A while back I posted that Pats Pre ECNL’s mercurial rise through the rankings looked odd. I suspected that Soccer ranking had an algorithm that give more weight to a team with ECNL or MLS in their team name. You chewed me up for it. But looks like this is it.
I remember this incident differently. There was a clear glitch with the rating, you were apoplectic about it, several people (including me) said wait a day or two for it to resolve itself, and that's exactly what happened.
There is never anything to do with any name of the team, bracket they are playing in, or any other factor. I don't know how many times people need to be told - the only thing that matters, is the game scores - and the team entity that they are attached to.
The ranking app doesnt care about the name of the league. With Pats there is a little going on that does give them a boost. In Socal ECNL league theres several highly ranked teams. If they beat or tie these teams it will push them up in rankings assuming they get more than the expected number of goals. However, once Pats are highly ranked themselves the ranking app weights games with similar ranking over those that arent similar. What this means is ties against an equally highly ranked team will help to maintain a high ranking more than they normally would. This is the feedback loop I was talking about. It artificially keeps Socal teams ranked high because of the number of highly ranked teams they play every week in league. Just to be clear I'm not saying that they wouldn't be ranked highly if the ranking app didnt weight playing similar ranked teams. This isnt the case. I am saying that they're ranked a little higher than their ranking would be without weighting similar ranked games.
Well, sure - but again - it's not a bonus or artificial boost for similar-rated teams, that makes the future predictions less accurate. It's set up that way because of the opposite - it instead makes future predictions as accurate as they can be.
I think we need to separate the definition of "better rating" into meaning higher rating / higher quality team, and "better rating" - more impactful on a team's rating. They have completely different meanings, and I think interchanging them may be causing much of the lack of clarity in these discussions.
Take 3 games, for the same team. They play 3 different teams. First one they are 4 goals stronger, and they win 4-0. 2nd one they are very similar, and they win 4-0. 3rd one that are 4 goals weaker, yet they still win 4-0. The first game will have an effective differential of 0 goals. The rating for the team will not be expected to change much at all - both because of the differential - but also because the game won't be weighted heavily. The 2nd one, the effective differential is 4 goals - which will be weighted heavily as the teams ratings are close. The 3rd one, the differential is an astounding 8 goals (congrats to the team), but while it will likely be helpful to the rating, the result will not be weighted as highly to their rating, compared to the 2nd game. Even though it's a fantastic win - it's not as predictive on how future games will go.
If a team consistently plays teams that are similar to themselves, the weightings for each individual game (and their results) will be similar as well. If a team instead plays teams all over the board - it can take a longer time for results of games against dissimilar teams, to make a dent in the rating earned by playing similar teams.
If the ranking app didnt weight games with a similar ranking more than those without a really good team could manipulate the ranking app by blowing out terrible teams 10+-1 over and over.
Sort of. If a team is 5 goals stronger than its opponent, and it keeps shellacking them 10-0 over and over again, they will bend to 10 games apart in strength. It would just take a very significant amount of games to do so. Again - even in this case I still wouldn't define it as manipulation - as what would be happening to the rating is a fair representation of how games are being scored, and the best prediction for how those teams would be expected to perform in the very next game.
Teams in leagues that have a mix of ranked teams in their league get screwed a little in ranking. But this isnt always a bad thing because in playoffs or tournaments they'll be better than their ranking and likely seeded against not as good teams. (initially) Once they blow out a good team the ranking app will quickly adjust their ranking up because they scored several more than the expected number of goals.
The ratings can move quite quickly, yes. If a team isn't getting game results that align with their current rating, one might be surprised how quickly they can either move up or down in rating (and therefore ranking). It doesn't take half a season or more for things to adjust. More recent games are much more heavily weighted than older games, even from only a few weeks back.
Yeah, we just lost a final on PKs to a highly ranked team and the app unfortunately recorded the 5-4 penalty score, not the 1-1 tie, and we also beat another highly ranked team in the group stage, but that result ended up in the Unranked section due to a slight name change for the opposition, so we didn't gain anything in the app from either result.
That's crappy that the tournament entered the score that way. Does the web page of results make it clear that it was a 1-1 game with 5-4 penalty kicks, or does it just show a 5-4 score? If there is a problem there, and you think the app should have scraped it differently - please send a link of the web page results to support. They have to clean up that engine continuously as results pages continue to morph, and often they only are alerted to do so when these anomalies are raised to them. If it is a problem instead with how the tournament entered on their own site and you can get them to correct it, just ping SR and they will rescan it. Some tournaments are more responsive to these changes/corrections, some (maybe most?) couldn't care less and aren't likely to ever correct errors. Some of them care about the electronic results so little that the championship game results never actually get posted to the app after the tournament ends.
For the second case, are you a pro user in the app? If so, and you are sure that the name change for that team (or any team), still represents the same team that is already in the database - add the data source yourself. You can correct anything you see across the board - not just your own team. But be sure, and be careful - as if there are changes made that someone else then objects to and reports - you may need to defend why you made that change, and ultimately lose editing privileges if it happens enough times.
That said, while I find the Ranking App rankings a fun distraction, it's just a ballpark guide, particularly for tournaments where inter-league play is in effect. The real value is in looking at the previous results with an educated eye.
A useful tweak might be to also offer a ranking for your team's current league, though I guess we have that already and it's called the League Table, LOL.
It's not perfect - and all it is able to ever to is to try and predict the future based on what's already happened. It does a darned good job of that, and people tend to underestimate how accurate it turns out to be. Just look at any of the threads here for things like Surf Cup predictions, State Cup predictions, or even just the game history of one's own team.
We create the tables you are suggesting for several team ages in the club, and update them every month or two. It provides a clearer picture than just the current season's standings bracket, about how each team is growing compared to its peers in bracket and beyond. It really helps all parties understand how a team is developing from a performance standpoint.
One way to think of it - is that if a set of teams, across their full history, ever played a single game outside of bracket (no tournaments, no playoffs, no anything), at the end of the season the standings bracket and an SR ratings bracket would look very, very similar. SR is just a representation of that team's performance, just as the standings bracket is. It would be somewhat more predictive than the standings bracket, as that is ranked purely by W/L/D, while SR ratings instead incorporate how significant each win or loss tends to be - but they would be awful close in most scenarios. It's just that in the real world, teams are not limited to a single bracket, and you probably don't want to wipe the slate clean every season to determine how strong an opposing team is - game history across all events and prior seasons can be helpful, and SR gets use out of ones up to 12 mo old, though the weighting of the older ones gets lower and lower.
