How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Chessqueen
Posts: 1445
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by Chessqueen » Mon Feb 22, 2021 5:15 pm

MonteCarlo wrote:
Mon Feb 22, 2021 3:58 pm
Yeah, a few quick points.

1. As I said clearly already, Stash 27.0 is still quite a bit weaker than Komodo, so I know it will lose some games.

2. Some of the moves in that game Stash 27.0 switches from quickly for me, so either your machine is very slow and/or the TC was faster than the Naka match, both of which would likely decrease the weaker player's chances.

3. As Larry mentioned, that was not one of the odds from that match.

4. Nowhere have I ever said that there will be no example of something a human might do better than Stash 27.0, especially as a one-off.

My only points have been these, which are in no way affected by the game you posted.

1. Stash 27.0 is ~3000 CCRL blitz, so it shouldn't be surprising if it can outperform strong humans at some chess tasks

2. A human like Naka is a big underdog in blitz against ~3000 CCRL engines; his 3000+ chess.com blitz rating is not relevant.

For that matter, a single game is also not "proof" that Naka is stronger than Stash 27.0. I've won individual games against a lot of much stronger players, and even that isn't what happened here. :D

Cheers!
The time control was NOT quicker and my machine is an Intel Core i3 8350K which is just slightly slower than the one used by CCRL to test Stash 27.0, but in reality it does NOT matter since both Komodo and Stach 27 used the same machine. Another point that you should consider is that in the game that I pointed out GM Nakamura Vs Komodo Nakamura was playing with just the f7 pawn Odds, whereas, I gave Stash 27.0 an Extra pawn the c7 + f7 pawn Odds.

CCRL 40/15 Rating List — All engines (best versions only) (Quote)

Ponder off, General book (up to 12 moves), 3-4-5 piece EGTB
Time control: Equivalent to 40 moves in 15 minutes on an Intel i7-4770k.
Computed on February 20, 2021 with Bayeselo based on 1'218'554 games
Color legend: Commercial, Free, Open source, Private.
Bold font - tested with 200 games or more. Normal font - less than 200 games.


67 Stash 27.0 64-bit 2992 +31 −31 50.2% −1.5 44.2% 328
67.4%
68‑69 FabChess 1.16 64-bit 2983 +20 −20 56.8% −47.7 42.8% 811
50.8%
68‑69 Scorpio 2.8 64-bit 4CPU 2983 +25 −25 48.8% +7.0 46.3% 492

MonteCarlo
Posts: 146
Joined: Sun Dec 25, 2016 3:59 pm

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by MonteCarlo » Mon Feb 22, 2021 5:28 pm

I'm not going to belabor the point after this, since I prefer not to get sucked into an xkcd Duty Calls situation. :D

First, it was f7 and an extra move in addition to being white (so pawn and two moves in standard odds-speak), not just f7.

Second, I explained why speed and TC would matter; I didn't say it was unfair. In general, faster TCs benefit the stronger player, so I'd expect more losses from the weaker player receiving odds.

Finally, I never said Stash 27.0 couldn't lose to Komodo 13.3 at odds. I said the opposite, that I was sure it would lose some games.

None of this changes my two main points.

Cheers!

Chessqueen
Posts: 1445
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by Chessqueen » Mon Feb 22, 2021 6:41 pm

MonteCarlo wrote:
Mon Feb 22, 2021 5:28 pm
I'm not going to belabor the point after this, since I prefer not to get sucked into an xkcd Duty Calls situation. :D
https://www.youtube.com/watch?v=oS24f6e ... orlandoard :roll: :mrgreen:

Finally, I never said Stash 27.0 couldn't lose to Komodo 13.3 at odds. I said the opposite, that I was sure it would lose some games.

None of this changes my two main points.

Cheers!

