Komodo 14 released.

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Nordlandia
Posts: 2628
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Komodo 14 released.

Post by Nordlandia » Sat Jun 27, 2020 4:56 pm

What happens if i select "Armageddon White must win" if Komodo is black.

In selfplay do it make more sense to select White must win for both parties than White must win against Black must win ?

mjlef
Posts: 1451
Joined: Thu Mar 30, 2006 12:08 pm
Contact:

Re: Komodo 14 released.

Post by mjlef » Sat Jun 27, 2020 5:18 pm

Nordlandia wrote:
Sat Jun 27, 2020 4:56 pm
What happens if i select "Armageddon White must win" if Komodo is black.

In selfplay do it make more sense to select White must win for both parties than White must win against Black must win ?
If Armagennon -White must Win is selected and Komodo is playing the black side, then Komodo will consider any draw a win for itself. For auto play it makes sense to select White must Win for both sides, but give the white side more time to compensate for the need to win. I am.not sure what chess GUIs allow that. Humans are normally given more time for the must win side.

Mark

User avatar
Nordlandia
Posts: 2628
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Komodo 14 released.

Post by Nordlandia » Sat Jun 27, 2020 5:48 pm

Mark! i know Fritz allow to adjust more time for the white side.

To offset the draw odds. I suppose it make sense to give white more time than black even for engine games.

Any recommended time controls for Armageddon. White get extra time, i.e. a minute or so (no extra increment).

lkaufman
Posts: 4222
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Komodo 14 released.

Post by lkaufman » Sat Jun 27, 2020 7:54 pm

Nordlandia wrote:
Sat Jun 27, 2020 5:48 pm
Mark! i know Fritz allow to adjust more time for the white side.

To offset the draw odds. I suppose it make sense to give white more time than black even for engine games.

Any recommended time controls for Armageddon. White get extra time, i.e. a minute or so (no extra increment).
I did some tests on that. In bullet chess double time for White is almost enough to balance things, in blitz White needs triple time at least. You can use a small increment, but of course White also gets double (or whatever is applied to main time) increment time. If you use no increment at all there is always the risk of time forfeits or games decided by the arbitrary minimum number of milliseconds per move an engine might require. The newest Komodo 14 GUI now allows engine matches to be set with tenths of a minute and of a second increment, so if you use that you can test with just 0.1" increment to avoid problems, that's plenty to avoid forfeits and minimums.
Komodo rules!

Chessqueen
Posts: 975
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Komodo 14 released.

Post by Chessqueen » Sat Jun 27, 2020 9:11 pm

lkaufman wrote:
Sat Jun 27, 2020 7:54 pm
Nordlandia wrote:
Sat Jun 27, 2020 5:48 pm
Mark! i know Fritz allow to adjust more time for the white side.

To offset the draw odds. I suppose it make sense to give white more time than black even for engine games.

Any recommended time controls for Armageddon. White get extra time, i.e. a minute or so (no extra increment).
I did some tests on that. In bullet chess double time for White is almost enough to balance things, in blitz White needs triple time at least. You can use a small increment, but of course White also gets double (or whatever is applied to main time) increment time. If you use no increment at all there is always the risk of time forfeits or games decided by the arbitrary minimum number of milliseconds per move an engine might require. The newest Komodo 14 GUI now allows engine matches to be set with tenths of a minute and of a second increment, so if you use that you can test with just 0.1" increment to avoid problems, that's plenty to avoid forfeits and minimums.

Can Komodo be modified to first consider sacrificing a major piece like in this case with the Rook to chase the opponent KING
even if the position is a winning position and winning in many ways. Like a boxer he could beat you slowly with many jabs but
he could also find a fast way to knock you out with a solid punch, with a code that first look at any piece that can take
regardless if it is only a mere pawn like in this case and further after the sacrifice within 15 to 20 moves deep if it come
up with a positive score that will not let the other engine out of the knockout punch, then it will select that route first.

PS: So again how many other engines first consider to chase the opponent KING by sacrificing a major piece :?:

lkaufman
Posts: 4222
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Komodo 14 released.

Post by lkaufman » Sat Jun 27, 2020 10:07 pm

Chessqueen wrote:
Sat Jun 27, 2020 9:11 pm
lkaufman wrote:
Sat Jun 27, 2020 7:54 pm
Nordlandia wrote:
Sat Jun 27, 2020 5:48 pm
Mark! i know Fritz allow to adjust more time for the white side.

To offset the draw odds. I suppose it make sense to give white more time than black even for engine games.

Any recommended time controls for Armageddon. White get extra time, i.e. a minute or so (no extra increment).
I did some tests on that. In bullet chess double time for White is almost enough to balance things, in blitz White needs triple time at least. You can use a small increment, but of course White also gets double (or whatever is applied to main time) increment time. If you use no increment at all there is always the risk of time forfeits or games decided by the arbitrary minimum number of milliseconds per move an engine might require. The newest Komodo 14 GUI now allows engine matches to be set with tenths of a minute and of a second increment, so if you use that you can test with just 0.1" increment to avoid problems, that's plenty to avoid forfeits and minimums.

Can Komodo be modified to first consider sacrificing a major piece like in this case with the Rook to chase the opponent KING
even if the position is a winning position and winning in many ways. Like a boxer he could beat you slowly with many jabs but
he could also find a fast way to knock you out with a solid punch, with a code that first look at any piece that can take
regardless if it is only a mere pawn like in this case and further after the sacrifice within 15 to 20 moves deep if it come
up with a positive score that will not let the other engine out of the knockout punch, then it will select that route first.

