Komodo 10.1 has just been released at komodochess.com. It is the same version which won the World Rapid Championship with a 97% score, and is also the same as the one that beat a GM by 3 to 1 giving him two pawns handicap. It is slightly improved from the version that won the World Computer Championship and the World Software Championship.
We estimate the elo gain over Komodo 10 to be 10 points. While we normally wait until we estimate 15 elo gain for a new version, considering all the results of the past couple weeks we judged that we should release this version.
There are two interesting new options. One controls how aggressively Komodo uses Null Move. The other, "Progress Threshold" controls how many ply without pawn move or capture must elapse before Komodo starts to pull the score towards zero. We used 85 in the World Championship, but we made the default just 30 because this makes Komodo a better analysis tool and shortens boring drawn games. It also helps Komodo recognize fortress draws sooner. We estimate that using 30 costs only 2 elo points and so recommend the default for most users. If you only care about results you can revert to the 85 setting, or something in between.
Komodo 10.1
Moderators: hgm, Rebel, chrisw
-
- Posts: 5960
- Joined: Sun Jan 10, 2010 6:15 am
- Location: Maryland USA
Komodo 10.1
Komodo rules!
-
- Posts: 1494
- Joined: Thu Mar 30, 2006 2:08 pm
Re: Komodo 10.1
There is a third feature "Smart Syzygy". This is designed for people with 6 piece Syzygy on their hard disks. Accessing 6 piece from a hard drive causes a pretty big slowdown. You could lessen that by setting Syzygy Probe Depth to a higher number, like 10. But that keeps Komodo from probing at lower depths when the program gets to a 5 piece or less endgame. It turns out the part of Syzygy that is probed during the search is about 1/2 a gigabyte. Modern computers have enough memory to automatically buffer this in memory for fast access.
So if "Smart Syzygy" is on, it always probes Syzygy TB for 5, 4 and 3 piece endgames, and for 6 piece endgames, uses whatever depth you set the Syzygy Probe Depth too. This can greatly improve play on a computer with standard hard drives.
Of course if your Syzygy tablebases are on a SSD, you can usually probe at any depth since they are much faster.
So if "Smart Syzygy" is on, it always probes Syzygy TB for 5, 4 and 3 piece endgames, and for 6 piece endgames, uses whatever depth you set the Syzygy Probe Depth too. This can greatly improve play on a computer with standard hard drives.
Of course if your Syzygy tablebases are on a SSD, you can usually probe at any depth since they are much faster.
-
- Posts: 313
- Joined: Sat Apr 05, 2014 12:09 am
- Location: Neenah, WI, United States
Re: Komodo 10.1
Hi guys. Thanks for all the work. So you viewed this as a milestone version of successful results, so you wanted to release it before you reached 15 elo. Understandable. Still loving dynamism. Thanks very much for it!lkaufman wrote:Komodo 10.1 has just been released at komodochess.com. It is the same version which won the World Rapid Championship with a 97% score, and is also the same as the one that beat a GM by 3 to 1 giving him two pawns handicap. It is slightly improved from the version that won the World Computer Championship and the World Software Championship.
We estimate the elo gain over Komodo 10 to be 10 points. While we normally wait until we estimate 15 elo gain for a new version, considering all the results of the past couple weeks we judged that we should release this version.
There are two interesting new options. One controls how aggressively Komodo uses Null Move. The other, "Progress Threshold" controls how many ply without pawn move or capture must elapse before Komodo starts to pull the score towards zero. We used 85 in the World Championship, but we made the default just 30 because this makes Komodo a better analysis tool and shortens boring drawn games. It also helps Komodo recognize fortress draws sooner. We estimate that using 30 costs only 2 elo points and so recommend the default for most users. If you only care about results you can revert to the 85 setting, or something in between.
Sincerely,
Tim.
-
- Posts: 1766
- Joined: Wed Jun 03, 2009 12:14 am
Re: Komodo 10.1
great features, & as someone still obstinately holding out on getting another SSD for tablebases, really appreciate the smart syzygy implementation, thanks!
-
- Posts: 134
- Joined: Thu Jun 01, 2006 6:05 pm
Re: Komodo 10.1
But how can to upgrade to Komodo 10.1? I've to pay other euros? This 'ld be incredible!
-
- Posts: 1339
- Joined: Fri Nov 02, 2012 9:43 am
- Location: New Delhi, India
Re: Komodo 10.1
You got THAT right !keres wrote:This 'ld be incredible!
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis
-
- Posts: 5960
- Joined: Sun Jan 10, 2010 6:15 am
- Location: Maryland USA
Re: Komodo 10.1
Those who want every upgrade should buy the subscription, that's what it is for. Basically, it amounts to getting upgrades for only about $10 each.keres wrote:But how can to upgrade to Komodo 10.1? I've to pay other euros? This 'ld be incredible!
Komodo rules!
-
- Posts: 8514
- Joined: Thu Mar 09, 2006 3:25 am
- Location: Jerusalem Israel
Re: Komodo 10.1
When you say you estimate 10 elo, do you really mean it could also be much more, but at LEAST 10 elo?lkaufman wrote:Komodo 10.1 has just been released at komodochess.com. It is the same version which won the World Rapid Championship with a 97% score, and is also the same as the one that beat a GM by 3 to 1 giving him two pawns handicap. It is slightly improved from the version that won the World Computer Championship and the World Software Championship.
We estimate the elo gain over Komodo 10 to be 10 points. While we normally wait until we estimate 15 elo gain for a new version, considering all the results of the past couple weeks we judged that we should release this version.
Stockfish publicizers seem to be estimating a much greater increase of ELO.
-
- Posts: 5960
- Joined: Sun Jan 10, 2010 6:15 am
- Location: Maryland USA
Re: Komodo 10.1
Well, my latest test at relatively long tc (5' + 3") shows +12 elo, and I think the IPON test will show close to ten. Of course it's possible but unlikely that the gain is much larger at really long time limits. Five elo per month is about normal for us, and I think is about what SF averaged from SF 6 to 7. Maybe they just had a lucky month or two lately.S.Taylor wrote:When you say you estimate 10 elo, do you really mean it could also be much more, but at LEAST 10 elo?lkaufman wrote:Komodo 10.1 has just been released at komodochess.com. It is the same version which won the World Rapid Championship with a 97% score, and is also the same as the one that beat a GM by 3 to 1 giving him two pawns handicap. It is slightly improved from the version that won the World Computer Championship and the World Software Championship.
We estimate the elo gain over Komodo 10 to be 10 points. While we normally wait until we estimate 15 elo gain for a new version, considering all the results of the past couple weeks we judged that we should release this version.
Stockfish publicizers seem to be estimating a much greater increase of ELO.
Komodo rules!
-
- Posts: 8514
- Joined: Thu Mar 09, 2006 3:25 am
- Location: Jerusalem Israel
Re: Komodo 10.1
Well, so it looks like you are very open to coming second!