Simon
banner
analytic.football
Simon
@analytic.football
Stat nerdery, #lufc and miscellany

Football team ratings model: analytic.football

I chat about FPL @analyticf.pl
As you do fewer things at outlier level it gets harder to justify needing to carry a defensive deadweight
November 30, 2025 at 11:53 PM
So how does is impact the ratings? In general, it sees the division with roughly the same shape, though it does moderate positions on Palace, Brentford, Brighton, Bournemouth & Sunderland towards bookie/other model consensus
November 18, 2025 at 4:05 PM
About to push some decent upgrades to the model at analytic.football

Big change is model now uses recent average errors to move future ratings - so less fluctuation in the ratings but still moving as quickly when it is consistently wrong

Result is less bouncy ratings (new ones on the left)
November 18, 2025 at 4:05 PM
Judging from Mandelson, Rayner, this means he should be in his last 48 hours in the job
November 13, 2025 at 6:35 PM
Folks what are we doing here this is completely meaningless
November 12, 2025 at 3:13 PM
Big notes of caution on reading Okafor's FBRef bars, very few minutes and a bunch as striker, with massive sub effects.

Taking out the striker mins, and accounting for sub effects. Last 2 seasons

Moves the ball into the box a bunch, nice outlet carrier & dribbler, hope he can add more shots
August 21, 2025 at 7:23 PM
Model output and projections for the new PL season.

***Model has no off-season info, so is rating teams on their May 2025 level, it will adjust pretty quick as we get some matches played***

I expect Brentford, Wolves to be worse than model suggests
I expect Tottenham to be better
August 14, 2025 at 3:15 PM
All the football front offices conspiring to make Luke Edwards look like a right tit is so funny.

Seven minutes between these
July 25, 2025 at 9:50 AM
I don't think #lufc are signing a 10 this window.

Longstaff in 23/24 was v different to last season. Between this, Diarra/Sadiki on the list it's imo pretty clear we are looking for pressy box-crasher from that 3rd midfielder, not a creative passer (though the other 2 can ball carry for days too)
July 8, 2025 at 10:27 AM
Nah don't do this #lufc
July 7, 2025 at 10:25 PM
2nd Scottish referendum inevitable with something like this I think
July 7, 2025 at 12:51 PM
May 21, 2025 at 9:46 PM
I dunno, signing off with "far more likely" seems a lot more than 60% to me!, But YMMV
April 25, 2025 at 1:25 PM
Yeah model thinks Burnley have got the rub of the green in a big way, especially when your attack was complete shite first half of the season. Leeds probably best numbers in 10 years, and the worst chasing pack by the model, bookies similar I think
April 15, 2025 at 9:08 AM
We are close, really close, but 99% is not 100%
April 14, 2025 at 11:21 AM
2 more wins would be almost certain yeah, quite likely we could get away with less but I'd prefer not to
April 14, 2025 at 10:14 AM
2nd largest of Burnley, SHU & Sunderland lands here (so assuming better GD, this is what we need to be top 2) - cumulative right column (%)
April 9, 2025 at 9:26 AM
Massive massive massive win for us tonight, model has always been keen but we dropped to 87% before this evening

With Burnley & Sheff Utd to play each other, relative strengths of schedule, and model's view on the relative strengths of the teams it is VERY keen
April 8, 2025 at 9:58 PM
Current state of world trade
April 3, 2025 at 12:53 PM
Cannot get my head around how deranged it is to divide the trade deficit by imports and conclude/say that is the tariff rate charged to the US
April 3, 2025 at 12:49 PM
Yeah I have the gap roughty ~half what it is in reality. Arsenal were hanging much closer earlier in the season but the attack has fallen off quite a bit with Saka out - model has downgraded attack rating in every game since he went off injured
March 23, 2025 at 4:55 PM
Don't know if this will bum you out more or less, but I have a decent gap between the two, even with Liverpool getting lasrge positive variance
March 23, 2025 at 4:42 PM
Improvements to viz throughout - they should now pretty much all provide more detail on hover. Also added little bits of explanation in the relevant places.

There should be far fewer reload-triggering Streamlit clicks also, so it should be a lot snappier to get around using the tabs.
March 21, 2025 at 7:30 PM
Similar story for expected points, where you can now see the absolute difference, alongside the previous scatter diagram. Upcoming matches also got a glow up
March 21, 2025 at 7:30 PM
New views allow you to see relevant information grouped together - e.g. for current ratings:
March 21, 2025 at 7:30 PM