• Faceoff Ratios

    by Tyler Dellow • June 8, 2009 • Uncategorized • 18 Comments

    Faceoffs

    The 2007-08 Edmonton Oilers were a lousy hockey team at ES. They were outshot 1984-1710. They were outscored 181-157. Bad times.

    There was a great deal of discussion at Lowetide’s this past week about defensive and offensive zone faceoffs and whether or not they affect a player’s scoring numbers. There seems to be a general consensus that a wide swing one way or the other will have some impact, but nobody seemed to know what the impact would be. I wanted to get a sense of that, so I took a look at the splits from the 2007-08 Oilers after offensive zone faceoffs and defensive zone faceoffs.

    I was surprised at how extreme it is. In the thirty seconds following offensive zone faceoffs, the Oilers were dominant, outshooting the opposition 217-121 and outscoring them 18-10. After defensive zone faceoffs, they were the flip side, getting outshot 315-161 and outscored 26-16. There’s still a marked edge after 45 seconds, which is basically an entire shift. One’s like being the best team in the NHL, the other is like being the worst.

    Assuming that this is in some way typical of NHL teams as a whole and that’s its not an effect created by lousier players tending to get the defensive draws – I would think that it is typical and that offensive and defensive zone draws are spread widely enough – then guys who are taking a lot of defensive zone faceoffs are going to be taking a bit of a hit on their numbers, whereas guys who start a disproportionate amount of time at the other end of the ice are going to look better than they otherwise are.

    How big of an effect are we talking about? Tough to say. The numbers I’ve assembled say that the Oilers and their opposition were taking about 29.3 ESS/60 in the 45 seconds after taking an offensive zone faceoff and allowing about 18.6 ESS/60. Now, sometimes there would be situations where you’ve got multiple faceoffs in a 45 second span, which will drive my numbers down, but there does appear to be a substantial effect. If you figure a .920 save percentage, you can probably figure on something like 2.4 ESGF/60 and 1.5 ESGA/60 as the expected outcome in the 45 seconds following an offensive zone faceoff, with the reverse being true when you start with a defensive zone faceoff. These numbers do strike me as being somewhat low across the board, to be honest – I’m not sure why that is. It seems to me that, while the spread is probably about right, there should be more shots for/against and goals for/against.

    In any event, we know that this isn’t going to smear the results too much for the majority of NHL players. 550 guys were on the ice for at least 500 faceoffs last year. If you look at this as a ratio, offensive/defensive zone faceoffs, 80 of those were at 1.0 +/- .05, 200 were at 1.0 +/- .1, 342 were at 1.0 +/- .2 and 405 were at 1.0 +/- .25.

    If you look at someone like – for example – Shawn Horcoff, you see he took 466 defensive zone faceoffs and 310 offensive zone faceoffs. Compare that with Vinny Lecavalier, who went 239:343 or Evegeni Malkin, who went 238:411. While I’m not saying that Horcoff would have won the scoring title with more favourable ice time, this does seem to me to be pretty significant stuff at the margins, perhaps more significant than I’d appreciated. If my numbers are right, Horcoff had a slight tilt in favour of offensive zone faceoffs in 2007-08. I would bet that if I went and looked, Marty Reasoner did the real dirty work for the Oilers. What do Horcoff’s numbers look like if he takes 285 defensive zone faceoffs and 481 offensive zone faceoffs?

    Only twelve teams had players who posted ratios of 3:2 or higher in more than 500 draws: Tampa Bay (Adam Hall, Jeff Halpern and Matt Perringer), St. Louis (Jay McClement), Pittsburgh (Jordan Staal), Phoenix (Kurt Sauer, Martin Hanzal and Zbynek Mihaelek), Philadelphia (Mike Richards), the Islanders (Brendan Witt, Richard Park, Radek Martinek and Tim Jackman), Nashville (Radek Bonk, Vern Fiddler and Jerred Smithson), Montreal (Mike Komisarek), Minnesota (Nick Schultz, Stephane Veilleux, Cal Clutterbuck, Martin Skoula, Kim Johnsson, Eric Belanger and James Sheppard), Edmonton (Shawn Horcoff and Kyle Brodziak), Colorado (Adam Foote and Scott Hannan), Atlanta (Boris Valabik, Vyacheslav Kozlov, Garnet Exelby, Colby Armstrong, Johan Hedberg (the Thrashers suck) and Marty Reasoner) and Anaheim (Travis Moen). Generally speaking, you’re not going to put up that number without being on a horrible team, or an average team that really dumps its hard work on a few guys.

