Analytics Corner: Take Good Shots, Win More Games

<em>“You miss 100% of the shots you don’t take” Well, actually...</em>

Yesterday, the @statsrespecter discussed what was going wrong with the Nashville Predators—was it team defense or goaltending?  The answer—one I agree with—is largely on goaltending.  Nashville is doing very well defensively in both limiting shot quantity (10th in the NHL, at 54.8 shot attempts against per 60 minutes) and shot quality (fourth in the NHL, at 2.09 expected goals against per 60 minutes)—they allow a few more shots by opponents because they are forcing teams to take low-quality shots from farther away.

(All statistics going forward are adjusted 5 on 5 unless otherwise noted.)

So if the defense is playing largely at an acceptable level, and goaltending is a major issue, that leaves us with a question.

What can the players do better?

Offensively, the Predators are very good at generating shot quality themselves—they have about 59.3 shot attempts per 60 minutes, good for fifth in the NHL.  And while they have the second best goals per 60 rate in the NHL right now (3.13), they fall into the same trap that the defense lays for opponents: they take a lot of very low-quality shots.

Quality in this piece will refer to expected goals (xG, using the model by per Fenwick shot (FF, or unblocked shot attempt).  Expected goal values are applied to all unblocked shot attempts. Blocked shots are recorded at the spot they are blocked, not where the shot was taken, so this is not accurate, so they are excluded.

Expected goals use many factors in determining shot quality.  If you want to read further, you can check out this excellent write-up by Josh and Luke Younggren—but to summarize, there are a few important factors that make most of the contributions to an expected goal value: distance from previous event (was a shot a rebound that was close to the last shot?), angle of the shot (it’s easier to get a shot in front of the net then from on the red line), and time since last event (shots occurring quickly after the previous event have higher xG values).

However, by far the most important factor that can be included is shot distance. (Note: things like goalie stance, screens, passes, etc. that are not recorded by the NHL are not included in these models—this takes tracking by hand and is very time-consuming.  So, unfortunately, those things are not explained here).

“Bryan, this is weights in a statistical model,” you might say—and it’s not necessarily the best way to convince players to take a better shot.  So in this piece, we’re going to focus on shot quality and distance to better isolate the impacts they have in games.

Getting Up Close and Personal

Another way to look at expected goals is to consider them as a probability—based on historical information, what is the likelihood an unblocked shot attempt will become a goal?  That’s exactly what expected goals are. Rather than assigning arbitrary limits to low-, medium- and high-danger chances, you can know that a shot with an xG value of 0.03 means the shot has a 3% chance of going in.  That chance goes up the closer you get to the goalie.

This graph is broken into three parts (shots that turn into goals, shots on goal, and misses) and represents all 5 on 5 shots taken by NHL teams until December 8th.  The horizontal axis is distance—the further right you go, the farther away from the goaltender you are.  The Y axis (and color) are your expected goal values.  You can tell immediately that the closer a shot is taken to the goaltender, the higher the chance of it becoming a goal.  The majority of shots taken within 20 feet are high-quality shots—most have a 10% chance or higher of becoming a goal!  As you move further out, though, you start to see high-quality shots dwindle, and by 60 feet away, high-quality shots are few and far between.

Note, there are a few outliers—for example, those two orange dots at about 0.30 xG in the chart are shots by Mattias Ekholm and Dante Fabbro.  They were recorded as tip-ins for some reason, and historically tip-ins from that distance had a higher chance of going in.  This is likely just due to the irregularities of scorekeeping over the years, but, by and large, the closer you are, the higher chance it becomes a goal.

What about the Predators?

As you can see from the two graphics above, Nashville is the third-worst team in the NHL at average expected goals per unblocked shot, and in the bottom ten in average distance to goal.  But these two numbers alone can’t tell much, so let’s put them into context of games and the schedule:

The blue line graph represents the five game rolling average of shot-quality (only the current game and the immediately before it are counted - this is better for getting a sense of how a team is trending at any given moment), while the individual bars are the single-game shot quality. You’ll notice that they are colored: green for games that Nashville won, red for regulation losses, and orange for overtime or shootout losses.

The biggest thing that jumps out to me is how the Preds perform when they shoot higher than their own average—when the average shot quality is greater than 0.052 xG/shot (or 5.2%), the Predators are 10-1-1. That accounts for 21 of their current 31 points in the standings.  On the flip side, when they shoot below that, they are a dismal 3-9-4.

Note: I do want to mention at this point that this relationship works this way for Nashville this season, and several other teams to varying degrees, but as of right now, there is no evidence that this is a correlation that is statistically sound league-wide over any season.  I’m applying what we see right now and making judgments based off of that.

What about defense?

If we were to translate the above graph into a scatter plot where shot quality for and against are the axes, we can better visualize this:

This time I have set the average lines to be the league-wide average.  Each quadrant has the record indicated: games higher vertically are games where Nashville limits opponent shot quality, while games on the right horizontally are ones where Nashville shoots high-quality shots.  While playing great defense helps Nashville (they are 11-4-4 when holding teams below league-average quality), they are even better when shooting with above league average quality: 7-1-1.  And if they happen to take low-quality shots while playing bad defense?  That’s where goaltending comes in—they have yet to get even a single point from any of those five games.  Nashville can win many games playing good defense, but they can dominate when taking high-quality shots.

What about the rest of the league?

If you’ll recall the chart above with average quality and distance for each team, you’ll see both the St. Louis Blues and Boston Bruins near Nashville at the bottom of the list - in fact, Boston is one of only two teams to shoot worse on average than Nashville.  Yet both of those teams are leading their divisions?  Let’s look at their charts:

Obviously Boston has a lot more wins and fewer losses than Nashville, but the overall theme stays the same: they are 7-1-0 when shooting above league average, and 14-4-4 when holding teams below average on defense.  This is another team that has good results when playing great defense, but even better when they play well offensively.  The difference comes down to goaltending—they are getting tremendous effort from their tandem, and where Nashville was 0-5-0 when shooting low-quality shots and giving up high-quality shots to their opponents, Boston stands at 3-0-2!

Stop me if you’ve heard this before: the Blues, while not as good as Boston when not shooting well, still have been able to count on goaltending to bail them out of games where they don’t shoot well—a combined 12-7-3 (Nashville was 6-9-4 when shooting below average).  And like Boston and Nashville, their 7-0-2 record when shooting well allows them to dominate, while good defense still gives a very good 11-3-2 record.

The Bottom Line

Nashville is playing good defense, but the goaltending has been very shaky. The Predators can play great defense and allow a lot of goals, or they can play awful defense but Pekka Rinne or Juuse Saros can pitch a shut-out.  The graph below shows the rolling averages for save percentage and expected goals against. Nashville can play great defense, but still see stretches where everything goes in the net.

The Nashville skaters - especially the forwards - can’t do much more to influence the goaltending at 5 on 5, so they need to focus on something they have a direct influence on—taking high-quality shots.  We’ve seen examples of teams that are succeeding based on performing very well when taking high-quality shots—the differences is that those teams have very good goaltending behind them.  But Nashville, alongside St. Louis and Boston, still performs extremely well despite defense when they shoot well, and the biggest factor is to take shots closer to the net.

I know that the Nashville system—and by extension Peter Laviolette—encourages taking shots from the point and then going after rebounds, etc, but Nashville no longer has the luxuries it did in years past.  They thrived in years past taking low-quality shots by their defenders and from outside the circles, but before they had stellar goaltending to back it up.  This year they don’t.  And Pekka Rinne and Juuse Saros can no longer hide the flaws present in this offensive system.