Yesterday Tyler Dellow wrote about Morgan Rielly’s poor on-ice save percentage over the past two seasons. While it is an interesting read, the hypothesis at the end of the article left me wondering why it wasn’t explored more.
The article starts out wondering if shift start location may be a factor at play in Rielly’s poor on-ice save percentage. Dellow noted that generally defensive zone shift starts result in higher on-ice save percentages while offensive zone shift starts result in lower on-ice save percentages. This makes sense. We know that “rush shots” will result in higher quality scoring chances and you are more likely to have rush shots against if you start in the offensive zone.
Dellow goes through a lot of analysis looking at Rielly’s on-ice save percentages by shift start type and whether the Leafs won or lost the face off to start the shift. I don’t think there is any concrete evidence shown that this was the reason for Rielly’s poor on-ice save percentage. Dellow noted that Rielly was “smoked after defensive zone wins, neutral zone wins, on the fly shifts and offensive zone losses” but it isn’t clear to me how “defensive zone wins” and “offensive zone losses” are at all related. I am not quite sure what to make of it but QoT and QoC across the various situations may be an issue too but that wasn’t explored by Dellow.
After a lengthy discussion, some shot location charts,and some ‘video’ review of goals against while Rielly was on the ice Dellow proposed a hypothesis but never explored it and this is what left me feeling short changed by the article. There was some discussion about whether it was bad teammates or good opponents that were the reason for Rielly’s poor QoC but in the end it seems Dellow has hypothesized that bad teammates when facing opposing teams stars is the reason.
The thing about being on the ice against those players is that mistakes carry a higher price than they do if Johnny Feetforhands is across the rink from you. When the Leafs’ collection of cast-offs and new NHLers was on the ice over the past two years, the odds of Johnny Superstar being on the other side of the ice was a lot higher if Rielly was behind them than it was if Gardiner was behind them.
I think Dellow is suggesting that Superstars will perform better against average players relatively more than average players will. If we have 3 groups of players:
- Group A – Superstars (think first liners)
- Group B – Average players (think second/third liners)
- Group C – Sub-par players (think fourth liners)
What Dellow is hypothesizing is the ratio of A vs B to A vs A is greater than the ratio of B vs B to B vs A.
So in effect, the combination of Rielly’s high QoC and poor QoT is making him look a lot worse than Gardiner and his average QoC and poor QoT.
This is an important hypothesis that Dellow puts out that I think has never been tested. In fact Dellow claimed that there is no publicly available data to test this but I disagree. We can certainly explore this further with data that we have (all of the data presented below I pulled from stats.hockeyanalysis.com or puckalytics.com (some aggregating of data required).
I am going to try to explore this hypothesis further but first let me do a bit of a comparison of Gardiner and Rielly from last season.
Dellow tried to lay some of the blame on the abundance of rookies in the line up. Let’s see if that holds water.
It doesn’t appear that it is the rookies that are the problem. Among forwards the worst on-ice save percentages belong to veterans Kadri, Bozak, Komarov and van Riemsdyk.
Let’s look at Rielly’s and Gardiner’s on-ice save percentage when the are on the ice with selected Leaf forwards. Note that I am using Corsi Save% here because I use goal and corsi data in my WOWY stats but the effect should be the same as using real save percentage.
So, Rielly was worse than Gardiner with every one of these players. Rielly was “less worse” with Komarov, Kadri, Marner and Brown and significantly worse with Matthews, Bozak and Martin. I am not sure there is anything to read from those groups of players though.
We can also look at Rielly and Gardiner by situation (note: here I am using real save percentage, not Corsi save percentage as above).
Once again, Rielly is always worse than Gardiner however there are some interesting points to pull out of this chart. First, Rielly is most worse when the Leafs are leading and when the Leafs are on the road. When the Leafs are on the road the opponent is most likely to get the more favourable match up as they have last change. Also, Babcock relied mostly on Rielly and Zaitsev to play the shutdown role so this will be at play when the Leafs are leading which also happens to be when the opponent plays their top offensive players more frequently.
