Midge wrote:SpaceDandy wrote:Given all the confidence the system has in knowing this squad, it ought to know they are, um, not good. No offense, but just objectivly not as good. Some data stabilizes quicker than other in data sets like this but should result in a stabilized understanding that this squad isn’t deserving of this rank - not conflating its confidence in the data with what that data is actually saying. Or at least better balancing between the number of data points and what those data points say. If the Baltimore Orioles played 1,062 games last season and the Red Sox played 162, at their respective winning percentages, one should be more confident how bad the O’s are and not overly diminish the Sox’s excellence based on fewer games played. Especially when the other squads/teams play closer to, per the analogy, 162 game schedule. This is how algorithms should get refined, seeing a result like this, understanding why and tweaking it.
Apologies to non baseball fans trying to follow that.
It is refined regularly as it gains confidence with other squads.
The problem here is people conflating TrueSkill with RANKING.
The system is not RANKING teams. It is giving them a True Skill score and a certainty based around that score. Then simply sorting squads based on the combined metric.
They are CERTAIN of this squads number with NO uncertainty ... and I would venture to say they are pretty correct in that assertion considering it is a terrible squad with a terrible True Skill score.
Everyone here is getting caught up in the ranking, when the ranking of a squad like AncientWarriors can fluctuate immensely because they allow their sigma to fluctuate quite often with a week off or more. That adds UNCERTAINTY ... and the system processes it as not being certain of where it should be, while it is still certain, given the amount of data it has, that the other squad sucks ass.
Also, the assumption that being ranked 435 is somehow taking someone's spot is kind of laughable. It is not difficult in the slightest to get ranked in the top 300 considering there are barely 300 legit squads playing at this point.
So a "legit" squad is a squad which wars twice a week. I get it.
To get to top 500, you don't have to win. Just get a bot to submit for war twice a week, don't need to do a single attack, and you'd get in top 500.
btw, there are 26K squads out there. If you have good attendance record (war twice a week), you can be in top 2%, even if you lose all your matches.