• Shootout Replacement Level for Shooters

    by Tyler Dellow • April 13, 2009 • Uncategorized • 11 Comments

    This is something that I’ve kind of been wondering about, given that Steve Tambellini is about to make a decision about the appropriate price to pay Ales Kotalik, assuming that he wants him to come back. In order to understand how valuable Kotalik’s skill at putting the puck in the net during the shootout is, you need to understand how often NHL players as a group can do this. The particularly relevant group are the players who a coach will be choosing as his options in the event that he doesn’t have Kotalik available.

    David Staples linked to a piece on NHL.com the other day with some quotes from Tambellini and Jim Rutherford about the shootout that’s worth taking a look at. There were a couple of paragraphs in particular that caught my eye:

    NHL General Managers are in agreement that if a player wants to keep his job in the League, he can’t only excel as a shootout specialist, but it sure helps his market value if he can score consistently in the breakaway competition.

    In most cases, shootout prowess has become a small part of the player evaluation process.

    “I think you could carry a guy on your fourth line that you consider a specialist, a real solid fourth-line player that can be a difference-maker in the end,” Rutherford said. “It would kind of be like in baseball where you have the pinch-hitter. He doesn’t play every day, but he’s a guy that can come off the bench and make a difference in the game.”

    SO

    The chart at left summarizes the shootout results by spot (not round, although OBVIOUSLY 1 and 2 are Round 1 etc.) in the shootout since the introduction of the shootout in 2005-06. I haven’t seen this before but this, in my mind, is the critical piece of information in figuring out when a guy starts adding value to your team in the shootout. The last column, “Over”, deals with how the shootout is finished after that shooter in the rotation. So, for instance, you can see that 12.7% of shootouts are finished after four shots and 38.8% of shootouts are finished after five shots. Obviously, 26.1% of shootouts ended precisely on the fifth shot.

    There was some discussion in various places about the decision to shoot first and I suppose that I could see some argument that it makes sense to choose one over the other if your goalie is very good or very bad at it – if the shootout can be finished with your team having a disparate number of shots and your goalie is either way above the norm or way below, it seems to me that there’s some argument that you want to maximize/limit his exposure. As you’ll see from the chart, 26% of shootouts are ended with the fifth shot. With that said, I haven’t gone and checked to see if the numbers add up and I kind of suspect that there’s no difference, or that it’s negligible.

    In any event, what really interested me was the exceedingly small difference between shooters 1-6 and shooters 7-12. Shooters 1-6 scored 1153 goals on 3423 shots (33.7%); shooters 7-12 scored 234 goals on 708 shots (33.1%). There’s some dropoff after that – shooters after the twelfth shooter scored 42 goals on 182 shots (23.1%) but they appear pretty infrequently – only one game in twenty goes beyond twelve shooters. As the average team plays 10 or 11 shootouts annually, that’s one shootout that extends beyond twelve shooters every year or so.

    To come back to what I was really interested in though, my sense is that the replacement level for the shootout is probably pretty high. If I was trying to put a value on Kotalik’s contribution in the shootout – and I have some ideas about that – I’d probably be starting from the perspective that I could get 33% shooting from someone else. My preliminary analysis comes up with a 50% shooter being worth something like .7 points or so above what a 33% shooter is worth, assuming 10 shootouts a year and that he’s healthy and available for all of them.

    To come back to Rutherford’s point above, that there’s room for a solid fourth-liner type who can get it done on the shootout…I’m not even sure that he’d have to be a solid fourth-liner. A guy like Kotalik, as long as he can play on the fourth line at a replacement level and maintain his shootout performance, he’s a useful player.

    About Tyler Dellow

    11 Responses to Shootout Replacement Level for Shooters

    1. April 13, 2009 at

      “It’s huge on so many different levels,” Tambellini told NHL.com.

      That quote kinda baffles me. Stipulating that it’s huge, isn’t it huge on one level? Like, exactly?

    2. April 13, 2009 at

      Matt: You aren’t allowing for MOMENTUM and CONFIDENCE.

