BuzzerBeater Forums

everyone calm down...

i didnt write the ratings formula so i can't go look for a bug, as I dont know the fine details of how it works.. I will ask charles about it.

can I ask, stepping back from the quantitative aspect whether or not your relative world ranking is in line with what you think it should be.. not the rating number, but the rating position.. and likewise, do the other ratings make relative sense to one another? or is the bug in your mind that the quantitative number is off and you haven't thought or examined the rankings beyond that? if you feel the rankings are off qualitatively can you get a specific example of a team that dropped many spots, or rose many spots in a way counter to what you would expect from their recent results?

I am asking this specifically to assist in tracing back the behavior of the algorithm to see whether the results match with the way it was designed, not to be a pain in the butt.

Last edited by BB-Forrest at 3/24/2009 3:27:31 PM

My relative world ranking (or better european ranking) is ok, but there was quite big differences between our U21 (Ceska Rep. U21) and the team up and bottom of us at the end of previous season.
But the relative rankings of others team is hard to say because in my mind the quantitative number is off.

What data do you store in the database? Is it possibble for example to find points which every team gains for every particular match (or every particular week - sometimes there is more than 1 match in a week)? I my opinion this should help us a lot in loooking for the reason.
I think almost all U21 teams have strange points last two ranking updates.
For example
Chile U21 (win against Rep. Dominicana U21 118:89)
Malaysia U21 (win against India U21 114:93)
Bulgaria U21 (lost to Andorra U21 125:126)
Hong Kong U21 (win against Hanguk U21 104:78)
Uruguay U21 (win against Costa Rica U21 114:59)
Nigeria U21 (win against Tounes U21 93:74)
has -15 pts last update. It looks like they lost 15 old points and gain 0 new points (maybe Bulgaria is near to correct points, but the others not).

this is where i have to get charles involved because your approach to figuring out whether something went wrong makes a lot of assumptions about how the algorithm works that I'm not sure are true... i'm not sure they aren't true either... but it might not be as simple as ... look at the last game they played... add/subtract some points based upon some analysis of that game.

I thinks it works something like this (maybe I'm not right):
1) There are some inputs (stored in database - like teams, position and points in rankings before match, match data)
2) There is some algorithm
3) There are outputs - points for a single match or single week and cumulative points for last 28 weeks (some of this I supposed to be stored in database)

My idea is to look over the outputs stored in database, this could shows whether there is a real problem. (For this first step we don't need to know algorithm).
And if it shows there is real problem than we need to now algorithm to search for bug in it. Or maybe there is "only" problem with input data (but for this case we also need to now algorithm).

but the concept of "points for a single match" might have no meaning in the contest of the algorithm. If its makes some complicated comparison of relative win/loss records and strength of schedule it becomes a very complicated question how a single game contributes to the ranking.. certainly an straight additive effect is not predetermined. again, i dont know how it works, i didnt write it, but if i did write it I could imagine doing it in a way that wouldn't allow for the contribution of a single game to be analyzed in the way you suggest (looking at the isolated quantitative change of the ranking variable outside of the context of all the other rating variables).

Ok, we will wait for Charles. He is probably the only person who can help us.

But I'm quite sure you think the algorithm is more complicated than it really is. I'm quite sure there is something like points for single match. There is not recalculation of whole last 28 weeks every week.

From the rules:

The world championships aren't the only way that national teams can compete for status. For every national team game that gets played, some number of ranking points get distributed between the participants. Even scrimmages are worth something, but the bigger the game, the more points get handed out. All ranking points that national teams receive, they keep for 28 weeks (i.e., two seasons). Points expire on a rolling basis.

And all my NT and U21 Rankings analysis showed it really works this way.

Last edited by rwystyrk at 3/25/2009 5:02:42 PM

but might there not be some renormalization that happens right before the enthusiasm reset every season for instance?

What do you thought when you wrote "renormalization"?
Recalculation of NT Rankings at all? I don't think there is something like this. Or if there is than ratings after renormalization looks very close to this before it (no big points differences in that time I mentioned).
But I think there should be fixation of ratings which are used for whole next 2 seasons for calculation of team strength. If it was like that it explains why all matches in first 2 seasons of U21 was worth the same amount of points (all matches at the same stage), because all team were equal for algorithm for whole 2 seasons.

i stand corrected. charles and mark did discover an issue and they are solving it