• The Whos and Hows of 2013 5v5 Shotblocking

    by  • September 22, 2013 • Hockey • 11 Comments

    I think I may have screwed up something in my spreadsheet, as someone has pointed out below. I’ll clarify the issue later this week.

    I was talking to a guy about hockey and stats last week and he said something to me about how he wasn’t a math guy. Not in the sense that he’s opposed to it, just in the sense that he doesn’t really get math. Every so often I get described as a math guy and I kind of laugh at it – my sister-in-law, who has a university degree in math, she’s a math person. I know how to count things and think “If (thing that’s commonly said about hockey is true), what would the statistical record show? Does it show it?” It’s not a math thing, really: it’s a logic thing. People shouldn’t get hung up on the math with this stuff; very little math is actually required. It’s more about trying to think clearly and look for tracks in the snow than anything else.

    I’ve been fooling around with the shot data recently, trying to make sense of shotblocking, and I came across something interesting. I touched on this topic earlier this year and again expressed some skepticism about the value of all the shotblocking that the Rangers were engaged in:

    Over the course of an 82 game season therefore, the Rangers going about their business as they do is going to save them 150 shots relative to the average NHL team and about 260 shots relative to a really shot block averse team like the Kings. If you convert that into goals at the rate of 8%, you come up with the Rangers’ shot blocking saving them twelve goals against per 82 relative to an average team and 20 goals against relative to a team like the Kings. Those are fairly substantial savings. We’re talking about two wins more than an average shot blocking team and three more than the Kings. Buying a player who would make that kind of a difference isn’t cheap. In a capped league, that could be an incredible benefit if you can get your players to buy into it.

    However, I think that those are kind of out of outer limit estimates. To start with, while I’ve assumed that 8% of shots are going to go in, I have a strong suspicion that that’s not true of shots that get blocked. Just from watching hockey, it feels like those shots are more likely to be shots that are taken from the blue line – those shots probably have a less than 8% chance of going in.

    Cam Charron wrote a cool piece a few months back talking about a bunch of stuff, including which players get their shots blocked most. His post is well thought out and worth a read but here’s the money paragraph:

    Conclusion? Shots that are blocked are generally coming from well out, from shooters who score less than 4% of the time anyway. You’d need to block 25 shots before saving a goal, all while hoping that you aren’t screening your goaltender or deflecting a puck. I’m guessing that the players who get in shooting lanes for point shots to block them end up costing their teams more.

    I don’t know that I endorse Cam’s guess there about players who block shots costing their teams more but I strongly suspect his reasoning about the typical shot that’s blocked being one that’s less dangerous is correct. I would bet a lot of money that the 8% outer limit that I used is too high.

    I’m kind of intrigued by the Rangers and Canucks swapping coaches this year because they were dramatically different teams last year in terms of shot blocking. Let’s start with that. I prefer to think about shot blocking in terms of the percentage of shot attempts blocked because if we’re interested in a team’s tendency to block shots, the fact that some teams take 58% of the shot attempts in their games and some teams take 44% of the shot attempts in their games makes just comparing raw totals pointless.

    If we want to measure how some tactic impacts on the game, we need to allow for the circumstances in which that tactic is employed to get a true sense of it. If you don’t do that, your analysis is going to be flawed. So, what does the data tell us about the percentage of shot attempts (shots on goal, blocks and misses) that were blocked in the NHL last year?

    Screen Shot 2013-09-22 at 6.04.20 PM

    The Rangers were the most likely team in the NHL to block a shot attempt, coming in at 30.1%. The Canucks were amongst the least likely in the NHL to block a shot coming in at 23.5%. So the Rangers were 28.1% to block a shot attempt than the Canucks.

    The point isn’t that Tortorella is the only coach in the NHL who asks his players to block shots or that New York invented it. It’s that the Rangers did it way more than other teams. Particularly more than the Canucks.

    I wondered if there were more differences between the Canucks and the Rangers, so I dug into things a little bit more deeply. In his piece, Charron talked about how defencemen were more likely to have their shots blocked than forwards were. That jibes with my experience of watching NHL games. Let’s dig into this in more detail. My sense of things is that forwards are more likely to take shot attempts than defencemen; let’s see if it holds true.

