PVS extension up

[Michael Sherwin]

I have played with overlapping pairs counting (I am not finished) and sofar no benifit against other programs.

However, I have finished a full 100 game match vs Hamsters 0.2 at 1+1 time limit using a scheme similar to what I have already posted. The result was Romi, +60 -29 =11 for a new 1+1 record! The 1+1 record previous to the current NG series was 63.5/100 and the 4+4 record of the NG series is 64.5/100.

The key point to be made sofar is that the 1+1 performance of the NG series fell to about 50% (that is why I went to 4+4 testing), so 65.5/100 at 1+1 is a 115 ELO performance increase!

I am planning a few more test and if I can show an even better performance I will post the results.

This will be my last post on this subject for awhile. In about two weeks Swami will be conducting a quick tournament that RomiChess (with PVS extensions) will play in. If anyone is interested in my work on this subject then keep a watch out for the results of Swami's tournament.

I now have a very positive PVS-PAIRS extention. It extends a full ply, but only based on one pair of hash moves. It is a very rstricted , but powerful extension and does not explode the tree. Needless to say, it is much different than Ed's.

Sofar, this extension is also giving a performance boost of over 100 ELO and if it can be combined with the other one already developed then the ELO gain may be even more.

There is also an anti-PVS extension that I have discovered, that when a hash move that previously caused a beta-cut fails to cut, then extend the next n alternative moves. This also gives a good ELO boost.

THERE IS GOLD IN THEM THERE HILLS!

[Daniel Mehrmann]

Hi Michael,

hmmm, i think your long detail ELO reports about improving the play strenght of RomiChess is more interesting for a Blog instead of a public programmer forum.
Please image if anybody of us would do the same here if he finds a new idea or big improvment, or better he believes he has found something, which often happend i think, what happend here in this case.

It's great that you might found an such improvment for Romi, but instead of listing serval test results, poor Alessandro, it would be more interesting to read about your changes of Ed's idea.

So, where is the Romi Blog ?

Best,
Daniel

[Michael Sherwin]

Hi Michael,

hmmm, i think your long detail ELO reports about improving the play strenght of RomiChess is more interesting for a Blog instead of a public programmer forum.
Please image if anybody of us would do the same here if he finds a new idea or big improvment, or better he believes he has found something, which often happend i think, what happend here in this case.

It's great that you might found an such improvment for Romi, but instead of listing serval test results, poor Alessandro, it would be more interesting to read about your changes of Ed's idea.

So, where is the Romi Blog ?

Best,
Daniel

Hi Daniel,

Me thinks, that your criticism misses the mark somewhat.

First, I only changed Ed's ideas, because they were not working for me. I think that others are interested in the changes--maybe even Ed himself. If I were to move these descriptions off site, then many here would not see them at all. The reason that I am not giving more detailed descriptions is because I am changing my own changes from one minute to the next. Besides, people will have to find what works for them and all I need do is demonstrate that there is something to be gained. The ELO reports as you call them is just secondary information that may give an indication of how much there is to be gained.

It is yet to be seen whether or not it will hold up in practice. That is why I said in my last post that I will refrain from making any further reports until Swami's tournament results are in. So, I already implied criticism towords myself in that I would spare the readers an endless tyraid of reports on every little change that I am going to try. So Daniel, you are, 'standing in front of a parked train, telling it to stop'!

As far as interesting goes, this thread is very popular and I doubt that the readers whom are mostly programmers, do not find it interesting (from a programming point of view). And I do not believe that "poor Alessandro" is very upset or even much concerned about the fact that I am using Hamsters to test against. I have discovered that every time Romi achieves a new record against Hamsters it translates to a truely stronger version of RomiChess. That is not true for other oppoments. And besides, Alessandro has gotten some useful feedback from me about his engine, due to my persistent testing against it.

I can not afford a blog.

Best,
Mike

[Michael Sherwin]

It is not looking so good now. Against the first 3 test engines it was fantastic. Against the next 6 it was worse. So who knows! I have sent a best guess to Swami for his tournament that looks rather good. There is still a lot of work to be done here. It does look that the 100+ ELO has mostly evaporated.

[ed]

It is not looking so good now. Against the first 3 test engines it was fantastic. Against the next 6 it was worse. So who knows! I have sent a best guess to Swami for his tournament that looks rather good. There is still a lot of work to be done here. It does look that the 100+ ELO has mostly evaporated.

Chess programming AD 2007 is about the cruelest form of self torture, hence I quit, could not stand the pain any longer.

Ed

[Michael Sherwin]

It is not looking so good now. Against the first 3 test engines it was fantastic. Against the next 6 it was worse. So who knows! I have sent a best guess to Swami for his tournament that looks rather good. There is still a lot of work to be done here. It does look that the 100+ ELO has mostly evaporated.

Chess programming AD 2007 is about the cruelest form of self torture, hence I quit, could not stand the pain any longer.

Ed

I agree! But, only to a point. I think that having a new concept of how to write a chess program and then writing it, is fun and rewarding. The tourture is taking your old code and trying to squeeze more and more out of it. I wonder if Tord Romstad would have such a strong engine if he was still hammering away at Gothmog. Tord has taken what he has learned and created two new engines, each stronger than the last. He has improved his understanding of the problem, then concieved a better solution and finally created a better chess program. This seems like a more rewarding path to take than constantly tourturing oneself to make an old concept play better. $.02

[Michael Sherwin]

It is not looking so good now. Against the first 3 test engines it was fantastic. Against the next 6 it was worse. So who knows! I have sent a best guess to Swami for his tournament that looks rather good. There is still a lot of work to be done here. It does look that the 100+ ELO has mostly evaporated.

Wow, DanaSah made a huge second half comeback!

yes,Danasah is actually the first seeded engine in this division,the latest update added some elo

And Romi caught up with Bruja! Or should I say, that Bruja fell back to Romi.

I am very happy with fourth place.


Since, last time Romi placed one from the bottom, can I conclude that this was a better performance and that the PVS extension code was helpful?

Yes,Romichess did well and it has improved a lot and the previous division had engines like pepito,Green Light chess,Yace etc,All these engines are one division ahead of Romi and so it was okay for the previous beta,Don't know how this beta would perform.

I need to report any performance gain or loss, back to the programming forum to clarify that discussion.

This beta is probably the best Romi beta, no doubt about it and keep on improving.

Well there you have it. A good indication that the PVS extension is worth looking into. But, if Ed's code or my code does not work for you then maybe some variation of your own will work!

The code that made it into this beta did the following:

if(Ply > 3 && depth > 2 && !inNull ... etc
Loop through the last four moves and if they were all hash moves that are also all flagged as EXACT then award a one half ply extension.

...

There are plenty of other promising variations--I just ran out of time for testing before I had to send the beta in for the tournament.

[F. Bluemers]

Hi Ed
email didn't work once again.
2 messages for you.

Best
Fonzy

[ed]

Hi Ed
email didn't work once again.
2 messages for you.

Best
Fonzy

I got one, will answer you soon.

Ed

[F. Bluemers]

And another pm!
after this i'll ask andres to relay the emails again

Best
Fonzy