Monday, April 17, 2006

Ken Hitchcock's Playoff PK Theory

Several years ago I remember hearing Ken Hitchcock talk about special teams in the playoffs. He said that familiarity benefited the defender. That the penalty kill would grow stronger as a playoff series progressed. This also because the PKers would have more time to practice and prepare to face the same players, and that practice time for the PP didn't make much difference.

I believed him. It just all makes sense. :-) In fact it gave me reason to believe that the Oilers might have more of a chance against the Red Wings powerplay in the playoffs.

Turns out there isn't much to support it at all. I went out to prove him right just now (albeit in a quick and easy way) and it didn't work out so well.

I wrote a script to grab the goals data from the gamesheets for the last four NHL playoff years (that is as far back as they go).

Here are the results for the last 60 NHL playoff series -->

Obviously there are more game 1s (15 per season) than there are game 7s (about 6 per season). Still, we should be able to spot a trend if there is one. And there just isn't. Unless I've made a glaring mistake somewhere ... there is nothing in it. :-(


Anonymous lowetide said...

That photo is payment enough vic. Was he doing one of those "Nostradamus predicts the world will end Tuesday" TV specials for this photo?


4/17/2006 12:50 pm  
Blogger speeds said...

Are PP opportunies roughly the same in each game as well?

Do they "let more go" as the series progresses"? "Call things tighter"? Or about the same?

4/17/2006 1:08 pm  
Blogger sacamano said...

Yeah, I had the same question as speeds. How do the numbers go if you look at pp efficiency rather than goals?

4/17/2006 2:23 pm  
Blogger Vic Ferrari said...

soeeds and sacamano:

Your thinking is that the refs keep their whistle tucked in their pocket more firmly later in a series, no? Personally I really don't think there is anything in it.

The way the NHL numbers their playoff superstats is a little odd. They use the number 30156 as an example. Where "30" means playoffs, "1" means round 1, "5" means series #5, and "6" means game 6. And just to be awkward the final game of the series is always renumbered to "7", even if it is a four game sweep.

My point, and I do have one:

It would be easy enough to write a script to pick the total PP time in each game for all 60 playoff series since the spring of 2001, but a genuine pain in the ass to sort it by proper game number.

Running the totals for the 60 "1" games against the 60 "7" games should only take 15 minutes or so though. Will that suffice?

4/17/2006 2:53 pm  
Blogger speeds said...

I don't think there's necessarily anything there, but it's tough to evaluate efficiency without evaluating both parts that affect efficiency.

4/17/2006 3:06 pm  
Blogger RiversQ said...

Just to add to sacamano's line of thinking...

Use efficiency as PP goals/hr or PP min/goal because the NHL's PP efficiency isn't worth much IMO.

I think PP minutes will definitely matter here Vic.

4/17/2006 3:13 pm  
Blogger Vic Ferrari said...

To add:

Worse than I thought, this would be a pain. Unless I, or someone out there, feels inspiration I think this "fewer panalties late in the series" thing goes unproven.

4/17/2006 3:21 pm  
Blogger mudcrutch79 said...

Even if the refs put their whistles away-that doesn't do anything for Hitch's point. If anything, it makes it less likely to be true because PPGPG increases. In order for his point to be true, wouldn't there have to be a big increase in game 7 PP?

4/17/2006 3:38 pm  
Blogger Vic Ferrari said...

mudcrutch79 said...

Even if the refs put their whistles away-that doesn't do anything for Hitch's point. If anything, it makes it less likely to be true because PPGPG increases. In order for his point to be true, wouldn't there have to be a big increase in game 7 PP?

Good point. And damn obvious now that somebody has said it.

4/17/2006 3:49 pm  
Blogger Dennis said...

Something else to consider when looking for playoff trends...does season series results matter?

I mean I'm thinking the Oilers will win at least two games from the Wings just because of how closely we played them in the regular season and two of those games didn't feature our upgrades of Tarnstrom, Spacek, Roloson and Samsonov.

So I'm wondering if there's anything to be found from checking season series results?

4/17/2006 4:23 pm  
Blogger RiversQ said...

