Ideas for Future Studies

[Topshelf88]

Always wanted to know if there was ever a study done to find a correlation between second period goal differential and wins/losses. My thought with that is sort of a measure of team speed...is there an advanced stat for that? I'm a noob to all of this.

 

[ChibiPooky]

We were just discussing (briefly) how shot-attempt-based metrics lose their validity if teams are coached just to throw anything and everything on net. Out of that discussion, I started wondering if the paradigm changes if teams were coached to just throw the puck on net rather than continuing the cycle. It seems to me that there's a potential pressure point there.

So first, I think it would be interesting to look at expected number of shot attempts and goals in offensive zone shifts (until stoppage of play in offensive zone or clear) after a decision to shoot versus cycle (from outside the "scoring chance area" - I don't need to know about when a player decides to take a good shot).

Second, I wonder if all of those decisions were converted to one choice or the other (looking at multiple different timespans), how that would affect shot-attempt-based metrics and their correlation to winning.

 

[fasterthanlight]

Hi Folks,

I am very new to the world of hockey analytics, but I have some experience with data analysis. I am home from uni over the holidays and I am already bored

So I am interested in the entry draft. It seems like the place where teams need to make the most decisions, and a very pivotal time for franchises in general. Is it possible to pick out "diamonds in the rough" in later rounds?

My general first pass strategy is as follows: define a binary response variable "has played in more than X NHL games Y years after they were drafted" and train a basic classifier, and see my accuracy. The end goal, however, would be to create a ranking, and compare how well my rankings perform compared to the actual draft rankings.

I have looked over some of the stickies and done some forum searching, but I still have a few questions.
My questions are:
1. Data Sources: Do I need to build an HTML scraper, or has someone already compiled this information?
2. Previous studies: What related previous studies have I missed? I didn't really find much.

Thoughts?

 

[Top 6 Spaling]

Hi Folks,

I am very new to the world of hockey analytics, but I have some experience with data analysis. I am home from uni over the holidays and I am already bored

So I am interested in the entry draft. It seems like the place where teams need to make the most decisions, and a very pivotal time for franchises in general. Is it possible to pick out "diamonds in the rough" in later rounds?

My general first pass strategy is as follows: define a binary response variable "has played in more than X NHL games Y years after they were drafted" and train a basic classifier, and see my accuracy. The end goal, however, would be to create a ranking, and compare how well my rankings perform compared to the actual draft rankings.

I have looked over some of the stickies and done some forum searching, but I still have a few questions.
My questions are:
1. Data Sources: Do I need to build an HTML scraper, or has someone already compiled this information?
2. Previous studies: What related previous studies have I missed? I didn't really find much.

Thoughts?

I am looking to break into hockey analytics as well, and I also want to research the draft. What I am most interested in, however, is what variable is most predictive of NHL performance. Has anyone done a regression test on how independent variables known at draft time (height, weight, points, league) affect dependent variables reflecting NHL performance (GP, G, P, P/G)? I'd be very curious to see if over/under sized players tend to do better or worse, particularly in relation to the average for their draft range (i.e. if undersized players picked between 10 and 20 tend to do well, or something like that).

Similarly, has anything been done that tracks the use of certain buzzwords relating to players and relates that to performance? For instance, players with the term "dynamic" used in their profiles play an average of x NHL games, while players with the term "safe" play an average of y NHL games.

EDIT: Have there been any studies done on how "rushing" players affects their long term development. Maybe for top 5-10 picks you could compare total NHL TOI to career GP or points or something. It wouldn't be perfect, since the really really good players can make the NHL early, but it might be interesting to see if there is statistical evidence that patience pays off. It might even be better to compare total NHL TOI in the first 2-3 seasons after their draft and GP or points to the average for that draft position to normalize talent difference.

 

[kwhyte]

Just a random thought tonight while trying to explain shot based statistics to a relative. Look only at games where the score is tied after 1 period ( or after 2) - how often does the team with the shot lead win?

