Andscacs - New version 0.80

[cdani]

Hi.

I published the new version, 0.80:

http://andscacs.com/

Changes:

* Added option to use various simultaneous threads of analysis to profit from the power of the CPU (multi-thread, SMP).
* Added option to analyze various alternatives simultaneously (multi pv).
* Ponder mode now seems to work.
* Little LMR improvements.
* Reduced a little more the worst moves at root.
* Small improvements on move ordering.
* Changed developing environment to Visual Studio 2013 Community.
* Added hash of refutation of 3 and 5 moves.
* Slight improvements on time management.
* Slight improvement on rook, knight and bishop evaluation, and king safety.
* Improvements in search function.
* Solved various bugs.

All in all, not sure the strength gain, may be 30 elo.

[Graham Banks]

Should be interesting to see the changes you've made in action.

[Charly]

Many thanks for the release.

The last version was qualified for the ACCC League Karpov. There will be strong opponents, but Andscacs will have good chances to qualify for the League Kasparov. I can't wait to test it !

[cdani]

I think it will give you no problems

[Ozymandias]

I guess you mean that it'll be around 30 ELO points stronger than v0.72, because you released a test version last week. Even as it is, this is one of the fastest improving engines in the world, 30 ELO in a week would be OUT of this world.

[Henk]

I guess you mean that it'll be around 30 ELO points stronger than v0.72, because you released a test version last week. Even as it is, this is one of the fastest improving engines in the world, 30 ELO in a week would be OUT of this world.

Yes one ELO point per week is already something.

[cdani]

I guess you mean that it'll be around 30 ELO points stronger than v0.72, because you released a test version last week. Even as it is, this is one of the fastest improving engines in the world, 30 ELO in a week would be OUT of this world.

Yes, 30 from 0.72

I really had no special success this time... I must improve it a lot specially the evaluation. This time I succeeded less I think because the search is already very difficult to improve, and for the moment I don't have a serious system to improve the evaluation.

Anyway I don't know if even those 30 are real at long time control, may be are even a regression... I hope no

But of course the real motivation for this version is all the added features, mainly multi pv and multi thread. I will have time to improve it more.

[cdani]

I guess you mean that it'll be around 30 ELO points stronger than v0.72, because you released a test version last week. Even as it is, this is one of the fastest improving engines in the world, 30 ELO in a week would be OUT of this world.

Yes one ELO point per week is already something.

I'm sure you will be able to do it a lot better than this!

[Ozymandias]

I don't know if even those 30 are real at long time control, may be are even a regression... I hope no

But of course the real motivation for this version is all the added features, mainly multi pv and multi thread. I will have time to improve it more.

I'm sure some of those points will walk the distance, to LTC.

You were right to spend time with SMP, in this release. At the level Andscacs is competing, you need that to be on an equal footing.

[Henk]

I guess you mean that it'll be around 30 ELO points stronger than v0.72, because you released a test version last week. Even as it is, this is one of the fastest improving engines in the world, 30 ELO in a week would be OUT of this world.

Yes one ELO point per week is already something.

I'm sure you will be able to do it a lot better than this!

No Skippers ELO cycles between 1800 and 2050 all the time. Due to bad testing, impatience and continuously rewritings.