      When you have Ales Kotalik on your bench, players are CONFIDENT so they don’t feel any pressure to, umm, score.

      And there’s a huge difference in team MOMENTUM when you win over when you lose, so Ales Kotalik could single-handedly steer the team clear of the “vortex of death”.

      Plus, there’s the actual, you know, shootout. So that’s three areas ;)

    3. lowetide
      April 13, 2009 at

      I think team should just choose shooters with names like random, chance and luck.

    4. Showerhead
      April 13, 2009 at

      The shoot first/second debate: are you suggesting that shooters and goaltenders alike will have their abilities vary dependent on what the shootout’s current score is? IE that a shooter could be better when ahead or behind as opposed to the same regardless of state? Are you suggesting that the guy who shoots 3rd might be better or worse than guys 1 and 2? I just can’t see it mattering in the end personally is all.

      Re: Kotalik, are you saying that a below average 4th liner isn’t capable of costing his team 0.7 points in a season as well?

    5. mc79hockey
      April 14, 2009 at

      Re: Kotalik. I prefer to think of things in terms of replacement level. The replacement level fourth liner absolutely costs him team points as opposed to, say, someone wiht first line talent playing a fourth line role. My points was that even if he’s a replacement level player, which is pretty bad, so long as he’s got his SO skills, he’s a useful player.

    6. April 14, 2009 at

      he’s a useful player

      Define useful, 2.5? Another point for the Oilers is that as long as you are winning under 60% of your games in regulation (as the Oilers likely will), going to OT as many games as possible will maximize your points. If your comfortable in SO, like the Oilers were with Garon, the team will end up getting more points without being a better team.

    7. David Staples
      April 14, 2009 at

      How much do you pay a replacement level player in the NHL? The league minumum? Twice that?

      My sense is that Kotalik is going to be asking for a helluva lot more than that . . . Will his shoot-out skills be worth, say, a million a year?

      If a replacement guy gets, says, $1 million a year, is Kotalik worth $2 million (I have no idea how much he is asking for, but I’ve seen some bloggers/posters speculate he might get $2-$3 million a year, which strikes me as way too much . . . but then I’m not the biggest fan of this particular player, and certainly not at that price in this market for a team with these cap issues)?

    8. mc79hockey
      April 14, 2009 at

      How much do you pay a replacement level player in the NHL? The league minumum? Twice that?

      Replacement level is the expected level of performance the average team can obtain if it needs to replace a player at minimal cost. Generally, you assume that you’d pay a replacement level player the league minimum. Any more than that is inefficient, because there should be other options that are cheaper.

      My sense is that Kotalik is going to be asking for a helluva lot more than that . . . Will his shoot-out skills be worth, say, a million a year?

      I figured that, in a league where the players were getting $1.4742B, the average cost of a marginal point in the standings was $883,243.45. Marginal points are those above replacement level.

      Now, the CBA screws things up, because you basically have tiers of players. Guys on entry level deals are capped and you’re going to get better bang for the buck with their deal than you will with a UFA like Kotalik (or Horcoff, for all of the people who insist on pointing out what Mike Richards or Ryan Getzlaf makes when discussing him). I am virtually certain that the $/marginal point for UFA years is going to be lower than it is for entry level or RFA years.

      So, in Kotalik’s case, if he was a replacement level player and you then expected him to add .7 marginal wins in the shootout to your team, you can justify league minimum + (.7)*($883,243.45) = $1.1MM, if you’re valuing points in the standings at the average cost/marginal point. The number almost certainly increases with UFA’s, although I’m not in a position to say by how much.

      That, of course, assumes that he’s a replacement level player otherwise, which he’s almost certainly not. I could easily see him getting $2MM to $3MM annually. Whether he makes sense in Edmonton, with the sheer volume of smallish (whether in stature or in game) forwards that the Oilers have, is another question. If I was running the team, I’d probably be inclined to let him go. The Oilers have a number of guys who look like they have some potential as shootout artists in O’Sullivan and Nilsson, they have bombers from the point in Lubo and Souray…I just don’t see where Kotalik’s skills make sense.

