Komodo Dragon vs IM Andras Toth knight odds match.

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: 2217
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by Chessqueen » Fri Jun 11, 2021 9:03 pm

Uri wrote:
Fri Jun 11, 2021 4:22 pm
Chess engine authors are way too arrogant when they claim that their program is better than Kasparov or better than Magnus Carlsen because it's obviously just a marketing lie.

Kasparov is still at least 200 elo stronger than Komodo dragon 2. Even Komodo Dragon 5 still doesn't stand a chance against the very best human chess players.
Lets assume that Komodo Dragon2 MCTS with one pawn Odds at standard time control of game in 90 minutes drop about 550 Elo if GM Nakamura or Carlsen ever play against Komodo Dragon2 MCTS at time control of at least 90 minutes per game. The question is which pawn Odds can Komodo Dragon 2 MCTS give GM Nakamura or any top 10 GM to get at least a Draw. I do NOT believe that Komodo Dragon2 at TC of at least 90 Minutes per game can give the f2 pawn.
Arizona will have to learn from the Chinese to regain Greenland from desert ==>
https://www.youtube.com/watch?v=KTpaJn22w4I

lkaufman
Posts: 4884
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by lkaufman » Fri Jun 11, 2021 9:08 pm

AdminX wrote:
Fri Jun 11, 2021 5:30 pm
Uri wrote:
Fri Jun 11, 2021 4:22 pm
Chess engine authors are way too arrogant when they claim that their program is better than Kasparov or better than Magnus Carlsen because it's obviously just a marketing lie.

Kasparov is still at least 200 elo stronger than Komodo dragon 2. Even Komodo Dragon 5 still doesn't stand a chance against the very best human chess players.
I don't think GM Kasparov or GM Carlsen would agree with you. :lol: However at Knight odds you might have a point, a very questionable one. LMAO
I asked Kasparov in person in early 2009 if he would be interested in a match with Rybka. He put up his hands as if to say "hell no"! But at the present time Kasparov or Carlsen or Nakamura or So would crush any engine, including Dragon 2, in a knight odds match at Rapid or slower time control. Whether that will be true with Dragon 5 or even Dragon 10 (I'll probably retire before then, but assuming someone continues it) is very hard to say. It seems impossible, but often the impossible becomes reality.
Komodo rules!

Chessqueen
Posts: 2217
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by Chessqueen » Sat Jun 12, 2021 1:21 am

lkaufman wrote:
Fri Jun 11, 2021 9:08 pm
AdminX wrote:
Fri Jun 11, 2021 5:30 pm
Uri wrote:
Fri Jun 11, 2021 4:22 pm
Chess engine authors are way too arrogant when they claim that their program is better than Kasparov or better than Magnus Carlsen because it's obviously just a marketing lie.

Kasparov is still at least 200 elo stronger than Komodo dragon 2. Even Komodo Dragon 5 still doesn't stand a chance against the very best human chess players.
I don't think GM Kasparov or GM Carlsen would agree with you. :lol: However at Knight odds you might have a point, a very questionable one. LMAO
I asked Kasparov in person in early 2009 if he would be interested in a match with Rybka. He put up his hands as if to say "hell no"! But at the present time Kasparov or Carlsen or Nakamura or So would crush any engine, including Dragon 2, in a knight odds match at Rapid or slower time control. Whether that will be true with Dragon 5 or even Dragon 10 (I'll probably retire before then, but assuming someone continues it) is very hard to say. It seems impossible, but often the impossible becomes reality.
You are correct to assume that Komodo Dragon 2 MCTS can give any top player like Carlsen the f7 pawn with the time control of 90 minutes, I know that one game is barely nothing but against stockfish 13 NNUE, Shredder 8 which is to be compared with Carlsen simply could not win.

