5 more games, and now we have 80 games in the books. In the last 5, Komodo has cut Houdini's lead from 7 games to 5, and is now also playing the white pieces. I don't like the odds of Komodo making up 5 games in the last 20, but one thing for sure- it doesn't look as if Komodo plans to go quietly into the night.
Let me just mention ONE MORE TIME that you can see I am not applying elo figures and it is doubtful I will until the final update. This match has nothing to do with elo, or the probabilities that "this engine" or "that engine" should do "this or that." This match is about one thing only- when it came down to the nut-cutting, one engine managed to come thru and the other did not. The engine that wins will be, in the end, the engine that played the best chess in this 100 game match. And that is ALL THAT CONCERNS ME. Here we go:
Alienware AURORA_R4 Intel i7 w/6 True Cores Factory overclocked
Fritz 11 gui 6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40 Match=100 games
Carlos777 wrote:I also heard he wanted to complete the 100 games match with Stockfish.
Actually an idea that has a lot of merit, and I certainly see your point. Problem with that is 3-fold. First, since that match took place, the particular Stockfish version has been beaten by Komodo. 2nd- just for the sake of argument- let's assume that Stockfish (at least that develop. version) has a style of play that gives Houdini a lot of trouble. I plan to hopefully have 2 to 3 qualifiers a year, which means the winner that comes out of the qualifiers gets a title shot at No. 1. Stockfish in the qualifying rounds will likely be facing engines that "do not" have that particular trouble with his style of play. And he would have to win at minimum 2 matches to come out with a chance against No. 1. And third, tho I did bill that Stockfish-Houdini match as "a shot at the title"- in reality it is unlikely a develop. version would ever get allowed in a qualifier- much less just a shot at the title without having to prove anything first. I think you would agree that qualifying matches and championship matches should be restricted to "officially released versions." And NOT officially released "retroactively."
The only news that could be bad for Houdini is that it is a lot of work to come out with a new version every 9 months or so. But the way Stockfish is handling their testing, there is a new version almost daily. You have to pick and choose- some won't show much if any elo increase- but they are already 35 to 50 elo above their last official release. And where will they be 60 days from now?
Having said all that- you still had a damn good thought. I just don't want to use my CPU time giving Houdini a chance to win something back that technically he never lost.
All the best-
george
I agree about official releases. You could do something similar like the candidates tournament. An 8 player RR tournament to decide who is the challenger. Although to be sincere there are no more than 4 or 5 engines that could give Houdini some trouble in long time controls, multicore conditions.
Now aside that, it would be interesting to run a rematch between that version of Stockfish vs H3 with more games to verify if it really is a "H3 Killer". Unfortunately, I don't have the resources to run that match, maybe someone else could.
Thanks for all these entertaining games/tournaments.
Carlos777 wrote:I also heard he wanted to complete the 100 games match with Stockfish.
Actually an idea that has a lot of merit, and I certainly see your point. Problem with that is 3-fold. First, since that match took place, the particular Stockfish version has been beaten by Komodo. 2nd- just for the sake of argument- let's assume that Stockfish (at least that develop. version) has a style of play that gives Houdini a lot of trouble. I plan to hopefully have 2 to 3 qualifiers a year, which means the winner that comes out of the qualifiers gets a title shot at No. 1. Stockfish in the qualifying rounds will likely be facing engines that "do not" have that particular trouble with his style of play. And he would have to win at minimum 2 matches to come out with a chance against No. 1. And third, tho I did bill that Stockfish-Houdini match as "a shot at the title"- in reality it is unlikely a develop. version would ever get allowed in a qualifier- much less just a shot at the title without having to prove anything first. I think you would agree that qualifying matches and championship matches should be restricted to "officially released versions." And NOT officially released "retroactively."
The only news that could be bad for Houdini is that it is a lot of work to come out with a new version every 9 months or so. But the way Stockfish is handling their testing, there is a new version almost daily. You have to pick and choose- some won't show much if any elo increase- but they are already 35 to 50 elo above their last official release. And where will they be 60 days from now?