Since the score is tied, the team with fewer shots has a better goalie, better quality shots, or luck on their side. If it's either of the first two then you would expect them to at least hold their own in the rest of the game. If it's luck, then you'd expect the team with more shots to have an advantage. What do the stats say?

This may sound biased, but we can rephrase it this way - if two teams are tied after one period, how often does the team with the higher save percentage win?

Is this something easy to check on one of the stats sites?

 

[chapulana]

I had read a lot about baseball and basketball analysis, but I'd never been interested in hockey ones until now, reason why I started a blog a week ago or so to delve into some questions. The first thing that crossed my mind (applicable to multiple sports) is how to measure streaks and the regularity of the teams behind them, so I applied some basic calculations and gathered some interesting results. Then I went to explore regularity for the players individually, getting to see who's who in terms of average performance and the regularity with which they get their average numbers, getting to know if they are over or underplaying during the season (sign of possible regression or clear improvement of stats).

I was wondering if anyone know about someone who has been working on stuff like that, as I may try and read their works for further expand mine in the future. If anyone is interested in entering the analytical world of hockey (as I read some of you in the last post here in this thread), we may get in contact and put some thoughts together to get into studying something (I've worked with data visualization too). I'm widely open to work with some other people, even more now that I'm starting, so just let me know here or by PM if you want, so we can start doing something.

 

[normand16]

I would love something like WAR (wins over average replacement) for hockey. There's just so many variables that make that difficult. Doing it with points only wouldn't be hard, but the defensive part would be hard to add in.

I'm not pretending to solve this whatsoever, but I few years back I tried to quantify production amongst forwards using more than just points. Would love to hear everyone's feedback.

http://https://fantasticfore.wordpress.com/2013/09/23/nhl-metric-takeover-forward-production-rating-fpr/

 

[hockeyr5]

I am interested to see how Goals Per Game in a players draft year predicts their success at the NHL level. I was easily able to download draft data from hockeyreference that shows the draft results with career nhl statistics but I am unsure how to easily find a way to get a players draft year statistics without going one by one.

Any ideas?

I also thought it would be interesting to try and see what are the best predictive factors for nhl success based on what league a prospect is coming out of. Example, maybe goals per game in the NCAA but being the team's leading scorer is the OHL or something like that.

 

[notdave]

I am interested to see how Goals Per Game in a players draft year predicts their success at the NHL level. I was easily able to download draft data from hockeyreference that shows the draft results with career nhl statistics but I am unsure how to easily find a way to get a players draft year statistics without going one by one.

Any ideas?

I also thought it would be interesting to try and see what are the best predictive factors for nhl success based on what league a prospect is coming out of. Example, maybe goals per game in the NCAA but being the team's leading scorer is the OHL or something like that.

Here's a look at CHL Draft Year PPG vs NHL PPG. The data used in the graph is for 2001 to 2011 draft years, and only displays players who have played at least 82 NHL games.

Let me know if you'd like the raw data.

 

[Dr Quincy]

I've been in an ongoing discussion on the Bruins board about the addition of Zack Rinaldo. Now, I've found that no matter what evidence I present that he just isn't a good hockey player, I'm continually met with a remark such as:

"Well, as long as he hits and brings energy, he'll help the team."

What I'd love to see someday is a study on "energy" a player brings to a team. For instance, if one could track the shift AFTER a player is on the ice. If such a thing as "energy" exists and if it is a sustainable trait, you would expect to see an increase in scoring or corsi or something in the shift following a player. I suspect that a study wouldn't find such an effect, but I'd love to be able to see what a player's statistics look like following a specific "energy" player and what they look like when following a line without that energy player.

I don't think the proponents of the energy effect will put much stock in results that go against their belief, but it would be nice to have some data.

