Fruit-061115a vs. Rybka Beta, Blitz Match

[Tulevaisuuden]

If I remember correctly, a long long time ago Ryan Benitez told in this forum, that his Fruit beta-versions had been beating the free rybka beta for a long time already.

To put this into test, I ran a fast blitz match with the new Fruit against the free Rybka, Timing 1min + 1s/move, Nunn-openings, intel E6300-processor. Here are the results:

Fruit-061115a - Rybka v1.0 Beta.w32 : 900/1655 654-509-492, 54%
Rybka v1.0 Beta.w32 - Fruit-061115a : 755/1655 509-654-492, 46%

This does show a statistically significant difference between the two in a head-to-head match under these conditions, Fruit being approximately 20-40 elo stronger.

[Marc MP]

If I remember correctly, a long long time ago Ryan Benitez told in this forum, that his Fruit beta-versions had been beating the free rybka beta for a long time already.

To put this into test, I ran a fast blitz match with the new Fruit against the free Rybka, Timing 1min + 1s/move, Nunn-openings, intel E6300-processor. Here are the results:

Fruit-061115a - Rybka v1.0 Beta.w32 : 900/1655 654-509-492, 54%
Rybka v1.0 Beta.w32 - Fruit-061115a : 755/1655 509-654-492, 46%

This does show a statistically significant difference between the two in a head-to-head match under these conditions, Fruit being approximately 20-40 elo stronger.

By curiosity, which nunn openings did you use? Easyway.epd? Looks like I'll try the 061115a version quite soon!

[Dariusz Orzechowski]

No loses on time? In my tests Rybka 1.0 beta does not like 1/1 timecontrol and loses occasionally on time.

[Tulevaisuuden]

By curiosity, which nunn openings did you use? Easyway.epd? Looks like I'll try the 061115a version quite soon!

I used Nunn.pgn, which seems to contain all the games by J. Nunn, and I limited it to 9-10 first moves. The openings were randomly selected from the file, same position repeated twice with different colors for different players.

To be honest, I'm not sure if that even counts as a real Nunn-opening repertoire, but it should be better than nothing .

[Tulevaisuuden]

No loses on time? In my tests Rybka 1.0 beta does not like 1/1 timecontrol and loses occasionally on time.

A very good point, there seem to be many losses on time for rybka. This pretty much invalidates my result, time to put up a new tournament then.

[Ryan Benitez]

Look at the score of the games that where lost on time. Rybka loses more games on time when down than not because of how its time control works. Also other engines including Fruit will also lose on time at fast time control from time to time.

[Dariusz Orzechowski]

I've checked it. In my 1/1 tournament, Rybka 1.0 beta lost on time as much as 10 out of 38 games! In only one of these 10 games it showed a negative score, in 2 games about equal and in the remaining 7 games Rybka according to its score was totally winning (+1.61 - +8.97). I haven't noticed so far any loses on time by Fruit 2.3.

[Marc MP]

By curiosity, which nunn openings did you use? Easyway.epd? Looks like I'll try the 061115a version quite soon!

I used Nunn.pgn, which seems to contain all the games by J. Nunn, and I limited it to 9-10 first moves. The openings were randomly selected from the file, same position repeated twice with different colors for different players.

To be honest, I'm not sure if that even counts as a real Nunn-opening repertoire, but it should be better than nothing .

Is the sampling done with or without replacement? I'm asking because I'm not sure how Arena (is it what you used?) does it. It might be better to have a large opening book (say 14-moves max) instead of sampling with replacement early opening positions (9-10 moves) to avoid duplicates. Good luck for your testing, and one last point: maybe Strelka 1.8 would be a better choice as it doesn't seems to loose on time like Rybka beta (at least out of 700 bullets games I ran) and support underpromotion, so that there are less chances for an annoying engine hang.

[Michael Sherwin]

If the new Fruit betas are stronger than the free Rybka beta it is not at blitz. I have checked that already. And Fruit plays ugly in blitz!

[Michael Sherwin]

By curiosity, which nunn openings did you use? Easyway.epd? Looks like I'll try the 061115a version quite soon!

EasyWay, did not start out as an epd file. That was done by someone else. EasyWay.pgn was never intended to be used directly by any engine. It was meant as a resource for programmers and testers to cull their own positions from it for testing. It merely consist of all the colums from the book, "Chess Openings the Easy Way", by Nick de Firmian. If anyone wanted to use EasyWay.pgn directly, then the file would need to be edited first to get rid of all the 'bad' end points (lines). Just did not want anybody to have a wrong idea about what the EasyWay.pgn file was all about!