    Thirteen teams saw guys post a ratio of 3:2 or higher in more than 500 draws. It’s a list full of the young, the stars and third pairing defencemen: Christian Backman, Kris Russell and Jakub Voracek in Columbus, Jim Vandermeer and Mark Giordano in Calgary, Brian Campbell, Jonathan Toews, Patrick Kane, Patrick Sharp, Troy Brouwer and Ben Eager in Chicago, Jordan Leopold in Colorado, Jiri Hudler in Detroit, Marc-Andre Bergeron in Minnesota, Steve Sullivan in Nashille (I assume that the Masterton voters will dock him accordingly), Markus Naslund, Ryan Callahan, Nikolai Zherdev, Nigel Dawes and Lauri Korpikoski for the Rangers, Ed Jovanovski (Glendale should be pointing to him along with Gretzky as wastes of money in their bankruptcy filings) and Kyle Turris in Phoenix, Petr Sykora and Evgeni Malkin in Pittsburgh, David Perron and Patrick Berglund in St. Louis, Martin St. Louis in Tampa (Vinny L just misses), and Alexander Ovechkin, Nicklas Backstrom and Mike Green in Washington.

    This is all a bit of an information dump, but it’s fascinating stuff, I think. I was reading something the other day, about Coleman Analytics, and some GM was quoted saying something along the lines that there was nothing like OBP in hockey. I kind of wonder if an ability to get the puck into the right end of the ice isn’t that – looking at teams like Chicago, Pittsburgh and Washington, they have obviously benefitted tremendously from being able to put their stars on the ice in the right end of the ice. That doesn’t happen without other players making sure that the puck is there when they come off.

    About Tyler Dellow

    18 Responses to Faceoff Ratios

    1. Ryan
      June 8, 2009 at

      A hypothesis for why the overall goal rates after faceoffs seem low: the defending players are usually fresh when there’s a faceoff, because they’d change lines if they weren’t (or the coach would call a timeout if they iced it). A disproportionate number of goals may be scored when the defending players are tired after a long shift. An alternate hypothesis is that a disproportionate number of goals are scored off the rush instead of off the cycle. I suspect both are true to some extent. The first should be testable with shift times (where does the NHL keep that data?). The second isn’t testable with current data, although in the past, I’ve considered watching highlights and tracking it manually.

      And I think you’ve got the right idea in the last paragraph. Using Wings as examples because I know them best, there are guys who push pucks northwards into the offensive zone (Filppula), guys who push O-zone pucks towards the net (Samuelsson), and guys who put pucks into the net (Hudler). The really valuable players, like Zetterberg, do all three.

    2. RiversQ
      June 8, 2009 at

      Tyler, how hard is it to go break it down by FO W/L? It seems to me that offensive zone faceoff wins (def. zone losses) would give you the larger spread you’re looking for.

      I recall Matt took a crack at FO% a year or two ago, but didn’t find much. I don’t think he broke it down by zone though, so I wonder if it makes a big difference here or not.

    3. Marchantfan
      June 8, 2009 at

      Once again, great stuff here. I know it’s my own personal hobby horse, but MacT called all of this last offseason when he said he wanted Marty Reasoner back here. Can you imagine the difference to Horcoff and the Oilers’ counting numbers had he been doing the heavy lifting in the D zone? I’m starting to become concerned that Tambellini fired the wrong person.

    4. June 8, 2009 at

      Good work on an interesting question. Where do goals against come from?

      I suspect there is an effect on player stats who get far more defensive zone faceoffs, but also wonder how big it is.

      By my count, there were roughly 15 opposition scoring sequences against the Oilers this year that started with a defensive zone faceoff loss.

      Your analysis shows that awful team in 2006-07 had as many as 26 goals against after defensive zone faceoffs (a slight different standard, as I only looked at faceoff losses).

      That ’06-’07 team was noted for defenders who could not move the puck, such as Jason Smith. They had one or two pairs who were about as effective as Staios/Strudwick was this year.

      Of course, the 2008-09 Oilers had defenders who could move the puck quite well, such as Lubomir Visnovsky, so that might explain the difference in our two numbers.

      Overall, based on my (subjective) observations, the past two years, I’d say that a third of goals against start with mistakes in the offensive zone (most often bad giveways), another third started in the neutral zone, and the final third started with defensive zone mistakes.

