Why are things like missed shots being used, and do shots even matter?

[plusandminus]

I've been away from hockey stats and this board for 4-5 years. I know of "advanced stats", as they were in use back then, and they hopefully has evolved further. But I'm curious about the focus on shots taken/allowed.

Why are people focusing on shots taken, missed shots, etc.?
Is there any proven correlation between shooting as many shots as possible, and between winning games and/or scoring goals?

When I did deep studies, I rather came to the conclusion that the number of shots taken was not a reliable factor to consider.
I also get images of different styles of play. For example, according to old stereotypes, Canadians used to "shoot a lot from all angles", while the Europeans (perhaps mostly the Soviets/Russians) prefered to pass around more before taking quality shots. (For example Sergei Makarov had great shooting percentage during his first seasons in the NHL, although I don't know if it has to do with that.)

NHL and hockey is about scoring goals for, and to prevent goals against.
Taking lots of shots is in itself not necessarily good och adds a teams chances of winning. It's rather the quality of the shots that counts.

For example, quality scoring chances might be a better indicator. (Unfortunately that might involve some arbitrary decisions on what a quality scoring chance is, so that has flaws too.)

Perhaps I will be surprised by the advances of statistics? Maybe deep and skilled studies has actually shown a strong correlation between taking many shots and winning games?

Click to read more...

 

[abo9]

I've been away from hockey stats and this board for 4-5 years. I know of "advanced stats", as they were in use back then, and they hopefully has evolved further. But I'm curious about the focus on shots taken/allowed.

Why are people focusing on shots taken, missed shots, etc.?
Is there any proven correlation between shooting as many shots as possible, and between winning games and/or scoring goals?

When I did deep studies, I rather came to the conclusion that the number of shots taken was not a reliable factor to consider.
I also get images of different styles of play. For example, according to old stereotypes, Canadians used to "shoot a lot from all angles", while the Europeans (perhaps mostly the Soviets/Russians) prefered to pass around more before taking quality shots. (For example Sergei Makarov had great shooting percentage during his first seasons in the NHL, although I don't know if it has to do with that.)

NHL and hockey is about scoring goals for, and to prevent goals against.
Taking lots of shots is in itself not necessarily good och adds a teams chances of winning. It's rather the quality of the shots that counts.

For example, quality scoring chances might be a better indicator. (Unfortunately that might involve some arbitrary decisions on what a quality scoring chance is, so that has flaws too.)

Perhaps I will be surprised by the advances of statistics? Maybe deep and skilled studies has actually shown a strong correlation between taking many shots and winning games?

It'd be great to add distance from net and something like angle from the center of the net. Then shot quality could be quantified instead of qualified

Actually, could the question be asked of "How many low quality shots does it take to match a high quality shot?"

 

[Filthy Dangles]

- Corsi or Shot Attempts were originally established as a way to measure possession, that's why Missed Shots and Blocked Shots, shots that have no chance of resulting in a goal, were used as inputs.

- Nowadays, there's things like Expected goals and shots from the slot/homeplate area which obviously take into account shot quality.

- A big problem with looking at whether winning the shot battle wins games is that teams who are trailing will often force the puck on net more, inflating their shot totals. Add on top of that there's a goalie blocking the net (unlike in a sport ilke basketball) a goalie can completely steal a game the team had no business winning.

 

[plusandminus]

It'd be great to add distance from net and something like angle from the center of the net. Then shot quality could be quantified instead of qualified

Actually, could the question be asked of "How many low quality shots does it take to match a high quality shot?"

Perhaps it could.
Angle and distance are two factors to consider. But also things like the position of the goalie. Sometimes it's basically of no use to shoot at all, if the goalie is covering basically all the net. (Yeah, rebounds may happen, but anyway.)

- Corsi or Shot Attempts were originally established as a way to measure possession, that's why Missed Shots and Blocked Shots, shots that have no chance of resulting in a goal, were used as inputs.

