FanPost

On the Importance of Context and Scope When Utilizing Metrics to Tell a Narrative

Or why being clear isn’t all that matters

It was brought to my attention that our site and its viewers would benefit from an article discussing the pros and I suppose cons of utilizing different types of baselines when, among other things, reviewing and/or comparing the metrics for a given player(s). More specifically, the distinctions between "Per Game", "Per 90", and "Per Season".

When you look up metrics, at the club, or team level, you will see either Per Game or Per Season metrics. Per Game normally being an average of the specific metric over the course of the season. A perfect example would be goals scored. Let’s say we are 25 matches into the league season, and Chelsea have scored 72 goals. Fantastic. Now, we can say "Chelsea have scored 72 goals so far in the season", or even "Chelsea have scored 72 goals 25 matches into the season", but this is a tough thing for a lot of folks to instantly comprehend. What does it mean? Is it good? People tend to think either in full seasons, or more commonly, in the scope of a single match. And so we average that total, 72, over the range of matches played, 25. Literally 72 / 25. 2.88 goals per match is the result, and yes, that’s fantastic. So we now have the same metric, but utilized over two different scopes, either a Per Game average, or Per Season (to date).

This might be boring for a lot of you, and that’s fair, because most folks already tend to understand and comprehend this level of statistics when it comes to sports, as we see it all the time. However, when we start looking at players, it’s an entirely different ball game. If every player only ever played 90 minutes, aka, the full match, at a time, metrics would be a lot easier. Of course we’d never see substitutions, and injuries would magically go away, and who would need a bench? But the reality is that players do often play less than 90 minutes at a time. Some start and don’t finish, others don’t start but do finish, the latter being a lot more common than the former.

Often, we see players that tend to substitute a lot as one of three cases:

1) Youth being bled in at the end of runaway matches

2) Aging veterans giving a blow to a tired starter

3) Specialists

4) Future Stars (youth getting serious minutes)

What I care about most, and I think most would agree with me, when talking about metrics, would be that third option, and to a lesser extent that 4th option. It’s fun to speculate on how youth perform, but the actual time spent on the pitch is usually so small, and the situations so polarized, as to become no more than a peak into the eventual reality of a youth player. And the aging veterans, we really only care about enough to make sure that they actually get on to the pitch. But that third group, what I’m calling the "Specialists", they are a hot commodity, and a very important subset of players that are outliers when it comes to looking at metrics. All of this really applies to the 4th group as well, so it’s a similar discussion, but the expectation is that they will become starters at some point so it’s not quite the same. I’ll be sticking to discussing from the perspective of the impact on "Specialists" for this article.

These Specialists often average about 40-60 minutes per appearance because they usually start about as much as they appear as substitutes. A few key names will come to mind, for the type of player I’m referring to:

· Cesc Fabregas

· Alvaro Morata

· James Rodriguez

· Willian

· Olivier Giroud

· Marouane Fellaini

All of these players averaged between 41 and 57 minutes per appearance last season in league play, and had between 11 and 18 starts, and between 9 and 18 substitute appearances. They all likely started at smaller clubs at some point, and are now at larger clubs, where they tend to be rotational players, filling gaps where needed, or providing something different as a substitute, a unique tool in the manager’s box.

The reason I’m going into such detail here, is because when we use the metric scopes we talked about at the club level, Per Game, and Per Season, these types of players tend to be shadowed out, and their true impact on the pitch not nearly as highlighted and signified as it should be. I’ll use Cesc for my examples, because that’s where it came up previously, and he’s such an obviously excellent example for the discussion.

When you look at Cesc at a Per Season level, his numbers look rather pedestrian and subpar.

Cesc
5 Goals Per Season
12 Assists Per Season
62 Key Passes Per Season

Honestly, those last two metrics actually look pretty good. But then you say, hey what about that De Bruyne kid, he put up good numbers too!

