I’m going to answer a question here, provide a little additional information that I left out of the last post and then, in the following post, get into looking at some Oiler dump-ins with Hall on the ice over the past two seasons. First, the question, from Copper And Blue’s Scott Reynolds:
One thing I’m wondering about is still the size of the samples…Even in some of the areas you’ve highlighted where the change seems significant, we’re actually not talking about very big differences. The 57% of carries that the Oilers achieve zero SAF, for example, is something like 49 of 86 carries (I’m assuming you’ve mixed successful and unsuccessful carries in that table since I’m able to do a better job of reproducing your math that way). If 50% is normal, that would be 43 or 86 carries with zero shot attempts. That’s just not very many carries, and may well be the result of normal variance. It is, of course, possible that I’ve done something wrong in my calculations. Even if I haven’t, I know that increasing the sample means mountains of extra work and that you’re probably eager to move on to check your hypothesis, and that seems reasonable. I just think we need to be careful to remember we’re really only talking about a few shifts difference with this stuff.
Scott is referencing this table:
Let’s clear up this first:
The 57% of carries that the Oilers achieve zero SAF, for example, is something like 49 of 86 carries (I’m assuming you’ve mixed successful and unsuccessful carries in that table since I’m able to do a better job of reproducing your math that way).
I have 46 carries with 0 SAF out of 80 carries in 2013-14 (46/80 = 57.5%). In 2012-13, I have 45 carries with 0 SAF out of 91 (49.5%). Expressed another way, it’s 34 carries with at least 1 SAF out of 80 this year and 46 carries with at least 1 SAF out of 91 in 2012-13. I like to think of these things in terms of binomial distributions and one way of testing this is to ask yourself how likely it would be to get 1+ SAF on just 34 (or fewer) carries out of 80 if your probability of generating at least 1 SAF is .505. I come up with .093 – it’s unusual but it’s not that unusual.
That isn’t the only piece of information that we have though. In a piece a few weeks ago, I noted that Hall has a problem in terms of shifts this year with 0 SAF. Here’s the critical table from that post:
We already know that Hall has a problem in terms of shifts with 0 SAF this year. What I’ve done is basically drill into a selection of shifts and break them into tinier pieces. I think that we need to interpret what I’ve found in light of that piece of information about the bigger picture. If we didn’t have information suggesting that Hall had a big problem this year in terms of 0 SAF shifts (and yet, no problem once the first SAF is generated), I’d be more cautious about these findings. Because I know about the bigger picture though, I’m more comfortable assuming that there’s something to them.
The three trends that I found in the sub-set that I’ve examined – more dumping of the puck, worse retrieval of said dump-ins and an inability to generate a first shot attempt on carries are all consistent with the larger problem that I’ve identified.
* * *
In the course of writing the previous post, I neglected to throw up one other table that I’d generated, dealing with the average length of various types of entry into the offensive zone by the Oilers with Hall on the ice in the sample that I produced. I have included the numbers for the other team just to give us a sense of context. Here it is:
I don’t see a ton in there, for what it’s worth. It would be interesting if we had this data for all games, league wide, because we could get into some really fascinating stuff. I’d love to know if, for example, Chicago has significantly shorter average times for entries into the defensive zone and longer average times for entries into the offensive zone. I’ve no idea what the case is. I suspect that teams like LA and STL might run up the time on successful dump-ins, for example, because there’s generally a repeat of the Western Front from World War I along the wall when they do it.
I’ve kind of hammered on this point repeatedly but getting down to this level is really where this type of analysis is going, I think. Imagine if you had five years of data for every team in the NHL, and knew how teams approached entering the offensive zone/defended entries the defensive zone. There’s basically no discussion of this beyond a few backwater blogs and yet every time it comes up, it seems as if there’s some interesting point being raised with it – Eric Tulsky had a great post about the LA Kings a little while ago in which he explained their phenomenal shot differential by reference to what goes on in the neutral zone.
When this data finally becomes available on a leaguewide basis – and I’m sure that there are teams gathering some variant of it now and that there will be SportsVU data at some point in the next five years – we’re going to be able to get much more specific about why teams outshoot and get outshot. This is kind of a pet peeve of mine at the moment – people speak with great assurance about how to fix some team (“the Oilers need to get bigger” or “You can’t win with a bunch of the same type of players”, whatever that means) and there’s no effective way of challenging what they say. The space for evidence-free blather is nearly endless. It doesn’t have to be that way and likely won’t be for much longer.Email Tyler Dellow at firstname.lastname@example.org