PS: So again how many other engines first consider to chase the opponent KING by sacrificing a major piece :?:
Well, it's not very important what Komodo considers first at the root, it looks at all moves to depth 1, then 2, then 3, etc. It is true that moves near the end of the list, including moves that look like blunders, do get reduced more which makes it take longer to see their benefit. It's easy to turn that off, but it loses elo. There are things we can do to increase the odds of finding a forced mate or overwhelming win of material, but they usually lose elo. It's hard to tell the difference between a clever sacrifice and a stupid blunder without a lot of analysis. If you are looking for a way to make Komodo choose a move that wins less material than another move just because it looks pretty, well aesthetics aren't in the program now. Of course they could be, but that would also lose elo.
Komodo rules!

Cornfed
Posts: 35
Joined: Sun Apr 26, 2020 9:40 pm
Full name: Brian D. Smith

Re: Komodo 14 released.

Post by Cornfed » Sun Jun 28, 2020 11:43 pm

A question about Komodo MCTS and Chessbase (since all serious - and most other - chess players uses it).

In their Tactical Analysis feature, an engine goes to the end of a game and works backwards comparing each move vs an engines preferred move. If you look at the GUI, it seems as a default to automatically consider two lines (pv). So, for MCTS, should you turn on that engine first and then simply hit 'tactical analysis' or go ahead and turn on the engine to consider 2 lines - pv, (or 3? 4?) having already set the CPU's to 2 (or 3 or 4)?

Or is simply turning on Komodo regular and let the default settings take over - 2 lines being looked at under however many CPU you specify.

lkaufman
Posts: 4222
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Komodo 14 released.

Post by lkaufman » Mon Jun 29, 2020 1:28 am

Cornfed wrote:
Sun Jun 28, 2020 11:43 pm
A question about Komodo MCTS and Chessbase (since all serious - and most other - chess players uses it).

In their Tactical Analysis feature, an engine goes to the end of a game and works backwards comparing each move vs an engines preferred move. If you look at the GUI, it seems as a default to automatically consider two lines (pv). So, for MCTS, should you turn on that engine first and then simply hit 'tactical analysis' or go ahead and turn on the engine to consider 2 lines - pv, (or 3? 4?) having already set the CPU's to 2 (or 3 or 4)?

Or is simply turning on Komodo regular and let the default settings take over - 2 lines being looked at under however many CPU you specify.
I am not very familiar with that specific GUI feature, I'd have to check it out to advise on how it works, but if you want to know whether to use Komodo regular or MCTS mode with it, I can advise just based on your comments. If it is analyzing and displaying just the best two lines, it's pretty much a toss-up as to which to use, but if it is analyzing three or more lines, then it is better to use the MCTS version. I can also say that if you use the regular version, looking at more lines will lower the quality of the analysis, while with the MCTS version you might as well look at a large number of lines, the quality of the analysis will be the same. This should be the same whether you are analyzing live or using this GUI feature.
Komodo rules!

User avatar
Ovyron
Posts: 4251
Joined: Tue Jul 03, 2007 2:30 am

Re: Komodo 14 released.

Post by Ovyron » Mon Jun 29, 2020 2:17 am

lkaufman wrote:
Mon Jun 29, 2020 1:28 am
I'd have to check it out to advise on how it works
The idea of the feature is to visit future positions to fill up the hash and then visit previous ones so MultiPV=2 makes use of the hash contents to be faster.

In reality it's a poor designed feature and the user would get much better results (either the same results much faster or the same output with higher depth) by using the engines' searchmoves feature and exclude lines already checked.

It's a feature for lazy users that want to have automation, an "I click and have to do nothing and get something" option, and the best advice I can give is to stick with Komodo regular and use SinglePV and exclude the future moves you already have analyzed on future positions, because then it'll be like the improvement of move choices between Komodo SinglePV and KomodoMCTS SinglePV.

(I don't actually have KomodoMCTS but that's how these things work)

Cornfed
Posts: 35
Joined: Sun Apr 26, 2020 9:40 pm
Full name: Brian D. Smith

Re: Komodo 14 released.

Post by Cornfed » Mon Jun 29, 2020 3:49 am

Ovyron wrote:
Mon Jun 29, 2020 2:17 am
lkaufman wrote:
Mon Jun 29, 2020 1:28 am
I'd have to check it out to advise on how it works

In reality it's a poor designed feature and the user would get much better results (either the same results much faster or the same output with higher depth) by using the engines' searchmoves feature and exclude lines already checked.

It's a feature for lazy users that want to have automation, an "I click and have to do nothing and get something" option, and the best advice I can give is to stick with Komodo regular and use SinglePV and exclude the future moves you already have analyzed on future positions, because then it'll be like the improvement of move choices between Komodo SinglePV and KomodoMCTS SinglePV.
Colorful language...and I'll not take offense; but truly its use is in having an engine quickly go thru a set of games (maybe those you played online or raw games from tournaments), check for tactical refutations/missed combinations/combinational themes, divergence from theory, endgame classification , diagrams of key points with training questions, etc...and a handy little graph occurs so you can spot where a game trended good or bad. All handy stuff really.

Oh, and how would one use this 'search moves feature' you mention on a set of games to get this kind of information?

Post Reply