What is New in Toga II 1.3x4?

[Terry McCracken]

I gather it has tablebases added...any improvement over Toga II 1.2a in middlegame playing strength?

I noticed there has been many changes recently in finding these engines and I had to use a rar utility to extact the new program

Terry

[Mike S.]

The readme says, changes in eval and changes in search (and +egbb).

I am currently waiting for the final 1.3 version to run it's Blivorix test. I have decided not to test versions which are declared to be "beta." Nevertheless, I had hoped that the main (or only) Toga 1.3x problem, the egbb usage if not all 3-4-5 tables are available, is fixed more quickly... (Of course I'd like to use that interesting egbb feature with Toga, but with 3- and 4-piece files only.)

Thanks anyway! I will be patient.

[Terry McCracken]

The readme says, changes in eval and changes in search (and +egbb).

I am currently waiting for the final 1.3 version to run it's Blivorix test. I have decided not to test versions which are declared to be "beta." Nevertheless, I had hoped that the main (or only) Toga 1.3x problem, the egbb usage if not all 3-4-5 tables are available, is fixed more quickly... (Of course I'd like to use that interesting egbb feature with Toga, but with 3- and 4-piece files only.)

Thanks anyway! I will be patient.

Thanks for your informed reply.

Do you know when or roughly when the final 1.3 ver. will be released?

Terry

[Mike S.]

Do you know when or roughly when the final 1.3 ver. will be released?

Unfortunately not... if I didn't miss it, there were no messages announcing this after the infos about the egbb problem, yet. Anyway, I think it should be soon. I cannot image that it is such a big problem if the rest of the engine works ok (as it seems to, accoring to some rating list results I saw).

[Terry McCracken]

Do you know when or roughly when the final 1.3 ver. will be released?

Unfortunately not... if I didn't miss it, there were no messages announcing this after the infos about the egbb problem, yet. Anyway, I think it should be soon. I cannot image that it is such a big problem if the rest of the engine works ok (as it seems to, accoring to some rating list results I saw).

Has it exceeded Toga II 1.2a by any significant margin?

Terry

[Eelco de Groot]

Do you know when or roughly when the final 1.3 ver. will be released?

Unfortunately not... if I didn't miss it, there were no messages announcing this after the infos about the egbb problem, yet. Anyway, I think it should be soon. I cannot image that it is such a big problem if the rest of the engine works ok (as it seems to, accoring to some rating list results I saw).

Has it exceeded Toga II 1.2.1a by any significant margin?

Terry

Hello Terry,

It's not very clear yet if there is any improvement. The results of the testing teams are contradictory. I think some improvement is clear in Blitz and it has been shown at Blitz timecontrols to be stronger by CCRL, but the CCRL is waiting for a fix of the endgame bitbases probing code in Toga 1.3x4 before testing Toga 1.3x4 at longer timecontrols.
The CEGT has tested 1.3x4 to be sightly weaker than 1.2.1a but private results posted in the forums from Pavel and from Thorsten in his forum have been good at longer timecontrols as well. I really could not say which version of Toga is stronger.

At 5 minutes + 5 seconds Fischer timecontrols or anything faster, I think 1.3x4 is maybe marginally stronger with -even with - various Blueberry settings I have tried, and almost as strong as Rybka 1.0 Beta or Hiarcs 11.1. Depending on the settings maybe against which program the results are best At least Toga 1.3x4 feels almost as strong as these two with some of the Toga Blueberry settings I have, but I don't have any match yet that Toga 1.3x4 Blueberry clearly won against either of them, let alone against both.

Eelco

[Uri Blass]

Do you know when or roughly when the final 1.3 ver. will be released?

Unfortunately not... if I didn't miss it, there were no messages announcing this after the infos about the egbb problem, yet. Anyway, I think it should be soon. I cannot image that it is such a big problem if the rest of the engine works ok (as it seems to, accoring to some rating list results I saw).

Has it exceeded Toga II 1.2a by any significant margin?

Terry

In one word the answer is no

based on the CCRL list it is only the best blitzer.

In CCRL 40/12 there is no significant difference between Toga1.3 and Toga1.2.1a(only 6 elo when the error is clearly bigger than 6 elo)

Toga II 1.3x4 2839 +23 −23 61.1% −81.4 30.2% 635
60.2%
Toga II 1.2.1a 32-bit 2833 +19 −19 57.2% −47.6 34.1% 905
60.4%

In CCRL 40/4 the difference is also not very high and it is only 18 elo so even after thousands of games we see no significant difference considering the error margin.

Toga II 1.3x4 2878 +16 −16 53.8% −26.9 33.1% 1368
55.5%
Toga II 1.2.1a 32-bit 2860 +9 −9 48.7% +5.4 32.7% 4965
50.8%


Uri

[Leto]

Do you know when or roughly when the final 1.3 ver. will be released?

