Let's go to the phones

by Dirk Hoag on Jul 31, 2007 7:25 AM EDT

Now that the bulk of the free agent signings are behind us, and we face a dearth of hockey news leading up to the beginning of training camps, I'm working on formalizing the data that I use into fixed database structures, rather than the ad-hoc spreadsheets I've been using previously. While I'm doing that, I wanted to poll ye dwellers of this vast connection of tubes called the Internet and get your thoughts:

When it comes to statistical analysis, what areas of the game do you think deserve more attention?

My research assistant is standing by!

Remember that I can only deal with the information publicly available, so while it would be neat to know "time on attack" or "successful pass percentage" information, to my knowledge it doesn't exist in any public forum. To jump-start your brain cells, here are a couple things I'm thinking of:

1) Look at team performance in the period of time after individual players take a Fighting major, to see if there is any verifiable "momentum" endowed on a team when particular players fight as opposed to others. Lots of people assume such a boost exists, but can we identify it?

2) Back-testing the PythagenPuck model that was used in November to predict that the Senators would end up alright despite their sub-.500 record at the time, and determine at what time it starts to become a reliable gauge of a team's prospects. Is it 10 games? 15 games? 20?

3) Back-testing the playoff prediction model I used this spring to see if some of the parameters can be tweaked to make it more accurate, and also see how it might have performed during the regular season.

So please, fire away in the comments and let me know where you think the spotlight should be directed...

Tweet Comment 8 comments | 0 recs |

Do you like this story?

Comments

I might expand on your fighting major and momentum analysis by looking at the other momentum swings in a game, like killing off a 5-on-3.

I'm a sucker for meaningful goalie stats, but as they are difficult to cull from the influences of the team in front of them, that may be tough. I might be interested in the # of times a team responded well to having a goalie yanked vs. comebacks with the same goalie in net. Might give some indication of coaches who have an itchy trigger finger not helping the teams in the way they think they are.

by Mike at MHH on Jul 31, 2007 9:39 AM EDT reply actions

I have always been interested in good defensive stats. I think you're post a couple months ago on goal cost to a team was a good start, but I'd be interested in coming up with a formula that measured the effectiveness of defensivemen on 5-vs-5.

Next week I plan on trying on coming up with a normalized version of my OIE stat (On-Ice Efficiency) that better takes into account a team's playing style. I wouldn't mind getting your thoughts on that when I do it next week (this week is international week, on my blog)

by Jibblescribbits on Jul 31, 2007 11:07 AM EDT reply actions

I vote improved playoff model. I took a crack at this but mine crashed and burned badly. You did much better.

by The Falconer on Jul 31, 2007 12:59 PM EDT reply actions

I think a problem with an improved playoff model, is that the playoffs aren't a big enough sample size to give a good indication of success.

It's a big enough sample size that weaker teams will usually play poor for a series of games, so you won't have weak teams winning the cup, but they can play well for long enough to knock off a good team or too. That's why you don't see bad teams winning the cup, but you see upsets happening a lot.

That's why I think you'll have trouble finding a playoff winning indicator.

by Jibblescribbits on Jul 31, 2007 6:38 PM EDT reply actions

Mike: As far as goalie stats go, I tend to think that if I can get the data into the form I want (something I am working on currently), it may pose the most fertile ground for comparative analysis, since we can use Shot Quality to see which goalies are facing close-in vs. more long range shots.

Jibble & Falconer: Whatever I do with the playoff model, I'll backtest it first against this year, and then also run it against 2005-6 and maybe fine-tune it further. You're right, the sample size is smaller, but we'll see if anything useful comes out of it. I won't know until I try...

by The Forechecker on Jul 31, 2007 8:05 PM EDT reply actions

That sounds interesting, but it still leaves the spector of defensive prowess in the mix. The number of close-in shots could well be a function of the system and personnel in front of a goalie, moreso than any relative posative/negative attributes to the goalie's game.

I may have suggested it before in the past, but I'd like to see a comparative analysis among goalies to see who coughs up the most rebounds on a nightly basis, and who are make first saves and safely putting the puck back into play or freezing the puck. It might be interested to see if there is any correlation between goalies that effectively freeze the puck and teams with good defensive zone face-off percentages. That is, are these goalies freezing the puck actually hurting or helping the team.

by Mike at MHH on Aug 1, 2007 10:08 AM EDT reply actions

Good point, you won't know until you try, and it is certainly a worthwhile analysis, even if the answer is : Anything can happen

I like Mike's ideas about goalie analysis too, interesting stuff

by Jibblescribbits on Aug 1, 2007 1:14 PM EDT reply actions

Mike, rebound analysis is definitely something I could push forward on. I did some of that last season, and have enough there to polish up and refine.

As far as goalie analysis goes, the idea with Shot Quality is that you separate the goaltending from the skaters. So if Team A gives up 20 shots from closer range than the 20 shots that Team B gives up, the goalie from Team A isn't penalized within the analysis. The trick there is to figure out a way to work around potential scoring biases in various arenas, as Alan Ryder pointed out over at Hockey Analytics.

by The Forechecker on Aug 1, 2007 6:52 PM EDT reply actions