Engine and database use in Correspondence Chess

Discuss anything you like about chess related matters in this forum.
Brian Egdell
Posts: 49
Joined: Sun Apr 15, 2018 2:38 pm
Location: The Netherlands

Engine and database use in Correspondence Chess

Post by Brian Egdell » Sun Sep 09, 2018 4:10 pm

Apologies in advance if this subject has been hashed out elsewhere, or if it is inappropriate for the English Chess Forum, or if I am posting it in the wrong place (in which case please move as appropriate!). I am a fairly new, and so far almost completely inactive, forum member. But this is a subject on which I would welcome people's views if that is possible.

----

I'm a former enthusiastic Correspondence Chess player, but after years of playing by post followed by a few years with the then-newly-possible Email format I stopped because I was not particularly interested in running chess engines to do the analysis work for me. As is well known, the ICCF allows engine use and the game has completely changed as a result compared with 25+ years ago.

However just a few years ago I took Correspondence Chess up again online because the site chess.com supports it (it's called "Daily" chess there) with a ban on engine use, backed up by a highly expert cheating detection team which is very effective at finding players who are illegally using software help in their games and banning their accounts. Even at the top levels far above my own much more average ability, players can generally enjoy correspondence chess without too much fear that they are playing an engine rather than a human opponent. This is really appreciated by many members there. Up until two days ago, the only help permitted on this site which was unavailable to the postal chess players of decades ago was consulting online databases of games in order to choose opening moves. Former postal players had to rely on periodicals like "Informator" or books and articles about openings.

But on 7 September, two days ago, the site suddenly announced that in its correspondence ("Daily") games players would now, with immediate effect, be permitted access to Endgame Tablebases in their games. This, of course, has the effect of immediately cancelling any need to show any technique in defending K+R vs K+R+B for example, or winning K+Q+P vs K+Q or K+Q vs K+R. These endings have been exhaustively analysed, as have all other endings with 7 or fewer pieces on the board, and for every position it is known what the correct play is and what the result of the "game" should be.

Here is a link to the article announcing the changed rule:
https://www.chess.com/blog/monitor/rule ... ence-chess

I'm deeply unhappy by this development (well, I'm biased, because I find non-trivial endings with very few pieces left amongst the most interesting parts of the entire game). I don't understand what chess on that site gains by the new permission to use tablebases when they were perfectly successfully outlawed before. What do other people think?

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

Re: Engine and database use in Correspondence Chess

Post by Roger de Coverly » Sun Sep 09, 2018 5:30 pm

Brian Egdell wrote:
Sun Sep 09, 2018 4:10 pm
I don't understand what chess on that site gains by the new permission to use tablebases when they were perfectly successfully outlawed before. What do other people think?
I think they are falling in line with the ICCF who allow them. Once you allow chess with an "open book", there seems little point in allowing some look ups but not others.

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

Re: Engine and database use in Correspondence Chess

Post by Ian Thompson » Sun Sep 09, 2018 6:39 pm

Their rules seem to be based on the premise of "if you cheat by using a computer we'll catch you." One reason for allowing the use of tablebases is that it's impossible for them to reliably detect their use - anyone illegally using one would just choose any one of the winning/drawing moves each time, avoiding picking the move that wins/draws most quickly every time.

Nick Grey
Posts: 1838
Joined: Thu Dec 01, 2011 12:16 am

Re: Engine and database use in Correspondence Chess

Post by Nick Grey » Sun Sep 09, 2018 7:44 pm

What is confusing is a governing body trying to remove all forms of cheating (getting help from a machine or others) and another which actively encourages it.

Either form is trying to sell us a lot of products with FOMO.

Brian Egdell
Posts: 49
Joined: Sun Apr 15, 2018 2:38 pm
Location: The Netherlands

Re: Engine and database use in Correspondence Chess

Post by Brian Egdell » Sat Sep 29, 2018 2:00 pm

Thanks for the reactions.

To address a couple of them:

@Roger de Coverly: The site chess.com sets itself apart as an place where correspondence chess is possible without the use of chess engines. Engines are, in my opinion, an entirely different issue from looking up opening lines or techniques in types of endgames. The site attempts to allow what correspondence chess used to be before circa 1990. I think this fills a serious need for some players, myself included.

@Ian Thompson: Without knowing which anti-cheating techniques the site uses exactly (they're kept secret with the exception that it is known that they use statistical analysis), I don't share your opinion. It is true that many positions in endgame tablebases show multiple methods to win or draw, but that is equally true of general chess positions where engine analysis, rather than database look-up, would be in order. The same statistical methods of cheating detection should be equally applicable to detection of tablebase use in my opinion.

