[Albion] 52 points prediction on FiveThirtyEight

AlbionInUs said:

“ Before a season begins, a team’s SPI ratings are based on two factors: its ratings at the end of the previous season, and its market value as calculated by Transfermarkt”
“ As a season plays out, a team’s ratings are adjusted after every match based on its performance in that match and the strength of its opponent. Unlike with the Elo rating system we use in several other sports, a team’s rating doesn’t necessarily improve whenever it wins a match; if it performs worse than the model expected, its ratings can decline.”
“ we use three metrics to evaluate a team’s performance after each match : adjusted goals, shot-based expected goals and non-shot expected goals”

FiveThirtyEight uses xG and transfermarkt values in order to determine a teams SPI. The SPI value is basically a “team quality” metric, which is used to predict probability for future matches. Although FiveThirtyEight as a site is reality accurate, its football predictions are flawed. Expected goals, as we found out, doesn’t tell the entire story. We would have been 5th last season on xG. xG accounts for a teams opportunities, not actually the quality of strikers. Transfermarkt, on the other hand, gives greater value to younger players. Since we have a younger squad, our market value, and therefore, our SPI rating, is artificially higher than other teams.

Not to say we won’t finish 9th by the way. Just saying that FiveThirtyEight isn’t some magical football predictor (there predictors for other sports are actually pretty good though).

Click to expand...

Superphil said:

For a sanity check I thought about this, and I reckon anything above 17th will be doing well, so this is a great prediction. I do think they’ve been on some sort of mind bending drug or 90% proof liquor though.

Click to expand...

Mtoto said:

Not so, on 538 at least. From their "How This Works" explanation:

"Shot-based expected goals are an estimate of how many goals a team “should” have scored, given the shots they took in that match. Each shot is assigned a probability of scoring based on its distance and angle from the goal, as well as the part of the body the shot was taken with, with an adjustment for which specific player took the shot.x

All players who have enough shots in our database to qualify are given a modifier based on their historical conversion rates (the number of goals they’ve actually scored, given the quality of the shots they’ve had). For example, Lionel Messi has historically converted a shot into a goal about 1.4 times as often as expected, so the probability of any shot he takes is multiplied by 1.4."


I'm not sure that anyone really claims that xG tells the entire story, it's an estimate of expectation and so "would have been fifth" should really be "could have been fifth", which is not the same thing.

As for their predictions being flawed, if you plot their predicted pct win chance against actual outcomes for the major leagues over the last five years, it's an impressively tight scatter plot (well, I think it is, at any rate), particularly when you're looking at the bulk of matches in the middle where the favourite is odds-against. And as they concede themselves, there are obvious areas where the modelling comes up short, which they are working to address as it's an ongoing process.

https://projects.fivethirtyeight.com/checking-our-work/

Click to expand...

AlbionInUs said:

I agree, yes. FiveThirtyEight is an American website, and seems to focus on a lot of other sports. In the link you showed me, they mostly analyzed the MLB, not really the Premier League. Either way, I like it as it represents a different perspective, but I feel that for the Premier League, the bookies are more accurate than FiveThirtyEight. For other sports, though, FiveThirtyEight is sometimes even more accurate than the bookies.

Click to expand...

Uptown Dunk said:

Long time reader of NSC but first time posting so please don’t give me stick for not having thousands of posts (yet).

I found this interesting prediction for season end positions:

https://projects.fivethirtyeight.com/soccer-predictions/premier-league/

Albion to end the season in 9th on 52 points. This corresponds closely to bookies predictions. Might be worth keeping an eye on along with staying up tracker.

Click to expand...

Mtoto said:

For the most part, their predictions are very similar. I take a look at the six biggest European leagues every weekend and there's rarely more than 3pc difference between the Betfair price and 538.

Long-term forecasts on 538 don't tend to change as rapidly as the odds can, which I tend to think works in 538's favour - betting markets can often over-react to one or two results. When Liverpool won 7-0 at Palace last December, I think they were cut to around 4-7 for the title and lots of pundits were saying that the title race was as good as over. 538, on the other hand, still had Man City as favourites, at a shade of odds-on (about 52pc chance iirc).

Click to expand...

Machiavelli said:

Thanks for your input on this thread. Is this more than a hobby for you?

Click to expand...