Well, either Hitch is wrong or this is the wrong way to look at it.

My guess is that his theory comes from counting scoring chances against at 4-on-5 for his teams. Of course, this becomes near impossible to prove and the simple approach of just looking at DAL/PHI results over the past four-five seasons reduces the sample size to uselessness.

4/17/2006 10:23 pm  
Anonymous choppystride said...

VF: Worse than I thought, this would be a pain

Not sure if you meant calculating the actual pp times from the boxscores (yes, that would be a pain...) or downloading the individual sheets of boxscores. If it's the latter, the following might help:

This gives game by game ppg & pp opps:

Trick is in this part: "fetchKey=20043ALL"

2004 = year
3 = playoffs (2 = regular season)
ALL = all teams (instead of CGY, TBL, etc)

Same construct for other views, e.g. RTSS reports (this gives sched-gameId mapping):

Note: beware of rows that point to the same games

4/18/2006 7:46 am  
Blogger Dennis said...

Before I forget...why would Hitch pose for a picture like this?:) Or did they just introduce the shadows in the darkroom?

Of course he does have that stupid look on his face and there's no absolving him of that..

4/18/2006 10:05 am  
Anonymous Anonymous said...

Long time reader, first time commenter.

Just out of curiousity, I looked up some stats from Hitchcock's glory days in Dallas. Lokking at his 1998 and 1999 playoff runs, here is what I came up with for those 2 years:

His PK % in games 1-3 was 87%, while in games 4-7 it jumped to 91%.

Power play goals allowed in games 1-3 was 0.667 per game, while in games 4-7, he had it down to 0.389.

So,it looks like what Hitch says is actually true, for himself. It's just that the rest of the teams aren't able to follow his example. Maybe he should've qualified his statement "that the penalty kill would grow stronger as a playoff series progressed... for well coached teams."


4/18/2006 3:27 pm  
Blogger Vic Ferrari said...


Terrific stuff, thanks. We're stuck with small sample sizes for sure, but it does exonerate him for the purposes of this blog post. :-)


Thanks, it's easy to pull off all the data with that. I was thinking of grabbing the goalie PP icetimes from the event sheets ... but it's not practical to do for a few reasons.

Going by that the PP% by series game number:
1 - 14.0%
2 - 13.7%
3 - 17.1%
4 - 12.2%
5 - 16.3%
6 - 19.2%
7 - 14.4%

Again, nothing in it for the league as a whole.


It is freaky, isn't it. :D I just used google image search and that one popped up on the front page.

4/18/2006 4:25 pm  
Anonymous choppystride said...

I was thinking of grabbing the goalie PP icetimes from the event sheets ... but it's not practical to do for a few reasons

Interesting that you brought this up...

I had the same idea a while back but found those goalie times a bit questionable. In many cases (approx 60/800 where each case is one team per game), those times exceed the PP times in the game summaries, which as you know are not overlap-aware.

I emailed the NHL to ask them why. That was back in December. I'm still awaiting a reply :)

And of course there were several games with completely missing data sheets...

Oh well, I guess that's My NHL.

4/18/2006 6:49 pm  
Blogger Vic Ferrari said...


Yeah, I don't know what the thinking is there. It seems that way with everything, just a little bit missing, or they change columns for PP and SH icetime for a stretch. Or they change the font on the shiftcharts for a two week period. Same old same old. It's like they are cool with putting this info out for everyone to see, but that want to discourage people from writing a script to download data en masse.

I swear the NHL must hire a guy to sit at the end of the production line where they make NHL souvenirs, his sole purpose being to add a small flaw to every NHL coffee cup with a diamond cutters chisel. :D

On the TOI sheets, that seems clean though. I remember you commented on OilFans about it. The shift IDs for whether it was EV or PP ... that's iffy because they don't seem to distinguish when a shift overruns from special teams into EV minutes. I wrote a scipt that peels the info off of them and into an array, same for the play-by-play. I wrote it in Excel VBA. If that works for you just let me know and I'll email it to you. Occasionally they restructure the layout for a game so a player might disappear, but it works the vast majority of the time.

4/18/2006 7:10 pm  

Post a Comment

<< Home