Beautiful spider loses its way 2

A double post today.

In the previous post, I talked about NFL.com's visualization of football player statistics. In this post, I offer a few different views of the same data.

***

The first is a dot plot arranged in small multiples.

Notice that I have indiced every metric against the league average. This is shown in the first panel. I use a red dot to warn readers that the direction of this metric is opposite to the others (left of center is a good thing!)

You can immediately make a bunch of observations:

• Alex Smith was quite poor, except for interceptions.
• Colin Kaepernick had similar passing statistics as Smith. His only advantage over Smith was the rushing.
• Joe Flacco, as we noted before, is as average as it goes (except for rushing yards).
• Tyrrod Taylor is here to remind us that we have to be careful about backup players being included in the same analysis.

***

The second version is a heatmap.

This takes inspiration from the fact that any serious reader of the spider chart will be reading the eight spokes (dimensions) separately. Why not plot these neatly in columns and use color to help us find the best and worst?

Imagine this to be a large table with as many rows as there are quarterbacks. You will able to locate the red (hot) zones quickly. You can also scan across a row to understand that player's performance relative to the average, on every metric.

I like this visualization best, primarily because it scales beautifully.

***

The final version is a profile chart, or sometimes called a parallel coordinates plot. While I am an advocate of profile charts, they really only work when you have a small number of things to compare.

 

 

 

 

Comments

[Jon]

I think the heat map display is the most promising, as its grid layout more easily allows simultaneous comparison among players or among indices. The other two formats make you choose to favor one comparison or the other, which is probably useful if you need to make an editorial point (ie "so-and-so is the best"), but less so if you want to allow easy digestion of all the data.

Separating the two rushing stats by a few pixels might clarify that those are qualitatively different dimensions of performance.

At the risk of complication, the grid dimensions could be scaled by importance, so that columns that are more important (passing yards) get more ink than less important factors (rushing touchdowns), and rows with players with more playing time get more ink than players with less.

It might also be more effective if the KPIs were normalized to represent performance more directly; "pass attempts per minute" and "pass completions as % of pass attempts" might be more indicative of performance than the raw data here, in which some of the raw figures are skewed by playing time. To make the display clearer, the Interceptions figure should be flipped around to measure "avoidance of interceptions" so that good performance goes to the right. Perhaps "pass attempts per interception" or "completions per interception" or "minutes played per interception"..."

[Jon]

(Apologies for repeating some of your points. I should have read your first post on this topic first! Interesting example.)

[derek]

Wow, lots of meat today. For the first chart, you don't have to index the metrics to just one level, you still have the freedom to index them to a second level as well. These two levels can be the first and second tertile, the first and third quartile, the mean plus and minus 1 s.d., etc.

I use a display a bit like yours as a teaching tool to show managers that while they claim to love their RAGs, they've actually chosen an amber range so absurdly narrow they're really red-green in disguise. I try to convince them to broaden their amber thresholds so that between 1/5 and 1/3 of their performance data ends up in the amber.

[derek]

For your heat map, you can shuffle the metrics and the players around as I describe here, in the manner of the late Jacques Bertin. I also think bars or spots let you see the quantities a little better than colours do, though that's not so big a deal if there are only say five levels like a Likert scale (and if there are more levels than five, spots struggle to show the difference anyway)