Tuesday, May 16, 2006

With a Commodore64 and a Couple of VCRs

There are some good comments in the Ron Wilson post below, which sort of got me going on Roger Neilson. The point was raised that maybe some of these guys have been into the stats thing a lot longer than we thought, and I think that is absolutely true. The NHL's fans may not be heavy into stats like many MLB fans, but many of the coaching staffs seem to have been a mile ahead of their MLB counterparts for decades, in terms of hard core analysis.

Things like the TOI sheets (the raw data used to compile the shift charts) has been produced by the NHL for 15 or 20 years. Though they just started publishing them publicly this January.

There was a cool academic paper entitled The Development Of A Data Base Management System For A National Hockey League (NHL) Team published in 1994. It was good because the guy who wrote it knows buggerall about hockey (as is obvious in his reasoning for why the NYR wanted this data) but writes the paper about his involvement in developing the software. This is the only thing he ever wrote on the subject of hockey as far as I know.

Although nobody is named, this is very obviously driven by Roger Neilson. Going by the content, and I think he coached there during the era they did the work during ('89 to '92).

Anyhow, some excerpts below in green, just on the specific info that NYR wanted compiled, my comments in black:

An example of some of the reports that are generated frequently are:
(a) game summary (listing of all events coded)
(b) scoring chances report (listing of all chances for and against
(c) defensive error summary (broken down by player and by reason,i.e., man unchecked)
(d) individual summary (which will range from the simplistic listing of all chances for and against to the very complicated, generating reports and ranking players with respect to their involvement as a subsidiary player on a rush on a power play divided by the average time they spend at the power play).

This is the stuff they had been compiling for yonks, but didn't have the data stored in a useful way.

This section concludes with a partial listing of the types of individual player data that are considered relevant by this NHL team. These are:
(a) number of minutes played
(b) number of scoring chances (both for and against, broken down by type of chance, such as on a power play, and by reason for chance, such as man-unchecked)
(c) the opponent on the other team who is involved in any highlight
(d) performance on face-offs broken down by place on ice and man advantage situation
(e) game particulars from schedule (such as the name of the opponent team and where the game took place)
(f) performance of player ranked by any combination of the above statistics (i.e., number of "scoring chances" a player has during power plays against certain teams as well as against certain players on that team categorized by whether the game was played at home or on the road and by whether they were successful, goal or not, all divided by minutes of playing time in that game for that player).

. . . . . .

Second Generation Statistics
These include the types of data that could be captured through the use of video replay and subsequently analyzed further. Although most of these statistics are now captured in real time through the use of dedicated personnel, it was video analysis that formed the impetus for their initial capture. These statistics can also be found in official league documents and specialized journals They include: face-off performance (by opponent), minutes played, power play minutes, shorthanded minutes, scoring chances, plus/minus, shots on net (by player).

The comment in bold is damn fascinating, or at least I think so. How long had some of these guys been tracking these things if they were into real time capturing of data 15 years ago? Damn.

Third Generation Statistics
Sophisticated computerized database systems and video editing software have enabled the capture and subsequent analysis of a third generation of statistics. For the most part, these statistics are not available for public consumption. These include: scoring chances in first two minutes of the game, players on ice two minutes before a goal is scored, amount of time between bodychecks and goals, face-off performance by location on ice and manpower situation, type of defensive error made (i.e., beaten to the outside on a rush), type of scoring chance that results most frequently in goals, number of scoring chances by player per minute played at even strength, number of scoring chances by player per minute played in game where number of minutes is less than ten compared to games where minutes played are greater than ten.

I've got a lot to say on all of this, for starters that last sentence is fascinating, something magic about that 10 minute threshold ... at least for the Oiler players in the last two years, that's all I've checked. Of course I'm not crazy enough to be counting scoring chances, so can just use EVgoals, EVpoints, EV+ and EV-. Still, it's compelling I think.