 

[Ryker]

I've been in an ongoing discussion on the Bruins board about the addition of Zack Rinaldo. Now, I've found that no matter what evidence I present that he just isn't a good hockey player, I'm continually met with a remark such as:

"Well, as long as he hits and brings energy, he'll help the team."

What I'd love to see someday is a study on "energy" a player brings to a team. For instance, if one could track the shift AFTER a player is on the ice. If such a thing as "energy" exists and if it is a sustainable trait, you would expect to see an increase in scoring or corsi or something in the shift following a player. I suspect that a study wouldn't find such an effect, but I'd love to be able to see what a player's statistics look like following a specific "energy" player and what they look like when following a line without that energy player.

I don't think the proponents of the energy effect will put much stock in results that go against their belief, but it would be nice to have some data.

You'd have to look at more than just the next shift, but you'd also have to restrict it to the shifts where the "energy" player actually does something that's supposed to spark the team.

 

[flountown]

I've been in an ongoing discussion on the Bruins board about the addition of Zack Rinaldo. Now, I've found that no matter what evidence I present that he just isn't a good hockey player, I'm continually met with a remark such as:

"Well, as long as he hits and brings energy, he'll help the team."

What I'd love to see someday is a study on "energy" a player brings to a team. For instance, if one could track the shift AFTER a player is on the ice. If such a thing as "energy" exists and if it is a sustainable trait, you would expect to see an increase in scoring or corsi or something in the shift following a player. I suspect that a study wouldn't find such an effect, but I'd love to be able to see what a player's statistics look like following a specific "energy" player and what they look like when following a line without that energy player.

I don't think the proponents of the energy effect will put much stock in results that go against their belief, but it would be nice to have some data.

I've done some cursory googling but was unable to find it. I think Broad Street Hockey did something along these lines. Nothing too in depth, but I think they did put together an article.

Also if you need more ammo, just read through their Zac Rinaldo articles. They were, rightly so in my opinion, not the biggest fans of Rinaldo.

 

[Michael Farkas]

Can any of yous think of a way currently and/or historically to scrape data to judge the timeliness of the goals a goalie has given up...? Without watching them all, I'll never be able to score them on quality, but timeliness of goaltending makes a huge difference in a game...

If it were in like several categories: Goals in the first minute, first goal of the game in the first 10 minutes, goal against within ~90 seconds of a goal for, goals in the last 10 mins of the 3rd/OT (close games)...and then throwaway goals, when games are well in-hand (3+ goals, X amount of time left or whatever)...

How easy or hard would something like that be, ya think?

And can it be done using the Hockey Summary Project (or other means) for older games...?

 

[Top 6 Spaling]

Has anyone heard of a study done on rebound control? It seems like it wouldn't be too difficult to get a percentage by dividing shots that were followed by a freeze within five seconds by total shots faced. It wouldn't be perfect, but it'd give an idea.

 

[Top 6 Spaling]

I'm think I'm going to try to do something along the lines of the above study, where can I find a database of all shots and who the goalie was for that shot?

 

[Matt Ress]

I'm think I'm going to try to do something along the lines of the above study, where can I find a database of all shots and who the goalie was for that shot?

Not sure where to find that data but I'm also interested in better evaluating goalies. If there is a place to find gaa vs short handed gaa? It's tough to quantify puck tracking, consistency, rebound control... and how all those things relate to gaa.

 

[Czech Your Math]

Can any of yous think of a way currently and/or historically to scrape data to judge the timeliness of the goals a goalie has given up...? Without watching them all, I'll never be able to score them on quality, but timeliness of goaltending makes a huge difference in a game...

If it were in like several categories: Goals in the first minute, first goal of the game in the first 10 minutes, goal against within ~90 seconds of a goal for, goals in the last 10 mins of the 3rd/OT (close games)...and then throwaway goals, when games are well in-hand (3+ goals, X amount of time left or whatever)...

How easy or hard would something like that be, ya think?

And can it be done using the Hockey Summary Project (or other means) for older games...?

