RandomSoccerFan
PREMIER
By SR ratings alone, Blues and Surf are as stated - but the ones missing include CDA Slammers & Rebels, with very similar ratings to Surf. Blues show two teams in top 5, for what it's worth.
Big year coming up 2012 ecnl
- Thread starter Striker562
- Start date
Striker562
BRONZE
How much do we believe these ratings? Feel like I’ve seen some of these teams that are not in top 5 play top 5 teams and hold their own, I feel like you won’t always get consistent results every game/tournament, if you guys could make your own rankings for these 2012’s what would you guys ranks from 1-10?
RandomSoccerFan
PREMIER
Been argued to death in threads like this (link) over and over again. A higher ranked team will beat a lower ranked team ~ 83% of the time for entire data set, and 75% of the time if the teams are in the top 100 nationally. It's going to choose the wrong winner 100% - that percentage each time - but on the whole, its accuracy can be measured and tracked, and betting against it isn't particularly wise. Remember, at this age level, #20 in CA is only 1 goal off of #50 in CA, which is only 1 goal stronger than #100. So there are absolutely going to be games where a team in the higher double digits holds their own and even beats a team in the lower double digits, or even single digits. But it's not going to happen most of the time.
irishvirus
BRONZE
The ranking don't account for attrition or player movement (Blues Oliveros does not exist anymore and Joga Bonito is a tournament team) A true top 10 in my opinion is: 1a&b. Blues, 2. Surf, 3. Slammers HB Elite, 4. Slammers HB Koge, 5. Rebels, 6. Beach, 7. Slammers Boatman, 8. City SC, 9. Eagles, 10. DMCV Sharks
You dropped the Rangers...
This is why you all need someone like we had in @Technician72. He brought updated news on player movement, and he had good sources for the latest news and his info was spot one. For example, someone from Legends wanted their 03/04 ranked hire because of a big tournament win. However, Tech Specs got the skinny that the Legends team was actually a combo team and mixed bag of guest players and not the real registered team. He knew who was guesting and who was leaving for greener pasture. He called it Tech Specs and I sure miss those days.
RandomSoccerFan
PREMIER
Yep. It is only looking backward at games that have actually been played by team entities that existed at that time. If teams no longer exist, combine, or are unlikely to play tomorrow similarly to how they played yesterday, it takes the data only after tomorrow's games can be played. For the most part, the immediate past predicts the future, and the data supports that, but there are certainly plausible examples where that's not the case.
This I agree with regarding the ranking app.
Where it falls short is near term challenges/changes that potentially affect the outcome. Things like a coach quitting the team, an impact player not able to play, the entire team has the flu, a big social event like homecoming, the team is highly ranked but is playing kickball + not posession, etc, etc, etc.
Not suggesting people bet on youth sports. But what I'm describing are events that cause gamblers to quickly bet big one way or the other. When this happens bookmakers move the line to incentivize betting the other way so their books are even. Bookmakers make a small amout per bet. As long as wins and losses wash each other out they make $$$.
Above is how professional Books are so good at predicting the outcome of games. They're using previous results to define a line. Then the line potentially changes based on the bets being placed. The net outcome is a predictive mix of previous results and current challenges/changes.
Above + the fact that often when top 50 ranked teams play each the game is a coin flip. This is why I don't like the rankings app for predicting individual games. Also some people look at the ranking app + think that if xyz team predicted to win that its a 100% that they will. But, that's not always going to be the case. Espicially if the predicted win is by a very narrow margin. (I agree predicted blowouts will almost always be correct in the ranking app)
The rankings app is VERY good at providing a high level view of all teams + how they stack up against each other independent of the league they play in.
RandomSoccerFan
PREMIER
Yes, yes, and yes, to all. But..... it's the same argument people make when they are completely and incorrectly discounting the predictions from the app. The data is the data, the results are the results, and it will pick the right winner that expected percentage of the time - which happens to be a very high percentage of the time.
