Dump-and-Chase is Killing the Wild


The Wild have the reputation as a boring team.  This isn't news.  Why are they boring?  Probably because they can't score.  How long has this been an issue?  Forever.

This isn't hyperbole.  They've literally had a terrible offense nearly every year in existence.  Their average goals per game has been about 24th in the league.  It's not just bad luck either.  Wild players have always been reluctant to shoot.  Their average shots per game has been 29th in the league.  Second to last.  They've been 29th or 30th NINE seasons.  Dead last FOR THE LAST FOUR YEARS RUNNING.


Thirteen years, three head coaches, and 183 different players.  Complete, systematic incompetence offensively.  Well, except the two seasons under Jacques Lemaire where he was given a good roster where the Wild averaged 18th in scoring and 19th in shooting.

A common criticism among Wild fans is the team's reliance on the dump-and-chase strategy.  The puck carrier crosses the center red line and shoots the puck along the boards deep into the other team's zone and chases after it.

Dump-and-chase was developed as a counter-attack to the neutral-zone-trap (NZT).  In the NZT, the defending team crowds their blue line and neutral zone with players, forcing puck carriers to the outside and preventing rushes.  In theory, teams can dump the puck past those defenders and grab it on the other side.  In theory.


Assistant Coach Ackbar


Wild beater writer Mike Russo wrote about the dump-and-chase strategy late last month, defending the Wild's implementation.  In the story, Zach Parise is quoted as saying "Everybody dumps the puck. The difference, though, is they dump it and forecheck the right way to get it back."

Are the Wild built for puck retrieval?  Not quite.  If the Wild are to dump the puck and retrieve it consistently, they need aggressive, speedy forwards who can hit defensemen or absorb a hit, then set up in the offensive zone with a cycle.  That describes Parise perfectly, but practically nobody else on the team.


The prototypical forechecker


Zucker and Cullen are close in terms of skills, both showing speed and tenacity.  But Koivu, Granlund, Bouchard, and Heatley lack the speed to retrieve pucks (little wonder that the latter three were on for three goals against last night).  Setoguchi, Brodziak, and Clutterbuck lack the puck skills.

The Russo story got picked up by Grantland writer Katie Baker, who linked a recent research paper studying dump-and-chase versus carrying the puck across the blue line.

The study was done by a team of researchers out of Berkely, California, who analyzed over 300 games from the 2011-12 NHL season.  Fortunately for us Wild fans, Minnesota was one of the teams analyzed, along with the Philadelphia Flyers, Buffalo Sabres, and Washington Capitals.

According to the researchers, "The team’s shot differential – which has been shown to be a strong predictor of wins – is determined almost entirely in the much less-heralded neutral zone. Neutral zone success involves more than getting extra zone entries; since carrying the puck across the blue line generates more than twice as many shots, scoring chances, and goals as dumping the puck in, gaining the zone with possession is a major driver of success."

Specifically with the Wild, there are a few damning observations:

  • The Wild frequently dumped the puck when they could have carried it across the blue line, which "exacerbated their deficit of talent with an inefficient system."
  • "The Flyers forwards were much better than the Wild at keeping possession of the puck as they entered the offensive zone."
  • "The Wild were giving the most playing time to the defensemen who play the worst in the defensive zone" (although the researchers thought this didn't make sense and questioned the data).

From what I can gather from the paper, it seems that talent on the ice will result in higher quality shots, but the strategy to carry vs dump-and-chase will result in higher quantity of shots.  Even with the lowly Wild, carries resulted in over twice as many shots and nearly twice as many goals.

In addition to these findings, I've observed a couple games this season tracking dump vs carry zone entries.

The first game was Minnesota vs Calgary on Feb 26th:

Wild dumped the puck 47/96 times = 49%
Flames dumped the puck 28/58 times = 48%
Wild outshot the Flames 30-21
Wild won 2-1 in overtime

The second was Minnesota vs Chicago on March 5th:

Wild dumped the puck 38/74 times = 51%
Hawks dumped the puck 34/85 times = 40%
Hawks outshot the Wild 32-23
Hawks won 5-3 in regulation

Two games is a small sample size.  But we see a tight, low-scoring game when two teams dump the puck about half the time.  And we see a near blowout high-scoring game when one team is carrying the puck quite a bit more than the other (even as the Hawks dumped the puck more in the third and gave up two goals).

