Shot quality, expected goals and the Royal Road

[DeluxeMoustache]

There has been a lot of talk over the year about goaltending, and invariably save percentage is brought up as the stat and interpreted as an indicator of quality of goaltending.

When people talk about the Hurricanes, one of the reasons for the poor record was that Scott Darling had an awful year. And he sure did.

Ward 23-14-4 2.73 .906
Darling 13-21-7 3.18 .888

Did Ward play well? Lights out? Ok? Poorly? Difficult to say. Generally .906 is below league average and regarded as not great. But he had a young team, weak roster, etc. Was he ok given what he was playing behind?

Now when Kipper played, from 2005-2010 he put up these numbers:
2005-06: 42-20-11 2.07 .923 (D Sutter)
2006-07: 40-24-9 2.46 .917 (Playfair)
2007-08: 39-26-10 2.69 .906 (Keenan)
2008-09: 45-24-5 2.84 .903 (Keenan)
2009-10: 35-28-10 2.31 .920 (Brent)

Did he forget how to play goal in 07-09? Not in my opinion. The team definitely played a more wide open style and elected not to prioritize defense.

Meanwhile, in Minnesota, where Lemaire had his team trapping well, you had guys like Fernandez and Roloson putting up save percentages in the .930s at times.

How do we know when a goalie is actually good or bad, or when their team is propping them up or dragging them down?

Intuitively, everyone knows that shots vary in quality. Using one number, save percentage, treats all shots as equal and does not distinguish how many saves are from easy or tough shots.

So the real challenge is how you can establish shot quality, and the probabilities of stopping shots of various quality.

How are goals scored, and what constitutes dangerous shots?

____

There was a guy who did some work for Sportsnet a few years back , Chris Boyle.
He had what was called the Shot Quality Project. He watched a large number of games and focused on shot types. The basic observations with segmented shots include the following observations:
- Clean shots, where the goalie has time to get set and square to the shot, have an expected save percentage of .949
- Shots preceded by a pass had expected sv% of .651

He took data from 2011-14, reflecting 6700 shots handled by the Bruins and Montreal (extrapolated from 4800 for Mtl). From this data, he found that Boston over 6700 shots allowed 371 more clean shots, 243 less shots preceded by passes, 85 less tips and 43 less shots on rebounds.

Bruins expected sv% was .923
Canadiens expected save percentage was .908

Actual results were that Boston maintained a sv% of .926 and Montreal .917

Price is a goalie with great fundamentals and I believe that his performance above expected reflects that, despite being a few mere points above league average of .914



https://www.sportsnet.ca/hockey/nhl/...s-a-team-stat/

The changing environment theory is best exemplified by goaltenders who move from team to team. Ben Scrivens
...
Pre-Edmonton, he had a respectable .917 NHL SV% over close to 1,400 shots. During his small-sample stint in L.A., Scrivens led the league in save percentage. But since his trade to Edmonton he has cratered, posting a disastrous .905 SV% (.887 this season). And it’s entirely environment based.

... Scrivens is a respectable +.001 above expected save percentage for his career. But it also shows what’s actually been expected on his different teams—922, .919, .909—and helps illustrate the massive difference in workload he has now compared to his first two NHL stops. Scrivens’s career .957 SV% on clean shots coupled with the Kings’ ability to limit high quality pre-shot movement opportunities was a perfect match of styles. Even in Edmonton, his save percentage on clean shots is +.011 above expected, but his exposure to high-quality opportunities in the form of cross crease passes has buried him. In Edmonton, Scrivens is facing 30 to 40 more difficult shots per 1,000 than he was in Los Angeles, or even Toronto.
Ultimately the Oilers’ struggles cost goalie coach Frederic Chabot his job. But the real reason Scrivens’s save percentage is below .900 is porous defensive zone coverage. When running the data for just 2014-15, Scrivens has an expected success rate of .903. Only 82 percent of the shots he has faced are ones in which he could set his depth and angle. The Bruins register 88 percent

Another of his findings:

I discovered that while only 15% of all NHL shots I reviewed involved pre-shot movement, it accounted for 50% of all goals scored.#

Steve Valiquette, a former goalie, adds another layer to the analysis.

He is interested in categorizing how goals are actually scored. He describes what he calls the Royal Road. Basically it is a straight line starting at the net and going down ice in the direction of the opposite net, to the top of the circles.



