Lead

Dec 19 11 9:47 AM

Tags : :

Roon Ba compared to eloratings [Matches of Top 100 teams (in Roon Ba)]:
both teams Top 100 (about 43% of the results): eloratings .375 / .378 Roon Ba
one team Top 100 (26% of results): Roon Ba .317 / .323 eloratings
at least one team Top 100 (69% of results): .357 each

(only official competition matches: eloratings is calculated by .001 front - which means next to nothing)

None of the teams Top 100 (31% of results): Roon Ba .347 / .354 eloratings

Total (about 1,220 results from 11/17/10 to 12/16/11): Roon Ba .354 / .356 eloratings.

What the numbers mean? Measured were the standard deviations between the predictions (based on ratings) and (then) actual results. In order to compare the ranking systems fairly with each other, for each ranking system the standings were used, which was at the time of the last publication of the FIFA rankings. For example: for matches after 11/23/11 the standings from 11/23/11 … etc.

The smaller the standard deviation the more accurate the ranking system. A (empiric) difference of .002 marked a distinct advantage, differences of >=. 005 a difference in quality. Please note: we have here is a sample (albeit a very large).

Incidentally, I have developed a polynomial regression model - at the suggestion of an interested. With this model, a rating of a certain rank to be assigned. The use of the model allows an analysis of all the ranking systems. The model provides projections for the standard deviations of the systems. Thus ranking systems can also be verified for accuracy, which will elude a review.

[The basic idea: system A says it is playing the No. 28 against No. 29, according to system B, No. 1 against No. 200. If the actual result then a draw, system A was in this match (much) more accurate ...].

Sorry, my English is not on the same level as my knowledge of statistics ... I hope that the text can be understood anyway.