Chessqueen
Posts: 1445
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by Chessqueen » Mon Feb 22, 2021 11:09 pm

MonteCarlo wrote:
Mon Feb 22, 2021 5:28 pm
I'm not going to belabor the point after this, since I prefer not to get sucked into an xkcd Duty Calls situation. :D

First, it was f7 and an extra move in addition to being white (so pawn and two moves in standard odds-speak), not just f7.

Second, I explained why speed and TC would matter; I didn't say it was unfair. In general, faster TCs benefit the stronger player, so I'd expect more losses from the weaker player receiving odds.

Finally, I never said Stash 27.0 couldn't lose to Komodo 13.3 at odds. I said the opposite, that I was sure it would lose some games.

None of this changes my two main points.

Cheers!
This is the statement that I do NOT agree with you, and I would bet $1000.00 that GM Nakamura can beat Stash 27 in Blitz which is only rated around 3034 which I believe the CCRL Blitz rating is inflated by 150 in comparison to Top human Blitz ratings, so they would be very close . You Mentioned that Top humans like Naka or MVL would likely lose by very large margins in blitz to most engines that are around 3000 CCRL blitz; their high chess.com ratings would not save them.

The time control was NOT quicker and my machine is an Intel Core i3 8350K which is just slightly slower than the one used by CCRL to test Stash 27.0, but in reality it does NOT matter since both Komodo and Stash 27 used the same machine. Another point that you should consider is that in the game that I pointed out GM Nakamura Vs Komodo Nakamura was playing with just the f7 pawn Odds, whereas, I gave Stash 27.0 an Extra pawn the c7 + f7 pawn Odds.


CCRL Blitz Rating List — All engines (best versions only) (Quote)
Ponder off, General book up to 12 moves, up to 6 piece EGTB
Time control: From Jan 2020, equivalent to 2'+1" on an Intel i7-4770K. See the "about" page.
Computed on February 20, 2021 with Bayeselo based on 2'464'260 games


70 Stash 27.0 64-bit 3034 +19 −19 54.4% −28.7 37.5% 915
50.8%
71 Bagatur 2.1 64-bit 4CPU 3033 +25 −25 47.7% +15.1 46.9% 475
56.2%

Chessqueen
Posts: 1445
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by Chessqueen » Tue Feb 23, 2021 1:01 pm

Chessqueen wrote:
Mon Feb 22, 2021 11:09 pm
MonteCarlo wrote:
Mon Feb 22, 2021 5:28 pm
I'm not going to belabor the point after this, since I prefer not to get sucked into an xkcd Duty Calls situation. :D

First, it was f7 and an extra move in addition to being white (so pawn and two moves in standard odds-speak), not just f7.

Second, I explained why speed and TC would matter; I didn't say it was unfair. In general, faster TCs benefit the stronger player, so I'd expect more losses from the weaker player receiving odds.

Finally, I never said Stash 27.0 couldn't lose to Komodo 13.3 at odds. I said the opposite, that I was sure it would lose some games.

None of this changes my two main points.

Cheers!
This is the statement that I do NOT agree with you, and I would bet $1000.00 that GM Nakamura can beat Stash 27 in Blitz which is only rated around 3034 which I believe the CCRL Blitz rating is inflated by 150 in comparison to Top human Blitz ratings, so they would be very close . You Mentioned that Top humans like Naka or MVL would likely lose by very large margins in blitz to most engines that are around 3000 CCRL blitz; their high chess.com ratings would not save them.

The time control was NOT quicker and my machine is an Intel Core i3 8350K which is just slightly slower than the one used by CCRL to test Stash 27.0, but in reality it does NOT matter since both Komodo and Stash 27 used the same machine. Another point that you should consider is that in the game that I pointed out GM Nakamura Vs Komodo Nakamura was playing with just the f7 pawn Odds, whereas, I gave Stash 27.0 an Extra pawn the c7 + f7 pawn Odds.