Great video where he describes it here.
https://www.omha.net/news_article/sh...never-heard-of

The theory is that crossing that line makes a goalie move laterally and therefore not be able to get set, increasing shooting percentage.

What he found was the following.

First, he categorizes shots as red or green

Red - low percentage shots, goalie can see the shot and has more than a half second to get set. This is 22-23 of a typical 30 shots.
These 75 percent of shots taken account for 18 percent of all goals.

Green - these are opportunities where puck movement results in the goalie not having a half second to get set.

Of all goals reviewed:

22 % : cross ice passes
Shots off pass across the Royal Road, below the tops of the circles
(From Shot Quality Project, Average shooting percentage on these shots is 30 percent vs. League average of 8.5 )
10 % : screen shots
9 %: one timers from behind the net (don’t necessarily have to cross the Royal Road, but the goalie can’t get fully set)
9 %: broken plays - redirections or deflections that force goaltender east/west movement
8 % - crossing Royal Road with possession to make goalie move laterally. If a player can cross the slot, chance of scoring goes up from 3% to 33%
8 % - deflections
8 % - rebounds off of the above


I found a lot of this interesting. I don’t think they are monitoring the right things in real time to have predictive data for all teams.

What are your thoughts?

Are there any better analyses you are aware of out there that make sense and seem to effectively analyze what shot types and what a goalie can reasonably be expected to stop?

In light of this, and having watched the Flames this year, does it seem to support the good shot count metrics / poor scoring results?

Consider the types of goals the Flames have given up. Do these analyses point to any obvious room for improvement?

If you have seen Canes games, and how Peters approaches the game, do you see potential for the Flames to improve particularly their O or D?

[Calgary4LIfe]

This has been a late season criticism of mind in how the Flames generate offence now. Thanks for the find! Really interesting, and it is good to know that people in the analytical community are actually looking into data much further than what is being trotted out.

Fischy, please send this to your source

[Matt Reeeeead]

There is some great content presented here.

But let's not forget how bad darling was this season and how below average can ward has been for 5-7 years now.

There could be something to the style Carolina is playing that is increasing their quality of shot against, but not to he degree of how poorly the goaltending was this season.

Fwiw, darling faced one of the highest degrees of difficulty in Chicago the year before and was dominant. For whatever reason he simply was not up to the task this year (there were rumours before the season that I can't substantiate that he was back on the sauce). Anyways,that difference was not system driven... it was performance, one way or the other.


Sent from my iPad using Tapatalk

[tkflames]

Great Post!! This was a great read first thing in the morning.

The team that I find exemplifies this theory is Boston. Last night (watching only the 3rd period), almost every shot attempt they had on net came from movement in the offensive zone. They were constantly looking for the cross ice pass in the high slot and when they missed the net had a high retrieval rate. I suspect that depending on Boston's success this year, we will see a lot more of this in the next couple of years. This theory passes the sniff test (goalies are blamed for a disproportionally large amount of goals against), the eye test in that teams like Boston look downright terrifying in the offensive zone and clearly the stats test as you so eloquently laid out above.

[gunnner]

Originally Posted by tkflames

Great Post!! This was a great read first thing in the morning.

The team that I find exemplifies this theory is Boston. Last night (watching only the 3rd period), almost every shot attempt they had on net came from movement in the offensive zone. They were constantly looking for the cross ice pass in the high slot and when they missed the net had a high retrieval rate. I suspect that depending on Boston's success this year, we will see a lot more of this in the next couple of years. This theory passes the sniff test (goalies are blamed for a disproportionally large amount of goals against), the eye test in that teams like Boston look downright terrifying in the offensive zone and clearly the stats test as you so eloquently laid out above.

The flames low shooting % could quite simply be because of how little the opposing goalies need to move laterally from flames shots. And this is a result of the flames slow transistion, always time for backcheckers to get back and block the cross ice pass.

[Lanny_McDonald]

So to summarize, goaltenders that have time to get setup and square to the shot are more successful than those who are moving, not square, or not ready for the shot. Kind of explains the Flames lack of success and blows a gaping hole in the turnaround based on Corsi argument.

[Enoch Root]

Thanks DeluxeMoustache - great write-up.

