Komodo 4 on long time control

[MM]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

[notyetagm]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

More details, please.

Why exactly do you believe this? And when will Komodo 4 be available?

[MM]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

More details, please.

Why exactly do you believe this? And when will Komodo 4 be available?

Hi, scrolling many sites (ccrl, cegt, ipon chess,http://www.amateurschach.de/) and reading the authors i have the impression that Komodo improves its performance at long time control. Right now it should be around 20-40 elo weaker than Houdini but its authors talked recently of about an improvement of around 20 (or 30) elo so i think that with a good optimization the overtake should happen. Note that Komodo has a positional style, not a tactical style; in these cases, usually the engine performs better with long time control.
As i know it should be released just before the end of 2011 or very early in 2012.

Regards

[Kingghidorah]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

It'll be close

[Don]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

We believe that Komodo does better at longer time controls based on significant data at longer and longer time controls, but the effect seems to be diminishing with depth. We also believe Houdini 1.5 scales better that Houdini 2.0 at depth, but we are not interested enough in Houdini to study this in any detail.

There does not seem to be a major difference in any program once realistic time controls have been reached, but that is a theory we haven't proved for certain. It's difficult to get the massive amounts of hard data on that to make an empirical proof. But it can clearly be seen at levels like game in 1 or 2 minutes vs game in 10 seconds that there can be a statistically significant difference in strength for the same program relative to others depending solely on depth.

The difference in strength between Komodo and Houdini 2.0 will be close when we release but it's not clear how strong Rybka 5 will be either - we think it likely that it will be stronger than Houdini by a non-trivial margin.

[MM]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

We believe that Komodo does better at longer time controls based on significant data at longer and longer time controls, but the effect seems to be diminishing with depth. We also believe Houdini 1.5 scales better that Houdini 2.0 at depth, but we are not interested enough in Houdini to study this in any detail.

There does not seem to be a major difference in any program once realistic time controls have been reached, but that is a theory we haven't proved for certain. It's difficult to get the massive amounts of hard data on that to make an empirical proof. But it can clearly be seen at levels like game in 1 or 2 minutes vs game in 10 seconds that there can be a statistically significant difference in strength for the same program relative to others depending solely on depth.

The difference in strength between Komodo and Houdini 2.0 will be close when we release but it's not clear how strong Rybka 5 will be either - we think it likely that it will be stronger than Houdini by a non-trivial margin.

Hello,

talking in general i think that every mainly strategical engine (or human) needs more time to produce its best level of play. A highly tactical engine like Houdini has an advantage in short time control games. This advantage dimishes when the TC becomes longer, it is a natural fact for humans too.

Regards

[mwyoung]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

It'll be close

Lonnie

I don't think close will cut it. If Komodo 4 is only close to Houdini 2. Then Komodo 4 is nothing more then a also-ran. And Komodo's sales will suffer.

I read from the Komodo team that Houdini does not scale well at long time controls. I have not seen this in my testing. And I now think that was wishful thinking. Since they said at one point that Komodo was shooting for an October or November release.

This tells me Komodo is having trouble overtaking Houdini. If Komodo 4 does not make the release for the Christmas buying season, it will be most telling.

And I find it funny for people to assume that Houdini is some how a static target....

[lkaufman]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

It'll be close

Lonnie

I don't think close will cut it. If Komodo 4 is only close to Houdini 2. Then Komodo 4 is nothing more then a also-ran. And Komodo's sales will suffer.

I read from the Komodo team that Houdini does not scale well at long time controls. I have not seen this in my testing. And I now think that was wishful thinking. Since they said at one point that Komodo was shooting for an October or November release.

This tells me Komodo is having trouble overtaking Houdini. If Komodo 4 does not make the release for the Christmas buying season, it will be most telling.

And I find it funny for people to assume that Houdini is some how a static target....