    Screen Shot 2013-09-22 at 6.04.32 PM

    Yep, that generally seems to hold true. Anaheim allowed the lowest percentage of SAA by forwards – 63.4% of the shot attempts at 5v5 against the Ducks came from forwards, which means that 36.6% of the SAA came from defencemen. Flipside of the coin was Carolina, against whom 72.9% of the SAA came from forwards, leaving 27.1% for the defencemen. So, our theory that SAA are more likely to come from forwards is borne out, although there do seem to be significant differences between teams – one additional shot attempt in ten coming from a forward strikes me as significant. If you suspect – as seems reasonable – that shot attempts from forwards tend to be more dangerous than shot attempts from defencemen, then pushing that ratio as low as possible seems to me like a good thing to do, all other things being equal.

    If we think that shot attempts from a defenceman are, on balance, less dangerous than shot attempts from a forward, than it might be interesting to look at whether teams differ in the percentages of shots by defencemen and forwards that they block.

    Screen Shot 2013-09-22 at 6.04.47 PM

    Isn’t that fascinating? The blood and guts Rangers were actually less likely to block a shot attempt from a defenceman in 2013 than were Vigneault’s fey crew of foreigners and dive artists. The differences between best and worst are pretty massive here too – St. Louis was more than 50% more likely to block a shot attempt from a defenceman than the Stars were. That’s a pretty significant difference in how teams play the game that isn’t captured by just looking at whether a team blocks a shot or not.

    But wait – we know that the Rangers were amongst the league leaders in shot blocking and the Canucks weren’t. So how does that make sense? Well – take a look at the percentage of shot attempts by forwards that each team blocked.

    Screen Shot 2013-09-22 at 6.05.09 PM

    Whoa. So there’s your difference between Vancouver and New York. The Canucks didn’t block very many shot attempts from forwards – New York specialized in it. San Jose, Montreal, Anaheim and the Rangers are kind of in their own world in terms of doing this. I’m skeptical of the Shark numbers – their home scorer is trigger happy relative to the rest of the NHL but it’s interesting to me that MTL/NYR/ANA all score very high in blocking shot attempts from forwards and relatively low in terms of the volume of shot attempts from defencemen that they block.

    So when people call the 2013 Rangers a shotblocking team, a more precise way to describe it is that they’re a team that blocks a lot of shots from forwards. That’s what makes them unique amongst NHL teams and what results in them piling up a lot of shot blocks. To circle back around to Cam Charron’s piece, it’s probably fair to conclude that the Rangers block more high percentage shots than the typical team – a shot from a forward typically comes from a more dangerous place on the ice than a shot from a defenceman. So, in addition to having blocked more shots, it may be that the typical shot blocked by the Rangers was higher value than the typical shot blocked.

    There’s a kicker to all of this too. Earlier in the post, I mentioned that the Rangers allowed very few shot attempts from forwards relative to the NHL norm? Look at what happens if you graph the relationship between percentage of shot attempts against by forwards and percentage of shot attempts against by forwards that your team blocked in 2013:

    Screen Shot 2013-09-22 at 6.05.22 PM

    Isn’t that fascinating? Teams that blocked a higher percentage of shot attempts from forwards tended to see a lower percentage of the shot attempts against them being taken by forwards. I caution that this is one year’s data but it’s pretty easy to imagine a reason why that might be: you don’t just decide to block shot attempts by forwards, you do things in the defensive zone that put you into a position to block shot attempts by forwards. Maybe you’re pulling your players in close to your net, clogging the area in front of it.

    Forwards, finding little space to make plays and shoot, might quite reasonably lapse into sending the puck up high towards the blue line, where there’s more space. This might lead to defenders shooting the puck more. So it’s possible, if shots from defencemen tend to a have a lower expected value in terms of goals, that there’s an unseen value in what the Rangers, Ducks and Habs were doing.

    As these things tend to do, this leads to a bunch of questions. The first, and most important, is whether this phenomenon tends to repeat or whether I’m seeing numbers and inventing a relationship to explain the noise. If it is real though – and I’d have to check – it defines the debate about what the Canucks and Rangers did (and might do in reverse this year) a little more sharply. We can see definite differences in tactics and start to ask better questions and get better answers about whether those tactics make sense.