I am a big believer in the 'get the goalie moving' argument, and especially passes across the Royal Road (bad name though).

I would like to see these stats for the Flames offensively. We can sit here and say they are too static, too set, but we need data to know for sure. But it certainly wouldn't surprise me to hear that they had fewer of these types of chances.

I am actually even more interested about this type of data defensively. IMO, when you play a more passive defensive strategy, being concerned more with positioning and getting sticks in lanes, you are far more vulnerable to the cross-ice pass. Teams have time to set these plays up because they aren't being pressured as much. And again, I don't have the stats for the Flames, but it sure seems to me that they give up a lot of those. In fact, I think that is the source of their problems defensively (one major source, to be more accurate).

I haven't watched Peters' teams enough to know how he coaches, but I have read a couple comments that make me think he employs a similar passive defensive strategy to Gulutzan (and some of the stats from last year seem to back that up). If that is the case, I will be very disappointed. That would probably be the end of Treliving for me (and I am a big supporter of Treliving).

I wonder if there is any way we can get these stats for the Flames.

[Cali Panthers Fan]

This is one of those statistical analyses that matches perfectly with the eye test and lines up with offensive and defensive strategies across the league. I think this is one of those that people would say "well of course" but it's good to see just how quantifiable this reality is. It tells me that when Gaudreau and Monahan are sometimes guilty of overpassing the puck, it's because they realize this is how to best score goals. When I see Bennett fly down the wing and shoot it into the crest, I now know that it's just not an effective strategy to score goals, and he needs to work on finding his teammates more.

At any rate, thank you so much for that write up in the OP. I will be referencing that throughout the upcoming season.

[Nelson]

First, to get those cross-ice passes below the faceoff circles, most of the time it seems like you’d need to be on the PP (not run by Cameron) or rushing up the ice on an odd-man rush because of the space requirements.

It seems like what is needed is a ASQ% (ACQ%, AFQ%...) measurement. The ASQ% would come from someone categorizing each shot, designating each shot a percentage (ex: cross royal road pass below faceoff circles = 22%), summing up all the shot percentages and dividing by total shots to get the average shot quality percentage.

Then, by comparing that to SF%, you could get a much clearer picture of your team’s actual share of the offence for a given game.

And then you could do it all again for the defensive zone (ASQA% and SA%).

These stats would be solely team stats and would do a good job of allowing coaches to evaluate the performance of their systems.

[Enoch Root]

Another way to look at it is from the goalie's perspective. Instead of trying to catagorize shots that include a pass or whatever, track the goalie and whether or not they had to move for a given shot.

Fully set, moving, moving cross-crease, screened

Or, we just watch the games. For my eye, TAM and WAS immediately come to mind as teams that do a really good job of creating plays that cross the RR.

[Joborule]

It's helpful to have stats quantify the effectiveness of forcing the goalie to move to make saves. It's obvious, but shows that it doesn't matter if you have favorable puck possession stats, if your shots aren't good quality. This is a way to represent the quality of the shots.

Most of the goals from Johnny and Monahan this season that I could recall revolved around puck movement. The last game of the season, consisted of that as well. In order for the Flames to be successful, they'll have to generate more chances that forces the goalie to make tough saves where they have to move laterally, and not allowed to set up.

There was a real lack of one timers this season from below the point, as well not enough second or third chances of rebounds. And for the shots from the point (ideally for screens and rebounds), far to many where blocked. If those three are addressed, it should significantly increase the goals for. Need a increase in quality of shots, rather than just quantity.

[Bingo]

Corsica hockey has expected save percentage using similar metrics, and then from that a differential to actual save percentage that you can sort.

If you look at the negative side, that is goaltenders that have under achieved in save percentage vs the quality of the shots they are facing (1000 min minimum) you get ...

1. Chad Johnson -1.71
2. Craig Andersson -1.70
3. Matt Murray -1.59
4. Carey Price -1.55
5. Scott Darling -1.47

Some big names on there playing behind some good hockey teams and not having great seasons.

I can't vouch for the counting stats behind his methodology but they are very least consistent across the league and he breaks shots into low, medium and high danger.

Top five were Grubauer, Ryan Miller, Bobrovsky, Saros and Carter Hutton ... Crawford, Luongo, Gibson, Rinne, Kuemper rounded out the top ten.