Having said all that- you still had a damn good thought. I just don't want to use my CPU time giving Houdini a chance to win something back that technically he never lost.
All the best-
george
I agree about official releases. You could do something similar like the candidates tournament. An 8 player RR tournament to decide who is the challenger. Although to be sincere there are no more than 4 or 5 engines that could give Houdini some trouble in long time controls, multicore conditions.
Now aside that, it would be interesting to run a rematch between that version of Stockfish vs H3 with more games to verify if it really is a "H3 Killer". Unfortunately, I don't have the resources to run that match, maybe someone else could.
Thanks for all these entertaining games/tournaments.
Best regards,
Carlos
I will do my best to try and replay that match for you if I can- between Houdini and the Stockfish version he lost to. I only have a few problems right now. I have to finish this Houdini-Komodo match first. Then the system I ran it on- Alienware AURORA R4 Intel i7 with 6 True Cores- I bought another one EXACTLY like it. It came in 5 days ago, and I have been waiting till this match was over to set it up. Because I am going to do the same with the 2 Intel i7s that I have done with my 2 Intel i5 4-Core systems. I am going to network the 2 i7s together so they will both use the same monitor, and I can switch back and forth between them, transfer files easily, etc. And I can't be running matches till I get this done. After setup, I gotta transfer files and all the normal stuff that goes with setting a new one up- like disabling hyperthreading, etc. And I am Giancarlo's beta tester for Equinox, and I just got a new beta there to try out yesterday. Then I am beta testing Strelka 5.6 and 5.7MP for Yuri. It's just a lot of stuff going on, but if there is any way I can- I will do it for you. Because even if you had the system to recreate it, I or someone would have to send the Stockfish version to you. Because it was taken and re-compiled for whatever reason, so now if you download it- it won't even load with the same name. Again, I will do what I can. Just sort of keep an eye out over here in case I get a chance.
Komodo is down 6 games with 5 to play. Somehow I don't believe the math is on his side. However, depending on a person's POV, it might really be nice if Komodo could cut 1, 2 or even 3 games off Houdini's lead. He has shown he can win back to back games more than once.
Thing is, when you really get down and study the situation, who's going to come out on top in a qualifier that can have a legitimate shot at beating Houdini other than Komodo. The biggest "if" that I see is at what point will Marco and Tord make another "official release". Because their last one has no chance against Houdini, Critter has no chance at all, Rybka and 4.1 have been too long without some work on it. Those may very well be your top 5, except it is unlikely Stockfish, Critter or Rybka could at this point beat Strelka 5.6 or 5.7MP, and possibly not Ivanhoe 46h. Hell if I know.
Alienware AURORA_R4 Intel i7 w/6 True Cores Factory overclocked
Fritz 11 gui 6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40 Match=100 games
geots wrote:
I will do my best to try and replay that match for you if I can- between Houdini and the Stockfish version he lost to. I only have a few problems right now. I have to finish this Houdini-Komodo match first. Then the system I ran it on- Alienware AURORA R4 Intel i7 with 6 True Cores- I bought another one EXACTLY like it. It came in 5 days ago, and I have been waiting till this match was over to set it up. Because I am going to do the same with the 2 Intel i7s that I have done with my 2 Intel i5 4-Core systems. I am going to network the 2 i7s together so they will both use the same monitor, and I can switch back and forth between them, transfer files easily, etc. And I can't be running matches till I get this done. After setup, I gotta transfer files and all the normal stuff that goes with setting a new one up- like disabling hyperthreading, etc. And I am Giancarlo's beta tester for Equinox, and I just got a new beta there to try out yesterday. Then I am beta testing Strelka 5.6 and 5.7MP for Yuri. It's just a lot of stuff going on, but if there is any way I can- I will do it for you. Because even if you had the system to recreate it, I or someone would have to send the Stockfish version to you. Because it was taken and re-compiled for whatever reason, so now if you download it- it won't even load with the same name. Again, I will do what I can. Just sort of keep an eye out over here in case I get a chance.