    5. Vic Ferrari
      June 8, 2009 at

      That’s typical, Tyler. For the first half of 07/08, which is the last time I ran a script to gather these, and for the league as a whole at EV with both goalies in the net, :

      The ratio of SF/SA is 2.0 to 1
      The ratio of GF/GA is 1.7 to 1

      Your numbers for the Oilers are:
      SF/SA 1.9 to 1
      GF/GA 1.7 to 1

      The shots ratio was a smidge lower for the Oilers, but that should be expected because the guys with that gig, mostly Stoll but Reasoner too, were pretty good on the dot and good positionally after the faceoff was won or lost.

      Much more important is the fact that this is indicates that the Oilers were territorially dominated, which is a harbinger for future struggles. Which is of course what happened when the puck luck ran out.

    6. June 8, 2009 at

      Worth underlining here again, I think, that Vic’s math has previously shown that where your shifts end — Off:Def ratio, or +/- — is the best predictor we currently have for future EV outscoring. Better than past EV outscoring, better than EV shot diff, even marginally better than Corsi. Position is everything…

      And yes RQ, I did do a bit of (limited) arithmetic on faceoffs back in the day, w.r.t PP/PK. The gist was that:
      - an outstanding FO team would win one “extra” PP FO and one extra PK FO per game, relative to an average team
      - this would improve their goal differential by ~3 on each of PP and PK over the course of a season
      - which translates to ~ 1 extra W over the course of a season

    7. Showerhead
      June 8, 2009 at

      Ryan: another factor would have to be that during a faceoff, you start with all 5 guys in position. No odd man rushes and a whole lot less potential for blown coverage.

      A general thought on players who make sure the puck is in the right part of the ice – kind of makes a guy like Erik Cole, for a familiar example, a bit more valuable on a team with a soft minute killer like Eric Staal than one without like, say, Edmonton.

      This post also leads me to a thought that’s been bubbling under the surface in my mind for a while. I’m beginning to wonder if a player who can exploit the hell out of soft minutes and disproportionate zone time but gives up a lot if caught on defense is actually more valuable, used in context, than many intelligent folk currently believe.

      Now of course you would have to have an appropriate and varied roster build for a coach to be able to use such a player in context, but perhaps I’ll take some time to expand on this idea at some point over the summer. I also still have about 75% of an Andrew Cogliano comps post put together that I would like to get around to finishing as it is definitely a better look than my first attempt.

      Carry on! :)

    8. June 8, 2009 at

      Now, sometimes there would be situations where you’ve got multiple faceoffs in a 45 second span, which will drive my numbers down

      I’m not clear on your methodology here: if say a goal was scored within 45 seconds of a faceoff but there was another faceoff in between, did you count the goal against both draws, or just against the most recent one? In one sense the two plays are unrelated, but in another there can be a continuation, esp. if both faceoffs are in the same zone (i.e. an icing, goalie freezing the puck, etc.).

      In that vein, one niggling concern I have with faceoff data generally is that faceoffs that occur mid-shift are counted along with those at shift start. A player who has the play go the wrong way during his shift resulting in an own-zone faceoff gets “credited” for taking that draw, whereas a player whose line forces an O-zone draw and stays out for it is “debited” for that. While I’m certainly grateful for the stats we do have — faceoffs are an absolute gold mine IMO — I would love to see ShiftStart faceoffs isolated from InShift draws somehow. Among other applications, ShiftStart could be directly to ShiftEnd draws (xfaceoff) to determine a truer ZoneShift. It still wouldn’t account for changes on the fly of course, but in my view it would be closer to comparing apples with apples.

      Only twelve teams had players who posted ratios of 3:2 or higher in more than 500 draws: …
      Thirteen teams saw guys post a ratio of 3:2 or higher in more than 500 draws
      .

      I read those two paragraphs as 3:2 in the defensive zone for the first and 3:2 in the offensive zone for the second. Perhaps it would be a little clearer if you chose defensive zone as the standard, in which case the second list is better described as “2:3″.

      But that’s nitpicking. You’ve done excellent work here, Tyler.

    9. Jake
      June 8, 2009 at

      What do Horcoff’s numbers look like if he takes 285 defensive zone faceoffs and 481 offensive zone faceoffs?

