Mark, I wasn't stating a precise value for a doubling when I said 100 ELO. But it appears that it is actually worth a bit more than 100 ELO for a doubling. I did a massive study of this at one point a year or more ago with thousands of games with UCT based Lazarus program and the strength improvement per doubling was very clear and impressive.
You can usually see that the 2 or 4 processor versions of Mogo or CrazyStone jumps way up in strength - so it's very clear that speed is really quite important for any program that is search based as UCT is and is properly scalable. You will also notice that Mogo and Crazy Stone rarely test their strong versions on CGOS, they often have versions that play super fast or do tiny numbers of play-outs such as "3k" etc presumably because there is no competition. You can see that these versions are generally several hundred ELO weaker - so in every case it is pretty clear that doubling the number of simulations (or speed) is very important. I have often wondered if UCT and Monte Carlo play-outs would have even been discovered a few years ago. It could very well be that this technology HAD to wait for today. Mogo and CrazyStone would not be impressive on a 386. - Don Mark Boon wrote: > > On 16-jan-08, at 17:22, <[EMAIL PROTECTED]> wrote: > >> We can use math to shed some light on the topic: >> >> * Assume that doubling the speed of a machine >> increases the rank of a program by 100 ELO, >> as Don has previously concluded. >> >> * Then we have the following table of approximate >> costs, which comes from the equation y = 100 * 2^x >> cost -> lost ELO >> ---------------- >> 1% -> 1.5 ELO >> 2% -> 3.0 ELO >> 3% -> 4.5 ELO >> 4% -> 6.0 ELO >> 5% -> 7.5 ELO >> 6% -> 9.0 ELO >> 10% -> 15.0 ELO >> >> * In my program (which implements undo), the cost of >> for suicide detection is around 1%, which means it >> would lose 1.5 ELO points. >> >> * If I wanted to know whether it was worth it, I would >> want to measure the ELO benefit by making better >> decisions concerning suicide. It is a small but >> real amount, probably at least 1 ELO (using my >> finger in the breeze). >> >> * Thus the issue of whether you detect suicide may >> be a complete wash. >> >> * On the other hand detecting superko costs more like >> 6% or so, which costs 9 or more ELO. So a benefit >> of 1 ELO for doing superko right may not be worth >> the cost. >> >> Conclusions >> >> * The effect of suicide detection is *very* small in >> the scheme of things, and is probably not worth >> arguing over. Superko is also small, but might be >> worth a tiny amount of effort. >> >> * Some kind of study to measuring the ELO cost of bad >> suicide and ko decisions would be useful. >> >> * I plan to detect both suicide and superko on principle, >> confident that it doesn't make much difference. >> >> Michael Wing >> > > I have a few question-marks here. > > First, did Don really say that a doubling of the speed gains 100 ELO? > Or did he say adding a ply would add 100 ELO? There's a big difference. > > Secondly, you say the ELO benefit for not playing suicide is 1. > Admittedly you say you used your wet finger in the breeze. Thinking > more about it I'd say Don is right, not playing suicide should be a > considerable gain. I'd say that (putting my wet finger up) a random > player that doesn't play suicide beats a random player that does 2:1. > What's that in ELO? 100 points? Should be easy to verify. > > Mark > > > _______________________________________________ > computer-go mailing list > computer-go@computer-go.org > http://www.computer-go.org/mailman/listinfo/computer-go/ > _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
