OMG!!!! HOUDINI 2.0Z +59 elo points(DISCUSS)

[Strelkaman]

By Barnard Date 2012-05-27 22:16 Edited 2012-05-28 02:15

http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=24987

[gleperlier]

Hi,

In fact only +13 after 1000 games if you read the post until the end

Gab

[mar]

OMG the new version looks promising! +100 after a few games ... uh at least +50 after some more ... no wait ... no improvement at all.
But don't worry I also can't test

[Strelkaman]

Hi,

In fact only +13 after 1000 games if you read the post until the end

Gab

+12,13,14 Elo for a tuning adjustment is an excellent result

[Graham Banks]

Hi,

In fact only +13 after 1000 games if you read the post until the end

Gab

+12,13,14 Elo for a tuning adjustment is an excellent result

Hardly likely to set the world on fire though.

[Ajedrecista]

Hello:

Hi,

In fact only +13 after 1000 games if you read the post until the end

Gab

+12,13,14 Elo for a tuning adjustment is an excellent result

The 95% error margin after 1000 games is still around 20 Elo.

Well said Gabor! These are the results I get for that match of 999 games:

Code: Select all

Elo_uncertainties_calculator, ® 2012.

Calculation of Elo uncertainties in a match between two engines:
----------------------------------------------------------------

(The input and output data is referred to the first engine).

Please write down non-negative integers.

Write down the number of wins:

320

Write down the number of loses:

282

Write down the number of draws:

397

***************************************
1-sigma confidence ~ 68.27% confidence.
2-sigma confidence ~ 95.45% confidence.
3-sigma confidence ~ 99.73% confidence.
***************************************

---------------------------------------

Confidence interval for 1-sigma:

Elo rating difference:     13.22 Elo

Lower rating difference:    4.69 Elo
Upper rating difference:   21.77 Elo

Lower bound uncertainty:   -8.53 Elo
Upper bound uncertainty:    8.54 Elo
Average error:        +/-   8.54 Elo

K = (average error)*[sqrt(n)] =  269.83

Elo interval: ]   4.69,   21.77[
---------------------------------------

Confidence interval for 2-sigma:

Elo rating difference:     13.22 Elo

Lower rating difference:   -3.83 Elo
Upper rating difference:   30.34 Elo

Lower bound uncertainty:  -17.05 Elo
Upper bound uncertainty:   17.12 Elo
Average error:        +/-  17.08 Elo

K = (average error)*[sqrt(n)] =  539.98

Elo interval: ]  -3.83,   30.34[
---------------------------------------

Confidence interval for 3-sigma:

Elo rating difference:     13.22 Elo

Lower rating difference:  -12.36 Elo
Upper rating difference:   38.95 Elo

Lower bound uncertainty:  -25.58 Elo
Upper bound uncertainty:   25.72 Elo
Average error:        +/-  25.65 Elo

K = (average error)*[sqrt(n)] =  810.79

Elo interval: ] -12.36,   38.95[
---------------------------------------

Number of games of the match:                999
Score: 51.90 %
Elo rating difference:   13.22 Elo
Draw ratio: 39.74 %

**********************************************
1-sigma:  1.2265 % of the points of the match.
2-sigma:  2.4530 % of the points of the match.
3-sigma:  3.6795 % of the points of the match.
**********************************************

End of the calculations.

Thanks for using Elo_uncertainties_calculator. Press Enter to exit.

For avoiding regression (I mean, a negative gain of Elo), this is what I get for 1.96-sigma confidence (~ 95% confidence):

Code: Select all

Minimum_score_for_no_regression, ® 2012.

Calculation of the minimum score for no regression in a match between two engine
s:

 Write down the number of games of the match (it must be a positive integer, up
to 1073741823):

999

Write down the draw ratio (in percentage):

39.739739739

Write down k (for making confidence intervals of (mu) +/- (k*sigma) in a normal
distribution); k must be positive:

1.96

Theoretical minimum score for no regression: 52.4023 %

Minimum number of won points for the engine in this match:       523.5 points.

Minimum Elo advantage, which is also the negative part of the error bar:
 16.7065 Elo

End of the calculations.

Thanks for using Minimum_score_for_no_regression. Press Enter to exit.

Anyway, those settings do not seem bad.

Regards from Spain.

Ajedrecista.

[Sedat Canbaz]

By Barnard Date 2012-05-27 22:16 Edited 2012-05-28 02:15

http://rybkaforum.net/cgi-bin/rybkaforu ... ?tid=24987

Dear Dimitri,

Very soon i plan to start testing Houdini 2.0z x64 6c (a modified version by Salvador) under SCCT conditions

And i wonder a lot about what will be the Elo difference between the default

Btw,still i am waiting Strelka to be released as MP engine
Due to i have no much interest testing the engines with single core

Best,
Sedat

[Strelkaman]

LOL thinking that maby someone in the future build HoudiniX that it would be better than Houdini 3 using the config.I mean after all this post shows that Houdini Pro is a tested version of Houdini 1.5 changing a few paramaters.What a pity indeed.A clone inside the clone

[Hood]

Cloning that is Java

[Sedat Canbaz]

Dear Friends,

Clone,clone clone...

Nowadays i see mainly postings regarding clone engines issue
Really it's started to be boring, same comments - no more no less

Some people claim that Houdini is a clone
Some people claim that Rybka is a clone
Some people claim that Strelka is a clone
Some people claim that Ivanhoe is a clone
Some people claim that Critter is a clone
...

In other words,many people claim that the current Top strongest engines are clones
Another interesting note is that, the debated engines (the strongest ones) are not allowed in World Championships
And sometimes i am wondering:
-Is that can be the reason that 'not allowed engines' are approx.100-200 Elo stronger than the original ones ?!

Come on friends,lets concentrate more how to improve the strength of chess engines

Btw,as i mentioned before, nowadays its very hard to find 100% original work
In my opinion, almost all chess engines are each other relative
Actually to check and to find the true answer is very simple:just is needed to be tested the engines without books
And you will notice how many chess engines will play almost same/similar openings

To be honest, I am looking for a serious opponent against Houdini,and no patience to see a such new superior engine

But however with respect to all:
-There is one true,the Authors who create engines as 'Leaders' are real masters in chess programing (from the past to the present)


Best Wishes,
Sedat