Stats from the stone age (12 years ago)

For those of you who don’t know, I’ve been faffing about with football/soccer statistics for an actual 20 years now. Originally it was a smattering of player stuff after I read Moneyball. And then it was a lot of gambling stuff. And then it was a LOT of player stuff (StatsBomb the Blog/Brentford/StatsBomb the Company), and now it’s a mix of all of the above but far more casual.

I have at various points in my journey been one of the smartest people in the room about one topic, and one of the dumbest people in the room about everything else.

We live, and we learn, and if you stay determined you can make real progress.

Today’s a little reminder that all the cool shit you see out there today started somewhere very different.

Want to see something pretty innovative from 2013*?

*not an exaggeration

Media and pundits only wanted to discuss goals, scorelines, and pashun. Mainstream journos hated assists. Pretty much no one outside of Arsenal fans cared about the stat, and that was only because Thierry Henry held the all-time PL record for assists in a season.

I was fairly certain key passes and assists were crucial to evaluating attacking play and attacking players, but there was zero research on it.

How about this, also from 2013?

What was innovative? Using stats and data in the public scouting of players at all. It was done here and there, but just a smattering and people weren’t really looking at the breadth of what players in different positions would/could do at the time.

Then you got me, saying “Have an avalanche of numbers, abbreviations, and tables.” 🤢 

Another crucial element was attaching an estimated price to the value of the player. That was basically unheard of, but felt totally natural to me due to my background in sports betting.

Note: Ben Yedder would move to Sevilla three years later for 9.5M Euros.

It turns out I was pretty good at pricing players for the transfer market — a trait that continues to serve me well more than a decade later.

It was not all sunshine and rainbows: I said a lot of ignorant, poorly nuanced things in 2013. 

And I said them publicly.

That’s what happens when you are learning!

Going from dumb to relatively smart is a process. It’s uncomfortable. Sometimes you write your little blog, make your little predictions, and then have to listen to Liverpool fans (and Director of Research Ian Graham) bashing you for years because you were still searching for the right frameworks to make better predictions.

If you want to change the world, you’re going to need a thick skin.

You cannot be afraid of looking dumb or you will never progress.

…Just try not to do it that often.

Thankfully, I also said a lot of fairly intelligent things that were precursors to changing how players are analysed and how football is played, both directly and indirectly.

They (they being Evil Luke Bornn) even invited me to speak at conferences with real stats people in a talk I called, “Data Science with Your Hair on Fire”.

In case you have more time to kill and want my version of history, I did some navel-gazing on my early work in 2016.

How about something completely cutting edge from the start of 2014?

What was innovative? Lots of things, but using data to visualise and evaluate player stats, including important things players do by position was almost unheard in football. (I suspect StatDNA had been doing it for years then, but that’s a story for Sarah Rudd to tell.)

That first radar from January 2014 is terrible. But without it, none of the other versions of radars would exist. And who knows how long it would have taken pizza charts and all the other variations we currently see as alternatives to hit the public ecosystem?

Radars got a lot better over the years and are still evolving to this day by adding new models and metrics teams care about for specific player archetypes. When I came into Brentford/Smartodds in 2014, I wrote a primer for both the football teams and gambling traders explaining the visualisations.

If you’re interested, I’ve personally written tens of thousands of words about radars and design, most of which are linked here. They gradually (and unexpectedly?) leaked into the coaching spaces as more and more coaches requested them from the analytics groups in teams like Liverpool and Arsenal to help them better evaluate their academy players and future prospects.

Quants typically hate the vis format, point out the reasons why they can be bad that I largely solved during the initial design phase, and swore never to use them. It is 1000% Good Will Hunting meme for me at this point.

I told quant people at the time, “Radars are not for you — they are for everyone else.”

And then I laughed gleefully as years later, quants told me how they had been forced to spend time and money programming radar generation inside their own team ecosystems, because the football and ownership people around them had overruled their good sense. Scoreboard, motherfuckers!

I also introduced radars to American Football before I sold StatsBomb to Hudl, and we not only produced performance versions of those for every position, we also added physical data derived from frame-by-frame tracking data as an additional template.

The irony to all of this is you can see from my earliest work that I was perfectly content working in the database. Give me tables of numbers and SQL statements, and I was totally happy.

Thankfully I realised somewhere along the way that making pictures out of this stuff might be helpful in popularising the use of data in football. So if you hate them in all the places they now appear across sports, I am probably the person in the world most deserving of your blame.

How about something truly insane in this trip down memory lane?

I forgot I had written this until literally this morning. It’s there on my personal blog from before I created StatsBomb the website, and I had zero recollection of it existing until just now.

But it is the crystalised genesis of what would eventually become StatsBomb Data, and the company that changed how everyone in men’s and women’s football scouts and recruits players.

In my memory, 2013 does not feel like an uncivilised time in the world of data analytics. Looking back at actual posts from that time, almost none of what we take for granted in football these days actually existed. Not the data vis, not the stats we prefer to use, not the advanced models… not even the basic tactical knowledge that’s now widespread.

Back then we were all just stumbling along blindly, trying new shit, and learning as we went.

Twelve years from now, perhaps everything we see and use in football analytics will have changed again. I’ve probably done 90% of what I’m going to accomplish in life pushing the discipline and tools forward, and most of my future work will potentially remain private when I go back to work inside football teams.

The question at that point becomes: Which of you is next up to help change the world?

—TK

If you enjoyed this newsletter, we’d appreciate it if you would forward it to a friend. If you’re that friend, welcome! You can subscribe to The Transfer Flow here. We also have a podcast where we go in depth on transfer news and rumours every week. We’re on YouTube here, and you can subscribe on Apple Podcasts or Spotify by searching for “The Transfer Flow Podcast.”