Komodo...wonderful

[lkaufman]

This engine is very awesome for analysys. The evaluation of the position is always more realistic than the one of critter/houdini. The only problem is that as to run until 300.000.000 nodes to find some original moves and in my computer that runs at 500 kn/s it takes 7/8 minutes. The only problem is not very suited in complex tactical position. in those positions you have to use it a with a fast tactical program like houdini or critter.

A few questions for the developers. Are you working at a release 6? Do you think you want to implement gaviota tablebases? do you think there si a margin of improvement i the strenght of the program ?

Thanks

Actually we already have enough improvements that we could release a Komodo 6 tomorrow with enough elo gain to catch Houdini 2, at least at IPON level or slower. The only reason we don't do so is our commitment to Komodo MP; if we released K6 sp first many would doubt that MP would ever come out!
Here are some test results of latest version ON WINDOWS, without benefit of an optimized (PGO) compile.
Blitz (5' + 3" on each thread of Intel 16 core machine):
vs. Houdini 1.5 + 9.5 elo (2574 games)
vs. Critter 1.6 + 8.6 elo (3224 games)
vs. Stockfish 2.3 +57.8 elo (1463 games)
Rapid (30' + 18" as above)
vs. Houdini 1.5 + 16.7 elo (1621 games)

Since Houdini 2 is only a few elo stronger than 1.5, and since we will gain several more elo from an optimized compile, I feel pretty confident in saying that we have at least caught up with Houdini 2 now except for bullet chess. Note also the improvement with longer time limits (at 1' we are still behind Houdini 1.5). The same trend is apparent on Linux in tests vs. Critter.

[Houdini]

Sounds strangely familiar...

[MM]

This engine is very awesome for analysys. The evaluation of the position is always more realistic than the one of critter/houdini. The only problem is that as to run until 300.000.000 nodes to find some original moves and in my computer that runs at 500 kn/s it takes 7/8 minutes. The only problem is not very suited in complex tactical position. in those positions you have to use it a with a fast tactical program like houdini or critter.

A few questions for the developers. Are you working at a release 6? Do you think you want to implement gaviota tablebases? do you think there si a margin of improvement i the strenght of the program ?

Thanks

Actually we already have enough improvements that we could release a Komodo 6 tomorrow with enough elo gain to catch Houdini 2, at least at IPON level or slower. The only reason we don't do so is our commitment to Komodo MP; if we released K6 sp first many would doubt that MP would ever come out!

Hi Larry,

what does it mean? You mean that Komodo MP will be based on K5?
If i knew that Komodo MP is based on K5 and there's a new Komodo SP stronger of around 20/30 elo (K6) than K5, i would not be interested in MP.

I think, or at least my personal interests make me think that you should release an MP based on your latest and strongest version.

Best Regards

[lkaufman]

Hi Larry,

what does it mean? You mean that Komodo MP will be based on K5?
If i knew that Komodo MP is based on K5 and there's a new Komodo SP stronger of around 20/30 elo (K6) than K5, i would not be interested in MP.

I think, or at least my personal interests make me think that you should release an MP based on your latest and strongest version.

Best Regards

We haven't decided yet whether to release K5 MP and shortly afterward K6 MP or to release them at the same time. Whatever we do, it will be done in a way that is fair to our customers. Obviously, we don't expect people to buy K5 MP without some assurance that upgrading to k6 MP won't cost a lot.

[lkaufman]

Sounds strangely familiar...

I think this is the first time I've posted WINDOWS results. It seems that our earlier results on Linux overstated how we would do on Windows, at least vs. Houdini, but this is no longer an issue. Anyway even you predicted that K6 would catch H2, so you should not be surprised that we have done so.

[velmarin]

We haven't decided yet whether to release K5 MP and shortly afterward K6 MP or to release them at the same time. Whatever we do, it will be done in a way that is fair to our customers. Obviously, we don't expect people to buy K5 MP without some assurance that upgrading to k6 MP won't cost a lot.

Perhaps the possibility of paying a monthly fee.

About 50$ for example, entitled to an update every 15 days.

[Houdini]

Sounds strangely familiar...

I think this is the first time I've posted WINDOWS results. It seems that our earlier results on Linux overstated how we would do on Windows, at least vs. Houdini, but this is no longer an issue. Anyway even you predicted that K6 would catch H2, so you should not be surprised that we have done so.

I'm not surprised that you would eventually catch up with Houdini 2.

I am surprised that you continue to make claims about being superior to Houdini 2 at non-blitz TC without playing any games against Houdini 2 at non-blitz TC. Don't you learn anything from your past failures?

Robert

[Don]

Sounds strangely familiar...

I think this is the first time I've posted WINDOWS results. It seems that our earlier results on Linux overstated how we would do on Windows, at least vs. Houdini, but this is no longer an issue. Anyway even you predicted that K6 would catch H2, so you should not be surprised that we have done so.

I'm not surprised that you would eventually catch up with Houdini 2.

I am surprised that you continue to make claims about being superior to Houdini 2 at non-blitz TC without playing any games against Houdini 2 at non-blitz TC. Don't you learn anything from your past failures?

Robert

Now you two stop it or I'll send you both to bed with no supper!

Don

[bupalo]

yes Don and please stop this thread. I wanted to make compliments to komodo because I noticed its very good analysis and subtitles manouvre of pieces in quite positions. But even houdini is great (and critter is good too!). No matter who is the strongest of the two so you can't compare apples to pears it's the first thing you tell to children. it's better houdini or komodo, it's like to discuss if the bishop is better than the knight. Maybe the bishop is better but you know the in some position is better the kight. So please continue to develop these 2 monsters that's matter

[kgburcham]

Sounds strangely familiar...

I think this is the first time I've posted WINDOWS results. It seems that our earlier results on Linux overstated how we would do on Windows, at least vs. Houdini, but this is no longer an issue. Anyway even you predicted that K6 would catch H2, so you should not be surprised that we have done so.

I'm not surprised that you would eventually catch up with Houdini 2.
I am surprised that you continue to make claims about being superior to Houdini 2 at non-blitz TC without playing any games against Houdini 2 at non-blitz TC. Don't you learn anything from your past failures?
Robert

Now you two stop it or I'll send you both to bed with no supper!
Don

I dont know Don.
Most programmers can do without a meal or two, I would prefer to take away their computers for a month. Maybe their cell phones too.
kgburcham