***

The big news is that chess.com has, in the last few days, reversed its decision! Tablebases are not to be permitted after all. According to the updated article, accessible from the same link above, the staff had thought that there was a general wish amongst the players for tablebase use to be permitted but were then shocked by the disappointed reactions from many players after the original announcement.

Nick Burrows
Posts: 1705
Joined: Sat Aug 14, 2010 12:15 pm

Re: Engine and database use in Correspondence Chess

Post by Nick Burrows » Sat Sep 29, 2018 2:17 pm

The point Ian made is correct however. A player can easily use a tablebase in an undetectable way, by randomly choosing one of the many (slower) routes to victory, without ever leaving the margin of victory.

The same is true for a determined engine user, he can easily randomise his move selection, even making moves that let the advantage slip....before eventually winning. There is NO method to detect this.

The best guide is simply an opponents rating. As engine ELO's are so astronomically higher than human players, any rating close to yours is going to be a human player, apart from the odd cheater that you catch on their way up the rating ladder. You probably have to accept this as an unfortunate part of playing chess via the internet, learn what you can from the game, and move on.

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

Re: Engine and database use in Correspondence Chess

Post by Roger de Coverly » Sat Sep 29, 2018 2:40 pm

Brian Egdell wrote:
Sat Sep 29, 2018 2:00 pm
Engines are, in my opinion, an entirely different issue from looking up opening lines or techniques in types of endgames.
That depends on how you study chess. If you are trying to increase your understanding of positions and the possible moves that could be played, you can check what else has been played, what people have said in analysis and what you might glean from analysing positions with an engine. In endings, being able to recognise positions as won or drawn is a useful skill which tablebases can help cultivate. If you play from memory as in OTB chess, how could you satisfy chess.com's stipulation of not using computer analysis?

Nick Grey
Posts: 1838
Joined: Thu Dec 01, 2011 12:16 am

Re: Engine and database use in Correspondence Chess

Post by Nick Grey » Sat Sep 29, 2018 2:59 pm

I'm not sure why anyone will want to use an engine.
I'm not sure why we cannot realise that a 3 day per move human analysis is likely to be a lot better than a 2/3 minute decision over the board.
Looking forward to my next 2 games in the County Correspondence Chess Championship.

Brian Egdell
Posts: 49
Joined: Sun Apr 15, 2018 2:38 pm
Location: The Netherlands

Re: Engine and database use in Correspondence Chess

Post by Brian Egdell » Sat Sep 29, 2018 3:01 pm

Roger de Coverly wrote:
Sat Sep 29, 2018 2:40 pm
That depends on how you study chess. If you are trying to increase your understanding of positions and the possible moves that could be played, you can check what else has been played, what people have said in analysis and what you might glean from analysing positions with an engine. In endings, being able to recognise positions as won or drawn is a useful skill which tablebases can help cultivate. If you play from memory as in OTB chess, how could you satisfy chess.com's stipulation of not using computer analysis?
All of what you say is correct, in my opinion. Satisfying chess.com's rules is done as follows: you do not turn an engine on to analyse the position you have reached (or might reach further in your analysis), nor do you consult a tablebase for your position (or a position in your analysis). Consulting what people have said about a particular exact position (generally that will be an opening position) is fine, though asking others about it in person when you have reached such a position is not. Using tablebases to further your general understanding of which positions are won or drawn is fine, but putting the exact position from your game (or analysis of a future possibility in your game) is not.

I admit that there are some slight grey areas at the edges of what is permitted. I believe that the site acknowledges that the rules are not entirely water-tight too. That, however, is true for any set of rules in any circumstances, which is why independent arbiters are needed in all serious competition to interpret rules in fairest way they can.

Brian Egdell
Posts: 49
Joined: Sun Apr 15, 2018 2:38 pm
Location: The Netherlands

Re: Engine and database use in Correspondence Chess

Post by Brian Egdell » Sat Sep 29, 2018 3:04 pm

Nick Grey wrote:
Sat Sep 29, 2018 2:59 pm
I'm not sure why anyone will want to use an engine.
I'm not sure why we cannot realise that a 3 day per move human analysis is likely to be a lot better than a 2/3 minute decision over the board.
Looking forward to my next 2 games in the County Correspondence Chess Championship.
I agree with you. To the best of my understanding, engine use has always been permitted in most Correspondence organisations, the ICCF included, for the simple reason that it was thought impossible to prevent. The site chess.com is trying to detect and prevent engine use after all, and I believe it is making a good job of it.