Arizona will have to learn from the Chinese to regain Greenland from desert ==>
https://www.youtube.com/watch?v=KTpaJn22w4I

Chessqueen
Posts: 2217
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by Chessqueen » Sat Jun 12, 2021 11:33 am

Chessqueen wrote:
Sat Jun 12, 2021 1:21 am
lkaufman wrote:
Fri Jun 11, 2021 9:08 pm
AdminX wrote:
Fri Jun 11, 2021 5:30 pm
Uri wrote:
Fri Jun 11, 2021 4:22 pm
Chess engine authors are way too arrogant when they claim that their program is better than Kasparov or better than Magnus Carlsen because it's obviously just a marketing lie.

Kasparov is still at least 200 elo stronger than Komodo dragon 2. Even Komodo Dragon 5 still doesn't stand a chance against the very best human chess players.
I don't think GM Kasparov or GM Carlsen would agree with you. :lol: However at Knight odds you might have a point, a very questionable one. LMAO
I asked Kasparov in person in early 2009 if he would be interested in a match with Rybka. He put up his hands as if to say "hell no"! But at the present time Kasparov or Carlsen or Nakamura or So would crush any engine, including Dragon 2, in a knight odds match at Rapid or slower time control. Whether that will be true with Dragon 5 or even Dragon 10 (I'll probably retire before then, but assuming someone continues it) is very hard to say. It seems impossible, but often the impossible becomes reality.
You are correct to assume that Komodo Dragon 2 MCTS can give any top player like Carlsen the f7 pawn with the time control of 90 minutes, I know that one game is barely nothing but against stockfish 13 NNUE, Shredder 8 which is to be compared with Carlsen simply could not win.


Note: I just realized that IM Andras will be playing starting at 9 a.m. Eastern daylight time,11:00 PM his local time in Australia !
Arizona will have to learn from the Chinese to regain Greenland from desert ==>
https://www.youtube.com/watch?v=KTpaJn22w4I

Cornfed
Posts: 269
Joined: Sun Apr 26, 2020 9:40 pm
Full name: Brian D. Smith

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by Cornfed » Sat Jun 12, 2021 4:08 pm

Chessqueen wrote:
Fri Jun 11, 2021 8:50 pm
Uri Blass wrote:
Fri Jun 11, 2021 8:24 pm
herO wrote:
Fri Jun 11, 2021 6:09 am
Chessqueen wrote:
Fri Jun 11, 2021 3:35 am
It would be interesting and considered a promotion for Komodo Dragon2 if you can accept predictions from us such as I predict a 3.5 to 2.5 win for IM Andras and whoever predict t it correct to receive Komodo Dragon2 for free, and for those that did NOT predicted the outcome correct to get Komodo Dragon 2 for 1/3 of the price.
I do not agree with your proposal. As a paying customer, I don't want anyone to receive a Komodo engine for free! I see no reason why people who write an estimate on the forum should get a Komodo engine for 1/3 of the price (66% discount)! A 20% discount is given to PAYING customers who have purchased a previous version of the Komodo engine. I have long opposed Komodo from putting their engine for free on their website. Now Komodo 12 is offered for free, which is not right. It was far better if only Komodo 10 was available for free.

Komodo offers an over-up-to-date version of the Komodo engine for free, which is very unfair to paying customers. I buy an engine and then a troll who doesn't want to pay for professional work downloads the engine for free. Absolutely unacceptable! Your proposal is a mockery of the entire Komodo team. Your opinion clearly shows you absolutely don't appreciate the hard work of the Komodo team and you ideally want everything for free. Nothing in the world is free. Deal with it.
It is not correct that Nothing in the world is free.

Stockfish is free and most people who use chess engines prefer to use it and not to use komodo because stockfish is stronger.

Edit:If you think that getting strong engines for free is unfair for paying customers then I think that the first complain should be to stockfish developers and you should tell them to stop release free versions of stockfish because it is unfair for paying customers of commercial engines.
I honesly believe that the Stockfish team or developers should at least charge $30.00, for their effort to compensate for their time in developing such strong engine. The only reasonable commercial price that I see on the market for a strong engine is Ethereal 3.00 NNUE at the moment, and because it sell for a reasonable price at the long run it will make more profit than some of the engines that are sol for 75% more.
Things change so much/fast...only a $30 (whatever the price) for a lifetime 'subscription' would make sense to me for the individual consumer.

