Athlete performance data must be useful to coaches and players, not just interesting. This is what we hear from current and potential team customers all the time, and rightly so. Recently, we have provided the means to add QB throw ball metrics, such as spin and velocity, into the mix of athlete data available from games and practice. Whilst this data might be interesting for coaches and players, it won’t, without an understanding of its application, change how they play or practice or recruit. Hence, we are developing an ‘Arm Index’ concept, to give coaches and sports science something usable from this smart football data for evaluation, tracking and decision making.
What’s the ‘Arm Index’ idea?
The Arm Index is the norm zone of spin and velocity produced by an individual QB when throwing the football. It is visualised and quantified relative to aggregated elite QB data.
The following anonymised example should bring this to life. In the plot below, the blue dots show the spin/velocity region of an individual QB’s throws versus an aggregated (and larger) group of elite QB throws, shown by the open dots.
Figure 1: Spin rate v Release Velocity of 15 to 45 yards throws from an individual QB (blue dots), n= 623, versus a group of QBs(open dots), n= 2323.
The shaded redzone is the QB’s mean ± 1SD spin and velocity.
The shaded red zone is the individual’s mean ± 1 standard deviation of spin and velocity. This represents their normal performance, and it visualises that this QB performs in a slightly below average zone of the spin/velocity region, in comparison to the elite group. Visually, the QB’s norm zone is lower middle, which reflects spin lower, but velocity average. In a number, this QB’s Arm Index = -1.3. Or slightly worse than par, to use the golf metaphor. We derive this from the normative distribution of two variables combined, and simplifying the maths slightly to make the values more intuitive. Here, 0 would be exactly average relative to the group, or par, on a scale of -5 to +5. A number around +3 would be very high relative to the group (eagle), and around -3 would be very low (double bogey).
So What?
The reason to develop the Arm Index is a pragmatic one: To create an applicable output from the Smart Football data. How can we turn 10’s of QB throws per practice session or game, or even 100’s per week or per camp, into an understandable graphic and intuitive number? Something that tells you about how they are throwing, rather than solely what they have thrown.
The aggregation of throws and comparison relative to norms is the start point. We understand this choice ignores the various context factors that influence how a QB throws the football and instead treats all throws as a homogenous group. However, by taking this approach, we create an overview of QB throw performance that can be used to surface trends, relationships and rankings. It’s level one data science, but it’s meaningful.
The assumption we are making to treat throw data as homogenous is that most of the time QBs are attempting to throw a ‘tight spiral’ pass, which requires them imparting a threshold level of spin and velocity into the throw. To increase the chances this assumption is more right than not, we are only including throws between 15 and 45 yards. Figure 2 should graphically explain this choice, where a non-linear release velocity versus throw distance relationship suggests velocity is reasonably homogenous in the 15-45 yard portion of the plot. Therefore, if we have enough throws to draw upon, our aggregated analysis should work. Furthermore, our growing database of thousands of elite QB throws in the 15-45 yd band is described by a normal distribution, with a mean of 52±8 mph, supporting the assumption and approach.
Figure 2: Top shows a non-linear relationship between throw release velocity and throw distance over 5-65 yards,but Bottom shows no relationship between 15-45 yards.
We also decided to begin, that we should not assume that ever increasing velocity and/or spin is directly related to better arm performance, until we can prove otherwise objectively. QBs need to create sufficient velocity and spin, so the pass can be accurate and catchable. Logically, all elite QBs are going to have half way to decent velocity and spin numbers, otherwise they would not be elite QBs. What is more likely is that each QB’s arm will operate in a zone of velocity and spin. There is natural variation between their throws, but our method is assuming that ‘on average’ a QB will throw at x velocity and with y spin, plus or minus.
We formulated this hypothesis following the analysis of 15-45 yard throw velocity data (figure 2), along with discussions with QB experts. Indeed, initial data analysis showed individual QBs occupy a distinct zone of velocity and spin, following an aggregation of 100+ throws. The fact that different QBs perform in different spin/velocity regions indicates this method can be used to distinguish between players.
In sum, the Arm Index is useful because it quantifies a QB’s throwing performance zone or ‘sweet spot’, through their relative position within a large aggregated elite dataset. This is probably the first and most salient use of the Arm Index, as it provides the means of evaluating a QB’s spin/velocity sweet spot relative to a group of elite QBs. For example, high school prospects relative to a historical college roster dataset.
Going back to the example above, you can conclude that objectively this QB throws slightly lower spin, on average, compared to his peer group of QBs. This performance is summarised in a single Arm Index number with accompanying visualisation.
The Index can also be used to track progression of developing players, add objectivity to QB specific combine tests, and monitor if QBs display any deviation relative to their sweet sport over the course of a season. Related to this, there is an obvious and important use during return to play (RTP), as knowing a QB’s fully healthy‘sweet spot’ provides the objective benchmark to progress back up to, as they work through their RTP program.
Example Application of the Arm Index: Monitoring Across the Season
Monitoring changes of an elite QB’s ‘arm’ in practice and games is an obvious application of the objective quantification of their spin/velocity norm zone.
This approach is standard practice in sport science, where an athletes’ outputs for running distances, sprints or accelerations and speeds –for example - are monitored continually and compared against their norm values from game day or specific practice sessions.
The following 4 plots shows the aggregated throws of the same QB from the example above. Each plot shows a week’s worth of practice throws, relative to their entire season norms. The region drawn in red is the mean ± SD of that week’s spin/velocity zone and the dashed green line shows the spin velocity trend for that week compared to the solid green line of their aggregated season trend. Using the same method as above we can quantify the QBs deviation from their norm with a single± spin/velocity number. Again, a score of 0 would be ‘par’ for them.
The data below tells a story of perfectly normal arm performance in weeks B and C, with +0.03, and +0.09 Index scores respectively. The plots show throw sevenly spread above and below the trend line, positioned slightly up and right of the middle. However, this declines in later weeks; slightly in week D (-0.03) and then significantly in week E (-2.9). Some outlying low value throws may skew WeekE’s number, but the trend line clearly shows only 2 throws above the QB’s normand many well below, indicating the arm performance is down. Maybe it was the context of the practice that led to this drop, but this is kind of insight that could trigger a conversation between coach, sport science and player to diagnose why there is a change relative to their normal, and if something needs to be done about it, such as review technique or implement a recovery protocol.
Week B, in the sweet spot (almost exactly middle of norms)
Figure 3: Plots for weeks B, C, D andE show the QB’s current week’s throws as purple squares, with red region as
mean ± SD zone, and dashed green trendline, against all season throws in blue dots and overall black region, and solid green trendline.
We have presented initial workings of an Arm Index, to assist and promote the application of Smart Football data.
Currently we are engaging with users to gain feedback and refine the output. We will collate our initial elite QB 15-45 yard throw database and share these norms in future blogs and papers. We will also produce an automated method for establishing an Arm Index score, relative QB groups or individual QBs to themselves.
Long term, as we hope the use of our Smart Ball data continues to grow amongst elite QBs, we will build upon this work, using larger data sets to create context specific scores. This will strengthen the utility for coaches and sport science and help overcome the limitations of the aggregated approach.
We will be able to quantify the ‘catchability’ of the pass through ball state at arrival and explore the classification of different types of QB throw (for different distances and play concepts), which give coaches objectivity on the quality of the throw.
Please reach out if you have found this interesting and would like to give us feedback or want information about how to access these kinds of insights.