Striker562
BRONZE
Blues seems to lose players but still manages to stay atop, also hearing the HBK slammers is putting together a great team, will be interesting how it all plays out next year
RandomSoccerFan
PREMIER
Apologies, just caught this. IMO, this is completely and totally misrepresenting the data. For 2012G, #1 playing #50 is expected to be a 5-0 game, with #1 winning 85% of the time, tying 6% of the time, and only losing 9% of the time. #1 playing #20 is expected to be a 4-1 game, with #1 winning 79% of the time, tying 8% of the time, and losing 13% of the time. #1 playing #10 is expected to be a 4-1 game, with #1 winning 79% of the time, tying 9% of the time, and losing 12% of the time. #1 playing #5 is supposed to be a 3-0 game, with #1 winning 68% of the time, tying 14% of the time, and losing 18% of the time. Even if we go all the way to #1 playing #2, it's still expected to be a 2-0 win for #1, winning 55% of the time, tying 16% of the time, and losing only 29% of the time.
None of these are anywhere close to a coin-flip, and it completely misrepresents the probability of winning or losing for each game. Of course if the ratings are closer, the percentages are closer, and as the ratings diverge, the chances of predicting the win accurately also increase. But even in the top 50, and even in the top 10, the ratings are far enough apart that a true picture of each game can be predicted with a measurable amount of accuracy ahead of time.
I probably shouldn't have said "top 50".
Data is going to be clumpy (its just how things work) + there will be more coin flip situations when teams are closely ranked than when they're ranked farther appart.
In these situations even though one team will be a predictive winner. If you're talking a 51/49 win loss ratio or even a 60/40 win loss ratio I'd still consider the game a toss up because of the things historical numbers alone can't predict. (Coach left the team, impact player sick, etc, etc, etc)
More than anything my concern is that some people just don't understand numbers. So when they see XYZ team will win or lose they take it as 100% truth without understanding that the rankings app is just an educated guess. Sometimes a very strong guess + other times a very weak guess.
RandomSoccerFan
PREMIER
Agreed. If the ratings are so similar that it's close to a 50% guess - it's a coin-flip by definition. 60-40 is pushing it, but people can potentially consider that a coin flip, even if 1 team is 1.5 times more likely to win rather than lose compared to the other team. IMO, anything more lopsided than that is being described incorrectly if someone is describing the predictions as a coin-flip - it just isn't.
Whether it's top 20, or bottom 100 - two teams that come together with almost identical ratings are going to approach an equal prediction, or even chance of winning for both (in this case it's less than 50% for both, as there is a chance that they will tie). And in addition, the predictiveness of those in the top 100 is going to be slightly, but measurably, less than the predictiveness of the entire data set - including not only the top 100, but the entire amount of teams of that age/gender. This isn't only because the ratings for matchups are closer near the top (which would lower the predictiveness), but likely also representative of teams near the top being slightly less predictable in terms of individual outcomes, due to both better play and lower scores.
You're also right that all of this is just a prediction - and can certainly turn out to be wrong for a particular game, that's in fact what's being measured. If people think that it's a certainty about what's going to happen, they are mistaken. But people who think that an incorrect result means that the ratings must be wrong - are equally mistaken.
Whether it's top 20, or bottom 100 - two teams that come together with almost identical ratings are going to approach an equal prediction, or even chance of winning for both (in this case it's less than 50% for both, as there is a chance that they will tie). And in addition, the predictiveness of those in the top 100 is going to be slightly, but measurably, less than the predictiveness of the entire data set - including not only the top 100, but the entire amount of teams of that age/gender. This isn't only because the ratings for matchups are closer near the top (which would lower the predictiveness), but likely also representative of teams near the top being slightly less predictable in terms of individual outcomes, due to both better play and lower scores.