Chessqueen
Posts: 2217
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by Chessqueen » Sat Jun 12, 2021 6:38 pm

lkaufman wrote:
Fri Jun 11, 2021 12:09 am
On June 19 and 20 on chess.com, IM Andras Toth (FIDE 2377, born 1981) will play a six game match with Komodo Dragon giving him knight odds in the now standard Rapid time control of 15' + 10", on chess.com. Since he is in Australia and I'm in the U.S., scheduling was a bit difficult, but we settled on two roughly two hour sessions starting each day at 9 a.m. Eastern daylight time, which should be ideal for Europe. Komodo will be set to take only about half or less of the allotted time, so 60 move games shouldn't exceed forty minutes. Komodo will alternate between removing the b1 and g1 knights, playing White as is standard for knight odds. Armageddon scoring will be used to determine match victory, meaning that draws count as wins for Dragon, who will normally be happy to draw down a piece even with a large Contempt factor. There is prize money for the match plus money for wins and a little for draws, so Toth will always have an incentive to complete games as best he can even if they won't affect the match result. In case of a 3-3 tie score, a final game (still with Armageddon rule) will break the tie. We'll use the latest dev. version of Dragon, probably in MCTS mode, with a high Contempt score, running on my 32 core Threadripper, with a small opening book to insure variety and avoid really hopeless openings. Toth is a chess.com streamer, so he will provide commentary on the games (during, after, or both).
Past experience at this or similar time controls vs. humans at knight odds has generally been favorable for Komodo against players below 2300 FIDE, but Komodo did lose by 5 to 1 to the 2500 rated GM David Smerdon a year ago, although that was pre-Dragon. Recently Komodo Dragon 2 scored about evenly with these knight odds Armageddon conditions against GM Alex Lenderman in games with time controls averaging about 6' + 1" (I'm excluding 3' + 1" and faster games where Dragon mostly won and 8' or 9' + 1" where it mostly lost). Lenderman is about 250 elo above Toth, but the time control for Toth is about 3.5 times longer, which obviously greatly favors the player with the winning position (Black). It's not obvious whether a player with 250 less elo but 3.5 times as much thinking time will play better or worse, so I expect a close match. If we do win this match we can challenge a GM next, but that's a big "if". So, who (or what) will win this match?



Next Saturday you will be entertained by the most funny player on this Planet IM Andras Toth. Here are some of his comments while playing a practice game versus the free version of komodo 11. I will quote his comments " This is fun! I do not like to give up an inch Ha ha Ha not only I am smarter and have more chess knowledge but I am faster too. That would be goody (meaning pleasant exciting). Trade, trade, trade. I am going to trade Dude, because I have a piece more, that would be goody. I am giving Komodo too much Air. I am really tempted to take this piece or even that piece. You are dead buddy, you will die. Gee Gee on your Back, Oh Boy I am enjoying myself beating this silly Dragon. Okay give me that pony. I am soo reading the computer mind, that if I can predict the winning lottery numbers the way I can predict Komodo moves, I would be a Rich man." https://www.youtube.com/watch?v=jDmyrxWASOU


Arizona will have to learn from the Chinese to regain Greenland from desert ==>
https://www.youtube.com/watch?v=KTpaJn22w4I

User avatar
ThomasMiller
Posts: 14
Joined: Mon Nov 02, 2020 4:14 pm
Full name: Thomas J. Miller

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by ThomasMiller » Sat Jun 12, 2021 8:24 pm

Larry have you done any study about how much each piece/pawn(with file) odds are worth based on all the handicap matches Komodo played?

lkaufman
Posts: 4884
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by lkaufman » Sat Jun 12, 2021 8:46 pm