Unfortunately not... if I didn't miss it, there were no messages announcing this after the infos about the egbb problem, yet. Anyway, I think it should be soon. I cannot image that it is such a big problem if the rest of the engine works ok (as it seems to, accoring to some rating list results I saw).

Has it exceeded Toga II 1.2.1a by any significant margin?

Terry

Hello Terry,

It's not very clear yet if there is any improvement. The results of the testing teams are contradictory. I think some improvement is clear in Blitz and it has been shown at Blitz timecontrols to be stronger by CCRL, but the CCRL is waiting for a fix of the endgame bitbases probing code in Toga 1.3x4 before testing Toga 1.3x4 at longer timecontrols.
The CEGT has tested 1.3x4 to be sightly weaker than 1.2.1a but private results posted in the forums from Pavel and from Thorsten in his forum have been good at longer timecontrols as well. I really could not say which version of Toga is stronger.

At 5 minutes + 5 seconds Fischer timecontrols or anything faster, I think 1.3x4 is maybe marginally stronger with -even with - various Blueberry settings I have tried, and almost as strong as Rybka 1.0 Beta or Hiarcs 11.1. Depending on the settings maybe against which program the results are best At least Toga 1.3x4 feels almost as strong as these two with some of the Toga Blueberry settings I have, but I don't have any match yet that Toga 1.3x4 Blueberry clearly won against either of them, let alone against both.

Eelco

CEGT Blitz has also tested Toga 1.3x4, after 450 games it is only 10 elo higher than 1.2.1a. With more games it can potentially end up slightly higher or even below 1.2.1.a.
http://www.husvankempen.de/nunn/40_4_Ra ... liste.html

[Eelco de Groot]

Do you know when or roughly when the final 1.3 ver. will be released?

Unfortunately not... if I didn't miss it, there were no messages announcing this after the infos about the egbb problem, yet. Anyway, I think it should be soon. I cannot image that it is such a big problem if the rest of the engine works ok (as it seems to, accoring to some rating list results I saw).

Has it exceeded Toga II 1.2.1a by any significant margin?

Terry

Hello Terry,

It's not very clear yet if there is any improvement. The results of the testing teams are contradictory. I think some improvement is clear in Blitz and it has been shown at Blitz timecontrols to be stronger by CCRL, but the CCRL is waiting for a fix of the endgame bitbases probing code in Toga 1.3x4 before testing Toga 1.3x4 at longer timecontrols.
The CEGT has tested 1.3x4 to be sightly weaker than 1.2.1a but private results posted in the forums from Pavel and from Thorsten in his forum have been good at longer timecontrols as well. I really could not say which version of Toga is stronger.

At 5 minutes + 5 seconds Fischer timecontrols or anything faster, I think 1.3x4 is maybe marginally stronger with -even with - various Blueberry settings I have tried, and almost as strong as Rybka 1.0 Beta or Hiarcs 11.1. Depending on the settings maybe against which program the results are best At least Toga 1.3x4 feels almost as strong as these two with some of the Toga Blueberry settings I have, but I don't have any match yet that Toga 1.3x4 Blueberry clearly won against either of them, let alone against both.

Eelco

CEGT Blitz has also tested Toga 1.3x4, after 450 games it is only 10 elo higher than 1.2.1a. With more games it can potentially end up slightly higher or even below 1.2.1.a.
http://www.husvankempen.de/nunn/40_4_Ra ... liste.html

Hallo Leto, yes I think that is possible. I could not find in the CEGT list if the Blitz testing is done with Toga playing without bitbases? Toga 1.2.1a I see was tested by CEGT against a very wide range of programs so that rating should be very reliable and steady now. In the 40/40 list both versions of Toga 1.3x4 with and without tablebases have an exactly equal rating of 2790 but have had different opponents. That would suggest that rating will also not change very much, altough there have not yet been played too many games there either.

Zap!Chess was one of the worst opponents there in 40/40 for Toga 1.3x4 but so far in Blitz Toga 1.3x4 had the best TPR exactly against Zap!Chess 2 CPU with 2905!

Eelco

[Terry McCracken]

Thanks for all the responses, had I known yesterday what I know now, I wouldn't have gone through the trouble of downloading it

I ran a very small engine match with Toga II 1.3x4 and it was crushed in a six game match by Rybka 1.0 32bit by 5.5-0.5, and the only reason I see it wasn't a sweep of 6-0 was Rybka offered a draw on move 15 in it's second game

The time controls were G/60.

IMO looking at the games, Toga II 1.3x4 is significantly weaker than Toga II 1.2.1a.

Terry