Friday, May 15, 2009

The Impact of Puck Possession and Location on Ice Hockey Strategy

That is the title of this article by Andrew C. Thomas, a must read.

The author followed the Harvard varsity hockey team for a season and recorded some terrific data on possession and zone changes during 5v5 hockey. He then related these to goals scored.

This meshes extremely well with NHL data, but gives a finer look at the 'hows'. In terms of strategy, he found that 'dump and chase' was equivalent to carrying it over the blue line going forward (in terms of expected +/-), and that the same holds true for chipping the puck out of your own end, as opposed to trying to pass/skate it out. This suggests that the ivy league is well coached, and that the players, on the whole, are staying within their roles and abilities (this is my conclusion on the strategy matter, not his).

You can skim over the mathy bits if that's not your thing, as they aren't particularly important. The general idea of still keeping events on the board, even after others have happened ... that's very wise. So if a team gets the puck into the offensive end, and then takes a shot and loses possession, forces the defender into lobbing it back into the neutral zone ... etc, etc ... and a goal is scored 32 seconds later. He would be counting forward from every event to the goal (or to no goal if none were there), as far as 40 seconds. The thinking being that one event (such as getting offensive zone possession) affects the myriad possibilities to come, even if the chain of events doesn't point to that play directly or intuitively.

That's called a Semi-Markov process. Which would make Roger Neilson's stuff from 20 years ago Benjamin Button Semi-Markov (BBSM), where he was looking at the scoring chances and working back through time. The acronym would be better if Brad Pitt's character had been named Benjamin Dutton, but you can't have everything. In fact I would guess that there are only a couple of degrees of separation from Neilson to the person who advised this author on the methodology (the Harvard coach or video coach, maybe? Or an alumni from the professional coaching fraternity?). In any case, by and large he's looking at the right things here. Terrific stuff.

A strict Markov chain is different in that it makes one event alter the nature of the future events, disregarding the events that preceded it. Such as when the score in a hockey game changes from tied to a one goal lead in the early third period ... conventional coaching strategy causes the expected number of total scoring chances to drop from that point forward, and the probability of the trailing team outchancing increases. The score from the first period doesn't affect this a fig, that's the distinction. Or at least that's the way I understand the Markov and Semi-Markov chains. The way of thinking is more important than it's name, in any case.

There is much to be gleaned from this, but the big apple near the bottom of the tree is faceoff data, so I'll grab just it for now. Starting in your own end is a killer; even if you win the draw (much better than losing it, obviously) the prognosis isn't good for the rest of the shift. We all know that I think, but this puts a 'how' with a 'how much'. And the 'how much' is spot on the same for the NHL, or at least for the first half of 07/08, which is all that I checked it against.

Interesting to note also that he acknowledges George Lindsey, an amateur baseball stats pioneer in the 1960s, for his advice. Lindsey's work in baseball is rudimentary by comparison, though it makes me wonder if he would have gone this route with his analysis if he had stuck with his baseball stats hobby. It would have been a monumental task if he had taken it on, baseball is complex.

Thursday, May 07, 2009

Scoring Chances: Part VI of Many. The Problem with the PK

Shots metrics like corsi don't work worth a damn on special teams. But now we have scoring chances.

Dennis was sure that the problem with the PK was the skaters. I thought that Roloson was just letting in too many, or shooters were making their shots, luck and all that. It appears that Dennis was completely right.

The goals per scoring chance during even strength hockey was .22.
The goals per scoring chance during the PK was .22.

So like many other checks on Dennis' results, a few of which I listed in Part I, it's eerily consistent.

Roloson's PPsave% was terrible this year, but it looks like it's not his fault at all. The Oilers not only gave up too many shots, as we all knew, but that they gave up a disporportionate amount of scoring chances per shot. And that's on the coaching staff and the skaters, no way around it.

Contrarian goalie is currently revising a post seeing how EVsave% changed from season to season for goalies that changed teams. It's a clever idea. If he accounts for sample size (for both before and after seasons) I suspect he'll find that there is next to nothing in it, but we'll see. BUT, I think that if the same test is applied to PPsave%, we'll see some dramatic shifts.

Scoring Chances: Part V of Many. Home and Away

Much of the banter on Oilers talk radio revolved around the Oilers being a terrible home team and a good road team this season. By eye, this struck me as nonsense. The Oilers played better at home, just got fewer breaks. Of course if you called that thought into a sports radio show you would be roundly mocked and chided, so I didn't.

But Scott has tabbed up the home and away scoring chances for the Oilers this year, and I have the overall league home and away corsi numbers for 07/08 as used in this post last summer.

So, expressed as a percentage of the total, and reflective of the percentage of the territorial advantage, here is how the Oilers did in terms of scoring chances, home and away, in 08/09. The 07/08 leaguewide corsi numbers serve as a benchmark, the average for the league.

And here you are, just click to enlarge:

And the special teams scoring chance numbers were far superior at home. Both the PK and PP. I suspect that this is mostly because an increase in territorial advantage leads to more PPs and fewer PKs for every team. So home teams get a lot more PP opportunities.

So the Oilers were better at home than on the road. And in the right measure. The evidence strongly suggests that the relatively good road record was just luck.

Tuesday, May 05, 2009

Scoring Chances: Part IV of Many. Anomaly

If you select 38 games from the Oilers season, and then you compare them to 38 other games from the same season, patterns should emerge.

We expect things that are driven by ability to repeat, and using correlation is a convenient way to compare the two sets of numbers. In baseball, this simple repeatability correlation meshes really well with the more complex soultions to the problem of separating luck from ability. And the same should hold true for hockey or any other sport.

If you do this with on-ice EVsave% (using shots directed at net instead of shots on net here, for convenience). There is negligible correlation. 0.02. Nothing there.

