Hockey Analytics

[LastPommerFan]

So how long until players start taking terrible shots all the time during contract years to drive up their Corsi?

[Jeanbe]

Not long at all.

[TrueBlueGED]

So how long until players start taking terrible shots all the time during contract years to drive up their Corsi?

 

Not long at all.

 

Given that shot charts exist, I have a really hard time believing this would be effective in negotiations. Would probably get said player in a coach's doghouse pretty quickly too, which would be self-defeating.

[LastPommerFan]

Given that shot charts exist, I have a really hard time believing this would be effective in negotiations. Would probably get said player in a coach's doghouse pretty quickly too, which would be self-defeating.

Could make arbitration interesting again.

 

Do you think the measurement existing has had/will have no effect on the way the players play the game?

[TrueBlueGED]

Could make arbitration interesting again.

 

Do you think the measurement existing has had/will have no effect on the way the players play the game?

 

You've seen the typical athlete reaction when presented with stats, right? :p

 

I honestly don't think it has happened, nor do I think it will happen. For starters, players already preach about "getting pucks to the net" and "good things happen when you get shots" in hopes of good bounces, so I think a lot of the feared potential behavior from these stats going mainstream is actually already ingrained in players. Secondly, if there were to be an effect, I think it would only be with the marginal players who are trying to pump up their value...the really good to great players will already have a high value by, ya know, being good at playing the sport properly. And lastly, like I alluded to above, I really can't picture a coach tolerating such behavior on a regular basis. I'd say GMs wouldn't be fooled by that type of play either...but the history of NHL contracts tells me GMs are easily fooled :lol:

[carpandean]

Also, while Corsi might be an important stat for player evaluation, it's still going to be behind Goals/Assists/Points.  So, taking a lot of bad shots might not be worth it, since it will hurt the latter.  As TBP said, it will probably affect marginal players the most, since the won't be scoring anyway, so they might as well pump the Corsi.

Edited February 22, 2015 by carpandean

[rakish]

Thanks X, this is all good. My rule is the metric needs to be more than indicating what you are trying to achieve, it needs to actually represent what you are trying to achieve.  If the landowner can manipulate the data, the data isn't very actionable. So, for me, in the end, measuring window space may be accurate regarding your wealth, but it's not an important measurement of wealth because it's not actionable, you can't tax it.

 

So to hockey.  Shots are indicative of how you are playing, but they aren't representative.

 

So how long until players start taking terrible shots all the time during contract years to drive up their Corsi?

 

Lets take a look at 3 charts.  They are Drew Staffords 10-11 (the contract year), 11-12, and 12-13.  In 10-11, he gets to the net.  Following the green line, over 6% of his shots are from 9 feet, 5.5% from 10 feet, 5% from 11 feet, 5.5% from 12 feet, 3.3% from 13 feet.  This is well over league average, the blue line.  If you count up the goals, the green dots, Stafford scored 15 of his 31 goals between 8 and 13 feet from roughly 25% of his shots.

 

In 11-12 his game changes, he is shooting further away from the net.  In 12-13 he is even farther.  He stops scoring.

 

Now the reasons this happened are probably more complex than my analysis (for instance Stafford had very good results playing with Roy, he has terrible results with Hodgson), but I attribute Stafford's decline in production to him being coached to shoot more.

 

Now if Corsi works for you, fine, but Regier failed while being on the forefront of this mindset.

[French Collection]

It would be interesting to see players' possession stats in contract years vs other seasons. The 110% cliche might be accurate in those years.

 

I'm old school so the analytics seem like too much, but I am intrigued by some of it to contradict plain points and plus/minus.

 

I hope the chips will be more accurate for possession numbers because I don't see shots, attempted shots as the ultimate possession number. Some players and teams may have the puck for long periods of time in between shots, while others may pepper the opposition from everywhere. An example of time between shots would be a strong cycling line (Getzlaf,Perry) that can hem you in your end for an entire shift but only get one shot. They're setting up for a good opportunity and wearing you down. The Russian teams of the past played a stick handling and passing game until they had a high percentage shot. I don't have the numbers but these guys seemed to have the puck for 40 minutes but only took 30 shots. They didn't dump and chase but would retreat back in the neutral zone to try a carry in.

[PASabreFan]

It would be interesting to see players' possession stats in contract years vs other seasons. The 110% cliche might be accurate in those years.

 

I'm old school so the analytics seem like too much, but I am intrigued by some of it to contradict plain points and plus/minus.