ThomasMiller wrote:
Sat Jun 12, 2021 8:24 pm
Larry have you done any study about how much each piece/pawn(with file) odds are worth based on all the handicap matches Komodo played?
Well, it's hard to do a rigorous study since opponents and time limits are constantly changing, as well as the version of Komodo being used, but i have some general conclusions about the values (in the opening position) in these matches. The clearest conclusion is that the value of the (a1) rook is a bit less than 1.5 knights, let's say about 1.4 knights. Rook pawns are worth something like just half a (nonrook) pawn, primarily due to the immediate rook activity. f and g pawns are worth significantly more than the other four (bcde) (which are close in value). And knights are worth about four average pawns. That's about as precise as I can get with the varied conditions. Only a few games with queen for pieces odds, which suggest that the queen is somewhere between two rooks and three minors in value, so maybe about 2.9 x knight.
Komodo rules!

Chessqueen
Posts: 2217
Joined: Wed Sep 05, 2018 12:16 am
Full name: George Pichard

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by Chessqueen » Mon Jun 14, 2021 3:01 am

lkaufman wrote:
Sat Jun 12, 2021 8:46 pm
ThomasMiller wrote:
Sat Jun 12, 2021 8:24 pm
Larry have you done any study about how much each piece/pawn(with file) odds are worth based on all the handicap matches Komodo played?
Well, it's hard to do a rigorous study since opponents and time limits are constantly changing, as well as the version of Komodo being used, but i have some general conclusions about the values (in the opening position) in these matches. The clearest conclusion is that the value of the (a1) rook is a bit less than 1.5 knights, let's say about 1.4 knights. Rook pawns are worth something like just half a (nonrook) pawn, primarily due to the immediate rook activity. f and g pawns are worth significantly more than the other four (bcde) (which are close in value). And knights are worth about four average pawns. That's about as precise as I can get with the varied conditions. Only a few games with queen for pieces odds, which suggest that the queen is somewhere between two rooks and three minors in value, so maybe about 2.9 x knight.
I was wondering and also a little concern, if you feed the first 4 moves of this game to Komodo Dragon2 MCTS what % percentage of the moves would be identical to the moves made by Komodo 11 free version that IM Andras used for this practice game, would you say at least 20% of the moves would be identical, even if you use your best computer Mr. Kaufman?

Arizona will have to learn from the Chinese to regain Greenland from desert ==>
https://www.youtube.com/watch?v=KTpaJn22w4I

lkaufman
Posts: 4884
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Komodo Dragon vs IM Andras Toth knight odds match.

Post by lkaufman » Mon Jun 14, 2021 3:35 am

Chessqueen wrote:
Mon Jun 14, 2021 3:01 am
lkaufman wrote:
Sat Jun 12, 2021 8:46 pm
ThomasMiller wrote:
Sat Jun 12, 2021 8:24 pm
Larry have you done any study about how much each piece/pawn(with file) odds are worth based on all the handicap matches Komodo played?
Well, it's hard to do a rigorous study since opponents and time limits are constantly changing, as well as the version of Komodo being used, but i have some general conclusions about the values (in the opening position) in these matches. The clearest conclusion is that the value of the (a1) rook is a bit less than 1.5 knights, let's say about 1.4 knights. Rook pawns are worth something like just half a (nonrook) pawn, primarily due to the immediate rook activity. f and g pawns are worth significantly more than the other four (bcde) (which are close in value). And knights are worth about four average pawns. That's about as precise as I can get with the varied conditions. Only a few games with queen for pieces odds, which suggest that the queen is somewhere between two rooks and three minors in value, so maybe about 2.9 x knight.
I was wondering and also a little concern, if you feed the first 4 moves of this game to Komodo Dragon2 MCTS what % percentage of the moves would be identical to the moves made by Komodo 11 free version that IM Andras used for this practice game, would you say at least 20% of the moves would be identical, even if you use your best computer Mr. Kaufman?

Probably well over 50%. But why is this figure significant? Dragon2 mcts avoids some of the bad moves like 6.c5? and 11.Re1?, so even if 90% were the same it would be a completely different game. Sometimes I play games where the top engines approve of 90% or more of my moves, it doesn't mean I play at 3500 level. It only takes one bad move to ruin a game.
Komodo rules!

Post Reply