correspondence chess in the age of NNUE

[cdani]

High draw rate at top level is sort of an heuristic proof chess is a theoretical draw

Or that the next level of chess is out of sight for the moment, I hope and I think.

[Thomas A. Anderson]

At top level CC the appearance of NN-based engine are definitively "plugging certain hole", but not all. The ones still open are those, when positions are not better judged by NN than with traditional engines. Personally, I was always looking for 'NN-suited" positions and try to fight the engines at that field. This might mean for me, that CC will loose even more attractiveness, the near future will tell.

[Uri Blass]

Taking out really weak players (probably not even using a computer) the draw rate for the past three years is already close to 60%. If you take out weak players, like those who only use a computer to check some positions, it raises over 75%. We're still talking over 140,000 games for just three years, and the best ones from this period are likely to finish yet.

To get to 90%, you need to increase the strength of players to a point were only 65,000 games are left. To give you an idea, you still have close to 2,500 players at this level. These are serious players who care about results and have both access and willingness to use HW. This doesn't mean they use the latest, but that's not necessary to face a draw problem, and if you select the cream of the crop (more than 14,000 games and close to a thousand players) you surpass the 95% draw rate. This is why Arno wrote a proposal I submitted to you some months ago, you didn't need the advent of NNUEs to have a real problem in your hands.

That's true, even before NNUEs CC draw rate was already significantly above 90% at the highest level of play according to these stats pointed out by L. Ljubicic:

NNUEs will spread this effect into lower Elos and probably kill the current concept of CC within a couple of years.

lower ELos simply do not use engines or do not use engines for a long time.

I do not believe NNUE made a big change and I doubt if it is possible to get less than 90% draws in correspondence games
against something like Stockfish10 even if you use NNUE against it assuming you use 8 hours for every move.

[lkaufman]

Taking out really weak players (probably not even using a computer) the draw rate for the past three years is already close to 60%. If you take out weak players, like those who only use a computer to check some positions, it raises over 75%. We're still talking over 140,000 games for just three years, and the best ones from this period are likely to finish yet.

To get to 90%, you need to increase the strength of players to a point were only 65,000 games are left. To give you an idea, you still have close to 2,500 players at this level. These are serious players who care about results and have both access and willingness to use HW. This doesn't mean they use the latest, but that's not necessary to face a draw problem, and if you select the cream of the crop (more than 14,000 games and close to a thousand players) you surpass the 95% draw rate. This is why Arno wrote a proposal I submitted to you some months ago, you didn't need the advent of NNUEs to have a real problem in your hands.

I didn't have time in November to look into this, but now I do. I have completed my new book for New In Chess, "Chess Board Options", which has quite a bit of material relating to these issues of excessive draws, ratings, rule changes, etc. It is supposed to be available on Amazon April 19. We did a fair amount of testing with alternate rules in Komodo for this book; the rule change which would have the most dramatic effect is to forbid repetitions (or score them as something like 1/4 point for the repeating side).

[Branko Radovanovic]

lower ELos simply do not use engines or do not use engines for a long time.

I do not believe NNUE made a big change and I doubt if it is possible to get less than 90% draws in correspondence games
against something like Stockfish10 even if you use NNUE against it assuming you use 8 hours for every move.

Looking at Ljubicic's stats, yes, that seems to be the case. No upward draw rate trend in the <1801 Elo bracket at all, so I guess no change for these players. 2200+ looks like it's going to become unplayable in 5 to 10 years though.

[Stephen Ham]

Hi Larry,

That's an interesting question. First, my qualifications to opine: I'm America's top-rated ICCF Grandmaster with a 2580+ rating, 240 ICCF games of experience, generally playing in tournaments with like-rated opposition.

That said - it's evident that chess is a draw "with best play". Best play includes numerous alternative continuations - no singular "best play" exists. Best play need not be "brilliant" - it instead must be error-free. Since chess is intrinsically a draw, many top-level ICCF tournaments now finish with 100% draws. No meaningful errors were made.

However, ICCF draws are not specifically a manifestation of NNUE. It's instead the result of generally improved hardware/software, game databases, plus 7-man TableBases (perfect chess). NNUE is just one specific component of this overall advancement in technology. Top-level ICCF play approached 100% draws even before NNUE went public. Improved software/hardware gets us closer to error-free chess. No meaningful errors - no decisive games - all draws result.

Therefore, many top ICCF GMs are considering quitting correspondence chess. I'm not there...yet. I still enjoy exploring new ideas, pushing matters, and taking reasonable risks when positions can't be 100% analyzed out. Top hardware with NNUE is still not perfect. I think many Stockfish NNUE evaluations are faulty. After long and deep searches, I've seen evaluations flip/flop.

Computer software/hardware will likely kill chess, even OTB chess. However, it's effect upon the ICCF is more immediate since all top-level GMs utilize computers to analyze games. And we have enormous amounts of time in which to do so. While chess will never be 100% analyzed out, it's "practically" analyzed out now in the ICCF. Once a game is completed, it's stored in the ICCF's archived database. ICCF players now have: hardware/software for analysis, endgame TableBases, and a complete inventory of thousands of ICCF games.

Human input in top-level ICCF games is now 99% eliminated, other than personal preference in selection of openings. OTB GMs are now mining the ICCF's game archives for useful opening theory. Soon, the human element will be removed in OTB chess too, rendering it more of a memory contest - who can recall data better than their opponent.