      Using your numbers (2.4 ESGF/60 and 1.5 ESGA/60) and assuming Horcoff’s team performs at this rate while he’s on the ice:

      Horcoff actual
      Offensive zone: 310 faceoffs * 45 seconds = 232.5 EV min * 2.4 ESGF/60 = 9.3 goals
      Defensive zone: 466 faceoffs * 45 seconds = 349.5 EV min * 1.5 ESGF/60 = 8.7 goals
      Total: 18.0 goals

      Horcoff reversed
      Offensive zone: 466 faceoffs * 45 seconds = 349.5 EV min * 2.4 ESGF/60 = 14.0 goals
      Defensive zone: 310 faceoffs * 45 seconds = 232.5 EV min * 1.5 ESGF/60 = 5.8 goals
      Total: 19.8 goals

      Horcoff as posed above
      Offensive zone: 481 faceoffs * 45 seconds = 360.75 EV min * 2.4 ESGF/60 = 14.4 goals
      Defensive zone: 285 faceoffs * 45 seconds = 213.75 EV min * 1.5 ESGF/60 = 5.3 goals
      Total: 19.8 goals

      2 more ES points, assuming he factors in on both goals?

    10. Vic Ferrari
      June 8, 2009 at

      Of course faceoffs are just a snippet in time when the NHL happens to record every player on the ice, and we know for a fact which zone the puck was in.

      If a player was getting a high ratio of offensive zone draws (Malkin, Ovechkin) then in the overwhelming majority of cases they are also being shifted over the boards with the puck heading north as well, or against tired legs on the other team, etc. This is very obvously the case with the two guys listed above in brackets, I think we all see that. And they are by no means unique in this league.

      The vast majority of shift changes come on the fly, but the faceoff zones are largely representive of the role of the player had otherwise as well. Not always, but usually.

      How shifts end doesn’t get much type on the Oilogosphere, but it should. I’m sure that if we used Dennis’ scoring chances copied Roger Neilson circa 1989 and tracked back to who was on the ice prior … players on the whole did a lot better when they followed a Horcoff shift than when they followed a Cogliano, Gagner or Nilsson shift.

      Hemsky should look good in this light as well, Cole too. Visnovsky whould be the bomb and Gilbert and Grebeshkov decent also. Souray not so much. But that’s just my sense of it, and we all have bias. I haven’t seen any hard numbers on this yet.

    11. Vic Ferrari
      June 8, 2009 at

      Jake:

      That’s a swing in expected EV+/- of a bit over 9, no? That’s a whack in one season. And that’s JUST the faceoffs, no account for similar on the fly ice time.

    12. Vic Ferrari
      June 8, 2009 at

      Also Tyler, you can take a more general look at faceoff ratio and shots ratio by team.

      There is a strong relationship there for most teams. The pattern in a scatterplot is pretty obvious. And of course that would be more dramatic if coaches weren’t generally sending their better players over the boards for that gig.

    13. mc79hockey
      June 8, 2009 at

      Re: Vic’s 11 and Jake –

      Using the numbers I gave, I come up with expected change in goal differential as follows:

      3.875 hours of ESTOI following an o draw
      5.825 hours of ESTOI following a d draw

      About 1.95 hours to the bad, at a cost of .9 GD/hour, means he starts a hair under goals behind, by my math.

    14. Vic Ferrari
      June 8, 2009 at

      Ah, I had misread jake’s post as GF and GA. My bad.

    15. June 8, 2009 at

      Vic: Your post #10 is full of interesting observations, the large majority of which, upon reflection, I emphatically agree. One perhaps off-the-cuff statement did raise my eyebrows, however:

      The vast majority of shift changes come on the fly

      That’s suffienctly at variance with my own very rough guesstimate that shift changes are split roughly 50/50 between on the fly or on a faceoff. I don’t have your capability to stripmine play-by-play or TOI sheets which might be able to quantify that — and I’d be most interested to hear of any results derived that way. However I do have shift counts from NHL.com and your own page for “On the ice for shift ending in even strength faceoffs” at TimeOnIce.com.

      Choosing Shawn Horcoff as an example, on the season he played 2204 shifts in all situations. That source doesn’t say exactly how many shifts he played on special teams, but it does specify that of his 1708:48 TOI, 1145:30, or 67%, came at even strength. It won’t be exact but I think it’s a good estimate that about 67% of his shifts ended with the teams at even strength, so 2204 * 67% = 1477 shifts. We know that 786 of his shifts have ended in an ES faceoff, suggesting that only around 700, or slightly How shifts end doesn’t get much type on the Oilogosphere, but it should.

      I refer to that data quite a lot, and find it extremely informative and useful. Just used it again, in fact!

    16. June 8, 2009 at

      (Argh! Formatting error on the above, as I inadvertently inserted a “lesser than” symblo which wiped out a bunch of text. I’ll repost from the top, marking in bold the beginning of the dropped text. Tyler, feel free to delete #15 and leave this one intact. Sorry!)

