The airing of grievances
Posted By: Joe Russell In Response To: The airing of grievances (Bill Phipps)
Date: Thursday, 26 December 2013, at 11:50 a.m.
In Response To: The airing of grievances (Bill Phipps)
'1) It is ABT only. Why leave off perfectly good data from Masters events?'
It seems that data would be just as or more valuable, as they are often longer matches against the most skilled competition. To ignore that data is a mistake.
'2) Because the Giants list is looking at the last 2 years, and the sample over that time is small, there is a tremendous amount of noise in the results. To the point, that they defy common sense.'
Agreed! To get any read on how well someone plays requires a significant number matches with recorded PR's or a significantly larger number of match results. I will concede that PR is not everything and that there are some players that are better at playing the opponent than others, but there is probably no stronger indicator of strength than PR, even if it does not tell the whole story. To get an accurate read of a player's strength without knowing their PR's probably requires 100's of matches.
'I believe there is a danger to trotting out a model that has some pretty columns, uses some fancy weights, purports to be mathematically based and spits out a final rating. The risk is that people will look at the final rankings and somehow call this science. In fact, this is just a limited tool with a high margin of error. It is just a way of organizing ABT results and adding weights for time.'
It IS just restating ABT results with weights added. Even the weights seem absurd to me. They are taken from other games/sports where there are many more active participants and events than there are in backgammon. The weights would likely work if the average backgammon player played in 20 tournaments a year, but with most players making it to between 2 and 7 events a year, it takes several years to get a read on their playing strength. I can see using weights, but in backgammon they should be ever so slightly scaled because of the small sample sizes.
I think there are other lists that are more representative of player strength. For example; using the USBGF ELO rankings for the top 20 Open players that have played in 15 or more events since 2009, you would get 11 players that are on Ken's list and nine that are not.
Since there are 11 players on each list, comparing the differences might serve as an indicator as to which list is more accurate at forecasting player strength.
Ken's top 7 players, not on both lists: Jeb Horton; John O'Hagan; Rod Covlin; Alfred Mamlet; Mike Senk; Howard Markowitz; Ted Chee.
The top 7 from the USBGF ELO list, not on both lists( Open players, 15 events minimum), which includes Master events: David Rubin; Stuart Thomson; Bill Phipps; Falafel; Bill Robertie; Ed O'laughlin; Michi.
Granted both lists have many very strong players, but I do think one is highly superior to the other.
I think the work being done by Ken is a step in the right direction, but I think the weights are too graduated, due to small sample sizes, and Masters events must be included before the system will have much value. Otherwise, as Bill said, it is just a restatement of ABT points with an arbitrary weighting system.
Messages In This Thread
BGonline.org Forums is maintained by Stick with WebBBS 5.12.