      With that said, he’s probably a great deal as a UFA somewhere else if they can get him for $2MM. I assume that most teams don’t look at the problem this way (I don’t think that most teams try to put a dollar value on the contributions a player makes as opposed to looking at how the market values a player with similar statistics, which isn’t the same thing, and the Oilers don’t even do that), but (it seems to me) it’s the only really rational way to do it.

    9. April 14, 2009 at

      Interesting stuff, Tyler. Gabe Desjardins brought this up on his blog awhile back. Thanks for answering my question in the Comment (singular) section. There is a lot of information in your table, which allows general conclusions about the first 6 rounds and then very specific conclusions about what happens in the sudden death rounds. The numbers are still small enough that random noise may be the biggest factor, but they’re interesting nonetheless.

      Penalty shots have traditionally been successful about 1/3 of the time. That has hardly changed in the shootout era, at least among competent shooters, which you show pretty conclusively run about 6 deep per team.

      In any event, what really interested me was the exceedingly small difference between shooters 1-6 and shooters 7-12.

      That is interesting. Surely all the unexpected sure shots like Kotalik, J.Jokinen and both V.Kozlovs are stacked in the first three shooters. That said, since this study covers the entire history of the shootout, it may be that guys like that started off down the list and had to succeed a few times out of the 4-5-6 hole before getting promoted to the first team.

      In the case of the Oilers, the leading percentage shooters seem to be the guys who don’t shoot so much. Here are the career rates of current Oilers with 5 or more career attempts:

      Horcoff: 5/9, 55.6%
      Kotalik 20/38, 52.6%
      Nilsson 7/14, 50%
      Pisani 6/13, 46.2%
      O’Sullivan 8/18, 44.4%
      Hemsky 14/42, 33.3%
      Gagner 8/25, 32%

      … with the two guys picked most often by Craig MacTavish ranking at the bottom of the table, bang on the expected 1 out of 3. The third choices have actually outperformed them — Horc, Nilsson and Pisani are a combined 18/36 for 50%. In Fernando’s case in particular, I’d bet at least 10 of his 13 attempts have been in extra rounds; despite his success he’s never broken through into the top three choices.

      12.7% of shootouts are finished after four shots

      This too is interesting. Two hypothetical squads of 33% shooters should decide ~10% of all games. To win either team needs four consecutive results in its favour, so (1/3 * 1/3 *2/3 * 2/3) = 4/81 * 2 teams = 8/81 or 9.9%. The actual result of 12.7% is far enough from random chance to suggest skill has its effect in quick shootouts. Perhaps historically there have been enough outliers like the 2005-06 Stars, who used the formula (J. Jokinen + Zubov + Turco *2) to win a disproportionate # of games quickly. But that’s just a guess; one would have to do a painstaking review of the 79 examples to see if there were individual teams that outperform expectations.

      While we don’ have precise information, it seems highly probable that Team B (shooting second) wins the four-shot shootouts more often, given Team Two has an overall conversion rate of 445-408 in those first two rounds.

      38.8% of shootouts are finished after five shots. Obviously, 26.1% of shootouts ended precisely on the fifth shot.

      Fifth shots are the “wild card” result. All shootouts that end in four shots have a score of 2-0; all that end in 6, 8, 10 etc. have a one-goal margin. But a 5-shot shootout occurs when either Team A scores in the third round to go up by 2; or Team A fails to score and remains down by one. Possible scores therefore are 2-0, 3-1, 0-1, or 1-2. Applying the general rule of 33% success specifically to The fifth shot, we can estimate from the available data that 2/3 of the five-round games are won by the team shooting second.

      On the other hand, the six-round games are more likely to be won by Team A. The fact that the sixth shot is required means the score after 5 shots is either tied or Team A is ahead by one. Obviously for the shootout to end at that point, Shooter #3B either scores to break the tie or fails to score and cements the one-goal deficit. Without doing all the permutations it seems to me the initial state after 5 shots Team B is more likely to be tied than a goal down, but not twice as likely.