- Nowadays, there's things like Expected goals and shots from the slot/homeplate area which obviously take into account shot quality.

- A big problem with looking at whether winning the shot battle wins games is that teams who are trailing will often force the puck on net more, inflating their shot totals. Add on top of that there's a goalie blocking the net (unlike in a sport ilke basketball) a goalie can completely steal a game the team had no business winning.

I see, possession. And if so, possession can be linked to outscoring opponents?

I may have missed what "Expected goals" are. I may need to read up on that, unless someone explains it.
I have worked with ways of calculating expected goals, but am unsure of if I did it the standard way.

Absolutely, trailing teams tends to shoot more. And also, leading teams then may have a breakaway, which is usually a quality scoring chance vs the opposing goaltender.
And at the end of the game, the trailing team will perhaps shoot even more desperately. Then the leading team gets hold of the puck and sends it into an empty net.

When fairly even teams play each other, and one of them gets a lead by a goal or more, they often will play more defensively, making the trailing team take bad shots.

So I still think shots taken or against aren't a very reliable parameter.

 

[wgknestrick]

Raw shot stats aren't really being used by anyone other than "savy" fans now that the NHL somewhat tracks shot locations. As the NHL improves their data collection technology, the analytics will also improve greatly. Early shot based analytics were equivalent to primitive telescope based astronomy. The science is only going to be as good as your ability to observe and record data. At the time, they were the best we had.

One of the better (and few remaining) sites now. Love the location based graphs
HockeyViz

 

[supsens]

Perhaps it could.
Angle and distance are two factors to consider. But also things like the position of the goalie. Sometimes it's basically of no use to shoot at all, if the goalie is covering basically all the net. (Yeah, rebounds may happen, but anyway.)






I see, possession. And if so, possession can be linked to outscoring opponents?

I may have missed what "Expected goals" are. I may need to read up on that, unless someone explains it.
I have worked with ways of calculating expected goals, but am unsure of if I did it the standard way.

Absolutely, trailing teams tends to shoot more. And also, leading teams then may have a breakaway, which is usually a quality scoring chance vs the opposing goaltender.
And at the end of the game, the trailing team will perhaps shoot even more desperately. Then the leading team gets hold of the puck and sends it into an empty net.

When fairly even teams play each other, and one of them gets a lead by a goal or more, they often will play more defensively, making the trailing team take bad shots.

So I still think shots taken or against aren't a very reliable parameter.

Expected goals depending on who’s model is basically broke down into 2% of shots from the blue line go in, 16% of wrist shots from within 15 feet, 7% of top shelf shots on one times from 20-30 feet and so on. My numbers are not correct it’s just an example..
So at the end of the game they add up how many shots from where and decide how many goals you should have gotten compared to the 10 year average of the league.
The guys that came up with the model state it does not predict future goals nor does it predict the number of future chances.
It’s basically an after the game count that means nothing

 

[JaegerDice]

The biggest issue with measuring performance by goals is that goals are relatively rare events in hockey. This is why plus/minus is a clumsy, often unreliable measure of a player's performance. There's too much noise and variability, too many factors, too few actual measurable data points.

Corsi, which is essentially plus/minus for shot attempts rather than goals, was originally intended as a measure of possession. It turned out to be a fairly good performance stat as well, because shot attempts give so many more data points than goals do. It also proved to have a relatively high correlation with future goal differential... higher than present goal differential or goal scoring rates did. So people opted to use it over plus/minus for goals or raw goal production stats.

Honestly, a lot of it is rather intuitive if you put the buzzwords away and just think about the game of hockey. There is one puck. In order to score goals, you must have the puck. As long as you have the puck, the other team has little to no chance of scoring a goal. Any player (and team) that is successful at taking shots, retrieving the puck after their shots miss, are blocked, or are saved, and taking more shots, is not only going to increase their chance of scoring goals, but also greatly decrease the chance of being scored against. There will always be nights when a goalie stands on his head or your own goalie ****s the bed, but owning the puck and pouring shots on net is a good buffer for those unlucky nights or stretches.

