More on the advantages of puck possession over dump and chase
Eric T.
July 11 2012 10:12AM
A few days ago, I presented some of what we were able to learn from Geoff Detweiler's data tracking the Flyers' zone entries in 2011-12. One of the things I noted is that carrying the puck into the offensive zone generates a lot more offense than dumping the puck in.
In this article, I'll follow that observation up by answering some of the questions that a lot of people have asked and showing a little more about what we found.
Do odd-man rushes inflate the results of carrying the puck in?
Geoff noted whenever an entry was an odd-man rush because we expected them to bias the results. In the end, we found that only 290 out of 9680 even-strength zone entries were on an odd-man rush, so we didn't even bother pulling them out in our year-end review on Broad Street Hockey.
However, it's been a persistent question since then, so I might as well go ahead and work things out:
| Carry-in | Pass-in | Dump-in | Deflect-in | Misc other | |
| Shots per entry, overall | 0.57 | 0.56 | 0.25 | 0.26 | 0.53 |
| Shots per entry, odd-man rush | 0.80 | 0.74 | 0/0 | 3/2 | 6/6 |
| Shots per entry, rushes excluded | 0.56 | 0.55 | 0.25 | 0.26 | 0.53 |
From this, we see:
- Gaining the zone on an odd-man rush is distinctly more dangerous than gaining the zone with even numbers.
- However, odd-man rushes are infrequent enough in today's NHL that they don't skew the numbers appreciably.
- Gaining the zone with even numbers is actually closer to an odd-man rush than it is to dumping the puck in.
Are there factors deflating the results of dumping the puck in?
One example that people often bring up is when the offense dumps the puck in and goes for a line change. We didn't log dump-and-change plays at all, only offensive plays where the team made an active effort to recover the puck after dumping it in. So those shouldn't be a factor.
Another notion that comes up a lot is that dumping the puck in is often the last-resort option on a well-defended play, and of course you expect to see fewer shots on a well-defended play. The response to this is that a lot of the deflections are designed plays, where you have a long dump-in with a deflection near center ice to cancel icing, and those plays don't look appreciably better than dump-ins. Besides, if the defense is challenging the blue line hard to prevent the carry-in, one might reasonably expect that to be a play where there is more room for a dump-in to succeed, not less.
It's also often suggested that bottom-six players are coached to just dump the puck, and so dump-ins look worse than they are because those plays are predominantly executed by lesser players. It is definitely true that the less-skilled players dump the puck in more often. However, the thing that was shocking about the first pass through this data was that the skilled players do not do appreciably better with the entries they get:
| Player | % with possession | Shots per carry/pass | Shots per dump/deflect |
| Jagr | 74% | 0.53 | 0.28 |
| Giroux | 67% | 0.52 | 0.30 |
| Voracek | 66% | 0.60 | 0.29 |
| Briere | 66% | 0.54 | 0.17 |
| van Riemsdyk | 60% | 0.54 | 0.31 |
| Read | 59% | 0.57 | 0.21 |
| Hartnell | 56% | 0.54 | 0.25 |
| Schenn | 55% | 0.67 | 0.26 |
| Rinaldo | 49% | 0.56 | 0.28 |
| Simmonds | 48% | 0.54 | 0.19 |
| Couturier | 45% | 0.60 | 0.30 |
| Talbot | 43% | 0.52 | 0.28 |
The Flyers' fourth line (Couturier-Talbot-Rinaldo) averaged 0.29 shots per time they dumped or deflected the puck in, while the top line (Giroux-Jagr-Hartnell) averaged 0.28. The fourth line averaged 0.56 shots per time they carried or passed the puck in, while the top line averaged 0.53.
The reason Giroux has a better shot differential than Rinaldo isn't that he does more with each entry; it's that he wins the neutral zone more often (more total entries) and does so more decisively (gaining the zone with possession).
If it's true that the less-skilled players are being coached to just dump the puck in -- and I suspect it is -- then the coach might be doing more to limit their offense than their own lack of skill is. This is the kind of inefficiency that can be identified, fixed, and exploited to gain an advantage over the rest of the league.
Regrouping
Yes, carrying the puck in is better than dumping it in. But if a rush is well-defended, is a team better off trying to regroup or dump the puck in?
This is a question that we can't truly answer, since no NHL teams are running that experiment for us. We know that when the Soviets tracked zone entries, they decided that regrouping was the right strategy, but this is a different era and the same strategic considerations might not hold in today's game. Let's try some math and see what we can come up with.
Dumping the puck in immediately is worth about 0.24 shots. Backing up and trying again gets you about 0.44 shots if you get the next entry (this is the average gain across all types of zone entry at 5-on-5), but if you turn the puck over it might cost you 0.44 shots at the other end. So your net shots from regrouping will be (0.44) * (odds you get the next entry) - (0.44) * (odds your opponent gets the next entry). As a result, regrouping turns out to be unfavorable if there's a 23% chance your opponent will eventually get the next entry after you back up.
That sounds grim for the regrouping strategy, but it's not as bad as it sounds:
- If you turn the puck over, the opponent won't automatically advance it into your zone; they might still turn it over again or need to dump-and-change, giving you another chance.
- The 0.44 shots per entry figure is the average for everyone, but some lines carry the puck in on more of their entries and therefore get more shots per entry.
- The lines with the puck skills to carry the puck in more are also the lines least likely to turn the puck over in the neutral zone after they regroup.
