From: Larry Kaufman comcast net> Date: 8 nov 2004 Subject: Re: Rating transformation Your free subscription is supported by today's sponsor: ------------------------------------------------------------------- Go on a Cruise for Free! http://click.topica.com/= caacOlEa2i6YsbnuqMaa/PermissionData ------------------------------------------------------------------- I think you may not be too familiar with chess/shogi rating theory.= =20 There is no need to concern yourself about variance of ratings, though you= =20 might choose to average your own rating over a period of time rather than= =20 using your latest rating, since the coefficient of change used by SC24 (16= =20 points up or down for a win against an equal) is much too high, which cause= s=20 excess volatility of rating but shouldn't affect its average value. You=20 speak of expressing your strength in terms of 1/N of Habu's strength. This= =20 sounds quite meaningless, unless perhaps you mean the expected ratio of win= s=20 you would score against Habu in even game play. That can be calculated by= =20 the formula: take ten to the power (rating difference/400). That number to = 1=20 is the expected score. For example, my SC24 is usually close to 2000, while= =20 Habu's is thought to be 3000. So, 1000 difference /400 is 2.5, and ten to= =20 that power is 316, so I should lose to Habu by 316 to 1. You might then say= =20 that I am 1/316 the strength of Habu, but I think that's a rather inaccurat= e=20 way to express this statistic. For technical reasons, there is some reason= =20 to think that SC24 ratings are spread further apart than they should be, so= =20 if we reduce the difference by 20% to compensate for this I would be "only"= =20 800 below Habu and therefore only a 100 to 1 underdog. A much more meaningful way to state a player's strength relative to= =20 Habu is by the handicap he would need to score 50%. Of course few of us hav= e=20 the opportunity to play multiple handicap games with Habu to determine this= ,=20 but it can be estimated. Based on research done to establish the=20 Pan-Atlantic system and allowing for the differences between that scale and= =20 the SC24 scale, I would estimate the following as fair handicaps vs. Habu i= n=20 the 30" per move games that dominate SC24, based on SC 24 ratings: A 2400= =20 player would take a bishop, a 2200 player would take a rook, a 2000 player= =20 would take rook & lance, a 1600 player would take two piece, a 1200 player= =20 would take four piece, a 600 player would take five piece, and a 200 player= =20 would take six piece. If anyone has a reason to think that this scale is no= t=20 reasonably fair, please let us know which side you think it favors and what= =20 are your reasons for your belief. Larry Kaufman ----- Original Message -----=20 From: "as20728" decef elf stuba sk> To: topica com> Sent: Monday, November 08, 2004 5:49 AM Subject: Re: Rating transformation Your free subscription is supported by today's sponsor: ------------------------------------------------------------------- Amazing Diet Patch The fastest - Easiest way to lose weight! Try it now FREE! http://click.topica.com/caacOlIa2i6Ysa7fLLna/MyDietPatches ------------------------------------------------------------------- Hallo everybody, Thank You very much for Your replies. If "The Rumors" ^_^ are true than we may directly compare anybody at shogidojo with Mr.Habu*. All we need is to employ a little statistics. Current rating of player is composite of two variables: true absolute ratin= g (that we don't know:-) and sum of all errors caused by selecting Your opponents and fact that all players start with rating 0** (and we don't kno= w this either:-). Luckily we may model our strength by mean and variance. We= may obtain these two by gathering rating over some period and applying well= known forumlas [ mean(rating)=3Dsum(rating[i])/(sizeof(rating)), var(rating)=3Dsqrt( sum((rating[i]-mean)^2)/(sizeof(rating)-1) ) ]. This interval estimation should work well for both low rating and high rating players. Strong players have high mean and low variance, weaker player vice versa. But for comparation we also need our gauge's rating expressed as mean and variance. It would be impractical to track all games of Mr.Habu*** here at shogidojo. Several possibilities may arise: 1) we may set Mr.Habu's rating as absolute value (variance=3D0 and mean=3D3= 000). But this is not suitable for comparation (see below). 2) we may use his rating from lists of nihonshogirenmei (even if it's only for fun we may use them:-) and add 130 (see original mail below) to final mean. Even if it's from completly differnet enviroment Now we need to compare them. I m not realy sure about this... First of all= we need to define what should result of our comparation. I think that statement like "I m one Nth of Mr.Habu's strength (with probability of p).