Sunday, December 16, 2012

Looking back before we look forward part II

Following on from yesterday's post comparing expected points to date (calculated as of today) with actual points, we see some players who have either significantly our or underperformed their underlying stats. We looked at Michu as the classic example of a player whose production looks a touch unsustainable, and below I have highlighted a couple more players to try and highlight some of the tools we can use to suggest a change in a player's future success.

Gareth Bale
Bale has enjoyed an excellent season by any measure to date, but there's strong reason to suggest it could have been even better. His six goals from 25 shots on target (SoT) is below the average conversion rate we'd expect (28% for the average midfielder, 33% for forwards, who might better compare with Bale given his goalscoring credentials). While that rate could regress there's not much downside in his data to suggest an accompanying adjustment elsewhere, though his on target rate of 47% could be argued to be a touch high given his farly high rate of shots outside the box (SoB). Furthermore, just two of Bale's 32 chances created have been converted to goals (6%) which is significantly below the average rate for TOT players of 15% and the league average of  14%.

Eden Hazard
Hazard is an interesting case as he sees factors pulling both ways, which in the linked chart essentially cancel out to make his xP very close to his actual points. On the goalscoring side, his two goals (excluding penalties) from 10 SoT suggests he's been a touch unlucky and with three total shots in three of his past five games, I'd feel okay backing him to top his three goal haul during the second half of the year. However, his ten assists look way too high, given his 32 chances created (32% conversion rate), when you consider that as a whole Chelsea are converted chances at a 17% rate (which is second in the league to only MUN). So while we might forecast a goal or two more, we're likely to see a comparable decrease in assist points and thus I'd suggest any upside Hazard may have is already baked into his price tag and so it will be tough to squeeze much more value out of that particular stone.

Marouane Fellaini
I've been suspicious of Michu's ability to continue his success so in fairness I also need to touch on Fellaini who I've been quick to praise. Fellaini's eight goals from 20 shots looks a touch high and in a vacuum we would expect some regression on that rate. However, one compensating factor we don't see from Michu (and which isn't currently accounted for in the model) is Fellaini's ability to hit the target. Of his 51 shots to date, 42 have been from inside the box (82%), yet he's only hit the target 20 times (39%). That isn't a bad rate but is materially below Michu's 44%, despite the fact that only 73% of Michu's efforts have come inside the box. So while we wouldn't expect another eight goals from Fellaini, there are some circumstances which might suggest his drop off could be less steep than some of the other regression candidates.

Two further complications though are his assist potential and his pending suspension. With four assists on 23 chances created (17%), he's outperformed the average Everton rate (12%) which is particularly concerning considering the Everton rate includes Fellaini's own good conversion rate. Again, it looks like most of Fellaini's upside is built into his price (and ownership numbers) and thus alternative options should at least be considered. I would have considered that a non-pressing issue but with a possible suspension upcoming a decision is going to be thrust upon us. Even if you aren't keen to sell, thousands of others will and so any profit you've built up could be wiped out. The fixtures through GW25 remain solid so it would be a shame to sell now, but three weeks is a long time to miss an 8.0m asset. More on this to come in the coming days.

Hopefully the above brief descriptions describe the kind of thought process that goes into identifying regression candidates and how we will use the data to forecast projections for the remainder of the year. I will soon post the results of the model for the last eight weeks and will again identify players whose attributes help explain some of the successes/failures/areas for improvement.

1 comment:

vanilla said...

Re: Fellaini and a possible 3 game ban. remember that there is a January wildcard to use. Many could sell him, take the profit, let his value fall, and buy him back without taking much of a loss (or any) when wildcard is activated.
I still like Fellaini, though. Even with the high ownership, it's difficult to pass up the consistent shots/chances. I'd have a hard time justifying a team free of he or Michu.
There is too much value in these two for my 8 million, differentiation be damned, when you consider that consistency in defense (like the Bosingwas, Vidics, Lescotts of old) is nowhere to be found, a massive midfield is the only way to maximize points, which means you likely own the alternatives to Michu/Fellaini already (Bale/Cazorla/Hazard/Mata/Silva).

Great blog.