions will be determined in large part by what the fair panel decide to do with the results of the excluded players.

[Matthew Turner]

Had you asked the players whether they wanted a balance of probabilities, a comfortable satisfaction, or a beyond reasonable doubt approach to banning cheats I would suggest that zero would have gone for the first of these options.

But weren't they asked, in effect, whether they were willing to risk being disqualified if it was suspected that they were cheating, because that's what the regulations said? Do strong players really sign up for important online events without checking that they are content with the anti-cheating measures? The regulations don't refer to "balance of probabilities", but if mere suspicion is sufficient then balance of probabilities must also be sufficient, a fortiori.

Jacques,
People tend to be poor at wading through regulations and terms and conditions and strong chess players are no different in that. Strong chess players (much more so than the general population) tend to believe that the anti-cheating mechanism run by platforms are very reliable. In those circumstances, one could reasonably scan through a regulation like 5.20 and believe it was a legal cover all, whilst still expecting a high bar for exclusion from the event. I say all this as a native English speaker.

Having seen the way the system has operated in the Universities Championships, players and Federations will have to think very carefully before committing to play in FIDE online events.

[Ian Thompson]

But weren't they asked, in effect, whether they were willing to risk being disqualified if it was suspected that they were cheating, because that's what the regulations said?

I don't think so. The regulations are still linked to on the ECF web site. Perhaps the regulations were changed before the event began, but, if someone entered based on the version the ECF linked to, they agreed to:

5.14. Based upon the results of the anti-cheating algorithm or/and other evidence deemed sufficient by FPP to establish a cheating incident, FPP is entitled to disqualify any player for a suspected fair play violation during the course or after the conclusion of any of the Championships event.

I can't see anything in the regulations saying what might be, or might not be, sufficient to establish a cheating incident. In my opinion, "establishing" something requires a higher level of proof than "suspecting" something.

Although it postdates this event, the FIDE Arbiters Manual explicitly says what standard of proof is required in online events:

The standard of proof shall be whether FPL has established an online cheating offence to the comfortable satisfaction of the hearing panel bearing in mind the seriousness of the allegation which is made. This standard of proof in all cases is greater than a mere balance of probability but less than proof beyond a reasonable doubt.

[Jacques Parry]

5.14. Based upon the results of the anti-cheating algorithm or/and other evidence deemed sufficient by FPP to establish a cheating incident, FPP is entitled to disqualify any player for a suspected fair play violation during the course or after the conclusion of any of the Championships event.

I can't see anything in the regulations saying what might be, or might not be, sufficient to establish a cheating incident. In my opinion, "establishing" something requires a higher level of proof than "suspecting" something.

Good point. I read it as a rather clumsy way of saying that FPP may disqualify if they suspect a violation, and that that suspicion may be based on any evidence they deem sufficient. On this reading, the word "establish" doesn't really add anything. I prefer this reading because, if the word "establish" were intended to imply that the incident must be proved, the words "a suspected fair play violation" would not make sense: you would expect it to say simply "FPP is entitled to disqualify the player for that incident", i.e. the one they have deemed to be established. But I agree that the drafting is unfortunate.