      Tyler, you show the % of shootouts that are over after each # of shots, but can you break that down into wins for Team A and B after each of 4, 5 and 6 shots? It appears Team B has an advantage in the early rounds, largely through performance in Round 1, and surely that translates into actual wins. However it’s not so simple as goal differential because of the possibility of two-goal margins in the truncated shootouts.

      In the extra rounds we have sufficient data to see that the advantage flips to Team A. Exact percentages can be derived; e.g. in Round 4 Team A has a 72-61 outscoring advantage, and we know 93 of those games are decided. This forces 52 wins for Team A, 41 for Team, 20 rounds where both guys score, and 79 where neither finds twine. Here’s the entire breakdown, with apologies in advance for the formatting:

      Round . # . A only . B only . both . neither
      ———————————————
      4 ….. 192 … 52 …. 41 …. 20 …. 79
      5 ….. 99 …. 19 …. 17 …. 12 …. 51
      6 ….. 63 …. 18 …. 15 …. 4 ….. 26
      7 ….. 30 …. 9 ….. 0 ….. 3 ….. 18
      8 ….. 21 …. 3 ….. 5 ….. 2 ….. 11
      9 ….. 13 …. 2 ….. 1 ….. 0 ….. 10
      10 …. 10 …. 2 ….. 1 ….. 0 ….. 7
      11 …. 7 ….. 1 ….. 1 ….. 0 ….. 5
      12 …. 5 ….. 2 ….. 0 ….. 0 ….. 3
      13 …. 3 ….. 0 ….. 2 ….. 0 ….. 1
      14 …. 1 ….. 0 ….. 0 ….. 1 ….. 0
      15 …. 1 ….. 0 ….. 1 ….. 0 ….. 0
      —————————————–
      Totals 445 …. 108 … 84 …. 42 …. 211
      Percent ……. 24% … 19% … 9% …. 47%

      Of the 192 games that go beyond three rounds, Team A wins 108 (.563), Team B just 84 (.438), or if you prefer fractions, 9/16 to 7/16. That’s statistically significant surely. Here are the success rates in single-round eliminations (round 4 and beyond):

      Player A – 150/445 = 33.7%
      Player B – 126/445 = 28.3%

      While Player A faces the same situation each time (score tied, shooting first) and continues to produce at about a 1 in 3 rate, Player B faces one of two possibilities:

      Player B with chance to win – 84/295 = 28.5%
      Player B must score to extend – 42/150 = 28.0%

      … Player B being equally unsuccessful in either case, about 5% below the “expected” 33%. Which in a sample of this size is over 20 goals below expectation. Can this just be random noise, or is there something else at play? Can it possibly be a psychological thing, that the guy who has a chance to win outright, or who has to score to keep it alive, is under a bit more pressure than the first guy, who can relax slightly in the knowledge that the goalie has his back?

    10. April 15, 2009 at

      Maybe I’m the only one, but I wouldn’t hate to see Kotalik back on a Sykora-type deal ($2M for a short term). The guy played well enough with the Horcoff-Hemsky combo that, barring a MacT-Penner love-in or a high-end acquisition, I wouldn’t mind seeing it continue.

      The underlying 27-10-83 numbers were stronger than the 21-10-83 numbers to be sure, but those numbers included Visnovsky. Even then, the difference in Corsi wasn’t worth an extra $2.25M.

      Of course, at the UFA stage and coming off a decent showing, I’m going to guess a one or two year deal isn’t in the cards.

    11. Hawerchuk
      April 17, 2009 at

      Bruce -

      I wrote up a post for you on the distribution of shots a while ago –

      http://www.behindthenet.ca/blog/2009/03/shootout-length-model-vs-actual.html

      One thing I found that was interesting was that guys who have taken 3 or fewer shots have a much lower scoring rate (~20%) than guys who’ve taken more. PPG doesn’t correlate very well with SO%, but not getting selected to take them does…

    Leave a Reply

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