In the time since, more data has become available, and there are more stats out there that measure THREAT rather than simple possession. Expected goals (xGF) is popular right now, though CF% is still useful simply because most of the rule-changes since the lock-out benefit teams that dominate possession and generate shot volume. There are outliers. The 2018 Washington Capitals are a good example of such an outlier. But for the most part, the teams that succeed are great at BOTH CF% and xGF%, while the teams that are good at one but not the other tend to be more scattered around the standings.

 

[morehockeystats]

The problem with Corsi is that if you really begin to judge subjects based on it, then coaches and players start sacrificing passing play in favor of getting shot attempts, and passing defense in favor of suppressing shot attempts.

 

[Doctor No]

Corsi, which is essentially plus/minus for shot attempts rather than goals, was originally intended as a measure of possession.

Minor quibble (I agree with most of your thesis overall) - Corsi was originally intended as a proxy for goaltender workload, since all shots aimed at the net (including missed and blocked shots) require the goaltender to react.

 

[oilerbear]

Do shots matter!


As a high function autistic child I was raised around the greatest thinkers of the 20th century. Friends of my father. Who worked on the design of:
Avro Arrow: (50’s)basis of all modern delta jets.
Area 25: (50’s/60’s) magnetically guided nuclear powered rockets ( deep space travel)
App0llo program: Lunar lander/ service module
Hawk missle retro fit ( to self guided radar)
ERTS: latter called Landsat Satellite program design.
Mag lev train: Commisioning.
Canada arm: Commisioning
ETC.......

They are Triarchial thinkers. Design, Build (Maintain), operate.
They taught me Multivariable thinking.

At a very young age I realized the binary differentiation is a fools game. 50-51 yrs ago around age 4-5 I started to observe my surroundings.

Like my father & friends
I started to take a multivariable Machine action Sequence of events ( SOE) look at sport outcomes. (proprietary theory)

2 critical paths of Action. (Proprietary theory)
Corsi to Corsi: multivariable results.
Corsi ( puck release) to Outcome: multivariable results.

Their can be varying differentiated outcomes in varying Human machine situations.

In 07-08 I presented a grouping of expected performance averages based on 3 diffrent Series of Situation permutations based on Desjardins Team, comp, ZS.
All the binary Permutation people were resistant to it.
However was able to establish expected performance origional based on 8 team x 8 comp x 8 ZS = 512 difrent expected averages.
Later reduced to 4 team (1st,2nd,3rd, 4th line), 4 comp and 6 ZS
96 diffrent expected averages.
Evp/60 varied from 3.15 to .34
+/- goal dif from -26 to +30
Their is no regression to a common average.


They said their was not enough data to regress to a mean.
That is a binary thought process.

You do not reduce analytical resolution to find answers. ( proprietary theory)

You increase The number of variables affecting SOE to come to greater differentiation to first see if their are large affect actors. Some as great as 500%.( proprietary theory)

When you look at a sport like Baseball a largely binary SOE sport. Most teams perform high actors the same way as best as Personel allows.( proprietary theory)


Having shot squirrels ( eat what you kill) at a young age (4-5)
I understood targeting.
I Visually Realized the concept of success rate based on distance and hitting target space. 2 seperate factors.
One a density
and
one a location differentiation.
This is understood by man 10s of thousand of years ago.

49+ yr ago I observed the violence of Flin Flon bombers being about attacking and defending the area in front of the net.