Mike Smith was a +.53 on the season.

[Itse]

Originally Posted by Bingo

Corsica hockey has expected save percentage using similar metrics, and then from that a differential to actual save percentage that you can sort.

If you look at the negative side, that is goaltenders that have under achieved in save percentage vs the quality of the shots they are facing (1000 min minimum) you get ...

1. Chad Johnson -1.71
2. Craig Andersson -1.70
3. Matt Murray -1.59
4. Carey Price -1.55
5. Scott Darling -1.47

Some big names on there playing behind some good hockey teams and not having great seasons.

I can't vouch for the counting stats behind his methodology but they are very least consistent across the league and he breaks shots into low, medium and high danger.

Top five were Grubauer, Ryan Miller, Bobrovsky, Saros and Carter Hutton ... Crawford, Luongo, Gibson, Rinne, Kuemper rounded out the top ten.

Mike Smith was a +.53 on the season.

Have to say, the names on that list make me seriously doubt the methodology. Sure, Murray didn't have the best regular season, but he's no Scott Darling.

[ricardodw]

Originally Posted by Bingo

Corsica hockey has expected save percentage using similar metrics, and then from that a differential to actual save percentage that you can sort.

If you look at the negative side, that is goaltenders that have under achieved in save percentage vs the quality of the shots they are facing (1000 min minimum) you get ...

1. Chad Johnson -1.71
2. Craig Andersson -1.70
3. Matt Murray -1.59
4. Carey Price -1.55
5. Scott Darling -1.47

Some big names on there playing behind some good hockey teams and not having great seasons.

I can't vouch for the counting stats behind his methodology but they are very least consistent across the league and he breaks shots into low, medium and high danger.

Top five were Grubauer, Ryan Miller, Bobrovsky, Saros and Carter Hutton ... Crawford, Luongo, Gibson, Rinne, Kuemper rounded out the top ten.

Mike Smith was a +.53 on the season.

Not sure but I think that the base statistics for their analysis is based wholly on position of the shot (available from the NHL game logs)

The main thing that the OP is presenting is that the quality of the scoring chance is based on prior puck movement.

Ovechkin and Laine have tons of high scoring chances from outside of the quality scoring area ( shooting from the left of the left face off dot) because the puck movement.

The corsica data would say the goalie is letting in goals from a bad shooting area when in fact they are making great saves when these shots do not go in.

[Travis Munroe]

Great post... was looking forward to this from your comment yesterday! This is why I argued so heavily against those who wanted to agree when GG would come out and talk about "bad luck, look at the stats".

I am sure there is a similar argument to be had for PPs

Despite 1,000,000 shots on goal, very few were cross-crease/ice forcing a goalie to move left to right while now having to find the puck (which has already been shot) through multiple layers of bodies. Watching the Capitals in these playoffs has been so entertaining as they are the lowest shooting team in the league but when they do let a shot go, it has a very good chance at scoring. Watching the 17/18 flames it was as if there were rules on no 1 timers, passes of no more than 5 ft and hold the puck for at least 2 to 3 seconds before you decide what the next move is. Its no wonder we had lots of shots and the visiting goalie was in a position to stop 99% of them.

[Travis Munroe]

Originally Posted by ricardodw

Not sure but I think that the base statistics for their analysis is based wholly on position of the shot (available from the NHL game logs)

The main thing that the OP is presenting is that the quality of the scoring chance is based on prior puck movement.

Ovechkin and Laine have tons of high scoring chances from outside of the quality scoring area ( shooting from the left of the left face off dot) because the puck movement.

The corsica data would say the goalie is letting in goals from a bad shooting area when in fact they are making great saves when these shots do not go in.

Exactly and 2 perfect examples.
IMO there is no such thing as a high / medium / low scoring zone on the ice.
Give me 10 one time shots from the point with 2 quick passes beforehand and 4 guys screening the goalie over 10 shots in point blank with nearly every hole covered by bodies on the ice and a perfectly square goalie.

[gargamel]

Originally Posted by ricardodw

Not sure but I think that the base statistics for their analysis is based wholly on position of the shot (available from the NHL game logs)

The main thing that the OP is presenting is that the quality of the scoring chance is based on prior puck movement.

Ovechkin and Laine have tons of high scoring chances from outside of the quality scoring area ( shooting from the left of the left face off dot) because the puck movement.

