Re: Komodo 10.4 released
Posted: Wed Mar 15, 2017 12:48 am
64 cores is right around the corner with the AMD Naples:
https://www.youtube.com/watch?v=-gv-TeAvlVs
https://www.youtube.com/watch?v=-gv-TeAvlVs
A bit disappointing. Assuming Intel's HT gives 20-25% NPS boost (not very much a strength boost in the case 8 --> 16 threads), their speed is almost equal, while in that French benchmark Ryzen 1800X was 20% faster than i7-5960X, both on all 16 threads.Ron Langeveld wrote:Preliminary results I got are:
About 12.000 kN/s in start position with a Ryzen 1800X (HT)
versus 9.5000 kN/s with a 5960X Intel 8-core desktop (no-HT)
Maybe comparing apples and oranges to you but hey, we all know Intel's HT sucks, right ? At least I diabled it as long as I can remember.
P.S. Neither setup was OC-ed.
I fooled around with this "Variety" UCI option, and saw something surprising: Komodo 10.4 Variety=0 already randomizes quite a lot on 1 thread at 10''+0.1'' TC. Very different from Stockfish on 1 thread, where I found almost no randomization. I must have missed an important aspect of Komodo, or is it something new with the default Komodo? Here are my results for the number of different final positions got in 5 moves from basic KID Saemisch at 10''+0.1'' in 400 self-play games:mjlef wrote:Sure.
"Variety" modifies something in the evaluation function. A user can enter a number from 0 (no variety--always play the best move) to 20 (most variety, play more different moves). The value of "Variety" defines the magnitude of this evaluation change. At the start of each new game, Komodo randomly selects a value in the range of 0...Value, and modifies this evaluation term. For the rest of the game it uses this new value in computing the evaluation. These smallish changes are enough to have Komodo play a variety of different games, without it changing elo much.
This means a human play will get some different move choices (not one forced moves but in places where several moves have a close eval). It is possible it is even useful in computer match play, since games will vary instead of Komodo always choosing the same move.
We were surprised to find that even the higher Variety settings do not hurt elo by much. In fact, some test matches did a little better than default. But for best play Variety=0 is probably best.
One last use: prevent someone with a copy of Komodo from "booking up" against your Komodo.
Mark
Code: Select all
Kom Var 20: 229/400
Kom Var 0: 218/400
SF: 6/400
Actually, I am talking with New in Chess about an updated version later this year!Frank Quisinsky wrote:Hi Larry,
to my feeling after Stockfish and Komodo and the first Houdini analysis for my opening book project:
To many of "critical" or better "unclear" openings, not important closed or open positions, Komodo is playing more safty with for humans more logical main lines. Houdini maybe a bit more speculative in comparing to Stockfish (can be good and interesting too).
But my resume is more and more clear:
Komodo is for opening analysis my personal favorite.
If I am looking in B01 analysis ... my favorite opening I have much knowledge ... opinion should be very clear because very often much more logical what Komodo can give me.
Question to Larry:
You wrote 2012 a very interesting written chess book:
The Kaufman Repertoire for Black and White:
ISBN-12: 978.90-5691-371-7
Engines used:
Komodo 3, Houdini 1.5, Rybka 4, Critter 1.2, Stockfish 2.1
I think with stronger hardware & engines today maybe opinion changed to different openings. Much better material can be produced. OK, all will be a lot of work ... means to update the written book.
I am sure with the current Komodo version many nice things can be found. In combination with "human like" engines ... Booot for an example, Wasp for an example (must not be the strongest engines) and of course Houdini and Stockfish you will find a lot.
Maybe analysis I do with my opening book project will be help a bit.
Best
Frank
Thanks for Komodo 10.4!
A quick test at 12 ply on one thread (714 games) showed no changed games with Variety set to zero. So I suppose it must be due to timing variation, but then the difference between Variety 20 and Variety 0 is way too small, and the difference from Stockfish is implausibly large. I guess I'll have to investigate further.Laskos wrote:I fooled around with this "Variety" UCI option, and saw something surprising: Komodo 10.4 Variety=0 already randomizes quite a lot on 1 thread at 10''+0.1'' TC. Very different from Stockfish on 1 thread, where I found almost no randomization. I must have missed an important aspect of Komodo, or is it something new with the default Komodo? Here are my results for the number of different final positions got in 5 moves from basic KID Saemisch at 10''+0.1'' in 400 self-play games:mjlef wrote:Sure.