Who's got the best zone entries? I do!


My observation is this:

Yes, the Wild are not doing a great job retrieving the puck.  But they are too comfortable dumping the puck in the first place.  They also put themselves in an unfavorable situation by being slow up the ice, allowing the other teams to set up in the neutral zone.  I believe a big part of this is playing too deep in their own zone, delaying the breakout.  And on the power play, the drop passes in the neutral zone delay movement as well.

The common line is that the Wild are still adapting to head coach Mike Yeo's system.  But is Yeo's system a good one to follow?  Or should Yeo adapt his system to fit the players?  Will the Wild be able to make the playoffs and prove the critics wrong?  Or will their be changes made to the coaching staff and front office soon?

17 comments:

  1. Freaking brilliant stuff here.

    ReplyDelete
  2. I haven't read the full paper yet, but the problem with their "simple, stupid math" is their "simple, stupid" decision on what plays to include. They went to great lengths to exclude dump and change as well as defensive retreat with a turnover from the data, but they kept the most obvious situation that 100% needs to be excluded, odd man rushes.

    Unless they've omitted those and simply chosen not to acknowledge that fact through the extensive data collection section, their findings are essentially worthless.

    The logic here is quite obvious once considered. There's two major factors here. First (and less important), is the fact that no decision is being made on whether or not to dump the puck into the zone. If there's no decision being made, it's not a data point that should be included in the analysis. These entries are not representative of the entries being analyzed, and therefore their outcomes cannot be used as predictors for the outcomes of the wholly unrelated entries being examined in the research. To make a (bad) analogy, this is the equivalent of a study researching the average volume of crickets after a rainstorm including measurements from sub-zero days in January. Technically the data being measured is the same, but it's obvious that including those data does nothing except throw off the results.

    The second, and more important, factor is the quality of those chances. Breakaways and odd man rushes almost always result in at least one scoring chance and often results in more than one. If I had to hazard a guess as to the average number of shots on goal generated by a breakaway or odd man rush, I think we could all agree 0.75 is a conservative estimate. In the case of a breakaway, there will always be at least one shot on goal unless the man with the breakaway misses the net. In the case of odd man rushes there can be missed nets, bobbled passes, or passes defensed. However, while there is the occasional rush that fails to generate a shot, many rushes generate 2-3 or more shots on goal. If we assume 0.75 shots on goal per odd man rush/breakaway and examine the study's results, we'll see that those opportunities generate 30%~40% more shots than the average "carry" entry into the zone.

    I'll acknowledge that a case can be made for including these situations in the study. However, if included in the study they absolutely must be placed in their own category and tracked independently. If 1 out of 3 "carry" entries to the zone are breakaways or odd man rushes (this is just for example to show the math with easy numbers, not an actual estimate) and we assume that 0.75 shots per entry rate (I suspect that's low), we see that the "real" shot figure for PHI on carry entries was 0.45 per entry as opposed to the 0.55 they have reported (dump and chase generated 0.24). If we up that shot average to 1.0, the "real" figure drops to 0.325. That figure dips extremely close to the league average of 0.28 and considering the data subsets provided in the paper, likely falls below the highest rate for an individual team.

    ReplyDelete
    Replies
    1. "Unless they've omitted those and simply chosen not to acknowledge that fact through the extensive data collection section, their findings are essentially worthless."

      There's a third option -- that we tracked and dealt with those and you missed that part of the paper. The paper reads as follows:

      "One question to address is the extent to which this difference arises from odd-man rushes, plays where the offense has a sizable advantage and nearly always carries the puck into the offensive zone. As part of the zone entry tracking, observers recorded which entries were odd-man rushes so that their impact could be evaluated. It was found that their impact was negligible, both because they are relatively infrequent in today’s NHL (representing less than 3% of all 5-on-5 zone entries) and because the difference between an odd-man rush and a standard carry-in is significant but not overwhelming (0.78 shots versus 0.57)."

      For what it's worth, the shot numbers in the paper include shots that missed the net, so your conservative estimate of 0.75 shots on goal is actually way too high.

