The Happy Accident of Shot Dominance

You know those times where you hear about the concept of something cool and you get so excited that you run off and start doing it, even though you haven’t heard all of the details yet? Then, when you find out how you are supposed to be doing things, you kinda like your way better? This is one of those times.

Playing around with TSR. Or not?
Last November, I was diagnosed with testicular cancer. The treatment was surgery followed by one giant bout of chemo. It’s not a fun process, but I came out the end just fine. Anyway, about the time of the surgery, I needed something to keep me busy, so I started reading up on the current state of football analytics.

One of the first things I ran across was a stat called TSR or Total Shots Ratio. “Cool,” I thought, “I’ll just toss that in a spreadsheet and look at the Premier League that way.” This is what I produced.

Team	ShotpG	SOTpG	ShCpG	TSR
Tottenham	17.9	6.3	9.7	1.85
Liverpool	19.4	6.1	11.4	1.70
Manchester City	17.4	6.1	10.4	1.67
Arsenal	15.7	5.4	10.6	1.48
Everton	16.7	5.4	11.9	1.40
Chelsea	16.5	5.6	12.2	1.35
Manchester United	14.8	5.6	12.9	1.15
Southampton	13.6	4.4	11.9	1.14
Newcastle United	14	4.6	13.4	1.04
Wigan	13.2	4.5	14.1	0.94
Swansea	13.3	4.4	14.9	0.89
West Bromwich Albion	13.4	4.5	15.7	0.85
Queens Park Rangers	13.2	3.8	15.9	0.83
West Ham	12.9	4.2	16.5	0.78
Norwich	10.9	3.5	14.1	0.77
Fulham	12.1	4.2	16.8	0.72
Stoke	10.2	3	14.4	0.71
Aston Villa	11.5	3.7	16.8	0.68
Sunderland	11	3.7	16.8	0.65
Reading	10.4	3.4	18.6	0.56

Despite what the label says, that is not TSR. TSR is something else. In TSR, all of the numbers are below 1. Obviously, not all of these are.

Too dumb to realize I was doing it wrong, I kept on donking around with these numbers, testing out some simple data transforms based on Shots on Target, home field advantage, etc.

About two months later, I realized that my numbers looked totally different to everyone else’s.

Awkward…

I blame the pain killers.

Not better, just different
Once I realized my error, I started to play with correct TSR numbers, and realized that I didn’t like TSR quite as much as my screwup. My accidental metric was more intuitive to me and required less work to do basic data transforms. I introduced it to Ben Pugsley and he seemed to like it too. The other thing that I like is that it shows team differentials much more clearly than the ratio numbers. It’s not like I did all this work and came up with something amazing – I didn’t know what I was doing, but ended up with a metric that seemed a little easier to immediately grok and work with.

Here, look at it this way. Bayern Munich finished the season with a shot dominance of 2.09, meaning they took 2.09 times more shots than their average opposition. That number was far and away the best in Europe this season.  Reading’s shot dominance number was .56. They took about 56% as many shots as they gave up to average opposition, and that number was the worst in Europe (though Pescara were close).

So against an average team, you expect Bayern to utterly dominate and you’d expect Reading to generally get pasted. No team in Germany was nearly as bad as Reading.

Bayern vs. Reading

Pitting Bayern against Reading? You expect Lennox Lewis vs. a twelve-year-old.

Simples, right?

Needing a name that is not TSR or even Total Shot Percentage, I guess we’ll just attach the “shot dominance” tag that’s been floating around and see if it sticks.

Anyway, the reason I am even wasting your time writing about this is because Ben wanted to start using it in his work but was kind enough to let me reveal it first. Now it’s revealed, Ben and everyone else can use it whenever they like, and we can go back to talking about interesting topics like Schrodinger’s Goals.

[Note: @WillTGM wrote about this a year before me, so feel free to give him credit, etc.]