Box scores go a long way to influence our perceptions of an NHL game. Because the season is so long and we can’t watch 1200 games, we tend to make conclusions based on box scores. Shots and saves are influential because they create a significant statistic called save percentage. One that has become overrated in our evaluation of goaltenders.
The NHL real-time tracker has added depth to our observations, but shot metrics continue to lag behind. These numbers provide a general overview, but lack context and fail to provide the accuracy that can be obtained by merging the eye test with better data.
To highlight these discrepancies, we took a deeper look at the Rangers-Islanders game that took place on Feb. 16. Immediately we noticed errors in the recorded data.
Interestingly enough, both data sets record a dominant first period by the Islanders, but the delivery of the message is where the conflict arises. The official record states that the Islanders maintained a 22-11 shot advantage. Further review showed six phantom Islander shots for a more manageable 16-11 advantage.
True shot metrics establish this as a more even period, but this is where possession metrics help us to identify how the Islanders maintained their dominance as they registered nine green shots (high-quality scoring chances) to the Rangers’ five. So while there has been a strong correlation to possession and dominance, the same correlation is lacking when connecting it to save percentage.
If we review the period based on save percentage, by allowing a goal on a red shot, Cam Talbot actually recorded an .875 SV% instead of the credited .909 because of the six phantom shots. When assessing his performance, this is a more accurate reflection of reality even though he bounced back to make saves on 8-of-9 green shots.
After the first period, the Rangers began to dictate the pace of the game and maintained close to a .550 possession number. This shows up in the enhanced shooting data as the Rangers out chanced the Islanders 13-1 in green shots.
The only reason it wasn’t a blowout win was based on the performances in the crease. Cam Talbot only faced 13 green-level shots in the game, but surrendered three goals on low-probability red shots. Typical save-percentage stats penalized him as he registered an .884, but the reality is the Islanders were credited with 12 extra shots, his real-save percentage should have been .839.
Jaroslav Halak on the other hand faced 40 shots, 18 of which were high quality. Five of the six goals he surrendered were of the green variety. Neither goaltender played particularly well, but Talbot was the beneficiary of a major stat boost courtesy of the NHL shot count.
These are the types of discrepancies that Steve Valiquette and I are fighting with our projects. Tuuka Rask made light of this recently in an NHL.com piece “As Rask is quick to point out, it’s not just where a shot is coming from that matters. It’s also the type of play which generates the shot.”
Identifying pre-shot movement is the key to unlocking shot quality and with it, we will be able to provide more context to the box scores that we heavily lean on for our perceptions.