Let’s split up the leading/trailing/tied situations by home and road and look at the difference between Rielly and Gardiner’s save percentages.
I have ordered these situations according to the difference between Gardiner and Rielly’s save percentages. Once again Gardiner is always better than Rielly.
What is interesting here is that Rielly is always closer to Gardiner at home than on the road in the same score situation. This makes you think that opponents may be really leveraging Rielly’s weaknesses by dictating favourable matchups. Also, the smallest difference is tied, followed by trailing and then leading which matches previous observations.
One of my favourite quality of competition measures is average opponent GF60 (OppGF60 on my stats pages). I like this because it tells you who is facing the better offensive opponents (typically the first liners). Let’s look at how Rielly’s and Gardiners OppGF60 compare by home/road and score situation.
|Rielly OppGF60||Gardiner OppGF60||Gardiner/Rielly|
|5v5 Trailing Home||2.31||2.32||1.00|
|5v5 Tied Home||2.34||2.26||0.97|
|5v5 Leading Road||2.19||2.09||0.95|
|5v5 Leading Home||1.97||1.88||0.95|
|5v5 Trailing Road||3.51||3.25||0.93|
|5v5 Tied Road||2.28||2.07||0.91|
(Note that the OppGF60 is the opponents GF60 in the matching situation, so for 5v5 Trailing Home it would be the opponents 5v5 Leading Road GF60).
Generally Rielly has tougher QoC except in 5v5 trailing home situations where Gardiner had marginally tougher QoC. Interestingly, the greatest difference in QoC is 5v5 Tied Road where we didn’t see that much of a spread in save percentage. However the next 3 greatest spreads in QoC (Trailing Road, Leading Home and Leading Road) were the three greatest spreads in Rielly’s and Gardiner’s save percentages.
Another way to look at this is to look at how Rielly and Gardiner performed against opposing teams top players. To do this I chose one player from opposing teams to represent that teams top line and then looked at how Rielly and Gardiner performed against them. I did this for each of the teams within the division plus Sidney Crosby and Alex Ovechkin, the two offensive stars in the eastern conference. Here are Rielly’s and Gardiners ice times against these players.
|vs Rielly TOI||vs Gardiner TOI|
So, it seems clear that Rielly does in fact line up against the opposing teams top line more often than Gardiner does. Now how do their results compare when playing against this group of players.
|‘Stars’ vs Rielly||‘Stars’ vs Gardiner|
Ouch, this isn’t looking so good for Rielly. A GA/60 of nearly 5.00 is horrendous and that corsi save percentage is well behind Gardiner’s.
So, when not facing this small list of first line players, how did Rielly and Gardiner do against the rest of the league?
|Others vs Rielly||Others vs Gardiner|
Gardiner still appears better than Rielly, however it is much closer. If I took the time to pull out the rest of the first line players on the remaining teams it is certainly possible that it would wipe out any difference. I’ll leave doing this as an exercise for the reader.
Takeaways and Conclusions
So, lets summarize some of our findings.
- It isn’t the ‘rookies’ that are the reason for Rielly’s poor on-ice save perentage.
- Almost any way we divide up the ice time (with various Leaf forwards, by score and home/road situation) Gardiner always has a better on-ice save percentage than Gardiner. This makes it difficult to claim randomness because we wouldn’t expect true randomness to be so evenly distributed across groupings.
- Rielly gets absolutely slaughtered by opposing teams top lines.
Now, getting back to Dellow’s hypothesis where he puts out the idea that superstars will dominate weaker players and the Leafs are full of weaker players and Rielly plays more against the stars and that this is the reason for Rielly’s poor on-ice Sv%. I have only really looked at Rielly and Gardiner but it may very well be that the first part of Dellow’s theory holds true. It may very well be true that star players can absolutely dominate weaker players. The unfortunate thing for Rielly is that it is looking more and more like Rielly is one of those weaker defensive players. It is not the rookies, it is Rielly (and probably Bozak too but that is for another post).