Correspondence ICCF Champion Vs Top Engine ?

[Milos]

The only really important thought is that a human...say with a 2000 elo (probably less come to think of it, maybe a lot less) can take engines and 'be' a really highly rated "correspondence" player. It has bastardized the game to the point that 'human'(unassisted) vs engine 'unassisted' is really irrelevant.

It's not that simple. ICCF tournaments typically have over 10 matches each. Unless you have a large room full of computers each one focused on each match you're going to have to figure out how much computer time to assign to each match, assuming you're going to go with the unassisted engine strategy. Requires a lot of patience as well, each tournament can last a couple years. This rating period (March 1 to May 31) I completed 17 games and only gained 16 elo.

It's far simpler than that, 10 games at the same time, is instead of 10 moves in 50 days, 100 moves in 50 days (and all of this assuming your opponents are gonna instantly play their moves which is never gonna happen). So you let your engine analyze each move in each game for 12 hours and you are still good to go. Someone more skillful can write a script to automate all of this. And you are almost guaranteed to have at least 50% of points in a tournament. And if there are some ICCF dinosaurs that haven't even heard of SF-NNUE or that believe it's enough to run engine till depth 30 to validate their "bright" ideas or just have trust in stupid and weak books, you are almost certainly gonna have more than 50% of the points. Yes it's really as simple as that.

[Uri Blass]

The only really important thought is that a human...say with a 2000 elo (probably less come to think of it, maybe a lot less) can take engines and 'be' a really highly rated "correspondence" player. It has bastardized the game to the point that 'human'(unassisted) vs engine 'unassisted' is really irrelevant.

It's not that simple. ICCF tournaments typically have over 10 matches each. Unless you have a large room full of computers each one focused on each match you're going to have to figure out how much computer time to assign to each match, assuming you're going to go with the unassisted engine strategy. Requires a lot of patience as well, each tournament can last a couple years. This rating period (March 1 to May 31) I completed 17 games and only gained 16 elo.

It's far simpler than that, 10 games at the same time, is instead of 10 moves in 50 days, 100 moves in 50 days (and all of this assuming your opponents are gonna instantly play their moves which is never gonna happen). So you let your engine analyze each move in each game for 12 hours and you are still good to go. Someone more skillful can write a script to automate all of this. And you are almost guaranteed to have at least 50% of points in a tournament. And if there are some ICCF dinosaurs that haven't even heard of SF-NNUE or that believe it's enough to run engine till depth 30 to validate their "bright" ideas or just have trust in stupid and weak books, you are almost certainly gonna have more than 50% of the points. Yes it's really as simple as that.

I believe most correspondence players play more than 10 games at the same time but I also think no need for 12 hours per move and one hour per move is usually enough not to lose even if the opponent use 100 hours per move.

[jr66]

The use of the phrase "kill us" makes me think I've missed something.

I didn't think anybody was claiming that unassisted engines would win against a strong ICCF player (human+engine centaur), which is what the "kill us" suggests to me.

I thought the claim was that unlike the before-times, unassisted top engines can't be regularly beaten by strong centaurs at correspondence.

Cheers!

Sorry, i said "kill us" because i read "hundreds of Elo more" with big tlme calculation elsewhere, which normally lead to many losses ?
I agree with Uri , expect for some positions where i must admit 1h of study ( i choose the word because a player who wait only the result of engine is not a CC player for me but we have this kind of player with good results now i know ) is too short of course ???
Sometimes, i really needed several seances of about three hours before to decide the move because the position was simply difficult.....

[jefk]

Monte Carlo wrote

"There was one recent test in these forums with some pretty strict restrictions for the unassisted engine, operated by me, against jefk"

yes indeed; now i'm not ICCF 'champ' but CCM, but anyway i tried to play a strong game
(i think my play was correct, and i showed in early midgame could have got a -drawish- endgame advantage with one pawn more, only thereafter i slightly messed up and got an equal endgame thus draw), and before that there was a thread (by LK) about the draw problem in (ICCF) corresp. chess since the Nnue engines.
As for playing more boldly, and trying to win against computer operator who knows what
he's doing (eg conditions you had but then even more comp time/move), forget it
that you can win. My impression is that wins only occur against players who experiment
with bold openings (usually Black defenses), for whatever reason, and then end up in a loss.
i payed 1.d4, coz RL Berlin is impossible to crack, Giuco Piano is a draw, then you can try 1.c4 which i tried sometimes in the past, but even then i'll estimate the winning chance as less than 1 pct (against an ICCF player >2400 or so, that means someone who knows what he's doing).

In the meantime, LK and me already discussed some possible rule (especially endgame)
modifications, also as result of some GM Nickel proposals in the past, i thought
a bit more about it revently, and wrote this last week on the ICCF facebook forum

we need to change the (endgame) rules, some minor modifications have been suggested by A.Nickel and L.Kaufman, but i think a more fundamental improvement would be to change the point system (0/0.5/1), depending on material imbalance in the endgame, so first, (to keep it simple) if White gets one more pawn (even in a drawn endgame, he gets 0.5+1/10 = 0.6 and Black gets 0.4 points.
Then we also should make a correction for the inherent advantage of White, with about 0.2 -0.25 points thus in the above example White would get 'only' 0.58 and Black 0.42.
Two pawns, similar, eg. 0.68 fvs 0.32. Count knight/bishops as 0.3 (although bishop is a bit more imho), etc,. Count all the points in a tourn and you get a winner. Changing stalemate rules etc. are more radical and would lead to (other) chess variants. Even with the change i suggest above, in future, advanced endgame bases may take such stuff into account, but who cares. For the time we get some competition back.