= " would be nice. [few integrals and google queries later] But creating such expresson is beyond my skills. Is such comparation possible? Can anybody help me? I m still looking for it but help would be appricated. ---- * I'm not sure, but Mr.Habu seems to be a fine gauge. ^_^ ** Sometimes You meet player with lower rating who pulverize You with "ease". I m at bottom of rating list thus I meet them regulary. *** Still assuming that player with best SD rating is Mr.Habu. Please correct me if I made some mistake there... most likely I did. Best regards, Adam Skalny ps: i m sorry some terms might be wrong but i m not used to write about mathematics in english. plus all my knowledge about this great domain is limited by one term of mathematical statistics at my school :-) So maybe someone with better knowledge may add more sense to it. pps: whops looks like i had too much free time once again;-) ----- Original Message -----=20 From: "Larry Kaufman" comcast net> To: topica com> Sent: Monday, November 08, 2004 7:39 AM Subject: Re: Rating transformation > Your free subscription is supported by today's sponsor: > ------------------------------------------------------------------- > Go on a Cruise for Free! > http://click.topica.com/caacOlEa2i6Ysa8edaea/PermissionData > ------------------------------------------------------------------- > > If we assume that the scales are comparable except for a constant, then > we can compare the SC24 ratings with the pro ratings easily. The top five= > SC24 ratings are presumably top pros (it is rumored that the top one at 3003 > is Habu, and Moriuchi and other top pros are known to play there), and th= e > top five ratings on SC24 average about 130 less than the top five on the Pro > rating list. So simply adding 130 to a SC24 rating should be a good first= > guess as to the comparable rating on the pro list. However for technical > reasons it is likely that the SC24 list is more "spread out" than the pro= > list, so probably this somewhat understates the equivalent ratings of > ordinary amateur players. But I think it's fairly accurate for strong > players. > > Larry Kaufman > ----- Original Message -----=20 > From: decef elf stuba sk> > To: topica com> > Sent: Saturday, November 06, 2004 2:56 AM > Subject: Rating transformation > > > Your free subscription is supported by today's sponsor: > ------------------------------------------------------------------- > Like music? Get a CD Player here > http://click.topica.com/caacOk8a2i6Ysa7fLLna/PermissionData > ------------------------------------------------------------------- > > Dear Mr.Kaufman and all other shogi funs, > would You mind to publish some formula for transformation shogidojo ratin= g > to pro rating or nihonrenmai ama rating. It d be interestig to compare one's > strength with "real shogi players" :-))) > > Best regards, > Adam Skalny > > Your free subscription is supported by today's sponsor: > ------------------------------------------------------------------- > Save up to 67% on Omaha Steaks + Get 6 FREE Burgers and a > FREE Cutlery Set + Cutting Board! > http://click.topica.com/caacOlKa2i6Ysa7fLLnf/OmahaSteaks > ------------------------------------------------------------------- > > Your free subscription is supported by today's sponsor: > ------------------------------------------------------------------- > Free recipes to get you in shape > http://click.topica.com/caacOlCa2i6Ysa8edaef/PermissionData > ------------------------------------------------------------------- > > > > > > > Your free subscription is supported by today's sponsor: ------------------------------------------------------------------- Get 4 Free low-fat, low-carb recipes http://click.topica.com/caacOlya2i6Ysa7fLLnf/PermissionData ------------------------------------------------------------------- Your free subscription is supported by today's sponsor: ------------------------------------------------------------------- 4 Free Low-Fat recipes http://click.topica.com/= caacOlza2i6YsbnuqMaf/PermissionData ------------------------------------------------------------------- --^---------------------------------------------------------------- This email was sent to: = shogi-l shogi net EASY UNSUBSCRIBE click here: http://topica.com/u/?a2i6Ys.= bnuqMa.= c2hvZ2kt Or send an email to: shogi-unsubscribe topica com For Topica's complete suite of email marketing solutions visit: http://www.topica.com/?p=3DTEXFOOTER --^----------------------------------------------------------------