In case you weren’t paying attention, earlier this month future NFL Hall-of-Famer Tom Brady threw for over 500 yards in a single game for the first time in his career.
His 517 yard tally against the Miami Dolphins on September 12th was over 100 yards more than his previous career best (a 410 yard performance against the K.C. Chiefs on September 22nd, 2002 – almost 9 years earlier).
Not only was it a personal best, he also set the Patriot’s team record with his performance.
But was this a statistically significant event, and if so, to what can we attribute it? Did he complete a higher percentage of his passes? Or maybe he just threw more passes than he usually does. How can we break it down?
Control Charts with Tableau to the rescue:
Breaking it Down:
Unsurprisingly, The answer to the first question is that yes, 517 yards is an “outlier” (or, beyond a control limit). Choose “Yards” in the upper left and you will see that the orange circle representing the game on 9/12 is above the upper control limit (UCL) – a statistically significant event according to Shewhart.
But what was different about this game than any other?
Scroll through the various quarterback statistical variables in the upper left and you will notice only one other value is an outlier for the record-setting game and that is “Long” (longest single pass of the game).
In the 4th quarter, with less than 6 minutes left, Brady threw a 99 yard touchdown strike to Wes Welker, racking up the most yards possible for a single play from scrimmage.
All of the other statistics are in control – meaning Brady’s performance was not abnormal (for Brady, anyway) in respect to attempts, completions, completion percentage, average yards per pass, or even quarterback rating (Brady has delivered superior performances in terms of quarterback rating 27 times in his career).
How is this Helpful?
This analysis illustrates the usefulness of Control Charts, and the awesome ability Tableau can give us to create and visualize many such charts in a convenient way.
These charts use standard table calculation functionality of Tableau along with the Shewhart calculations for Control Limits (different than what Tableau shows how to use here, which uses standard deviation and so is more sensitive to the normal assumption).
How to construct a Shewhart Control Chart in Tableau.