In the first article in this series on advanced statistics, I challenged the myth that Corsi is a ‘magic bullet’ that can be used as a universal hockey analysis tool. As I explained Corsi is not able to distinguish the quality of scoring chances. Nor is it a good metric to predict a team’s place in the standings or success in the playoffs.
In this piece, I will present you a brand new category of advanced statistics that came to mind as I was debunking Corsi. Here, I will introduce you to something I call shot attempt scoring efficiency or SATSE.
What will my new tool of measurement bring into the world of advanced stats?
SATSE is designed to evaluate the quality of the offensive drive. It relies on three data points: Corsi, on-ice shooting percentage and on-ice save percentage.
You may be asking why I am using Corsi at all, after describing it as a mostly irrelevant metric in the previous article. Corsi provides a measure of all shots directed at the net to capture the total number of attacks. Even if a shot is blocked or misses the net, it could have been good opportunity to score.
This information will help provide an idea on the number of times a dangerous attack was created or permitted while a particular player is on the ice.
That said, how do you distinguish a good opportunity from a lesser one? This image below shows clusters of shots versus goal percentage from 2010-11 through 2016-17.
The cluster in blue is mostly situated in the danger zone in front of the net. Those shots have an average chance to score of 16.8 percent. Further out, shots taken from the orange cluster have a 4.9 percent chance to score.
The green, red, and purple clusters are very rare occasions, which mostly end in goals. When it come from behind the net, it will mostly be a deflection off a skate or another body part ending up the net. They can be considered lucky goals.
The high percentage over the blue line has a easy explanation too. Most of them are simply accomplished with an empty net. If a goaltender stops it, it was done over a puck drop in the zone. However, if it goes past him, he was chasing butterflies.
With that in hand, the on-ice shooting percentage enters into action. Based on the probability from the information above, the higher the percentage is, the better the shot location should be.
Supported by these two stats, a simple formula is established to estimate the offensive shots that were from a dangerous area while the player is on the ice:
CF x oiSH% = SATSE F
(Corsi For x on-ice Shooting Percentage)
We can do the same estimation with the shots that were allowed from a dangerous area while the player is on the ice:
CA – (CA x oiSA%) = SATSE A
Corsi Against – (Corsi Against x on-ice Save Percentage)
Those two metrics will give us the opportunity to subtract them and obtain a differential. I call it SATSE Diff.
To verify whether or not those numbers can be determined as reliable, I started by calculating SATSE Diff for all players on each team, starting at the 2007-’08 season through the 2016-’17 season. Then, I put these totals in order from high to low for each year.
Two observations were made. The difference between that new order and the standing is at an average of +/- 2.7 rank through a ten year span. Plus, the teams that did the best in their division or conference had 88.8 percent chance of making the playoffs. Since the new playoff format, the number increases to a 92.2 percent chance.
The is a very interesting correlation, even though I used shot attempts. In theory, using Fenwick (unblocked shot attempts) should be closer to the standings than Corsi, since blocked shots will certainly not reach the net, therefore won’t result in a goal.
The review of this alternative shows closer results to the standings, and 95.3 percent chance of making the playoffs since the new format. This provides a stronger argument to say that my newly-created SATSE is on the right path.
However, the research goal is not to get close to the standings. For that reason Corsi stays put.
We can now continue with few alterations like the SATSE F/min and SATSE A/min.
For example, in 2016-17 Max Pacioretty had a SATSE F/min of 0.116 and a SATSE A/min of 0.101. In comparison, Sydney Crosby had a SATSE F/min of 0.141 and a SATSE A/min of 0.114.
With this variation, the SATSE FAR and SATSE AAR (AR = for above replacement) can be created, taking its inspiration from the goal above replacement metric. To build this one, the average SATSE F/min and SATSE A/min from all the player as to be done, for each season.
SATSE F/min / (SATSE F/min + League average) = SATSE FAR
1 – (SATSE A/min / (SATSE A/min + League average)) = SATSE AAR
The results are shown by percentage. The pivot point to be above or below average is at 50 percent. Doing this exercise with the same two players gives us:
SATSE FAR: 60.4 percent
SATSE AAR: 52.9 percent
SATSE FAR: 65 percent
SATSE AAR: 52.3 percent
All these measures could be adjusted to the zone start. Let’s take Alex Galchenyuk for example. Last season, he started 74 percent of the time in the offensive zone. His SATSE F was 114.22, but after the zone start adjustment, it is 86.81.
If a player starts more often in the offensive zone, it is obvious that he and his linemates will have more offensive opportunities. This variation is a bit trickier since we cannot evaluate line changes and it will be greatly influenced by special units. For these reasons, I will take this no further.
However, SATSE cannot be mistaken as a measure of individual player performance, because like Corsi, it is still a group effort stat. I’m not yet sold that a stat will be developed to individually single out players for the purpose of building a team.
However, this new metric might help to highlight some players. Additionally, since it is a good indicator of the success of the team, it might help to create new data such as metric establishing the individual influence of a player on his SATSE for or against.
I will end this by giving you the Canadiens’ table from last season.
What do you think of SATSE? Please leave your comments below.