Welcome to the first installment of the Friday Face-Off. This will be a recurring piece that looks into an issue in the world of Arizona Coyotes hockey and use statistical reasoning to offer a solution. Today's opening edition will concern the growing goalie controversy in the Coyote crease.
Mike Smith vs. Devan Dubnyk
The Coyotes have themselves an issue. Mike Smith has struggled this season, that is undeniable. Also undeniable is that Devan Dubnyk has performed fairly well in limited duty this season. Which has led to discussion that goalie number 40 should be starting as often if not more than goalie number 41 right now.
Is that a reasonable request? The traditional numbers would suggest so. All numbers before Thursday's game against the Los Angeles Kings.
As the chart shows, the numbers are clearly in favor of Dubnyk. He has as many wins as Smith in fewer than half the starts. His save percentage is much higher and his goals against is streets ahead. This should be a no brainer, right? Before giving Smitty the standing eight, let's dig deeper in the numbers and see if we can really determine what is happening.
The Advanced Stats
Everyone points to the save percentage and goals against as obvious reasons to put Dubnyk in the pipes more often, but we use those numbers because they are the stats we are used too. However, they may give an idea of a goaltender's performance, but they don't tell the whole story. That's where pulling out the advanced stats can be a huge help.
With that in mind, I have compared Smith and Dubnyk in five additional categories: shots faced per 60 minutes, save percentage in close situations, goal support, blocked shots and corsi percentage for. All stats were calculated only for starts (so Dubnyk's performance in the Lightning game is discounted).
Firstly, here's what they have seen in shots per 60.
- Mike Smith: 29.76 s/60
- Devan Dubnyk: 31.74 s/60
Why shots per 60 instead of shots per game? Well, it takes into account the fact that Smith was pulled in the Tampa game and it also takes into account time off ice for an extra attacker and extra time played in overtime. As we see, Dubnyk actually sees more rubber per game than Smith does (which is surprising). The difference isn't huge, but Devan sees 6.2% more shots per 60 minutes than Smith does.
Based on the save percentage numbers, not only does the Regina native see more pucks per game, he saves more as well, so shots per 60 is in favor of Dubnyk. But how do they compare making saves when it matters most?
The Saves that Save
The next stat I looked at was to compare their save percentages in close situations. I defined a close situation as any game situation where the goal differential was two or less (tie game, 1-0, 3-1, 5-3, etc.).
- Smith: 437 saves on 487 shots, .897 sv%
- Dubnyk: 194 saves on 210 shots, .924 sv%
The interesting thing about this stat is that it pretty much mirrors each goalie's actual save percentage. Smith is struggling to stop the puck in any situation and Dubnyk isn't racking up saves in blowouts either way. Put another check in favor of Dubnyk.
But those stats directly correlate to the individual's performance. Is there anything that can show what is happening with the team playing in front of the goalies?
Same Sweater, Different Teams
A goalie is graded decidedly on his own merits, but the way the team plays in front of him can directly affect their performance. How do the Coyotes play in front of their two starters and is there a significant difference was an important question to answer, but what's the best way to do it? I chose three different metrics to measure how the team plays in front of the goalies: goal support, blocked shots and Corsi for. Let's start with goal support.
- Smith: 2.16 goals per game (Coyotes scored 41 goals in 19 starts)
- Dubnyk: 3.14 goals per game (22 goals scored in 7 starts)
Devan Dubnyk gets a goal per game more in support than Smith does. The Coyotes have struggled to score all year, but, for whatever reason, have found it easier to score when Dubnyk is in the crease. This explains why Smith's win/loss record is fairly horrific. He can only allow one goal in a game to win. I don't know of many goalies that consistently post a 1.00 GAA.
Now that we know the Coyotes score more with Dubnyk in net, how do they do defensively in front of the netminders? While blocked shots aren't a great indicator of team defense, it can give you an idea of how the team does in trying to keep pucks from the goalie. Here are the blocked shot numbers for both goaltenders.
- Smith: 14.37 blocked shots per game (273 blocks in 19 starts)
- Dubnyk: 20.86 blocked shots per game (146 blocks in 7 starts).
This is the mindnumbing stat in researching this piece. There is a quantifiable difference in the Coyotes' efforts blocking shots with Dubnyk in net versus Smitty; in fact, they block 31.1% more shots with Dubs in net. Here's more to take away from the blocked shot stat. In Smith's 19 starts so far this season, the Coyotes have blocked 20 or more shots only once, November 7 against the Anaheim Ducks. In Dubnyk's 7 starts, the Coyotes have blocked at least 20 shots four times including a 31 shot performance against the Sharks November 22 in San Jose. I have no answer for why the team blocks more shots with Devan in net, but the numbers through the first quarter of the year speak loudly: more pucks are blocked with him in the pipes.
We now know that Dubnyk sees more shots per game and also gets more shots blocked in front of him, so lets take the next step and examine the corsi for % numbers for the Coyotes in each goalie's starts.
Corsi For One, Corsi For All
Why Corsi for percentage? It takes into account all shots attempted in 5 on 5 situations for both teams to show how one team performed in that game. For example, a Corsi for % of 55.5 means that 55.5% of Corsi events (shots attempted including blocked shots and missed shots) in that game were in favor of the team considered. Here are the Corsi for means for both goalies.
- Smith: 49.1 Corsi for%
- Dubnyk: 47.8 Corsi for%
On average, the Coyotes control possession slightly better with Smith in net than Dubnyk. However, there are two problems with using these numbers for a conclusive assumption. Firstly, the difference between the numbers is only 2.6%, which isn't statistically significant. Secondly, with Dubnyk only having seven starts, his numbers could fluctuate quite a bit from start to start for the average to accurately depict the performance. With those issues on hand, I also calculated the median Corsi for% for each:
- Smith: 47.1 (median)
- Dubnyk: 50.6 (median)
Using the medians creates a more accurate middle ground as the extremes get ignored, but even then the difference is only 6.9%, still far from a significant rate. What does all this mean?
The Coyotes score more and block more shots with Devan Dubnyk in net, but the possession numbers don't vary greatly from goalie to goalie. Arizona basically gives Dubnyk more margin for error when he is in net than they do for their so called number one netminder.
The Final Draw
Now that the numbers have been crunched and scrutinized, what conclusion can be drawn? Simply put, right now Devan Dubnyk gives the Coyotes the best chance to win on a given night. While small sample size can certainly be argued (which I would for the traditional numbers supporting Dubnyk's claim to more starts) you could also use small sample size to argue against Mike Smith being the obvious number one.
I don't believe that Mike Smith should become the backup for the rest of the season, but there is no arguing that the team would certainly benefit from Dubnyk playing more games. Who knows, maybe getting some time off could do Smith some good as well. At the very least, a 60/40 timeshare should certainly be entertained given the performances of the two goalies this season. The advanced stats and traditional ones agree on that.