    Email Tyler Dellow at tyler@mc79hockey.com


    11 Responses to The Whos and Hows of 2013 5v5 Shotblocking

    1. OilFanInYYC
      September 22, 2013 at

      I’d like to see the same data on the 05/06 Oilers. The reason being is that I remember Rolison mentioning the adjustment he needed to make when coming to the team. He said something like “I had to learn to look after only the top of the net because the shot blockers took away the bottom of the net.” I’d assume that he was talking more about the in close shot blocks more so than the point shot blocks.

    2. September 22, 2013 at

      Great article, Tyler! Not that this is news to you, but it might be interesting to look at Boudreau’s results in Washington to see if this is something he carried from team to team. You could potentially do the same thing with some of Therrien’s Pittsburgh teams. Might help us to get a sense of whether or not this is a repeatable coaching skill or if it’s much more dependent on the players a coach happens to have.

    3. September 22, 2013 at

      Is there any reason to worry about sample size and bias here? 48 games, shot attempts split by position, all games within conference?

      Since VAN and NYR were somewhat stable over the last few seasons, I’m curious if these patterns– particularly how NYR largely ignores D shots– persist year-to-year.

    4. palefire
      September 23, 2013 at

      As a card-carrying it-pays-for-my-mortgage-and-my-bad-habits math guy, I gotta say — maybe you aren’t a math guy, but “I know how to count things and think “If (thing that’s commonly said about hockey is true), what would the statistical record show? Does it show it?”” puts you way way way ahead of most people.

    5. RiversQ
      September 23, 2013 at

      Interesting bit about Anaheim allowing fewer shot attempts from forwards. I wonder if that is repeatable? If so, that’s interesting fodder for the shot quality issue.

    6. Saj
      September 23, 2013 at

      Nice work! I think this has the potential to be a fascinating series. Most hockey people have likely already made up their minds (with little more than their own anecdotal evidence combined with confirmation bias) about whether it’s better to try to block a shot or just let the goalie see it; but you may actually be able to show one tactic to be superior.

      While you’ve shown that the percentage of F SAA that are blocked varies considerably by team, it is also possible that the shots that do get through have a higher chance of scoring (thanks to more screens and deflections) on teams that block a lot of F SAA. You probably have the data to test this out, and therefore determine the net result: do more F/D SAA blocks result in more or fewer GAA?

    7. Pingback: Spectors Hockey | NHL Blog Beat – September 23, 2013.

    8. Vik
      September 24, 2013 at

      Are you sure about those numbers? I ran them a while ago and I’ve got Vancouver 3rd last in D shots blocked and the Rangers 10th and my totals jive with those on hockeyanalysis. And yours as a whole look really high (the most I’ve got is Montreal at 43.59%). The forward shots are also off from what I’m looking at but not as significantly so.

    9. donair poutine
      September 24, 2013 at

      An interesting question here is whether (or to what degree) this impacts Lundqvist’s perennially high SV%. If the NYR are blocking higher percentage shots than the other team, that would suggest that Lundqvist sees less high percentage shots than the other team, which would suggest that his stats are to some point inflated. While I think he’s certainly an above average goaltender I’m highly distrustful of “good goaltenders” as a thing over the long term and so I’m pretty likely to jump on this.

    10. Vik
      September 24, 2013 at

      So I decided to look into things and believe I found the problem. Basically you’re taking the number of shots a team’s forwards blocked and you’re dividing that by the number of SAA taken by forwards as opposed to just taking the % of blocks on those SAA by forwards. Similarly, you’ve taken the blocks by the team’s D and divided that by the number of SAA taken by D. The total % of shot blocks and the % of shots by forwards graphs both look good but it seems to me the other two are very off and, actually, rather useless.

    11. Tyler Dellow
      September 24, 2013 at

      Thanks Vik. I’ve been fiddling around with a spreadsheet, trying to figure what I screwed up – when I did the 2008-12 numbers, they were totally different.

      I’ll write about this.

    Leave a Reply

    Your email address will not be published. Required fields are marked *