report about my plans of releasing movei

[swami]

So perhaps you mean that version 423 is 50 points stronger than version 403?

Could not understand how version 423 can be stronger than version 432.

Bernhard

I'm fairly sure he means he'll release 423 when he has a version he thinks is at least 50 points stronger than 423.

Correct.

423 also seems to be 50 elo better than 403 based on CCRL results.

Uri

Hi Uri,

How much elo is it better than progress 10 10 10?

[Uri Blass]

Here is the data from the ccrl list

1 Movei 0.08.423 2691 +24 −24 52.3% −18.9 32.0% 575
90.6%
Movei 0.08.403(10 10 10) 2651 +25 −25 46.6% +21.2 26.0% 572
Movei 0.08.403 2636 +17 −17 52.2% −17.5 30.2% 1241
62.7%

[Uri Blass]

That's a wonderful result Uri, 50 elo don't come easy!

I think that half of the improvement in 423 relative to 403 is by deleting some counter productive part of the evaluation.

I think that getting another 50 elo should be easy
considering the fact that 423 does not use hash efficiently and considering the fact that 423 does not use principle variation search or internal iterative deepening.

Note that the name principle variation search is misleading and it is the main reason that I did not investigate what it is.

I thought that if I have pv array and start from the pv then I have a principle variation search when programs that do not use principle variation search are only programs that use methods like mtd and need to collect the pv from the hash.

Uri