The corsica data would say the goalie is letting in goals from a bad shooting area when in fact they are making great saves when these shots do not go in.

That's correct (I think), but I don't think there's any way to get data on puck movement without watching every game though, so it's probably not practical to incorporate that on a league-wide stat.

Originally Posted by gunnner

The flames low shooting % could quite simply be because of how little the opposing goalies need to move laterally from flames shots. And this is a result of the flames slow transistion, always time for backcheckers to get back and block the cross ice pass.

The first I heard of Valiquette's Royal Road theory was when it was posted on here to explain why the '14-15 Flames were so much better than possession stats suggested they should be. Under Hartley, the Flames were between 2nd and 12th in the league in shooting percentage each year (7th in his final year). Under Gulutzan, the Flames were 14th then 29th. The team didn't get less talented, so the system must have been at least partially to blame.

Not to turn this into another Peters thread, but hopefully that wasn't the cause of the Hurricanes' low shooting %. They finished 29th, 28th, 20th and 28th in shooting percentage in his 4 years there, though, to be fair, that wasn't much worse than their shooting percentage before he got there (25th).

[Nelson]

Originally Posted by ricardodw

Not sure but I think that the base statistics for their analysis is based wholly on position of the shot (available from the NHL game logs)

The main thing that the OP is presenting is that the quality of the scoring chance is based on prior puck movement.

Ovechkin and Laine have tons of high scoring chances from outside of the quality scoring area ( shooting from the left of the left face off dot) because the puck movement.

The corsica data would say the goalie is letting in goals from a bad shooting area when in fact they are making great saves when these shots do not go in.

That’s a very good point. And that is also why SC% and HDCF% are flawed. You could be standing 2 mm in front of the goalie’s pad and whack the puck into the pad (easiest save possible), but if you’re in the high danger area, then it’s a HDSC%.

We need a metric based on these SQ studies the OP is talking about. I personally would like a SQF metric and a SQA metric, although I guess the expected save percentage could substitute for the SQA metric, even though one measures what I’m interested in more directly. I want to know what kind of shots the team is giving up.

The other thing is that I think you need both shot volume and shot quality metrics. This should not be an either or thing but a both and thing. This is because, ideally, a team would have many quality shots for and few bad shots against.

I know ppl hate CF, etc. because of posters like me who overemphasize them, but they are part of the picture.

[mikeecho]

I came across the Royal Road concept last summer and spent most of the Flames games I watched looking for this. I didn’t sit with a pen and paper tracking it, but the eye test suggested there was very little lateral puck movement across the Royal Road and a lot of shots at squared up goalies.

All season long the narrative was how the system was driving scoring chances (even those “high danger” ones) and how the Flames were outshooting teams. It was only a matter of time until the Flames broke out and started to pile up goals and win games. When that never happened, the usual explanations were 1) a shortage of high end finish (if these guys could just hit the net, they’d score like crazy) and; 2) pure bad luck (low sh% isn’t sustainable... do nothing and it’ll revert back to the mean and the goals will pile up).

1) If you’re always shooting at a goalie that’s square, you’re going to shoot a lot of pucks wide. You don’t have much to shoot at and unless you want to hit him in the chest, you shoot for what little open space you have, and there wasn’t a lot.

2) The dismissal of low sh% as luck and something that will “return to the mean” because luck can’t stay bad. There are no hockey Gods. The shots that didn’t miss the net hit the goalie. Not luck. Once again, because there was very little lateral puck movement and the goalie was square to the shots.

The biggest concern I have with Peters coming in is that his teams traditionally have a lower ranked PP and low sh%. Both areas to me are red flags for lack of lateral puck movement. I haven’t watched enough Hurricanes games to know if that’s a coincidence (maybe his teams really lacked skill and had bad luck ) or a function of system that (like GG’s system) seemed to prioritize shot quantity over shot quality.

[Bingo]

All high danger metrics are based on an addition or subtraction equation based on ...

1) the shot has to be in home plate
2) and get a situational add to it to move it from a scoring chance to a high danger event

I have yet to find what the formula is, and what the adds and subtracts are, but it isn't as simple as every shot in a zone gets to be a high danger, and I suspect the royal road does come into the equation making more valid than many of you are suggesting.