 

I hope the chips will be more accurate for possession numbers because I don't see shots, attempted shots as the ultimate possession number. Some players and teams may have the puck for long periods of time in between shots, while others may pepper the opposition from everywhere. An example of time between shots would be a strong cycling line (Getzlaf,Perry) that can hem you in your end for an entire shift but only get one shot. They're setting up for a good opportunity and wearing you down. The Russian teams of the past played a stick handling and passing game until they had a high percentage shot. I don't have the numbers but these guys seemed to have the puck for 40 minutes but only took 30 shots. They didn't dump and chase but would retreat back in the neutral zone to try a carry in.

Excellent. The underpinning of Corsi as a measure of possession seems to be that when you're in possession (control might be the better term), you're supposed to be shooting it toward the goal.

[BagBoy]

I think the way we approach power play (PP) stats is overly simplistic. Instead of goals per PP attempt, why not goals per PP time? Better yet, why not goal differential per PP time (to penalize teams for allowing shorties)? Anyone seen or done any analysis on the merits of that metric vs. the traditional goals per attempt metric?

[PASabreFan]

I think the way we approach power play (PP) stats is overly simplistic. Instead of goals per PP attempt, why not goals per PP time? Better yet, why not goal differential per PP time (to penalize teams for allowing shorties)? Anyone seen or done any analysis on the merits of that metric vs. the traditional goals per attempt metric?

OMFG. You're a genius. Thumb through the Sabres Media Guide and there's your name. The Gilbert Number. You'll be famous.

[Dannepanne]

Don't know if this is irrelevant to the thread. But I remember someone said that Larssons advanced stats were good relative to the team. Is this true?

Edited February 23, 2015 by Dannepanne

[French Collection]

Curiosity got the best of me so I went to NHL.com enhanced stats for the first time.

I looked at team stats only. Looked at SAT as the main one, focusing on top teams and some anomalies (Habs, Flames, Canes and Ducks).

I also looked at traditional stats like shots on goal and goal differential. The shots led me to figure out how many actual shots were generated by SAT. Overall goals divided by shots led to team shooting and save percentages.

 

Interesting findings:

 

Preds are 9th in SAT 51.5%. Goal differential is 56.1%. Obvious factor is Rinne's SV% but they also have a 9.5 Shooting %.

Habs are 22nd in SAT 48.9%. Price is right because they have the best SV%. Outshot by 1.6 shots per game. 1 trick pony.

Ducks are 15th in SAT 50.3%. Shooting % is an excellent 9.9% while SV% is good. Can they keep it up?

Isles generate and give up tons of SAT. Good goal differential due to 9.6% shooting. .900 SV% is worrisome.

Bolts have a solid SAT 52.9%. Excellent goal differential may be due to whopping 11% shooting.

Red Wings are 53.2% SAT. They don't generate lots but have 300+ fewer SAT against than the next best club. Solid SH and SV %.

Blackhawks have top SAT differential. Goal differential is even better. Their tough division has them lower in the standings.

Kings are a beast at 54.1% SAT (top %). Low goal differential and standing due to low SH % and so-so SV%.Quick rebound?

Flames have 28th ranked SAT 44.5% yet sit in a playoff battle and rank 15th in overall points. Their goal differential is+15! SH% of 10.6 helps as does a decent SV%. Digging down showed that only 56.1% of SAT against led to actual shots on goal. The next best team was 63.2%!!! They are blocking shots and shooting lanes way better than others.

Canes sit 12th in SAT yet 27th in points. Poor goal differential. SH% of 7.4 was even lower than the Sabres. SV% was poor as well.

Sabres SAT is 37%. The only stat I found to be good was that 72.1% of our SAT end up as a shot on goal, tops among the 11 teams I studied.

 

This was actually a fun exercise for an original six era guy.

[rakish]

I think the way we approach power play (PP) stats is overly simplistic. Instead of goals per PP attempt, why not goals per PP time? Better yet, why not goal differential per PP time (to penalize teams for allowing shorties)? Anyone seen or done any analysis on the merits of that metric vs. the traditional goals per attempt metric?

Check out behindthenet.ca for this

[IKnowPhysics]

Great article on analytics on NHL.com today.  Maybe the the best summation of the utility and intentions of the emergent field that I've seen.  It doesn't go into any math, but rather a solid explanation of why hockey has headed this way.

 

http://www.nhl.com/ice/news.htm?id=754099&navid=nhl:topheads

 

It's worth the full read, but here are selected quotes for the lazy.