I guess the larger point is that most of these coaches are a helluva lot smarter than we fans give them credit for. And have been for a very long time. And though I'm sure that people think that mudcrutch is batshit crazy for compiling those second and third generation stats on his site ... really he's just helping the rest of us catch up with the things that Bowman, Johnson, Neilson, Keenan, Murray, etc have been monitoring for a couple of decades or more.

If nothing else, I hope this is a starting point for conversation.

8 Comments:

Anonymous lowetide said...

I think there may have been guys doing this but there were also guys who weren't. Orval Tessier as a for instance, I be shocked if he had used this kind of info when he was an NHL coach.

I have a bunch of HN's maybe I'll zip back to the 80s to see if there's any evidence of this (although as you state it isn't like they published it or it was common knowledge for the fanbase).

5/16/2006 8:41 pm  
Blogger speeds said...

Awesome stuff Vic, but what a tease with that 3rd last paragraph?

What makes the first 10 minutes so different? How does that difference mesh with the poisson goal distribution I've read a bit about from you, Oilswell, etc?

5/16/2006 10:47 pm  
Blogger Vic Ferrari said...

speeds:

I just ran a simple metric for player results in games where they played more than 10 minutes, or less than ten minutes.

With all the penalties called in the NHL this year that line has moved a lot, but the fact is that players put up much better results by any measure when they break that line, or something close to it.

Granted the last time I checked for this season was February, but Georges, Reasoner and Harvey all had very good numbers on 10+ minutes nights ... and tragic results on nights where they saw very little icetime. You could argue that one feeds the other ... they got more minutes on the nights that they did well because: they were playing well! But mostly I think it's just a case of guys not having a good track record of producing after being cold after 10 minutes (in actual time) or more of sitting on the bench.

5/16/2006 11:39 pm  
Blogger Vic Ferrari said...

LT:

You don't have to go back to Orval, our own Ron Low once remarked that one of the things he missed most about being an NHL head coach as "all the free time". Seriously. There's a guy who'd knock back a couple of beers and bring a bowling ball to a chess game. :D And full credit to the guy, he had his share of success with a pretty dodgy bunch of young Oiler players, credit where credit is due. Not my kind of coach, but there's more than one way to skin a cat.

I've been looking for the Bob Johnson dossier, but can't find it. A grainy pdf file of skewed photocopying, but it's good stuff. It's probably somewhere on my hard drive, but if someone has it please resend. There's some detail for you, pages and pages of what is expected of you as a Flame c/w countless on-ice diagrams ... makes you glad you weren't a fucking defenseman for the guy, sweet Christ.

Dress codes and play by play dissection aside though, two points drive home from Bob. "We win battles" and "Our D don't poach". Which are two pretty good things to have at the top of the list.

Then again, Badger also told us that running downhill would make you a faster skater, and it turns out that all it does is fuck up your knees even worse. So take it for what you will.

5/16/2006 11:51 pm  
Anonymous lowetide said...

Good stuff vic. There's some funny stuff in the NHL's past, suspect if we go back far enough we might find leeches.

One coach who thought the game a lot and had all kinds of rules was Shero. Don't know how sophisticated he was, but I do know he loved to share info and he used to have massive coaching clinics which were basically idea exchanges with coaches from Russia and Sweden and this would have been the 70s for sure.

5/17/2006 7:31 am  
Blogger sacamano said...

You guys kinda freak me out.

5/18/2006 10:21 am  
Blogger sacamano said...

Wait, that comment was meant for this post.

5/18/2006 10:22 am  
Blogger Vic Ferrari said...

LT:

Cool stuff on Shero. You've mentioned him quite a few times, with the classroom style learning ... the drawing out plays and then picking a player to answer "what he would do next". Seems ahead of his time. I remember reading that he decided that practicing the PP too much was killing the spontaneity and confidence of the PP guys, went two years without practicing the PP. And that was 15 or 20 yeasr before Keenan became famous for doing the same in the NHL.

Is there a good book on Freddy Shero, Lowetide?

5/18/2006 10:55 am  

Post a Comment

<< Home