So we would expect that the number of scoring chances against, per 100 shots directed at your net ... we'd expect that to not repeat at all. But it does a bit: r = 0.35. Go figure. It's not a big number but it's persistent. 1000 simulations is a lot.

And we would expect that the number of goals against, per 100 scoring chances against ... we'd expect that to not repeat at all. But it does a bit: r = -0.24. Go figure. It's not a big number but it's persistent. And it's negative. Meaning that if you were giving up more goals per scoring chance in one random half of the season (implying that you're giving up higher quality chances) it is likely that the pendulum will swing the other way for you in the other random selection of games. i.e. if a player is giving up more goals per 100 scoring chances in 38 randomly selected games, then there is a better than 50/50 chance that he will give up fewer goals per 100 scoring chances in the other 38 randomly selected games. That doesn't make any sense at all to me.

Use smaller game sets (20 random games vs 20 different random games) and the numbers get smaller but they still persist. r=0.13 for chances per corsi, and r=-0.13 for goals per scoring chance.

Maybe I've just made a mistake in my scripting, but I don't think so. And it probably doesn't matter much to end results. Fine brushstrokes I know. Still, there may be some truth about hockey, something that isn't obvious to the eye, living in this phenomenon.

I'm lost for an explanation.

I suspect that I am about to go on a posting hiatus, hopefully others pick up the scoring chance data from Scott and test some theories.

Scoring Chances: Part III of Many. Scoring Against Minnesota.

For years the Wild goalies have been stopping pucks at a better rate than average. A lot of the credit for this has been given to Jacques Lemaire and his systems play. I suspect that there is something to that, just not as much as most people. Sure, the Wild PK, and especially the fact that they take very few penalties, helps the goalies overall save percentage numbers. They rarely have to kill a 5v3 penalty either.

But at even strength hockey, I think that the goalies deserve mor of the props. Fernandez and Roloson have continued to be very good at stopping pucks at evens since leaving the Twin Cities, and I suspect that Backstrom would also.

On the season as a whole, the Oilers averaged 35.6 scoring chances for every 100 pucks that they shot with malice in the general direction of the other team's goalie. They averaged 65.7 scoring chances for every shot that was on goal (though a great many scoring chances were not recorded as shots on goal).

So if Minnesota really are giving up fewer scoring chances, relative to the number of shots they allow, we should expect the Oilers to have done far worse at generating scoring chances at EV when they played The Wild. Here is how it shook out:

vs League vs Wild
35.6 35.9
65.7 64.9

Pretty damn similar, no? Yeah, I know it's just six games and it might be coincidence. Still, methinks Backstrom deserves more props than he gets from most hockey fans around the league. And it is unimaginable that the Wild don't track scoring chances themselves, so I expect they knew what they had in Backstrom when they signed him to the long term deal this winter.

The Wild are an interesting team. They have been living off of good goaltending and excellent special teams for a while now, though. If they can land the Sedins this summer, that should send a ripple down the lineup, and they should be very strong in 09/10. They are a lot more fun to watch than they used to be, I think the reputation is larger than the facts in this case. The series against COL last spring was some of the best hockey of the playoffs to my mind.

Monday, May 04, 2009

Possession Is Everything

I think that Mike Babcock is right, possession is everything. Damn close to it at even strength, anyways. I also agree with his thinking that being a "puck possession(TM)" team has little to do with coaching style, and everything to do with how good your players are.

Scott sent me Dennis' scoring chances for 77 Oiler games this year, cleanly formatted in game by game fashion. So it was fairly easy to drop the scoring chances by game and player into an array, along with the on-ice goals and on-ice shots directed at net (corsi). All the information used in this post is for even strength ice time only.

From there it's simple enough to grab the top 20 skaters by ice time (Stortini being the cutoff point), randomly select 40 games from the season, and see how well EV scoring chances correlates with goals and corsi in those games.

Here I used percentage of the total, it saves buggering about with ice time and seems more relevant in any case. So if, in those 40 games you were on the ice for 20 goals-for and 30 goals against ... that would be a goals rating of 40%. i.e. 40% of the goals that happened were for your team. Same for scoring chances and corsi.

I wrote a script to randomly select 40 games from the season and calculate the correlation. And then to do the same 1000 more times. And as shown in the chart above, after 40 random Oiler games this past season; the correlation of scoring chances to corsi averages .84, and the correlation to goals averages .59. Both strong relationships.

And then I repeated the excercise for 10 game samples, 20, 30, 50 etc.

The idea is that the more games you're looking at, the more luck washes out. And you can see that over a small batch of games the player's scoring chance rates mesh reasonably well with results (goals) for these 20 Oiler players as a group. But as the sample of games grows larger, near the 77 total, the relationship becomes overwhelming and obvious. And if the season were 40 games longer there is every reason to think that the correlation would grow even stronger, even though there isn't much room to grow at that point.

And corsi, our best guess at territorial advantage using readily available stats, it becomes remarkably similar to scoring chances given enough games of data.

In this season, the on-ice shooting% was spread fairly evenly across the board, so the relationship of scoring chances to goals is stronger than it will probably be most years.

Bottom Line: Players that drive possession at EV drive results at EV. Maybe not the next game the Oilers play, or the next dozen, but eventually. It's unstoppable, ability trumps luck eventually, you just have to be patient.

Saturday, May 02, 2009

Ron Smith on Scoring Chances

Every once in a while I stumble upon a hockey article that is terrific, too good to quote, because everything is interesting.

This article on Ron Smith is an example of that. Former assistant to both Neilson and Keenan, current Hurricanes scout, and a man who has been tracking NHL scoring chances since 1980, this is a bright and articulate man. Just terrific stuff.