You're also right that all of this is just a prediction - and can certainly turn out to be wrong for a particular game, that's in fact what's being measured. If people think that it's a certainty about what's going to happen, they are mistaken. But people who think that an incorrect result means that the ratings must be wrong - are equally mistaken.
irishvirus
BRONZE
I do not rate them from what I have seen
irishvirus
BRONZE
The "somewhat" gap as you described is generous. Head-to-head wins since the 11-a-side transition have been one-sided games. The results against same opponents favors Blues.
I figured Diego's team would eventually absorb into one super Blues team. I would imagine a lot of the now bench players will look elsewhere for starting ecnl opportunities
Here's a couple of examples showing why if people don't understand the "predictiveness" of the ranking app it can cause problems for a team.
Example 1... Say a team has to travel for a game but the ranking app is showing a win. There's going to be parents that choose to stay home that weekend specifically because the rankings app shows a win. In coin flip situations by not traveling for the game you've just handed your opponent a win.
Example 2... Some parents want to play on the top team no matter what. If they don't understand how numbers work in the rankings app they also don't understand that if 3-4 teams have a nearly identical ranking it doesn't matter which one of the 3-4 teams you play for.
Example 3... If a team has 24 rostered but only 18 can suit up on game day coaches need to understand that predicting a win by 1 goal is an entirely different game than a win predicted by 8 goals. If coaches don't understand numbers + predictivemess they won't field the right players.
As you can see the ranking app showing a single game predicted win or loss isn't nessasarlly a bad thing. Unfortunately because many people don't understand how numbers work often poor decisions are made. Throw in potential real time challenges like coach leaving the team, impact player being sick, etc and predictions for very closely ranked teams isn't worth much.
BTW a way to address what I'm describing is to list closely ranked games (withing some defined variance) as a "coin flip" even if one team or the other is mathmaticaly the winner or loser.
Example 1... Say a team has to travel for a game but the ranking app is showing a win. There's going to be parents that choose to stay home that weekend specifically because the rankings app shows a win. In coin flip situations by not traveling for the game you've just handed your opponent a win.
Example 2... Some parents want to play on the top team no matter what. If they don't understand how numbers work in the rankings app they also don't understand that if 3-4 teams have a nearly identical ranking it doesn't matter which one of the 3-4 teams you play for.
Example 3... If a team has 24 rostered but only 18 can suit up on game day coaches need to understand that predicting a win by 1 goal is an entirely different game than a win predicted by 8 goals. If coaches don't understand numbers + predictivemess they won't field the right players.
As you can see the ranking app showing a single game predicted win or loss isn't nessasarlly a bad thing. Unfortunately because many people don't understand how numbers work often poor decisions are made. Throw in potential real time challenges like coach leaving the team, impact player being sick, etc and predictions for very closely ranked teams isn't worth much.
BTW a way to address what I'm describing is to list closely ranked games (withing some defined variance) as a "coin flip" even if one team or the other is mathmaticaly the winner or loser.
RandomSoccerFan
PREMIER
Right - but the results of all of these what-ifs are already fed into the ratings, because they are already affecting the game data. The end result is exactly the same. Someone overweighting what they mean and then taking action, even unnecessary action, may happen - and then the game results will be what the game results will be - and the rating is affected positively or negatively depending on those results. By your hypothesis, the individual game results should be less predictive than the expected - but the ratings themselves are based on the individual game results. If teams really changed significantly from game to game, their own results would show a bunch, maybe even a majority of green overperform and red underperform results - but it is a very rare team that will show that type of history.
In a perfect world, all of the parameters you listed would be accounted for in the ratings/predictions, and would shade the percentages for an upcoming game up or down depending on a variety of factors that it cannot today, by just looking at the scores. In that same perfect world, the predictiveness of such an app would be expected to be significantly better than the current one, and it sure would interesting if that were possible.
Your suggestion to address a non-existent problem already exists in the app - it is giving the exact percentages of the guess in advance of the game. People can use their own judgement whether 50% chance, 80% chance, or 95% chance means that a win/loss is likely.
TheStr1kerz
SILVER
What have u seen to leave them out of ur top 10? Just this year alone Rangers are 500 against ur top 10