We will almost surely release before mid-Dec. The delay has been in getting MP working, as Don gets sidetracked making program improvements. Today we made a nice gain that should put us too close to call with Houdini at 40/40 based on CCRL ratings, though we will probably not catch Houdini at blitz levels with this release. The CCRL and CEGT tests show us scaling much better than Houdini, and also show Houdini 2.0 scaling much worse than Houdini 1.5. In fact we may pass Houdini 2.0 on some lists without passing Houdini 1.5! This lower rating of 2.0 vs. 1.5 is why people assume that Houdini is static.

[Don]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

It'll be close

Lonnie

I don't think close will cut it. If Komodo 4 is only close to Houdini 2. Then Komodo 4 is nothing more then a also-ran. And Komodo's sales will suffer.

I read from the Komodo team that Houdini does not scale well at long time controls. I have not seen this in my testing. And I now think that was wishful thinking. Since they said at one point that Komodo was shooting for an October or November release.

This tells me Komodo is having trouble overtaking Houdini. If Komodo 4 does not make the release for the Christmas buying season, it will be most telling.

And I find it funny for people to assume that Houdini is some how a static target....

The versions of Komodo we have now are the same strength or stronger than Houdini. If you look at CCRL 40/40 you will see that Komodo is in second place, 27 ELO behind Houdini. We have gained over 30 ELO so we have caught Houdini 2.0 and probably even the slightly stronger Houdini 1.5. This is 64 bit single CPU engines. Rybka 4 is only slightly behind and in general there are 4 programs in a close pack for second place with Stockfish lagging a bit behind but after than almost 100 ELO gap.

The primary wait is that we want to get the SMP version working well and we would like to come out with a program that is significantly stronger than Houdini, not just slightly stronger. But we are not going to delay a long time just to do this as we can do it with a later release. Houdini and Rybka will no doubt respond, but we will follow up with something substantially stronger. Komodo is ripe for many optimization's that are already in other programs and we have never focused as much on that as we have been able to achieve ELO by smarter algorithms.

[mwyoung]

I think that Komodo 4 on tournament time control is going to overtake Houdini. That is my opinion looking at the different rating related at different time controls (http://www.amateurschach.de/)

Regards

It'll be close

Lonnie

I don't think close will cut it. If Komodo 4 is only close to Houdini 2. Then Komodo 4 is nothing more then a also-ran. And Komodo's sales will suffer.

I read from the Komodo team that Houdini does not scale well at long time controls. I have not seen this in my testing. And I now think that was wishful thinking. Since they said at one point that Komodo was shooting for an October or November release.

This tells me Komodo is having trouble overtaking Houdini. If Komodo 4 does not make the release for the Christmas buying season, it will be most telling.

And I find it funny for people to assume that Houdini is some how a static target....

We will almost surely release before mid-Dec. The delay has been in getting MP working, as Don gets sidetracked making program improvements. Today we made a nice gain that should put us too close to call with Houdini at 40/40 based on CCRL ratings, though we will probably not catch Houdini at blitz levels with this release. The CCRL and CEGT tests show us scaling much better than Houdini, and also show Houdini 2.0 scaling much worse than Houdini 1.5. In fact we may pass Houdini 2.0 on some lists without passing Houdini 1.5! This lower rating of 2.0 vs. 1.5 is why people assume that Houdini is static.

I tested Houdini 2 at long time controls and it was a bit better then the 25 elo Robert said it would be, and some others have shown the same results.

Don't know why CCRL stopped testing Houdini 2 at 40/40. The results they did post were very good and have Houdni 2 at 3341 at 40/40.


CCRL 40/40 results from the complete rating list tab.

Houdini 2.0 64-bit 4CPU 3341
Houdini 1.5a 64-bit 4CPU 3306
Rybka 4.1 64-bit 4CPU 3269
Critter 1.2 64-bit 4CPU3264

But my point in my post was Houdini does not seem to have a problem coming out with a new version when needed. I would not assume a new version is not in the wings waiting for you guys to show your hand.

So Komodo 4 being close will not cut it IMO. I would expect a Komodo 4 only a few elo better then Houdini would be trumped almost immediately.

Robert does not seem to be like Rybka. Who will only release a new stronger version after a year more. This is what I mean by static.

So I do not assume Komodo 4 will be safe from a new release from Houdini for very long.