CCRL Blitz Rating List — All engines (best versions only) (Quote)
Ponder off, General book up to 12 moves, up to 6 piece EGTB
Time control: From Jan 2020, equivalent to 2'+1" on an Intel i7-4770K. See the "about" page.
Computed on February 20, 2021 with Bayeselo based on 2'464'260 games

80 Invictus r305 64-bit 4CPU 3003 +13 −13 48.4% +10.2 45.3% 1857
75.6%
81 Scorpio 2.8 64-bit 4CPU 2995 +17 −18 42.4% +50.3 36.8% 1101
7

Here it is against Invictus 305 rated 3003, whereas GM Nakamura held a draw with the same f7 odds


Chessqueen
Posts: 1445
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by Chessqueen » Tue Feb 23, 2021 8:02 pm

Even with the Extra c7 pawn taken off just like Stash 27.0 these two engines were NOT a match for Komodo 13.3. Do NOT try this with Stockfish 13, since it will NOT beat Stash 27 nor invictus r305 :roll: :mrgreen:


lkaufman
Posts: 4529
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by lkaufman » Tue Feb 23, 2021 10:28 pm

Chessqueen wrote:
Tue Feb 23, 2021 8:02 pm
Even with the Extra c7 pawn taken off just like Stash 27.0 these two engines were NOT a match for Komodo 13.3. Do NOT try this with Stockfish 13, since it will NOT beat Stash 27 nor invictus r305 :roll: :mrgreen:

I think that you are confirming what I've found many times in the past, which is that normal (A/B) engines are much worse at exploiting a handicap than top human players are. Nakamura played a lot of blitz games with Komodo personalities recently, for which I had approximate CCRL blitz ratings based on playing them against CCRL blitz engines at suitable ratings. My conclusion was that at fast blitz, Nakamura would only get a CCRL blitz rating of something like 2200 (shocking, I know). Yet it seems that 3000 ccrl rated engines are not better than Nakamura at exploiting a handicap. Of course some of this is due to Nakamura being aware that he only needs to find good moves, not "best" moves, but still the difference seems hard to explain. I'm about even with Dragon at "fast rapid" (10' + 5") knight odds, but I think I'm only something like 1800 or so on CCRL blitz scale at that time limit, and I think an engine needs a lot more than 1800 to be competitive with Dragon at knight odds at 10' + 5". So I don't think this is anything specific to one person, it's a general difference between human and engine play.
Komodo rules!

Chessqueen
Posts: 1445
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by Chessqueen » Thu Feb 25, 2021 10:41 pm

Larry stated " I'm about even with Dragon at "fast rapid" (10' + 5") knight odds, but I think I'm only something like 1800 or so on CCRL blitz scale at that time limit, and I think an engine needs a lot more than 1800 to be competitive with Dragon at knight odds at 10' + 5". So I don't think this is anything specific to one person, it's a general difference between human and engine play."

Here is a game against Snowy rated 1978. I am almost certain that you can beat Snowy rated 1978 or even BikJump rated 2000 by CCRL but it play horribly the endgame without EGTB



Chessqueen
Posts: 1445
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by Chessqueen » Fri Feb 26, 2021 2:39 am

Here is a knight Odds very tactical game against BiKJump rated 2103



lkaufman
Posts: 4529
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: How many of you believe that Stockfish 12 or Komodo can give a rook Odds to a 1450 engine ?

Post by lkaufman » Fri Feb 26, 2021 3:28 am

Chessqueen wrote:
Fri Feb 26, 2021 2:39 am
Here is a knight Odds very tactical game against BiKJump rated 2103


Actually I have played Bikjump (on my super-fast i-9 10 core 20 thread laptop), and while I may have scored in Rapid, I have very little chance in blitz or near-blitz (10' + 5") against it. Yet I will score much better than Bikjump taking knight odds from Dragon. This is what I'm talking about. I know to "keep it simple" when up a knight, Bikjump doesn't know that.
Komodo rules!

Post Reply