Is public soccer analytics writing in 2019 too safe?
On the old cowboy days of 'do this and earn 6 extra points in a season.'
As ever, this is a reminder that this month is a free preview ahead of November, at which point this site will go down to one free post a week! So, if you want to get on board now, please:
Alright, on to the show…
Selling the bottom line
This post is kind of a sequel to last week’s post on understanding club workflow. It also came out of a conversation I had with a club analyst last week.
By today’s standards, the early days of the online soccer analytics blogosphere—the brief pre-xG era when people like James Grayson were referencing old Edmonton Oilers hockey analytics blogs on PDO to generate metrics like Total Shots Ratio—seem somewhat primitive.
To me, though, a writer at theScore who was trying to discover more about whether there was such a thing as ‘soccer analytics’, it was an exciting time, when it really felt like there were people thinking about football in a novel way.
Though no one would dare use TSR in polite company anymore, one simple yet effective thing Grayson and others did well was project final points totals based on current shot ratio numbers.
This approach was invigorating because it cut through the statistical science (lol, such that it was) to get to the chase: sure, Man United might be in second now, but they were posting numbers in relation to the rest of the Premier League that would likely put them in a fight for Champions League qualification. You can see it right on the table. And the inference to what they should do better in defence or attack was more or less obvious.
Even better, I think Ted Knutson’s effectiveness as an evangelist for soccer analytics has come in large part because in his meatier “this is what Statsbomb can do for you” posts, he explains the potential of his methods—in this case, better set-piece management—in terms of extra goals per season, and how those goals translate into x number of added points.
Though I think the metrics have largely improved in the past 7 years or so, we’ve lost some of that straightforwardness in public analytics writing.
I think I understand why.
Descriptive vs Normative analytics
For one, no one wants to write checks their butt can’t cash, and as the soccer analytics scene has grown up, analysts are rightly wary of doing anything more than developing a metric, showing a list of whatever that metric spits out (e.g. midfielders who add value, teams that are most effective in the final third etc.), and then vaguely musing on how this metric might be able to help a team do better.
This is to say that analytics seems to be stuck in making interesting descriptive claims about football (here is how to tell which team/player is good or bad), and shying away from the old cowboy “we know better than you” spirit of normative claims (if you do this, you can score x more goals a year and earn x more points in the table).
In normative analytics, the end goal isn’t the metric per se, it’s the concrete thing a team can do to get better that doesn’t involve spending more money on players.
For a completely fictional but at least plausible example, normative analytics would tell you that, if teams took throw-ins ‘this way,’ they’re likelier to retain possession, increasing the likelihood they will score x more goals in the season, translating into x number of extra points.
Normative analytics is exciting. It’s provocative. It gets to the heart of what makes the scene fun, competitive. If you’re a View 2 person, it’s part of what makes football itself fun to watch.
Of course, normative analytics comes with some fairly significant drawbacks.
The biggest one is that normative analytics requires analysts to toe a fine line—making their case publicly but also not giving clubs valuable info for free. This invariably leads to holding things back, which can undermine their case.
This paradox was a big deal in the soccer analytics scene for a while, but then kind of…went away, I think in part because we all sort of settled on soccer analytics to mean “let’s use numbers to better understand football” instead of “let’s use numbers to find ways teams could win more football matches.”
Give away the good stuff for FREE
Years ago, when this debate on what analysts should and should not give away to the general public, I made this same case, so I’ll make it again today.
I think analysts should give away their good stuff for free, that is, write freely and openly on normative analytics of the “do this and earn more points during the season for no added cost” variety.
Why?
Because once one club knows your secret spice, every club will know your secret spice.
Also, you can’t make a living as an analyst on the back of one or two neat things a club can do on the pitch to score a few extra goals.
If you want to provide value to a football club, you need to bring that entire mentality to every single aspect of a club’s operation, going down to the most granular level—everything from how a team takes free kicks to which hotels they stay in on the road, what the pre-match warmup is like, how loads are managed among players, which factors to calculate in transfer fees etc. etc.
But by writing exciting, normative pieces derived from the excellent descriptive work everyone has settled on these days, you make the case for analytics clearly, you generate interest in the field, you show that clubs still have a long way to go to harvest low-hanging fruit—you raise the tide that lifts all boats, including your own.
Just my two cents.