      Vic: Your post #10 is full of interesting observations, the large majority of which, upon reflection, I emphatically agree. One perhaps off-the-cuff statement did raise my eyebrows, however:

      The vast majority of shift changes come on the fly

      That’s suffienctly at variance with my own very rough guesstimate that shift changes are split roughly 50/50 between on the fly or on a faceoff. I don’t have your capability to stripmine play-by-play or TOI sheets which might be able to quantify that — and I’d be most interested to hear of any results derived that way. However I do have shift counts from NHL.com and your own page for “On the ice for shift ending in even strength faceoffs” at TimeOnIce.com.

      Choosing Shawn Horcoff as an example, on the season he played 2204 shifts in all situations. That source doesn’t say exactly how many shifts he played on special teams, but it does specify that of his 1708:48 TOI, 1145:30, or 67%, came at even strength. It won’t be exact but I think it’s a good estimate that about 67% of his shifts ended with the teams at even strength, so 2204 * 67% = 1477 shifts. We know that 786 of his shifts have ended in an ES faceoff, suggesting that only around 700, or slightly less than 50%, ended on the fly.

      Horcoff is a complex case, however, so let’s choose a player from the other end of the spectrum. Zack Stortini is a guy who played 99% of his ice time (375:13/379:03) at even strength, so let’s just assume all of his 562 shifts ended at even strength. We do know that 287 of them ended in an even-strength faceoff, which again is a little more than half.

      Marc Pouliot played 978 shifts on the season. 518 of them ended with an ES faceoff, which again is a clear majority even without accounting for the 10% of his ice time that came on special teams.

      Similarly, on the blue Ladi Smid for example played 1185 total shifts, and a clear majority of them, 612, ended in an even strength draw. Again, that’s without accounting for his special teams shifts (whh in his case was just 6% of his TOI, thus the adjustment would be minimal). I tested player after player with similar results, but perhaps there’s something wrong with my methodology. Please enlighten me if so.

      Might be interesting to test other teams in a similar fashion, those percentages might vary depending upon the line-matching proclivities of the coach along with other, muddier factors.

      Bottom line is that if a higher percentage of shift changes occur on faceoffs than you surmised, that the faceoff data gives us an even stronger proxy for a player’s overall role.

      How shifts end doesn’t get much type on the Oilogosphere, but it should.

      I refer to that data quite a lot, and find it extremely informative and useful. Just used it again, in fact!

    17. Vic Ferrari
      June 8, 2009 at

      Bruce:

      Good stuff. Intuitively I would have thought that about a third of shifts ended in a faceoff, maybe less. Clearly it is more than half of them.

      I first busted out the ZoneShift thing in a post comparing Zetterberg and Lecavalier over a year ago. Trying to illustrate that Zetterberg was starting a much larger share of his team’s in his own end and that his shifts, on the whole, ended in the offensive end more. And where your shifts end determines where the next line’s shift starts, and that matters. A lot.

      That post was widely linked and almost universally misunderstood.

      As it happened, the on-ice shooting percentage for Vinny (which was unsustainably high) fell into the tank after that post, and the opposite happened for Z, though not to the same extreme. T.B missed the playoffs and Z went on to win the Conn Smythe.

      If there is a measure in hockey that speaks to a player’s character, that is it imo. The fans won’t love you for getting the puck deep at the end of a shift and hustling down to make it tough for the other guys to change, setting up the shift afterwards for a good start to things. You can bring that every game and only a handful of fans in the stands will ever notice.

      On the other hand, end the shift by trying to beat one last guy one-on-one … if it works you’re adored by radio callers, if it doesn’t work nobody will remember anyways, not unless it leads directly, and immediately to a goal against, and that’s damn rare.

    18. June 19, 2009 at

      A few thoughts:

      1) Very much enjoying this discussion. It feels like the cumulative research is getting closer to a predictive model all the time.

      2) Random question directed at Vic (or anyone who knows the answer to this): Do the “end of shift faceoff” data at the TOI site automatically adjust for neutral zone faceoffs caused by a goal being scored at one end–or do I need to make that adjustment myself to get a correct measure of how many shifts really did “end” in the OZ or DZ.

      3) Yes, the Thrashers suck at zone shift, by I do want to thank the Oilers for Marty Reasoner and point out that he won the game when ATL faced off versus EDM last.

    Leave a Reply

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