Brian Egdell
Posts: 49
Joined: Sun Apr 15, 2018 2:38 pm
Location: The Netherlands

Re: Engine and database use in Correspondence Chess

Post by Brian Egdell » Sat Sep 29, 2018 3:13 pm

Nick Burrows wrote:
Sat Sep 29, 2018 2:17 pm
The point Ian made is correct however. A player can easily use a tablebase in an undetectable way, by randomly choosing one of the many (slower) routes to victory, without ever leaving the margin of victory.

The same is true for a determined engine user, he can easily randomise his move selection, even making moves that let the advantage slip....before eventually winning. There is NO method to detect this.

The best guide is simply an opponents rating. As engine ELO's are so astronomically higher than human players, any rating close to yours is going to be a human player, apart from the odd cheater that you catch on their way up the rating ladder. You probably have to accept this as an unfortunate part of playing chess via the internet, learn what you can from the game, and move on.
We must agree to disagree over this. Ken Regan has made huge contributions to the theory of detecting cheating by engine use, and those who think they can evade detection by such techniques as you describe are unpleasantly surprised when they get their account closed for cheating.

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

Re: Engine and database use in Correspondence Chess

Post by Roger de Coverly » Sat Sep 29, 2018 3:21 pm

Brian Egdell wrote:
Sat Sep 29, 2018 3:01 pm
Consulting what people have said about a particular exact position (generally that will be an opening position) is fine,
What if they wrote something like "here an engine suggests ... "?

Brian Egdell
Posts: 49
Joined: Sun Apr 15, 2018 2:38 pm
Location: The Netherlands

Re: Engine and database use in Correspondence Chess

Post by Brian Egdell » Sat Sep 29, 2018 3:33 pm

Roger de Coverly wrote:
Sat Sep 29, 2018 3:21 pm
Brian Egdell wrote:
Sat Sep 29, 2018 3:01 pm
Consulting what people have said about a particular exact position (generally that will be an opening position) is fine,
What if they wrote something like "here an engine suggests ... "?
Yes that's a good example of the sort of grey area I had in mind. Here I would say: "avert your eyes from what the engine suggests", but now we're getting into dodgy areas. :D

Kevin Thurlow
Posts: 5821
Joined: Wed Apr 30, 2008 12:28 pm

Re: Engine and database use in Correspondence Chess

Post by Kevin Thurlow » Sun Sep 30, 2018 11:04 am

ICCF allows engine use and you can make claims based on the tablebase, but it still bans advice from people, and you would think that is impossible to police as well. Even if they do enforce that (which they have), it is only for relatively low-rated players, not for World Champions (where they spectacularly failed to enforce it).

Books were always allowed, and databases are effectively books. The only issues revolve round things like Fritz, Stockfish etc. I think ICCF take the view that you will not become a CC grandmaster by relying entirely on computers, so it's not worth the effort banning them. I don't believe the computer detection software is that reliable anyway - it might be for blitz, but not for CC. I occasionally find good moves quickly, give me 6 days, I might find a better one.

Brian Egdell
Posts: 49
Joined: Sun Apr 15, 2018 2:38 pm
Location: The Netherlands

Re: Engine and database use in Correspondence Chess

Post by Brian Egdell » Sun Sep 30, 2018 1:22 pm

Kevin Thurlow wrote:
Sun Sep 30, 2018 11:04 am
ICCF allows engine use and you can make claims based on the tablebase, but it still bans advice from people, and you would think that is impossible to police as well. Even if they do enforce that (which they have), it is only for relatively low-rated players, not for World Champions (where they spectacularly failed to enforce it).

Books were always allowed, and databases are effectively books. The only issues revolve round things like Fritz, Stockfish etc. I think ICCF take the view that you will not become a CC grandmaster by relying entirely on computers, so it's not worth the effort banning them. I don't believe the computer detection software is that reliable anyway - it might be for blitz, but not for CC. I occasionally find good moves quickly, give me 6 days, I might find a better one.
Really appreciate your comments.

I'd like, with the genuine respect due to you, to pick up on two things you say.

(1) Databases are effectively books. Well, yes and no. They cannot be accessed in the same way that books are. An endgame tablebase such as the freely accessible 6-piece ones could not be printed out because it is far too big (I've heard it said that it would take thousands of tons of paper). Related to this, but also aside from it, is the practical difference in how consulting an endgame book as opposed to and endgame tablebase affects one's play in Correspondence chess. The book helps towards showing what technique is required, and teaches a lot about the given endgame type in the process, leading (hopefully) to an interesting and well played endgame. The database stops the play in its tracks, finishes the game, and teaches little or nothing.

(2) I'm in agreement that engine detection software is not reliable for correspondence chess. For that reason it is never relied on by itself on chess.com. It is useful for flagging suspicious cases, but no account on chess.com is banned without human input.

Post Reply