• The Golden Ratio

    by  • September 27, 2013 • Hockey • 5 Comments

    I’ve gotten a bit off track this week but last week I did some writing about SAF/100 and SAA/100, which are attempts to measure how good teams are at keeping the pressure on/stopping the bleeding once the first shot has been given up on a shift. Once you’ve started to look into multi-shot shifts, the other thing that you can start to look into counting is the number of shifts on which a player’s team records or allows at least one shot attempt.

    I assembled comparisons of the ratio of shifts with at least on SAF to shifts with at least one SAA. I separated forwards and defencemen and imposed a minimum of 200 open play Corsi events during the course of the year to be considered.

    Screen Shot 2013-09-27 at 12.57.43 AM

    Screen Shot 2013-09-27 at 12.58.00 AM

    Huh. If you think that that’s look pretty tightly connected, you’d be correct. It’s a correlation of 0.939 for defencemen and 0.952 for forwards. That’s incredibly tight – on par with things like the relationship between goal difference and regulation points. Let’s look at the relationship between open play Corsi% and SAF/100 and open play Corsi% and SAA/100, again for forwards and defencemen.

    Screen Shot 2013-09-27 at 12.58.11 AM

    Screen Shot 2013-09-27 at 12.58.20 AM

    Screen Shot 2013-09-27 at 12.58.29 AM

    Screen Shot 2013-09-27 at 12.58.38 AM

    If you aren’t schooled in looking at graphs like this, the more something likes like a line, the tighter the relationship is. The more it looks like a cloud of nothing, the looser the relationship is. There’s a certainly a relationship between SAF/100 and open play Corsi% ( a correlation of .537 for forwards and .453 for defencemen) and a relationship between SAA/100 and open play Corsi% (a correlation of -.404 for forwards and -0.480 for defencemen) but it’s not nearly as strong as the relationship that exists between shifts with 1+ SAF/shifts 1+ SAA and open play Corsi%.

    What does this mean? Well, it can inform our thinking I think. What this suggests to me is that worrying about giving up or generating multiple shot attempts should take a bit of a back seat to generating or preventing that first one. On the one hand, this seems sort of simplistic. “You’re saying that having more shifts in which you generate at least one shot attempt than where you allow at least one shot attempt is a good thing? Thanks Dr. Hockey Stats.”

    This seems to me to focus the inquiry though, in that it suggests that what we should worry about is how a player contributes to generating or preventing the first shot and, possibly, that coaches should worry more about generating or preventing the first shot. It also allows us to start weighting things in our brain a little better or, if we were doing it right all along, with some more evidence for how we thought about things.

    Take Patrick O’Sullivan as an example. I posted his ratios a few posts again but here they are again:

    Screen Shot 2013-09-19 at 11.01.34 PM

    If the Oilers wanted to know why they got that wrong and sent a scout back to look at video of O’Sullivan in Los Angeles, what might they find? My suspicion is that what they might find is that O’Sullivan was kind of a passenger in terms of doing the things that lead to shots. I assume that most people reading this are familiar with the work that’s been done on neutral zone play and how carries into the offensive zone lead to substantially more shots than pucks that are dumped in.

    What I wonder – and this is a theory that I’m not really in a position to test right now – is whether an analysis of O’Sullivan’s time in LA might find that the players with whom he was playing were doing the work in terms of getting the puck into the offensive zone in such a fashion that shots were likely to result and in preventing entries to the Kings’ zone in a fashion that was likely to lead to shot attempts. This leads us back to the zone entries/zone exits stuff, I think.

    If other people were the ones doing the work, then O’Sullivan would look really good but he wouldn’t be the one responsible for the results and, unless you had players like the players he was playing with, you wouldn’t be likely to get those results if you acquired him.

    I was puzzled about Andy Sutton’s good season in his last season with the Oilers. Here’s what his open play Corsi% looks like during that portion of his career which was played during the BTN era.

    Screen Shot 2013-09-27 at 2.38.11 PM

    So a good year in New York, the bad to horrible years and then a weirdly good year in Edmonton in 2011-12.

    What if we look at his SAF/100 and SAA/100?

    Screen Shot 2013-09-27 at 2.38.22 PM

    Huh. He managed to be over 150 in terms of SAA/100 for four years in a row, including in his good year in Edmonton. When you got a shot attempt with Andy Sutton on the ice, odds were pretty good that more were to follow. That’s not a good thing.

    Screen Shot 2013-09-27 at 2.38.34 PM

    And there it is – the ratio swung pretty dramatically for Andy Sutton in 2011-12 – for the first time since 2007-08, his team was more likely to get a shot attempt with him on the ice than they were to concede one. If you’re a member of the Oilers front office in the spring of 2012, trying to decide whether to re-sign Sutton or cut him loose, this is pretty valuable information to have, I think.

    You’ve refined the question that you need to ask yourself about Sutton’s performance down to: “Why did this ratio shift dramatically for the good in 2011-12?” Maybe the answer is that he happened to be out with Hall against scrubs a lot, I don’t know. If you were gathering good time stamped data on zone entries/exits, you could probably refine things even more. The point though, is that the first step to getting good answers is asking good questions and incorporating this type of data into your analysis lets you ask the good questions.

    Let’s look at the players who played on the 2011-12 and 2013 Oilers through this lens. Forwards first:

    Screen Shot 2013-09-27 at 2.37.44 PM

    Things that jump out at me: only RNH, Jordan Eberle and Teemu Hartikainen improved by this metric last year and Hartikainen improved from terrible to terrible, so I don’t even know that it really counts. Also, while I’m sure it’s coincidence, every single F who didn’t hit at least 0.9 in 2011-12 is gone (caveat on Ben Eager, who didn’t play enough last year to make my list but is back): Hartikainen, Petrell, Jones and Belanger – poof.

    The D? Same thing – ratio is down across the board, except for Whitney, who improved from terrible to terrible.

    Screen Shot 2013-09-27 at 2.37.58 PM

    What does it all mean? Well it’s a framework of sorts, a structure within which to process our observations about hockey players. Knowing what I now know about Andy Sutton and why he had a great Corsi% in his last season in Edmonton, I think I can ask better questions about whether that was likely to repeat and whether I’m paying Andy Sutton for someone else’s skill, in which case I can ask whether I’d be better off getting someone who doesn’t have Andy Sutton’s deficiencies to benefit from Taylor Hall’s (or whoever’s) skill.

    Email Tyler Dellow at tyler@mc79hockey.com

    About

    5 Responses to The Golden Ratio

    1. Bryan
      September 27, 2013 at

      Great read. I like the idea of the single number but wonder how this could be impacted by zone starts to allow for a true replacement value.

    2. Patrick
      September 27, 2013 at

      Very interesting analysis. As I understand it, the Oilers do not have anyone in-house but do use a third party for their analytical work. I believe MacT is a believer in this type of analysis, do you know how much it is used in the NHL today?

    3. speeds
      September 28, 2013 at

      Question for you Tyler:

      Do you think the getting the first shot itself is significant in and of itself, or might zone entries be a confounding variable?

      • speeds
        September 28, 2013 at

        “confounding variable” may be the wrong term there, I haven’t taken stats for awhile, so please forgive me if so.

      • Pierce Cunneen
        September 30, 2013 at

        Well getting the first shot is the outcome. Zone entries/zone exits are part of the process of getting that first shot.

    Leave a Reply

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