It would basically be the inverse of the "clutch" metrics that are proposed from time to time. I think some sort of expected win % differential would work best (i.e., Exp. Win % before GA minus EW% after GA... the higher the number the more "untimely" the goal allowed). It's coming up with that formula that would be the difficult part.

 

[Poogs]

Id like to see a stat where we include assists on shots, or even corsi attempts. If you set someone up with a wide open net and they miss the net, all that goes down as is a Corsi. Where as if the shooter just happened to shoot a few inches to one side, he scores, and one or maybe two guys get credited with an assist. So itd be cool to see who assisted on other events other than just goals. Dont think itd be that hard either.

 

[chico_hawk]

Is there a stat available that shows the percentage of rebounds given to shots on net?

One of the key basic metrics for a goalie, imo, is rebound control...obviously the 1st choice is to not give a rebound (of course by not allowing the puck to enter net).

The goalie who gives a lot of rebounds likely tends to face more shots (and probably gives up more goals)....would be interesting to see if the stats bear that out.

 

[BigRangy]

Does anyone have any data on the type of goals a team scores? (off the rush, off the cycle, etc) The skillsets for these types of plays can be a little different and I'd like to know if any players would be noticeably better at one or the other.

Also, it would be interesting to see what the sv% would be on shots off the rush vs shots from the cycle. Could provide good data for teams to optimize their zone entry and transition strategies.

 

[Michael Farkas]

This seems very data scrape-able...perhaps it's been done.

PP conversion after initial faceoff win or initial faceoff loss. Inverse for PK. Remove instances where another penalty wiped out the PP or a period ending.

Would be interesting to see.

 

[Finnflash]

Very new to the boards and this area interests me.

I've been in the area of human behaviour predictability for approximately 15 years. Mostly in the area of criminal behaviour (no not a criminal profiler) and in management (employee performance).

While there are some good metrics to predict future outcomes in these fields (if an offender has an addictions issue and gets out of incarceration then abuses his addiction there is a greater chance he will reoffend) and in employee performance ( if a new employee doesn't have clear role expectations greater chance he will not perform to what's expected) I havent seen to many psychological factors/metrics for athletes or specifically hockey players.

If there are some out there please let me know

 

[Stand Witness]

This seems very data scrape-able...perhaps it's been done.

PP conversion after initial faceoff win or initial faceoff loss. Inverse for PK. Remove instances where another penalty wiped out the PP or a period ending.

Would be interesting to see.

It would be. I have a feeling that it wouldn't be a random result when the stronger teams on the PP generally do score earlier in rather than later and the stronger teams on the PK do kill off more of the penalty.

Also, I wonder what the success rate for PKs/PPs is when the puck is dumped from the zone at certain points of the PP/PK? Like at what point is it essentially 95% certain the PP is killed off? I wonder if clearing the puck at the half way point is almost a certain kill. Could be beneficial to the way teams run their PK if they know that at the half way point (or just after) they are almost certain to have killed off the penalty if they get it out of their zone by that time.

 

[eperry]

Very new to the boards and this area interests me.

I've been in the area of human behaviour predictability for approximately 15 years. Mostly in the area of criminal behaviour (no not a criminal profiler) and in management (employee performance).

While there are some good metrics to predict future outcomes in these fields (if an offender has an addictions issue and gets out of incarceration then abuses his addiction there is a greater chance he will reoffend) and in employee performance ( if a new employee doesn't have clear role expectations greater chance he will not perform to what's expected) I havent seen to many psychological factors/metrics for athletes or specifically hockey players.

If there are some out there please let me know

It shouldn't be too difficult to develop a proportional hazard model for players' propensities to be suspended or fined, provided you can acquire the necessary biodata. I know it's been done in the context of recidivism.

 

[lifeisruff]

With Dan Bylsma as the coach of the Sabres, this question is more relevant then ever for me.

I've always wondered if the concept of "over ripening" is a thing about AHL level prospects.