Homeplate theory: ( Proprietary Theory)
Called ( Rickisbox) by blogger on Lowetide.
I realized this was a mass theory.
Shot success Density x shot volume = goal Mass

this lead to a multitude of theories listed below (all proprietary)
- Fwd Team, Comp, NZ def and Coach Face off ZS and Bench change with or without pocession is minimum resolution for determining expected CF and CA.
- Forwards have greater than 100% affect on zone penetration and dictate CF and CA
- avg HD shot ( homeplate) goes in 500% more than avg LD SH ( perimeter) at even.
-d pairs establish expected save% baseline for goalies to perform around based on release point of Corsi.
- goalie performance is a +/- ve save% measure relative to save% baseline established by Dpair.(SOE)
- Elite HD area Dmen.
- Dmen performance must be Measured to their defensive side (higher resolution)
- Flin Flon Bomber tough: Dmen who prevent penetration of Homeplate and forwards who Penetrate homeplate.
- Dmen have an individual expected save% baseline to their side that regresses to their own mean. Cause of multivariable affect.
- forwards regress to their own sustainable shot density based on homeplate.
- listing expected goals shows a complete lack of understanding of the 3 seperate ( not 2) sustainable individual averages for position groups.
Forwards ( expected goals)
Dmen ( baseline save%)
Goalies (+/- measure relative to baseLine save%)
- PDO is embarrassing and provides zero value by not differentiating dman and Goalie performance to each side.
- Offensive Dmen are a forward/ dman hybrid (Rover) who occupy a large% of Forward space.
-Rover off performance must be measured versus forwards.
- most Rovers ( all but top 5-10 each yr) perform at a 4 th line forward even off pace.
- Rovers abandon defence of Homeplate a lot resulting in a high% 3F - 1R - 1D - 1G structure.
- 7 of every 8 final 4 teams are top GA teams that run a higher % of 2D - 1G defence of Homeplate.
- High% 3-1-1-1 structure can be successful in reg season but will usually be weeded out by first 2 rounds in playoffs. See TML.

Do shots matter?
Ah yes?
but!
What about the second action of targeting

45+ yr ago I was playing road hockey with a younger Ron Gunville ( player Personel director of WHL Champ PA Raiders).
All theories below are proprietary theory based on this observation.
- Ron moved with the ball like a table hockey goalie. (
-The ball hit him the goalie at a high rate.
-the ball had 0% chance of going in.
- it’s shot density was 0
- it was a closed shot.
- Corsi that have 0% chance of going in are closed Corsi.
- Closed Corsi = ( blocks + misses + closed shots)
- open shots: shots that hit open open space I. Net elevation.
- you must exclude all closed shot data from analytics to get a true shot density map.
- the new open shot density map must be a combination of 2 densities x/y location from point of Corsi and x,y location in open space in net elevation.
-all current shot based analytics is highly in accurate.
-Elite closed Corsi Dmen: have the highest% of closed Corsi/ corsi faced( caused by fwds).
Dmen like Languay, Stevens, Kris Russell.
-Rovers yeild the highest% of free path open shot densities to their abandoned side.
- Rovers have the worst baseline save% to their side.
-Many Norris winning rovers are the worst defencemen. Having bottom 40 baseline save%

You must not exclude large affect factors in any analytical analysis.
Until all teams chase Personel to play large affect multivariable factors the same way, current small affect binary analysis ( table scraps) is minimal value and high inaccurate.

Open shots matter!
You want Dmen that reduce the open shot density:
Larsson, Benning, Russel, Chara, ........
Not brutal open shot density Dmen who chase offence like
Karlsson, Reilly, Barrie, Giordano, Burns, Klefbom, Bear,
You want elite +ve open shot save% goalies.

Analytics is largely coach & forward driven Corsi to Corsi.
Running NZ transition is a large actor.
Analytics from Corsi to outcome is a complex multi variant SOE.

these are just a few of the multitude of Proprietary Theories of mine.
That lead to my repetative cup core roster theory.
That Vegas selected 100% of the players available in expansion draft that fit that theory.
That I stated on this site in April, before expansion draft, would likely be 2D-1G cup final team.

 

[Filthy Dangles]

The problem with Corsi is that if you really begin to judge subjects based on it, then coaches and players start sacrificing passing play in favor of getting shot attempts, and passing defense in favor of suppressing shot attempts.

Yep. Goodhart's Law.