Ornaments to the game

It’s surprising to pick up the Age and find out, twice in a week, that talented AFL players are, in fact, wall ornaments. But, that’s what Matthew Lloyd would have us believe about Brett Deledio, and Matt Murnane about Jeff Garlett: they’re barometers.

I’m sure they mean well. Something like, “when Deledio’s playing well, the Tigers are too”, or “when the Blues win, Garlett usually kicks a bag”.

Well, which is it? Do the teams influence the players, or vice-versa? Murnane has a go at answering this:

‘It is an interesting thought for Blues supporters to ponder – do Garlett and Betts play well because Carlton are winning, or do the Blues win because Garlett and Betts are playing well? A group of former Carlton greats posed that question this week said the answer probably was somewhere in the middle. It’s the “chicken and the egg”, so to speak.’

So I guess that’s vice-versa-versa-vice. Well done to the group of Carlton greats for working it out for us. What they may be saying is that there is a feedback loop between the teams and players… which, possibly, makes Lloyd’s and Murnane’s articles redundant (potential headline: Players excels in team, team wins… Player even excels more! Team wins by even more! … continue until you run out of exclamation marks).

What about other leads? Researching barometers, sadly, is a dead end. Barometers measure atmospheric pressure, but have no impact on it; if the metaphor holds, when the Tigers are crushing their opponents, Deledio should be sitting on the sidelines taking careful notes.

Science would suggest that what’s required is a controlled experiment: after Richmond played Fremantle last weekend, they needed to briefly rest (until they felt as energetic as they did at the start of the game), and play again, in identical conditions, replacing Deledio with a player whose skills are demonstrably the precise average of Deledio’s teammates. In which match did Richmond do better? Hmmm.

We could just measure Richmond’s performances with and without Deledio, but that assumes that every match poses an equal challenge – an assumption discussed in earlier posts. Leaving that problem aside, sadly for us (but happily for Deledio) he hasn’t missed a game all season. Well, can we at least demonstrate that his performances have been different in the wins and losses? Lloyd provided a random grab bag of stats in which Deledio had a better average in wins than losses – inside 50s, scores and uncontested possessions (to give a bit more substance to this, I’ve added rebound 50s and contested possessions). However, with only 16 matches played this season, the sample sizes are small. If we accumulate all of Deledio’s stats into one big lump measuring “performance”, and assume a normal distribution, we can calculate two confidence intervals – one for Tigers wins, one for losses.

So, perhaps that will show us whether Deledio performs differently in wins than losses. But what if his teammates improve, too? How can you be the team’s (sigh) “barometer” if you’re no different to anyone else? So, for good measure, I’ve repeated the exercise for Deledio’s teammate Daniel Jackson.

It will come as no surprise to readers that in Richmond wins, Richmond performs better than in Richmond losses (you can look it up, it’s true), even by this potted measure of performance. Richmond’s overall performance score is about 9% higher in losses; therefore to see if there is anything remarkable in Deledio or Jackson’s performances in wins, a new metric can be created by reducing their performance scores in wins by that premium. The charts below include the confidence interval, calculated using this formula) – and they’re far from conclusive.

With 95% confidence, we can say that Deledio’s performance score is higher in wins, but there’s overlap once the premium is subtracted – we can’t be sure that the win minus premium score is higher. Meanwhile Jackson’s scores in wins and losses can barely be differentiated. Is Deledio the barometer? Is Jackson the anti-barometer, or some sort of malfunctioning barometer? How many other players have similar or different stats? Why am I trying to ask and answer these questions – isn’t that Lloyd’s job… before publishing?

The Garlett thesis is even more obscure: “since 2010, [Garlett and Betts have] combined for six goals or more in 14 games – and the Blues have won 13 of those matches.”

Checking this would have me trawling through the entire league – Murnane’s job – for combinations of players kicking 6 goals in a match (and also maybe for 5 goals – no-one is above trying a sneaky selective endpoint) and seeing if this is indeed remarkable. As a point of comparison, though, I can reveal here that the combination of Buddy Franklin and me for 6 goals or more has yielded 15 straight wins for Hawthorn since 2008. (He sooo leans on me. Chicken and egg and chicken.)

Frankly though, all of this data is pretty much garbage. If Lloyd and Murnane want to observe that Deledio and Garlett excel when their teams do, they should leave the data out of it. In all of this, the only thing that seems certain is that, before any player is anointed as a wall ornament, at least a few more calculations are needed.

Advertisements

Click here to like, or tell me whether this is genius or gibberish

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s