Composite Computer Rankings and Correlations

9 April 2019

We have 12 computer ratings again this week, and the analysis is essentially the same except the Condorcet Consensus is a better fit than the Bucklin. As might be expected as more games are played the rankings are slightly less different than last week - the largest cumulative distance for any composite is smaller than last week's least.

∑Dist ISR KLK MAS DII FMG MGS RT SAG DOK MOR NOL RPI

25840 Cond 946 981 1076 1280 1529 2132 2188 2241 2803 3129 3721 3814

25927 BMaj 968 1000 1157 1334 1488 2118 2239 2230 2770 3118 3758 3747

26557 Borda 1180 1223 1284 1388 1643 2379 2033 2332 2727 3004 3607 3757

26706 Mcomp 1198 1263 1274 1403 1677 2414 2034 2347 2738 3011 3570 3777

1-ISR 2-KLK 3-MAS 4-DII 5-FMG 6-MGS 7-RT 8-SAG 9-DOK 10-MOR 11-NOL 12-RPI

The order of computer rankings by distance from the Condorcet composite is a beginning, but having found that the ISR is "most like" each of the individual composites, we can re-order the computer rankings by how like they are to the ISR. Six of the rankings differ from the ISR by fewer than five per cent of the team-pairs. Among these seven ratings, only four of 21 pairs differ from each other by more than six per cent.

Avg ISR MAS KLK DII MGS SAG FMG

9506 ISR 9733 9675 9602 9567 9554 9550

9485 MAS 9733 9629 9612 9437 9516 9480

9497 KLK 9675 9629 9618 9491 9384 9624

9466 DII 9602 9612 9618 9395 9407 9504

9310 MGS 9567 9437 9491 9395 9239 9519

9304 SAG 9554 9516 9384 9407 9239 9231

9428 FMG 9550 9480 9624 9504 9519 9231

I report the extra "composite" rankings as Computer Ranking Composites including a table of distances between pairs of ranking after the report by team.