Thanks for your interest,
george
Thank you George. I'd appreciate it. Take your time, there is no rush.
geots wrote:Komodo is down 6 games with 5 to play. Somehow I don't believe the math is on his side. However, depending on a person's POV, it might really be nice if Komodo could cut 1, 2 or even 3 games off Houdini's lead. He has shown he can win back to back games more than once.
Thing is, when you really get down and study the situation, who's going to come out on top in a qualifier that can have a legitimate shot at beating Houdini other than Komodo. The biggest "if" that I see is at what point will Marco and Tord make another "official release". Because their last one has no chance against Houdini, Critter has no chance at all, Rybka and 4.1 have been too long without some work on it. Those may very well be your top 5, except it is unlikely Stockfish, Critter or Rybka could at this point beat Strelka 5.6 or 5.7MP, and possibly not Ivanhoe 46h. Hell if I know.
Alienware AURORA_R4 Intel i7 w/6 True Cores Factory overclocked
Fritz 11 gui 6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40 Match=100 games
geots wrote:Komodo is down 6 games with 5 to play. Somehow I don't believe the math is on his side. However, depending on a person's POV, it might really be nice if Komodo could cut 1, 2 or even 3 games off Houdini's lead. He has shown he can win back to back games more than once.
Thing is, when you really get down and study the situation, who's going to come out on top in a qualifier that can have a legitimate shot at beating Houdini other than Komodo. The biggest "if" that I see is at what point will Marco and Tord make another "official release". Because their last one has no chance against Houdini, Critter has no chance at all, Rybka and 4.1 have been too long without some work on it. Those may very well be your top 5, except it is unlikely Stockfish, Critter or Rybka could at this point beat Strelka 5.6 or 5.7MP, and possibly not Ivanhoe 46h. Hell if I know.
Alienware AURORA_R4 Intel i7 w/6 True Cores Factory overclocked
Fritz 11 gui 6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40 Match=100 games
One possible problem would be if the Stockfish team release a new version while the development version is playing in your match.
Have there been any indication of when the Stockfish team plans to release the next official version?
You bring up an interesting situation. Most likely I would finish the develop. match and then run a match with the official release and compare. But if I had not gotten past the halfway mark- it might be determined by how well the develop. version was doing. I think it is fairly certain that at this point in time Marco could release a version 35 to 45 elo stronger than the last official release. There are just too many different testers showing that. But what intrigues me is that they are mostly all running blitz, and the slower they make the control, the better Stockfish seems to do. So it could really be more elo than that. I guess Marco has a problem all authors wish they had. Most all programmers would feel like 35 to 40 elo was a good release time. But if he keeps gaining a bit more strength every week or so, at what point do you release? Maybe as long as you are having success, go for the top. Myself- I have not asked him about it, mainly because I'm not sure the best way to word the question.
Leto wrote:Have there been any indication of when the Stockfish team plans to release the next official version?
Some time ago Marco stated that he plans the next official release after the next TCEC season so unless he changes his mind, we are at least couple of months from that.
The match is finished so we have the 100 games completed. After the memory problems and a few bugs here and there, I think this is an excellent start for Komodo in the "MP world". You know he is going to improve, but at the same time I doubt Robert is going to sit around twiddling his thumbs either. And then we have Stockfish, which seems at times to be getting stronger by the day. I would say these are in the top five, and the last 2 spots will be fought over by Critter, Equinox and Strelka. It will be interesting to say the least. One thing- I have never lied to anyone here, and I never will. I would pull hard for Equinox and Strelka to finish out the top 5. It would then be nice to know that of the top 5 engines in the world, I was beta testing for 3 of them.
Alienware AURORA_R4 Intel i7 w/6 True Cores Factory overclocked
Fritz 11 gui 6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40 Match=100 games
Congrats to both Robert and Don and Larry. Houdini is Houdini- no further explanation needed there. But he can turn around and see Komodo now.
On to installing and loading a new system......................