"Variety" modifies something in the evaluation function. A user can enter a number from 0 (no variety--always play the best move) to 20 (most variety, play more different moves). The value of "Variety" defines the magnitude of this evaluation change. At the start of each new game, Komodo randomly selects a value in the range of 0...Value, and modifies this evaluation term. For the rest of the game it uses this new value in computing the evaluation. These smallish changes are enough to have Komodo play a variety of different games, without it changing elo much.
This means a human play will get some different move choices (not one forced moves but in places where several moves have a close eval). It is possible it is even useful in computer match play, since games will vary instead of Komodo always choosing the same move.
We were surprised to find that even the higher Variety settings do not hurt elo by much. In fact, some test matches did a little better than default. But for best play Variety=0 is probably best.
One last use: prevent someone with a copy of Komodo from "booking up" against your Komodo.
Mark
Starting position:
[d]rnbqk2r/ppp1ppbp/3p1np1/8/2PPP3/2N2P2/PP4PP/R1BQKBNR b KQkq - 0 5
Different final positions:
Looking at the distribution of openings in the resulting EPD files of different positions, it seems that Komodo Var 20 or Var 0 cannot yet play randomly reliably, some openings in 400 games (on 5 moves) are repeated 20+ times, while the majority are single occurrences. Still, Komodo Var 20 is probably the best tool to build opening "books" for a position we are interested in.Code: Select all
Kom Var 20: 229/400 Kom Var 0: 218/400 SF: 6/400
Hash not cleared at ucinewgame?lkaufman wrote:A quick test at 12 ply on one thread (714 games) showed no changed games with Variety set to zero. So I suppose it must be due to timing variation, but then the difference between Variety 20 and Variety 0 is way too small, and the difference from Stockfish is implausibly large. I guess I'll have to investigate further.Laskos wrote:I fooled around with this "Variety" UCI option, and saw something surprising: Komodo 10.4 Variety=0 already randomizes quite a lot on 1 thread at 10''+0.1'' TC. Very different from Stockfish on 1 thread, where I found almost no randomization. I must have missed an important aspect of Komodo, or is it something new with the default Komodo? Here are my results for the number of different final positions got in 5 moves from basic KID Saemisch at 10''+0.1'' in 400 self-play games:mjlef wrote:Sure.
"Variety" modifies something in the evaluation function. A user can enter a number from 0 (no variety--always play the best move) to 20 (most variety, play more different moves). The value of "Variety" defines the magnitude of this evaluation change. At the start of each new game, Komodo randomly selects a value in the range of 0...Value, and modifies this evaluation term. For the rest of the game it uses this new value in computing the evaluation. These smallish changes are enough to have Komodo play a variety of different games, without it changing elo much.
This means a human play will get some different move choices (not one forced moves but in places where several moves have a close eval). It is possible it is even useful in computer match play, since games will vary instead of Komodo always choosing the same move.
We were surprised to find that even the higher Variety settings do not hurt elo by much. In fact, some test matches did a little better than default. But for best play Variety=0 is probably best.
One last use: prevent someone with a copy of Komodo from "booking up" against your Komodo.
Mark
Starting position:
[d]rnbqk2r/ppp1ppbp/3p1np1/8/2PPP3/2N2P2/PP4PP/R1BQKBNR b KQkq - 0 5
Different final positions:
Looking at the distribution of openings in the resulting EPD files of different positions, it seems that Komodo Var 20 or Var 0 cannot yet play randomly reliably, some openings in 400 games (on 5 moves) are repeated 20+ times, while the majority are single occurrences. Still, Komodo Var 20 is probably the best tool to build opening "books" for a position we are interested in.Code: Select all
Kom Var 20: 229/400 Kom Var 0: 218/400 SF: 6/400
Assuming you have at least one GB memory on the machine free, what you want to do with 5 piece Syzygy is leave several GB free (so set Hash appropriately) and set Probe Depth to 2. Most operating systems will cache hard disk access in memory, so once enough have been loaded, Syzygy access should be very fast. In this example Smart Syzygy should be off.Nordlandia wrote:For Komodo 10.4 playing on older computers.
Is it recommended to activate "Smart Syzygy" in case probing 5-Men Syzygy from an 100 MB/s (read) HDD. If so what is the recommended settings for probe depth.
I've used 4-men syzygy but they rarely impact game compared to 5-men.
It seems probing 5-men from HDD add speed penalty.