KDB
6 Goals Per Season
18 Assists Per Season
103 Key Passes Per Season

He’s OBVIOUSLY better than Cesc. Well, one thing is for certain, De Bruyne had 36 appearances to Cesc’s 29 appearances over the course of the full season of league play. Now, you might say to yourself, "yeah, that means he’s obviously more durable, and delivered week in and week out", and both of those statements are in fact accurate. But we might question whether or not those extra appearances are skewing our assessment of who had the better impact for their club in those metrics.

So then we might utilize a Per Game level perspective instead. Surely that answers the question, right? That should get to the bottom of things, and averaging things out always makes things better. So would could say:

Cesc
0.17 Goals Per Game
0.41 Assists Per Game
2.14 Key Passes Per Game

KDB
0.17 Goals Per Game
0.5 Assists Per Game
2.86 Key Passes Per Game

And you then tell yourself, see, De Bruyne still had the better season, based on Per Game numbers! I knew it! That Cesc can jog on, I want KDBae on my roster! However, if what we were doing was looking for a starter, to play week in and week out, I would agree that at least considering Per Game scope has some merit, what we REALLY should be looking at, if we truly want to assess who was more productive on the pitch, is what’s called a Per 90 metric. Because in the above example, De Bruyne played 2885 minutes in the league last season, whereas Fabregas played only 1328 minutes, or less than half of De Bruyne’s total.

When we say "Per 90", what this means is that we take their full body of work, ALL of the minutes, and then divide those minutes up into 90 minute chunks. And instead of using appearances as the number to divide those full season numbers by, we divide by the number of 90 minute chunks instead. The goal here is to attempt to level the playing field for all players. And THIS is why it’s so important and relevant for those "Specialist" type players, that often come in to be game changes as a substitute. We don’t care about their performance per appearance usually, especially when comparing them to other players. We care about comparisons made at the most equal and level method possible. Again, if the goal is to evaluate starting players, while this is still an incredibly beneficial way to evaluate, it is still important to keep in mind the total number of minutes in a season a player has played, as well as their average per appearance.

If we break down the numbers for these three albeit simple metrics for Fabregas and De Bruyne at the Per 90 level, they would now look like this:

Cesc (14.76 90 minute chunks)
0.3 Goals Per 90
0.8 Assists Per 90
4.2 Key Passes Per 90

KDB (32.06 90 minute chunks)
0.2 Goals Per 90
0.6 Assists Per 90
3.2 Key Passes Per 90


At this point you can see that on a Per 90 scope, Fabregas outperforms De Bruyne on each of these metrics. This scope isn’t often used for things like goals and assists, because they are so infrequent, but where it often comes into play is when talking about things like passes, and tackles, and dribbles, things that happen numerous times throughout a match, but can often be very spiky when it comes to frequency. Using the Per 90 scope really helps smooth out outliers, and provide that level baseline for tracking these numbers.

Personally, I find very little use for the Per Game metric when talking about individual players. I leave it for team level analysis. I prefer to either look at season totals, and provide a context of minutes played, or more often use a Per 90 metric, and ensure that not only am I clear about what I’m tracking, I’m also making it clear to the reader both what they are looking at and why.

My goal is always to provide clear and easy to understand and interpret metrics in my analyses whenever possible, so that the broadest audience can not only appreciate what they are reading, but feel more educated about something in the process, too. And possibly share that knowledge with others, passing along what they’ve seen and read. So there’s a big integrity piece as well, when it comes to providing metrics and analysis to readers. I’m the source of that information, so if I’m misleading, or unclear, and then that message gets incorrectly transferred on, everyone suffers. And that’s the LAST thing I ever want to happen. It makes me look bad, it makes the reader who transferred the information on look bad, and it makes everyone who tries so hard to use metrics in a sport where most don’t care about metrics have a more difficult challenge in breaking down that barrier.

Hopefully this helped explain both the reasoning behind using the Per 90 metric as well as the benefits of doing so. As always, questions and feedback are appreciated.

This FanPost was contributed by a member of the community and was not subject to any sort of approval process. It does not necessarily reflect the opinions held by the editors of this site.