Sorry for the pessimistic prognosis, Larry. But that's the way I see matters.

All the very best,
-Steve-

[Uri Blass]

I disagree about the part of OTB games.
In OTB games it is impossible to memorize everything and the better player may get his opponent out of book assuming the better player change his choice and does not use a fixed opening.

memorizing 1,000,000 different lines in the opening is practically impossible for humans and I believe that for every smaller number of lines I can easily go not in one of your lines and still get a draw in a correspondence game.

Here is a simple test.
Give me some opening repertoire that you plan to play.
I am forced to get out of your opening repertoire in less than 10 moves.
What is the size of the opening repertoire that you need in order to get a winning advantage with white after I get out of your repertoire.

I guess it is more than 1,000,000 moves.

[lkaufman]

Hi Larry,

That's an interesting question. First, my qualifications to opine: I'm America's top-rated ICCF Grandmaster with a 2580+ rating, 240 ICCF games of experience, generally playing in tournaments with like-rated opposition.

That said - it's evident that chess is a draw "with best play". Best play includes numerous alternative continuations - no singular "best play" exists. Best play need not be "brilliant" - it instead must be error-free. Since chess is intrinsically a draw, many top-level ICCF tournaments now finish with 100% draws. No meaningful errors were made.

However, ICCF draws are not specifically a manifestation of NNUE. It's instead the result of generally improved hardware/software, game databases, plus 7-man TableBases (perfect chess). NNUE is just one specific component of this overall advancement in technology. Top-level ICCF play approached 100% draws even before NNUE went public. Improved software/hardware gets us closer to error-free chess. No meaningful errors - no decisive games - all draws result.

Therefore, many top ICCF GMs are considering quitting correspondence chess. I'm not there...yet. I still enjoy exploring new ideas, pushing matters, and taking reasonable risks when positions can't be 100% analyzed out. Top hardware with NNUE is still not perfect. I think many Stockfish NNUE evaluations are faulty. After long and deep searches, I've seen evaluations flip/flop.

Computer software/hardware will likely kill chess, even OTB chess. However, it's effect upon the ICCF is more immediate since all top-level GMs utilize computers to analyze games. And we have enormous amounts of time in which to do so. While chess will never be 100% analyzed out, it's "practically" analyzed out now in the ICCF. Once a game is completed, it's stored in the ICCF's archived database. ICCF players now have: hardware/software for analysis, endgame TableBases, and a complete inventory of thousands of ICCF games.

Human input in top-level ICCF games is now 99% eliminated, other than personal preference in selection of openings. OTB GMs are now mining the ICCF's game archives for useful opening theory. Soon, the human element will be removed in OTB chess too, rendering it more of a memory contest - who can recall data better than their opponent.

Sorry for the pessimistic prognosis, Larry. But that's the way I see matters.

All the very best,
-Steve-

Thanks for your input. I agree that correspondence chess already had a huge problem with draws before NNUE, but since I hadn't heard a lot about players wanting to give up or change rules I thought it wasn't yet bad enough for this to happen, now maybe it will be the majority view among strong players. I agree more with Uri regarding OTB chess; so many mistakes are made by humans that it is certainly no problem for White to avoid theory and still win games if he is the stronger player; maybe somewhat of a problem just for Black.
So the question now becomes whether top correspondence players will just give up, or whether they will favor some change or reform to address the problem. That could be a rule change, such as penalizing the player who repeats position, or it could be the computer chess solution of mandating bad openings with each player having a shot at the White side once. It would be nice if chess960 were the solution, but from the results we've seen here even chess960 is hopelessly drawn with top engines and long time limits. Have either rule changes or mandated (bad) openings attracted much interest among top correspondence players yet?

[Ozymandias]

So the question now becomes whether top correspondence players will just give up, or whether they will favor some change or reform to address the problem.

Many already gave up. By 2017, more than 10,000 players were still active in my DB. In the last three years, less than 6,000 remain.

[Stephen Ham]

Hi Larry,

The ICCF moves very slowly. Since it's an international organization, what appeals to country/culture X is dismissed as unacceptable by country/culture Y. It's thus difficult to get an international consensus on many things.

Presently, the ICCF is working on "low hanging fruit", yet still taking years to address obvious problems, such as: poor sportsmanship, excessively long time-controls, and "Dead Man's Defense" (i.e. taking forever to play out inferior/lost positions, almost to the point of waiting for the opponent to die). Yes, inferior/lost positions do occur. They're rare, but...they exist, usually at lower performance levels.

Really, what ICCF rule changes can fix the problem of chess being intrinsically a draw? Is it still chess if the ability to repeat a position is penalized? Regardless, rule changes inside the ICCF come very slowly.

Also, mandating playing inferior openings in the ICCF will NEVER be popular. Most top-level games/tournaments take roughly a year to complete. So, why willingly suffer the results of bad openings in multiple tournament games for a year? Putting forth effort and time, not to mention the cost for the latest PC hardware, is unpleasant enough when knowing one's best efforts still have a 99% likelihood of just being a draw. Surely it's even worse to expend the cost/effort/time playing bad openings one never chose in the first place.

Some top ICCF GMs have already quit correspondence chess for the reasons cited in my previous post. Other top GMs are considering it. Some day, I too will quit, but presently enjoy the delusion that I can still win games...somehow. The challenge still exists for me. However, the enjoyment of correspondence chess, with its "delayed gratification" of more than a year, is approaching zero for many of us.

All the very best,
-Steve-