      Anyway, I agree with your last paragraph that they needed to be tracked and tested. But your intuition seems to be way too high about how frequent and damaging they are.

      Delete
    2. As I noted, at the time of posting I hadn't made it through the entire paper (having read the first 2-3 sections and scanned the remainder). I had missed that part of the paper, but it's also far more damning than a defense as it raises ethical questions. Unless you're going to straight faced argue that 0.78 versus 0.57 (a 37% difference) is so insignificant as to be deliberately ignored, even though removing these data would have been trivially difficult, but the 0.57 versus 0.28 (a 49% difference) is so significant that you've produced an entire paper on the topic, I don't see how it's even possible to try defend that point. Yes, they were only 3% of all entries, but that's about 7% of all carries and an explicit decision was made to include these data. I'm not coming up with any reason why they would be included except for "it makes our result look bigger."

      Why would you include these data, especially after going to the work to identify them, determine they were significantly deviant from the rest of the set, and already having a mechanism in place where they could be properly discarded or segregated?

      Delete
  3. Ultimately, while the research is interesting to look it, its flaws are far too great for any reasonable conclusions to be drawn from it. Personally, I believe that carrying the puck into the zone does result in a higher number of shots as well as a higher success rate at putting the puck into the net. However, this paper can't be used to support that conclusion because of the aforementioned issues with their methodology. On top of that, while they did some minor analysis to evaluate the defensive "risk" of carry versus dump, they didn't do any analysis as to the current vogue excuse for dumping the puck early (tiring out the defense by forcing them to retrieve pucks) nor was any of the data controlled for score differentials. There's simply far too many issues completely overlooked in this analysis, and the fact that they didn't even acknowledge these deficiencies makes it impossible for their findings to be taken as anything more than a curiosity that might merit a quality analysis.

    ReplyDelete
    Replies
    1. Boy, you guys sure are quick to assume that we overlooked obvious possibilities.

      We did look at score effects. The data have been published on broadstreethockey and nhlnumbers, but space was limited for the SSAC paper and I had to cut that section.

      The effect of score is predominantly that teams that are leading are likely to dump-and-change on plays where teams that are trailing will dump-and-chase. This has a modest impact on the carry-in totals (though nowhere near as large as the gap between the Flyers and Wild), but has essentially zero impact on the number of shots or goals per entry.

      Delete
    2. As one should question the presence of other possibilities if it isn't in the article.

      Delete
    3. Agreed completely.

      But there's a difference between something like "I wonder if they looked at this" or "it seems like they should look at this" and something like "this article is worthless because they didn't look at this" -- especially when the majority of the concerns that are raised are, in fact, discussed in the paper.

      Delete
    4. Yes I agree. The criticism could have been handled more professionally on a technical paper such as yours. I just didn't like the unprofessional response with your first line. I know it's a blog and what not but when discussing technical papers professionalism should always be practiced.

      Delete
    5. And blogger ate my comment...

      Short version - Data Collection did not indicate the score was considered, but listed many other, occasionally minor things. List appeared to be exhaustive but was not.

      Delete
  4. Of course, the factor that wasn't considered was turning the puck over in/near the neutral zone - which is the primary reason given by players/coaches for dumping rather than carrying in. I'm pretty sure that dump and chase results in significantly more shots than turning the puck over.

    Also, one of the factors you left out in terms of players who are more successful in the dump and chase is technique - which after watching Parise for 20 games, I am pretty sure is just as important as speed and tenacity. Parise has always started his 'chase' before he completes his 'dump' which evens things out considerably with the defenseman.

    ReplyDelete
    Replies
    1. "Of course, the factor that wasn't considered was turning the puck over in/near the neutral zone - which is the primary reason given by players/coaches for dumping rather than carrying in."

      The paper has four paragraphs, a table, and a figure discussing what happens after failed entries and how dangerous the neutral zone turnover is and how that impacts the decision matrix. I'm not sure why you concluded that we didn't consider this factor.

      It's true that the paper doesn't discuss technique. But given that we couldn't reliably identify any players whose dump-ins resulted in more shots than their teammates, I'm not sure what that discussion would look like. Maybe something like "no matter what technique players use, the differences in outcomes on their dump-ins appear to be roughly the same."

