kujo wrote: ↑Wed Aug 22, 2018 10:10 pmThis could be because, the point differences in the initial ELO seeding of teams was all even (separated by 4 points each) - whereas in reality, there
should have been a much wider gap between Germany and Korea.... For example, before the new ranking there was more than 1000 point difference between them, while in the new one they had 212 point difference only...!!
That was my guess, but I didn't want to bias you before you looked at it. Their win against Germany should have given them so much more points than the two losses against teams ranked above them. In fact I saw somebody predicting earlier with a calculation using the points of the independent ELO ratings (that existed earlier) that S. Korea might move up to around #27 or something on that rating. That was probably a bit too much, but they deserved to move up at least to the 35-40 range. Korea is one team that got totally screwed. In fact, teams like Egypt dropped badly and fell behind teams that were inactive who might not be as good as them. All quite unfair, but people haven't figured things out well enough to complain, and so FIFA got away with murder!
Again the issue is with using a linear scale. I thought a steeper drop of points per ranking spots were needed in the first 10, next 20, next 50 and the next 100. Basically a nonlinear drop. They could have come up with a nonlinear scale similar to what was in the independent ELO ratings that existed earlier.
I find that the world cup matches caused a nonlinear curve at the top now (in the first 40 ranks) where the points plotted against ranks show that the new curve has risen above the
Pseed points curve. But seeing that the friendlies between India, Kenya, TPE and NZL hardly made a blip of ranking changes among them, I can see that moving up at the bottom is extremely hard, like you said!
I think this is pretty screwed up. I won't be surprised if FIFA drops this within a year or two and goes to something else. Or maybe they would do a re-curving of points (again preserving the ranks) for a reset of the ELO. I am guessing that ranking changes will be minimal month to month from now on. Am I right?
There lies the problem I have always felt with ELO (or with my understanding of it). We simply do not play enough football matches for ELO to work. I think a team needs to play some 50-60 matches for their rankings to move up by say 20 spots (i.e, get around 10 odd matches against teams that are significantly higher in rankings, say by 40-50 spots, so as to get 4-5 big wins, which is all tough to schedule). And that would take say 4 years. Over such a period, a team's quality could become substantially different. That is, ELO could be simply too slow in responding, for it to be used in football. If teams play 30 matches per year, it would be responsive enough to match the "quality changes" that happen in a team. That is my hunch. Again, am I right?