If you were to wake up in the morning and check the box-score of a game you missed the night before, what would you be trying to figure out? What would you be looking for? I’d bet you’d ask yourself questions like “how did they do last night?” or “oh, did they stay hot?” or even “I bet my friend they’d crack in crunch time, what happened?”. Whether you realize it or not, at that moment, you’re looking at sports metrics and using them as evidence to make judgements in two ways: you’re balancing how much the results were a function of performance and how much the results are a measure of ability.
Those two opposing ideas are the two most fundamental concepts in sports analytics. The idea of a metric being a function of performance refers to the fact that all sports data are the result of some action or set of actions by a player. A hockey player scores a goal 50 times, which means their Total Goals data point equals 50. The metric value is the output of a given action. The idea of a metric being a measure of ability refers to how well it is a representation of how talented a player truly is at completing some action. So for that same hockey player, that 50-goal data point is, to some unknown degree, a measure of how talented they are at scoring goals.
To start exploring exactly what those mean in relation to each other, let’s think about an example. Say there is a basketball game played and Player X scores 50 points, paving the way for their team to win in a blowout victory. Is this single 50-point outburst enough to convince you that this is an elite scorer? Probably not. There is no way for you to know whether this was just a really good night for Player X or if it was par for the course for them. You’d probably need some more information, right? But on the other hand, can you confidently say that this was a very good scoring night for that player? Absolutely. So, in this scenario, you’d be more confident that the 50-point mark is a function of performance than a measure of ability.
Fast-forward one month and Player X has averaged 50 points for the entire month of January. Are we now ready to say with some confidence that this data point is not only a function of performance, but is also a decent measure of scoring ability? Since this wasn’t an isolated outburst but a trend of elite scoring? I feel like most of us would say so.
But these changes don’t only happen over time. There will often be metrics that are collected over the same period of time that relate to the same base metric but are indicative of one concept substantially more than the other.
Say there are two hockey teams – Team A and Team B – that both play 20 games. In that time, Team A went on the power play 40 times but Team B only went on the power play 20 times. Of the 40 times Team A went on the power play, they scored 20 times. Of the 20 times Team B went on the power play, they scored 15 times. If your goal was to find out which of the two teams is better at scoring on the power play, would you want to know that Team A had 5 more power play goals than Team B? Or that Team B scored on 75% of their power plays while Team A scored on only 50% of their power plays?
When you think about that that question, you’re balancing the ideas that we’ve been discussing here. The fact Team A scored 5 more goals in that time is very clearly moreso a function of performance when making this comparison because they had 20 more opportunities to score those goals. The fact that Team B scored on 75% instead of 50% of their power plays is also a function of performance, but it is also much more effective as a measure of ability because it offers more insight into either team’s ability to convert on their chances.
Every metric in sports analytics will be a function of performance and a measure of ability to some degree. The amount a metric should be either of those is completely dependent on the question at hand, as is shown in the included examples. And these concepts will keep coming back in future chapters as we discover how they tie into more in-depth analysis.