- Market value
- Percentage of non-injury/suspended minutes used for previous club
- Relative quality of previous club
- Rate of use for new players at current club.
I had hoped for this week to have applied the rating to all 20 Premier League club transfers ahead of the the 2015-16 season, but time constraints meant I was only able to use it on the top 7 clubs.
Nevertheless, the results so far are roughly in line with what I had anticipated. Though I am loathe to use any kind of linear regression analysis as I don’t really have the expertise and it’s not really a good idea until all the data is in, the results are noisy but still statistically significant with an R-squared of .55. With some assistance, I will try to post the regression next week for all 20 clubs.
Some brief changes from last week: I changed the rating from a 1-3 to a 0-2 to scale, for obvious reasons, and to measure relative quality I used Lars Schieffer’s Clubelo.com.
It’s important to reiterate, the point here is not to produce a working model for scouts, but simply to demonstrate how a simple algorithm anyone could design can outperform completely subjective impressions about which new transfer is more likely get more playing time.
Moreover, a poor rating isn’t necessarily indicative of a bad transfer but rather the likelihood they’ll get a lot of playing minutes right away—many transfers are long term bets.
So how’d we do?
Where the Model is Doing Well
Overall, the results were fairly intuitive. If a team paid relatively little money for either a very old or a very young player from a lower league club where they only played about 50% of available minutes, it shouldn’t come as a big surprise if they don’t feature with their new first team very often. However, there have been a few other interesting findings.
Though he was much hyped as a major addition to Man United, the algorithm gave Memphis Depay a 5/10 rating, based on his age and the relative quality of PSV Eindhoven compared to Manchester United—Depay played 43% of available minutes last season. Similarly, it gave a 6/10 to Raheem Sterling based on his young age and the relative quality of Liverpool compared to City in Elo rankings. Sterling would go on to play 61% of available minutes for Man City.
Finally, though Gokhan Inler was apparently a “key signing” for Leicester City ahead of the 2015-2016 title winning season, the rating gave him a 4/10, mostly for his age and lack of playing minutes the year before with Napoli—Inler would only play 6% of available minutes. Contrast that result to Okazaki (8/10, 63%) and Kante (7/10, 90%).
The rating also seems to like some teams more than others—Man City’s players neatly fit expectations, with De Bruyne (8/10, 85% of available mins), Otamendi (9/10, 87%), and Delph (5/10, 41%) all lining up nicely. West Ham’s results were also quite close to expectation:
Where It is Not Doing Well
Centre-backs. Hoo boy. Van Dijk at Southampton (6/10, 100%), Huth at Leicester City (4/10, 100%) and Toby Alderweireld (6/10, 100%) were major outliers. There are some possible reasons; ‘peak age’ for this category seems generally older, so the scoring here breaks down. Furthermore, there doesn’t seem to be as much competition for these starting positions, at least in the sample of the teams I’ve looked at so far. Finally, teams like consistent CB pairings, so if you get five games in a row, you’re more likely to get 33 more.
There were also a few duds in other categories; the rating didn’t like Martial’s age, but it didn’t seem to stop Van Gaal from wisely putting his trust in him—he earned a 5/10 rating but played 94% of total available minutes. Similarly, the rating (and me, personally) liked Son Heung-min more than Tottenham—he earned a 6/10 but only played 41% of available minutes. There were at least two other misfires along similar lines.
In all I’m quite happy with this so far, though it will be interesting to see how it fares with the other 13 teams, particularly as we get toward the bottom of the table.
As of right now, would I use this model to scout footballers? No.
Would I, as a football fan in general, use it as a very simple way to gauge in a few seconds how a player might fare at their new club next season? Probably.
With transfers, sometimes it’s best to just look at the most simple factors before looking into any ‘advanced stats.’ Are they young or old? Do they play for a good team? Did their team use them a lot last season? What is their going market rate? Does my team suck at recruiting ready-made stars?
After that, you can drill down further. The aim here is simply to demonstrate that basically anyone can do this kind of assessment—it takes less than a minute wheeling around on transfermarkt.co.uk. A little systematic thinking can go a long(er) way. It can also reveal how much further you can go with some hired expertise to help out.