Having more than one machine to test

[Kempelen]

Hi,

I own now three different computers. Main one is a Core2 Vista, other two are 1 cpu-xp. All have different memory sizes and CPU GHz. I am happy because now I can make more test in the same time.

To start, I have launched the same tournament in all computers, and three tournaments show the same result, but I am not completely sure if this means that is a correlation between tournaments in different hardware.

Have you find that repeteating a tournament in different computers show the same result? is it safe to assume there is a correlation?

regards

[hgm]

What do you mean by 'same result'? Are the games the same? The final standings? Or the result of very game.

It depends a lot on what exactly you play. If I play Joker against Fruit, the result is always the same...

[Kempelen]

What do you mean by 'same result'? Are the games the same? The final standings? Or the result of very game.

It depends a lot on what exactly you play. If I play Joker against Fruit, the result is always the same...

I mean I obtein the same score. I launch a gountlet tournament with 12 opponent. I score 49%/50% in all machines. Independent game results could vary, but not the final score.
What I try to understand is if these results let me conclude that I can compare tournament between machines.

[hgm]

Most programs are not very senitive to what machine they run on. I don't know if you make attempts to correct for the hardware speed (if any), but the Elo difference of most engines is not very sensitive to the TC either, if the difference is not too extreme.

[Laskos]

What do you mean by 'same result'? Are the games the same? The final standings? Or the result of very game.

It depends a lot on what exactly you play. If I play Joker against Fruit, the result is always the same...

I mean I obtein the same score. I launch a gountlet tournament with 12 opponent. I score 49%/50% in all machines. Independent game results could vary, but not the final score.
What I try to understand is if these results let me conclude that I can compare tournament between machines.

Do you use a fixed set of opening positions? In this case the results are ought to be similar. If you are using a large generic book the results should show some variability. Even if you are using a book, did you check that they actually do not play the same games out of the book?

Kai

[Kempelen]

Do you use a fixed set of opening positions? In this case the results are ought to be similar. If you are using a large generic book the results should show some variability. Even if you are using a book, did you check that they actually do not play the same games out of the book?

Kai

I use the same EPD file for starting positions and the same time settings.

[Laskos]

I use the same EPD file for starting positions and the same time settings.

Ahhh, ok, so expect very similar games and scores. The sensitivity to CPU speed is very low in a fixed position. Better use three different EPD files, one for each comp, and adjust the time so that CPU speed (better to say Kn/s speed) multiplied by time is the same on all three. That would be a fair testing, you could take the average to get a more precise score.

Kai