 

 

 

The principles of "Moneyball" apply to hockey as much as any other sport, and in order to claim an understanding of the game, executives must be willing to look at it from every angle. That is where analytics come into play. The willingness to discard preconceived notions which don't stand up to strict scrutiny are already playing a role in separating the most successful teams from the rest.

The first step in embracing analytics is to understand them. Only with that base can the misconceptions and biases that skew hockey decision-making at the highest level be appreciated.

 

 

Any analyst would say a player's performance on Thursdays in March or a team's home/road power-play splits are largely meaningless. The misuse of statistics, therefore, is poor analytics, just as taking a player's quote out of context is poor journalism.

 

 

Analytics have a ways to go in differentiating variance from talent, but an inability to recognize clustered randomness in extreme cases has led to some of the biggest management mistakes in hockey history. Being wary of chaotic concepts can give a team numerous wins a season. Falling prey to them can waste millions of dollars.

 

 

"[Analytics] are about effective decision-making with a high reward," said Thomas, the War-On-Ice founder. "If hiring someone for $100,000 right now can get you a free-agent-value savings of a million dollars and more flexibility under the cap, they've paid for themselves right there."

Dallas Stars general manager Jim Nill told Travis Yost of TSN.ca this summer, "We are all trying to get 3-5 percent better. It's a cap world and we are limited. We are always looking for the next thing."

A team can win without analytics, and many will lose employing them, but the additional information and a scrutinized process provide a greater chance at success. As the metrics improve and attitudes shift, this will become more apparent.

 

 

Analytics have led hockey executives, journalists and fans to pose the same question DePodesta asked Beane prior to joining him in Oakland, and they have precipitated a massive shift in the way teams do business.

Rather than exclusively trusting the eye test, condemning players for misfortune in small samples, or labeling players as lazy or enigmatic based on reputation or hearsay, analytics has provided the opportunity to scrutinize decision-making and avoid those characterizations. Dismissing analytics means settling for an obsolete method of evaluation and using an inefficient business practice.

[That Aud Smell]

I read something over the weekend -- I wish I could find it, recall where I saw it -- but it was a writer (McKenzie?) relating a story from a GM (Coyotes?) who explained how #fancystats helped confirm the organization's understanding of two players between whom the organization needed to choose -- a so-called shutdown D-man and a puck moving D-man who didn't play overly physical. The analytics helped the team understand that the shutdown guy was always lowering the boom on people and mixing it up in tight space because he wasn't moving the puck effectively (and therefore his ratio of defending versus possessing was like 70-30), while the puck mover didn't do a ton of defending because he was so frequently to the puck first and then made a good play with it (and his ratio of defending versus possessing was more like 30-70).

 

I will try to find it. It was just an interesting food-for-thought read.

[dudacek]

I read something over the weekend -- I wish I could find it, recall where I saw it -- but it was a writer (McKenzie?) relating a story from a GM (Coyotes?) who explained how #fancystats helped confirm the organization's understanding of two players between whom the organization needed to choose -- a so-called shutdown D-man and a puck moving D-man who didn't play overly physical. The analytics helped the team understand that the shutdown guy was always lowering the boom on people and mixing it up in tight space because he wasn't moving the puck effectively (and therefore his ratio of defending versus possessing was like 70-30), while the puck mover didn't do a ton of defending because he was so frequently to the puck first and then made a good play with it (and his ratio of defending versus possessing was more like 30-70).

 

I will try to find it. It was just an interesting food-for-thought read.

Subtitle: why Mark Pysyk is better than most of the defencemen ahead of him on Buffalo's depth chart.

[TrueBlueGED]

I read something over the weekend -- I wish I could find it, recall where I saw it -- but it was a writer (McKenzie?) relating a story from a GM (Coyotes?) who explained how #fancystats helped confirm the organization's understanding of two players between whom the organization needed to choose -- a so-called shutdown D-man and a puck moving D-man who didn't play overly physical. The analytics helped the team understand that the shutdown guy was always lowering the boom on people and mixing it up in tight space because he wasn't moving the puck effectively (and therefore his ratio of defending versus possessing was like 70-30), while the puck mover didn't do a ton of defending because he was so frequently to the puck first and then made a good play with it (and his ratio of defending versus possessing was more like 30-70).

 

I will try to find it. It was just an interesting food-for-thought read.

 

I remember a fairly significant debate I was involved in about whether Regehr or Sekera was better...well, this is the core of the reason why I thought Sekera was better. Just very different perspectives on what is valuable.