      Delete
    2. Admittedly, I was only responding to the original blog post, and was not responding to your (meaning Eric T) previous (and referenced in the post) analysis – which I did not go read. My bad. But, since you brought it up… let’s take a look at the Section 6 that you referred to:

      The first thing that jumps out is that, by extending the reasoning that is being applied in the paper with respect to failed carries vs. dump-chase, the data is saying that no team should ever do a dump-change. Sure, the data shows that having fresh legs results in more shots if your team happens to turn the puck over in the neutral zone; and fewer shots for the opponent if you don’t, but the net overall value is the worst for all cases. Unless you agree with this presented ‘conclusion’, we have found a case where good old (HCW) hockey conventional wisdom (or, perhaps, un-considered variables) trumps this data.

      So, why couldn’t there be similar problems with the conclusion regarding failed carry-in situations, or other conclusions/assumptions made in this section? I am not saying that the conclusions are wrong; just that the methodology probably requires further evaluation/justification/explanation.

      This analysis is also based on approximately 40 games in which the same team was involved in all of them. No notice is made to the small sample size or the potential bias of the same team being involved in all of the entries on one side or the other. This seems like an oversight. And yet, the conclusions from this section appear to be given equal weight to those from the rest of the paper.


      p.s. Some other nit-picky notes: Full seasons from two teams in different conferences and 2 half seasons from teams in different divisions and a different division from the full-season team in their conference adds up to 228 games. Subtracting this from the 330 games tracked leaves 102 games. Dividing 102 by 7 (the low of the stated ‘7-10’ game range for ‘most other teams’) results in 14 (of 26 remaining) teams. While this technically qualifies as ‘most’, it is certainly misleading. Also, 96 only played 37 games for the Wild last year. 9 only played 55. The Wild were juggling lines for probably half of their games last year. To suggest that they had ‘common’ lines is a stretch; as is drawing any conclusions relative to those lines based on data from all 82 games.

      Delete
  5. Also, seems to me that if one really wanted to compare the merits (in relation to shots attempted) of dump and chase versus carry-in, the carry-in side (as well as the dump and chase) should be limited to situations where the defense is at no disadvantage at the beginning of their zone time; i.e. where no offensive player enters the zone before a corresponding defensive player.

    ReplyDelete
  6. In the digression of my post, I feel that the intent of my little rant is being missed.

    This study is interesting, and has produced data which appears to indicate the direction a fully funded, more thorough review's data would provide us.

    However, with the issues in the analysis, some of the assumptions used, and the (unfortunately necessary) small sample size, it doesn't actually "prove" anything. Regardless of whether or not the authors intended for it to be used to determine strategy (judging from the replies here, I'd suspect they'd rather not), the headline for this article alone shows that people are running with it in that regard.

    Both Katie Baker and now Jarick have attempted to tie this back to Russo's article about dump and chase, but the reality is that all Russo pointed out is every team in the NHL utilizes dump and chase as at least a part of their offensive system. If anything, this study further confirms Russo's article and disproves Ms. Baker's contention and challenge that she directed at Koivu. Not only do all teams dump and chase for a significant portion of the game, they all seem to dump and chase at nearly the same rate.

    ReplyDelete
  7. Sorry to be coming in late here, apparently I missed this originally.

    First off, thank yous to the researchers are in order. So thank you. While E=MC2 moments get all the glory, the reality is most research progress is made in much smaller increments, and personally I'm quite pleased to see this paper.

    It seems to come down to "yes, everybody dumps and chases, but those who do it least without turning over the puck get the best results". The question then becomes one of personnell vs system --is a team dumping and chasing too much when they don't need to (i.e. they've been ordered to)? Or are they dumping and chasing because they don't have the skill players to do anything else?

    I think the key analsys quetion is correct, "Given the extreme differences in offensive success for entries with possession and dumping the puck in, it is worth questioning whether teams are pressing hard enough to carry the puck in." But to get at it, we need to know more about turnovers in the neutral zone between the red line and the offensive blueline, and what correlations they have to other data.

    Anyway, fun stuff, and glad to see someone working on it in an organized way. That we can argue about it is fine --that's how improvement is made.

    ReplyDelete