One of the things I kind of can’t stand during Silly Season is the lack of good websites for evaluating transfer acquisitions. That sentence probably sounds awfully stupid considering we basically know nothing about what to look at in order to make accurate, predictive statements about player prospects.
But every time I read news of the latest transfer—in this case, Schalke’s Lewis Holtby going to Spurs—I think back on Adam Bate’s interview in November with Hamburg’s head of technical scouting, Steven Houston. Houston made brief mention of some of the little things scouts look at in making intelligent assessments of various transfer targets:
“You also have to factor in the team and you have to factor in the league itself. That’s where you really are starting to do data analysis and crunching data. I think as opposed to saying he made X number of passes per game you start to look at how many of his team’s passes did he make. After all, a player at Barcelona is always going to make a greater volume of passes than any player at a team that makes fewer passes.
“There are lots of examples. In Holland there seem to be more goals scored so very often you find a lot of strikers scoring over 20 goals per season. How does that translate into other leagues? You can look at historical trends for players who made that transition. That’s what a lot of teams are trying to find out – how you can effectively make that translation.
“You need to ask how many times we have seen them and have we seen them against top-quality opposition. Injury information is also very important. We took a player on a free transfer called Rene Adler who is now back in the Germany squad. He had a knee injury and he was out for a year but he was a fantastic goalkeeper before that so there are lots of factors to consider.”
All of these factors are intuitive and would require fairly basic computation for the amateur analytics person. You might be thinking, “Well what about whoscored.com?”
Who Scored? is great, but many of the measurements available on the site don’t translate into anything particularly meaningful. For one, player metrics only cover the most current season, which doesn’t afford much of a reasonable sample size to make any definite statements one way or another. Moreover, it’s crucial stat—key passes—is entirely subjective. You can make some reasonable conclusions based on a few things here and there, but it’s all pretty much through a glass darkly.
Then there is a site like the Castrol Index which is moderately better, but still disguises its methodology and doesn’t make any attempt at averaging, preferring instead to chart out points on a game-by-game basis which includes games the player did not start. Here’s Holtby compared to Moussa Dembele, for shits and giggles:
So this is what I’d like: an experienced sports statistics person to pair up with a skilled web designer. They could set up a fairly minimalist player database that includes information on average player actions per minute, with some sort of arbitrary weighting based on the club/league etc. Total percentage of total team passes attempted maybe, percentage of total team interceptions. Some of it might be bullshit, some not.
The idea though would be to move closer to the idea of isolating average player actions, or percentage of player actions out of the total team, weighted for league and club, over a career-long sample size.