Tinapa 1.01 based on Stockfish

[Ralph Stoesser]

...
I expect very small ELO difference (in either direction) if any.

A pity people finds easier to get a nice name then testing his pacthes

What are your test results? Or is it enough for us to know that you don't expect something?

[mcostalba]

...
I expect very small ELO difference (in either direction) if any.

A pity people finds easier to get a nice name then testing his pacthes

What are your test results? Or is it enough for us to know that you don't expect something?

Ok, I am not native speaker so perhaps I got it wrong, but for me "to expect" it doesn't mean I know for sure or I have tested or I strongly believe it is like this.

For me it is just a guess, nothing more, so could very well be wrong. The only practical consequence of my "guessing" is that _I_ will not test that patch because it doesn't look interesting to me (and I maybe very wrong, but that's my opinion), so I will dedicate my testing time to something else.

I don't know what is enough for you (nor that I particulary care) and what is enough for you, whatever it is, will not change the fact that I "guess" patch will have little impact on ELO (again, I may be very wrong) and so I am not very motivated in testing it. No more no less then just this.

[alpha123]

...
I expect very small ELO difference (in either direction) if any.

A pity people finds easier to get a nice name then testing his pacthes

What are your test results? Or is it enough for us to know that you don't expect something?

Ok, I am not native speaker so perhaps I got it wrong, but for me "to expect" it doesn't mean I know for sure or I have tested or I strongly believe it is like this.

For me it is just a guess, nothing more, so could very well be wrong. The only practical consequence of my "guessing" is that _I_ will not test that patch because it doesn't look interesting to me (and I maybe very wrong, but that's my opinion), so I will dedicate my testing time to something else.

I don't know what is enough for you (nor that I particulary care) and what is enough for you, whatever it is, will not change the fact that I "guess" patch will have little impact on ELO (again, I may be very wrong) and so I am not very motivated in testing it. No more no less then just this.

A non-native speaker's English tends to be better than a native speaker's English.... (we warp it a lot)

Peter

[gerold]

......people are being driven away from talkchess in droves......

Really? Where have they all gone? Most of those I can see posting in other forums also post here.
By the way, I get on well with Chris Formula, so this comment is not supposed to be a further dig at him.

Chris Formula is posting in the clonedyke forum.

[Kurt Utzinger]

Hi Swaminathan
Are you convinced that the STS is a relieble test? I have my doubts. Is it possible to say much about playing strength if 869 positions out of 1000 are solved using a 10s/move level? Maybe 120 positions of the remaining 131 positions are solved on a 3m/move level. If so, we have 11 (or 1,1 %) unsolved positions out of 1000 ? Or in other words: do you believe that the results of the STS-10s/m would give the same ranking list for the engines if using a thinking time of 3-5 min/move?
Kind regards
Kurt

[swami]

Hi Swaminathan
Are you convinced that the STS is a relieble test? I have my doubts. Is it possible to say much about playing strength if 869 positions out of 1000 are solved using a 10s/move level? Maybe 120 positions of the remaining 131 positions are solved on a 3m/move level. If so, we have 11 (or 1,1 %) unsolved positions out of 1000 ? Or in other words: do you believe that the results of the STS-10s/m would give the same ranking list for the engines if using a thinking time of 3-5 min/move?
Kind regards
Kurt

It's not intended for making the ranking list or ratings of the engines. It is intended to find information on areas that the engine is weaker in.

Many authors of <3000 rated engines found it useful to tune the evaluation function. It has been helpful to them which is what I'm happy about.

Ofcourse, you can say some engines solve some more in 3 minute per position just as they play better when given more time. Improving the performance of the engine so that it could find more and more in less and less time generally improves ELO. Ability to see it faster counts.

You can see the results page here:

http://sites.google.com/site/strategict ... st-results

*most* engine's results closely correlate with their actual ratings.

Best regards,
Swami

[Look]

[...]

Many authors of <3000 rated engines found it useful to tune the evaluation function.

[...]
I suppose Richard Vida (author of Critter) uses this suite plus a tactical one for sanity check for changes in his code, which I suspect is one reason of his fine improvements.

[swami]

[...]

Many authors of <3000 rated engines found it useful to tune the evaluation function.

[...]
I suppose Richard Vida (author of Critter) uses this suite plus a tactical one for sanity check for changes in his code, which I suspect is one reason of his fine improvements.

Yes, Critter is an awesome engine.

I had tested all the versions in both the games and in STS. The engine made a steady improvement, and consistent progress in each one of the EPD's.

[Kurt Utzinger]

It's not intended for making the ranking list or ratings of the engines. It is intended to find information on areas that the engine is weaker in.

Many authors of <3000 rated engines found it useful to tune the evaluation function. It has been helpful to them which is what I'm happy about.

[...]

You can see the results page here:

http://sites.google.com/site/strategict ... st-results

*most* engine's results closely correlate with their actual ratings.

Best regards,
Swami

Hi Swami
Good answer and I understand nowthe intention of the STS much better.
Regards
Kurt

[swami]

According to Luis Barutti (Who has actively tested Chessmaster personalities, and other engines)

Code: Select all

You can post the results, certainly.
 
2 Rybka 4 w32 3282 
747 games

3 Tinapa 1.01 1T 3257
223 games

4 Stockfish 1.7.1 JA 1T 3242
437 games

5 Rybka 3 1-cpu 32-bit 3238
298 games

6 Stockfish 1.8 JA 1T 3235
535 games

7 Naum 4.2 1T 3213
410 games

9 Critter 0.70 32-bit 1T 3189
341 games

10 HIARCS 13.1 SP 3141
225 games

11 Shredder 12 UCI 3139
303 games
 
All the best,
Luis Barutti