[MattPie]

I read something over the weekend -- I wish I could find it, recall where I saw it -- but it was a writer (McKenzie?) relating a story from a GM (Coyotes?) who explained how #fancystats helped confirm the organization's understanding of two players between whom the organization needed to choose -- a so-called shutdown D-man and a puck moving D-man who didn't play overly physical. The analytics helped the team understand that the shutdown guy was always lowering the boom on people and mixing it up in tight space because he wasn't moving the puck effectively (and therefore his ratio of defending versus possessing was like 70-30), while the puck mover didn't do a ton of defending because he was so frequently to the puck first and then made a good play with it (and his ratio of defending versus possessing was more like 30-70).

 

I will try to find it. It was just an interesting food-for-thought read.

 

 

I remember a fairly significant debate I was involved in about whether Regehr or Sekera was better...well, this is the core of the reason why I thought Sekera was better. Just very different perspectives on what is valuable.

 

The anti-stats folks hate the concept that "hits" aren't necessarily a good thing, because in order to make a hit you by definition don't have the puck (same for blocked shots). It's easy to say, "well that's stupid, of course hits are good", but the logic is there. As with many things, you shouldn't look at the raw numbers since those are easy to skew. Raw Blocked shots isn't a useful as, say, percent of shots blocked while player X is on the ice crossed with TOI, but the the anti-stats folks are still stuck on "blocked shots ARE always good!". I'll take the guy that blocks 2 of 4 shots while it's on the ice vs the guy who blocks 5 of 10 in the same amount of time.

 

I'm an engineer, I like to think that most of the physical world can be described by math in one form or another. Sometimes we don't understand the math or even know what to measure, but I really do think there are more ways to measure hockey than just goals and "that guy looks good tonight".

Edited February 23, 2015 by MattPie

[LGR4GM]

I read something over the weekend -- I wish I could find it, recall where I saw it -- but it was a writer (McKenzie?) relating a story from a GM (Coyotes?) who explained how #fancystats helped confirm the organization's understanding of two players between whom the organization needed to choose -- a so-called shutdown D-man and a puck moving D-man who didn't play overly physical. The analytics helped the team understand that the shutdown guy was always lowering the boom on people and mixing it up in tight space because he wasn't moving the puck effectively (and therefore his ratio of defending versus possessing was like 70-30), while the puck mover didn't do a ton of defending because he was so frequently to the puck first and then made a good play with it (and his ratio of defending versus possessing was more like 30-70).

 

I will try to find it. It was just an interesting food-for-thought read.

This sounds like a good explanation of Mark Pysyk versus Jake McCabe.

Subtitle: why Mark Pysyk is better than most of the defencemen ahead of him on Buffalo's depth chart.

exactly.

[That Aud Smell]

I remember a fairly significant debate I was involved in about whether Regehr or Sekera was better...well, this is the core of the reason why I thought Sekera was better. Just very different perspectives on what is valuable.

Regehr at the peak of his game would have been a more interesting comparison to Sekera. And I take, and agree with, your point.

 

There's a natural hurrish-durrish man-love response to the D-men whose game features the smearing of opponents into the glass. That response creates a pretty significant bias.  

[X. Benedict]

Regehr at the peak of his game would have been a more interesting comparison to Sekera. And I take, and agree with, your point.

 

There's a natural hurrish-durrish man-love response to the D-men whose game features the smearing of opponents into the glass. That response creates a pretty significant bias.  

 

Really you need both types in the mix. Brian Campbell and Teppo Numminen could flip the ice in a second with one or two touches, but neither was very accomplished at crease-clearing when things got tight down low. 

[That Aud Smell]

Really you need both types in the mix. Brian Campbell and Teppo Numminen could flip the ice in a second with one or two touches, but neither was very accomplished at crease-clearing when things got tight down low. 

 

Quite right.

[LGR4GM]

Really you need both types in the mix. Brian Campbell and Teppo Numminen could flip the ice in a second with one or two touches, but neither was very accomplished at crease-clearing when things got tight down low. 

I agree. I think we actually are seeing Bogosian as a hybrid.  He is a nasty peice of work but his outlet passing and ability to get to the puck are also quite useful. 

 

I think hits for defenders are more valuable than forwards because defenders are typically hitting to seperate someone from the puck who is entering for already in the zone.  Forwards through hits more often in the offensive zone.  This could be wild speculation on my part but I wonder if there is any correlation to hits by a defender and shot attempts by the opposing team. 

[That Aud Smell]

I agree. I think we actually are seeing Bogosian as a hybrid.  He is a nasty peice of work but his outlet passing and ability to get to the puck are also quite useful. 

 

As with most things, it's not generally all or nothing -- it's a continuum. Bogosian appears to be a nicely balanced D-man.