David Dahlem wrote:I have to admit, i haven't followed all this new Stelka5 stuff, therefore my question - Is there a Strelka 5 MP version?
Regards
Dave
Silvian R
Moderators: hgm, Rebel, chrisw
David Dahlem wrote:I have to admit, i haven't followed all this new Stelka5 stuff, therefore my question - Is there a Strelka 5 MP version?
Regards
Dave
Thanks Silvian. Your reply certainly seems definite!Sylwy wrote:David Dahlem wrote:I have to admit, i haven't followed all this new Stelka5 stuff, therefore my question - Is there a Strelka 5 MP version?
Regards
Dave
Silvian R
Why do I always have to answer questions about Robbolito? Ask Roberto Pescatore!geots wrote: Jury, have you heard anything about this new [in the last week] private version of Robbolito that will be released in a couple weeks. Supposedly it destroys Houdini 2. Couldn't get much info on it.
Osipov Jury wrote:Why do I always have to answer questions about Robbolito? Ask Roberto Pescatore!geots wrote: Jury, have you heard anything about this new [in the last week] private version of Robbolito that will be released in a couple weeks. Supposedly it destroys Houdini 2. Couldn't get much info on it.
http://translate.google.com/#it|en|Roberto PescatoreOsipov Jury wrote:Why do I always have to answer questions about Robbolito? Ask Roberto Pescatore!geots wrote: Jury, have you heard anything about this new [in the last week] private version of Robbolito that will be released in a couple weeks. Supposedly it destroys Houdini 2. Couldn't get much info on it.
I don't see how it can apply to a "difference". What if you just publish the one rating? Then the "difference" is not shown. Isn't this based simply on standard deviation? Which would mean the +/-8 applies to the actual value in question, not how it relates to other values???Laskos wrote:It's ELOstat, and +/- 8 does really refer to difference in an individual match.Michel wrote:I don't think so. The scores are not independent (if one of them goes up, the other goes down). So +/- 8 looks correct.I am not sure I quite buy that 10 +/- 8. It seems closer to 10 +/- 16, because each individual rating has a +/-8 error. When you subtract, both error bars should be included.
Even if the score were independent the error bar should only be multiplied by 1.4.
EDIT: Actually I am not sure. I thought that the way Bayes Elo is set up
is such that the +/- 8 really refers to the difference. But I am not sure.
Kai
Elo always relate to other values. A +/- in the elo case will always refer to the pool in which it was examined.bob wrote:I don't see how it can apply to a "difference". What if you just publish the one rating? Then the "difference" is not shown. Isn't this based simply on standard deviation? Which would mean the +/-8 applies to the actual value in question, not how it relates to other values???Laskos wrote:It's ELOstat, and +/- 8 does really refer to difference in an individual match.Michel wrote:I don't think so. The scores are not independent (if one of them goes up, the other goes down). So +/- 8 looks correct.I am not sure I quite buy that 10 +/- 8. It seems closer to 10 +/- 16, because each individual rating has a +/-8 error. When you subtract, both error bars should be included.
Even if the score were independent the error bar should only be multiplied by 1.4.
EDIT: Actually I am not sure. I thought that the way Bayes Elo is set up
is such that the +/- 8 really refers to the difference. But I am not sure.
Kai
LOL, sorry Bob, here I have erred, I meant 10 is the difference and 8 is indeed the error in that particular match (error being 2 standard deviations, 95% confidence, as it shown in both BayesElo and ELOstat by default). Think about an individual match as that: 10 +/- 8 without a base means that you take one engine at _fixed_ Elo say 3195 (has no errors, set by the average and the difference/2 alone), then, the second engine, has 3205 +/- 8 rating. Or the other way around. When you have a rating list of many engines, you either set a base, or their average itself sets a pretty stable base, therefore this sqrt(X^2 + Y^2) for the total error.bob wrote:I don't see how it can apply to a "difference". What if you just publish the one rating? Then the "difference" is not shown. Isn't this based simply on standard deviation? Which would mean the +/-8 applies to the actual value in question, not how it relates to other values???Laskos wrote:It's ELOstat, and +/- 8 does really refer to difference in an individual match.Michel wrote:I don't think so. The scores are not independent (if one of them goes up, the other goes down). So +/- 8 looks correct.I am not sure I quite buy that 10 +/- 8. It seems closer to 10 +/- 16, because each individual rating has a +/-8 error. When you subtract, both error bars should be included.
Even if the score were independent the error bar should only be multiplied by 1.4.
EDIT: Actually I am not sure. I thought that the way Bayes Elo is set up
is such that the +/- 8 really refers to the difference. But I am not sure.
Kai