We go again. And Again. And Again. And Again.
Consistency is the hallmark of the goalscorer.
I wasn’t actually going to post today - I think over the next month or so, I’ll probably do some season review posts, but right now, there feels like there are just too many moving parts, and too many feels to parse what’s happening. The highs of the late winner for Brian Graham last week, followed by Livi’s confident performance on Tuesday, and then the quotes from Graham in an article yesterday that makes it sound like he might be almost done at Thistle, all have added up.
I wasn’t sure I had something worthwhile to say this week.
It’s pretty clear that Thistle fans have been spoiled over the last decade and a bit - two centurian goalscorers in Doolan and Graham following one from another. Anyone who’s sensible realises that we’re unlikely to be able replace those goals all at once. We’re going to need build back up. And unless we’re very lucky, or recruit very well, the chance of having a third in that sequence seems relatively low.
Then Marc Lamberts reposted an article from November that he’d written on striker consistency, and I thought - as an appetiser for the possible end of an era game this evening - I might do a quick post using the same approach.
Striking consistency
I’ve looked in previous posts at shooting efficiency and attacking creativity, as well as the overall team aspects of building an attack.
But one of the big components of racking up the goals is consistency. xG describes the probability of a ‘normal’ striker scoring a particular chance: 0.2 xG roughly means that a normal striker would score this particular chance 1 in 5 times.
xG is not really a metric to use for a single shot. Or even a single game. The aim is to use it over many shots, and then the ‘probability’ kicks in. Averaging xG of 0.2 per shot and taking 20 shots should mean that a player will normally get 4 goals (0.2 * 20 = 4).
However, to judge how likely that is to happen, you can’t just look at the sum of all your xG. You need to know whether all 20 shots are similar.
When I was at school, they introduced a points system to encourage healthy eating. You got 1 point for a burger, 2 for a cheese pizza or fries, and 10 for a salad. At the end of the month, the kid who had the most ‘healthy eating points’ would get a prize - if memory serves, it was a really good prize too (I think maybe some sort of MP3 player - maybe even a first gen iPod shuffle).
Now, maybe those of you who went to a certain calibre of school wouldn’t see the problem here. But after the same kid had won two months in a row whilst putting on around 3 stone a month by having four burgers, three portions of chips, and 2 pizzas a day, and therefore outscoring the kids who ate a single salad, they decided to change the scoring system.
It’s the same with xG. You could rack up 4 xG over 20 shots by one sitter six yards out and nineteen long-range blooters. If you do it that way, you can’t realistically say it’s likely you’ll score 4 goals - maybe 1 or 2, but more likely none.
The strikers likely to perform or outperform at the xG predictions are the ones who have 20 shots that are all close to 0.2 each, not just the ones who manage to take 20 shots that add up to 4 xG.
Consistency matters.
Marc’s approach was to use a number called the ‘Coefficient of Variation’1. The most consistent players are those who have a low CV. The lower the number, the more consistent their shots are - they get an average of 0.2 xG per shot (or whatever) over 20 shots (or whatever) by hitting every shot at the same xG.
This chart shows some of the players at each team who have at least 10 shots. Players are further left if they are more consistent in their shooting, and more red if they have a higher quality average shot.
I’ve cropped out the more variable shooters for now, because otherwise the chart is unreadable. Basically - Brain Graham is very consistent as a shooter. In fact, he’s the second most consistent shooter in the league when compared to players with 40+ shots this year.
George Oakley and Brian Graham are the most consistent in the league of these high-volume brigade, with only Ben Wilson at Airdrie in a comparable tier. For comparison, Robbie Muirhead is 56% more erratic in his shooting quality - he’s scored a decent click of goals this year, but he’s very variable in shot quality. He takes long shots, difficult shots and ‘give it a whack’ shots, as well as poacher shots.
The most compelling striker in the league
When you plot those two features together - the average shot quality and the consistency of that shot quality - there are not many players that escape the pack clustered together in the middle.
Only Barney Stewart and Robbo (at Hamilton, rather than at Thistle) average a higher shot quality (and are also more consistent), but this is from a much lower volume of shots.
Chris Kane takes a similar volume and average shot quality, but he’s less consistent. Similarly Anton Dowds before his injury.
In this data, Oakley’s much further left - his average shot quality is lower - although he’s equally consistent.
The only player who is also high-volume, high-quality, and high-consistency is Ben Wilson at Airdrie. Who’s not normally in the conversation for the top tier strikers in the league (Graham, Muirhead, Oakley, Dowds, maybe Vaughan or Pollock, maybe Kane).
Everyone else at Thistle is more erratic (further up), a lot lower in average shot quality (further left), or much lower volume (smaller bubble). Or two of those. Or all three.
If we look beyond xG (the quality of the chance) and instead at post-shot xG (the quality of the shot, which is also a function of the player’s technical shooting ability), the picture is even starker.
Graham is (almost) out on his own at the right of this chart. Robbo’s out there too, but obviously with a much lower volume of shots. And again, Graham’s consistency is pretty high up there.
All this to say, Brian Graham is an anomaly. Almost no other striker in the league maintains high volume, high quality, and high consistency of shot quality over a season.
That’s why he’s still posting big numbers of goals, even at the long tail of the player age curve.
It seems obvious to say that you can’t replace him with another striker in the league, but based on the underlying numbers, you really can’t. When he goes - hopefully not after tonight, despite the quotes in the press - we need to work harder at finding goals from elsewhere in the team. There just isn’t a player in the league at his level.
Ricco Diack is very consistent in his post-shot xG (better than Graham, actually!), but at a much lower average quality. Ben Wilson is worth a look for a long-term replacement - he’s closer to Brian Graham’s underlying numbers than most other players in the league. But if this is to be the end of the era, then we need a striker who is not only in the right place at the right time … but who is consistent.
For now - let’s be grateful for what we’ve got. A rare and special striker, who has given us many, many memories to cherish. And let’s get behind the team again for another push.
If you care, it’s the standard deviation divided by the mean xG per shot. You can get a low CV by having a big mean xG or a small variation in xG per shot