Other major sports have quantifiable methods of breaking down situations and plays. For any base/out/inning/score situation in a baseball game, there is a corresponding value that determines how many runs ought to be scored through the rest of the inning and also a value of win probability. The same holds for football. The score/down/field position/time left gives the viewer a great gauge at how the outcome of the game will unfold. Knowing the win probability in each of these situations helps us to figure out the value of each event that happens during gameplay. The value of a double to the gap in baseball is quantifiable. The value of a 12 yard gain on 2nd and 4 from the 50 is quantifiable. Hockey is a little different.
War-on-ice.com has a win probability chart based on time left, score, and EV/PP/PK situation. That still leaves many discrete events during gameplay that occur without moving the win probability meter. How valuable was that cross ice pass? How valuable was that hit in the corner that forced the turnover? Even though they don’t immediately impact the win probability calculator there are recorded events that do have a hand in dictating the outcome of games.
Faceoffs, for example, have been studied and quantified. Schuckers, Pasquali, and Curro found out the faceoff differential required to yield a single goal differential for all situations and faceoff locations. Another noteworthy finding was from Gabe Desjardins who specifically looked at the value of winning an offensive zone faceoff. There are 13 unique events that can occur in an NHL Play by Play report. Today I’m going to attempt to do similar work to the faceoff pieces linked above to determine the value of one of those unique events: the Takeaway.
Disclaimer: I’ve been warned a few times over about the unreliability and subjectivity of how and when Takeaways are recorded. I’ve decided to proceed with my study anyway, simply to get a gauge for its value rather than accepting its value as final.
I looked at play by play logs from the past two seasons of regular season game and tracked how soon after a takeaway a shot for and shot against occurred. If there was a stoppage of play, I would “reset the clock” to zero and wait for another takeaway. Below are the per game SF and SA values for two seasons worth of data:
This graphs gives us a general idea of the disparity between SF and SA after a takeaway and how long it takes until the lingering effects of a takeaway are diminished. However, we need to convert the shot differential to be in terms of a “per takeaway basis”. There were 32,840 takeaways in regular season games over the last two seasons, which means there were 6.7 takeaways per game. If we sum the shot differential value (SF-SA) for each second on the graph above and divide by 6.7 we get the shot differential value for a single takeaway.
When we do that, we get 0.14 shots per takeaway. It’s certainly not much, but it’s something. To get its true value let’s continue converting this value. Since overall shooting percentage was 8.9% over the last two seasons, we get 0.14*.089 = 0.012 Goals per Takeaway (or 1.2 Goals per 100 TkA). Since an extra goal is worth about a 1/3 of a point in the standings and a win is 2 points, then (2 * 3) / 0.012 = 485.8 TkA per additional win. I would imagine this value would change based on which zone the takeaway occurs and also based on score effects (i.e. Do leading or trailing teams get more takeaways?)
To put our finding into context, Ryan O’Reilly led the league for the past 3 non-lockout seasons with Takeaway totals of 101 in 2011-12, 83 in 2013-14, and 98 in 2014-15 (I should note that Mark Stone also had 98 for this season). Additionally, teams that lead the league in this category hover around 800 total team takeaways per year. However, this is similar to looking only at Faceoff Wins without considering Faceoff Losses. Teams may get a couple extra wins over the course of the year based on their Takeaway totals, but we still need to consider Giveaways, which I’ll cover next time.