It works out that if Giroux's line is going to turn the puck over less than 1/3 of the time after regrouping, regrouping will probably be advantageous. If I were a head coach, I would definitely run some drills to see whether they can beat that number.
Pressing every opportunity
Even setting aside the idea of regrouping, our data can impact strategy. The difference between offense generated on a carry-in and a dump-in is quite large. This makes me suspect teams aren't trying hard enough to press their marginal edges and carry the puck in on borderline plays.
However, I can't prove that teams aren't being aggressive enough, since we only know the overall average for carry-in plays and not specifically how those marginal plays go. For example, a distribution like this would support the current approach:
| Hypothetical carry type | Hypothetical carry frequency | Hypothetical carry generated |
| Odd-man rush | 6% of carry-ins (known) | 0.79 shots apiece (known) |
| Poorly defended | 35% (hypothetical) | 0.7 shots apiece (hypothetical) |
| Typical | 40% (hypothetical) | 0.55 shots apiece (hypothetical) |
| Well defended | 19% (hypothetical) | 0.3 shots apiece (hypothetical) |
This is a plausible hypothetical situation -- it matches our known odd-man rush results and gives the known average shots per carry in overall (0.57). And it has those marginal well-defended situations close enough to what a dump-in achieves that if this were what actually happens in the NHL, there would be no reason to press harder on the marginal plays.
However, although this distribution is plausible, it doesn't look very likely to me -- it has the "poorly defended" group being so poorly defended that those plays are almost like an odd-man rush, and yet they are still twice as frequent as a well-defended play. Yet one has to invoke a distribution this skewed to defend the current strategic approach.
In other words, I can't prove that teams should be pushing the marginal chances to carry the puck in harder than they are, but it seems very likely. And it would be consistent with findings in many other sports that strategies tend not to be aggressive enough (stealing a base to try for one run instead of playing for a big rally, punting or kicking a field goal on fourth down instead of trying to extend the drive, hitting a soft second serve to avoid a double fault instead of trying to keep the opponent on their heels, etc).
Conclusions
Carrying the puck in is way better than dumping it in, more than twice as good -- and it's not because of odd-man rushes or player skill or any other external factor; it's just because having the puck in the opponent's zone headed towards the goal is a lot better than trying to outrace the opponent to try to get the puck in the corner.
Most people don't recognize just how big the difference is, and the data suggests that teams should be trying harder than they are to carry the puck in. If coaches are telling their third line to dump the puck, they are probably giving away scoring chances. If coaches are telling the players to dump the puck in borderline situations where they think carrying it might lead to a turnover, they are probably giving away scoring chances. Even regrouping and trying again might be better than dumping the puck in, especially when the team has their top line on the ice.
Finally, while Geoff's data has been extremely helpful in breaking down the Flyers' performance this year, our conclusions will be a lot stronger when we have multiple teams to compare. If you're interested in joining this project, let me know (bsh.erict -at- gmail or @BSH_EricT) and I'll forward the templates we've used so you can track games next year or make a summer project of going through last year's games on Game Center Live.
[Data on odd-man rushes are corrected from original version to remove six spurious entries.]
Recently by Eric T.
Excellent follow-up. Here's hoping you get some volunteers to track a wider range of teams.
Awesome, awesome stuff yet again. I definitely intend on tracking zone entries for the Sabres next season. I had planned on tracking scoring chances but that plan was smashed a couple weeks ago haha.
Yeah, I'm waiting for there to be more data to even begin to form any conclusions, but this is obviously a great start. I think you make a great point that coaches across sports tend to be incredibly risk-averse, and it wouldn't surprise me if this is one area where they're especially so. It's one of those hockey truisms to 'never turn the puck over at the blueline', but if you stick by that, you're probably giving up a lot of offense - a lot of the best in the world are or were notorious for blueline turnovers - Gretzky, Jagr, Malkin, etc.
Please tell me who dumped the puck in on an odd man rush? I guess its not out of the question, but I wonder if those were not really errant pass attempts?
@Interesting...
Thanks for asking.
It was just three entries all year, so I assumed they were just bizarre plays. But since you asked, I pulled those specific plays out of the data...
It turns out all three of them happened in the first game Geoff recorded this year. In that game, he used the "Odd man rush" column of the spreadsheet for a variety of notes like "Immediately broken up" or "Turnover", and the sheet interprets any text in that column as signifying that the play in question was an odd-man rush.
I've now fixed my data and will adjust the relevant lines in this article.
@Triumph44
Agreed on wanting more data -- and since most people have a tendency to draw strong conclusions on too little data, I can't fault you for holding off until we know for sure.
There are some things I'm pretty certain of at this point (most of this article) and some things are still pretty tentative (most of the previous article).
As this project moves forwards, we'll be able to firm things up one way or another.
Liked the article, thanx.
I think many coaching strategies are risk averse. I wonder if failed attempts to enter the zone with possession, may increase the chance of a goal (or successful entry) against?
@SCM
Perhaps.
Unfortunately, I can't answer that question. We didn't track failed attempts -- the line between "turned it over trying to carry it in" and "turned it over on the way through the neutral zone" is too blurry.
I do know that a successful carry-in reduces the likelihood that the next entry will go against you more than a successful dump-in does, but the failed carry-in's that we didn't track undoubtedly change that equation a little.
These data are great but we do need to make sure we don't make broad statements about this stuff. Are these data from Flyers games only? I've seen people take an article like this and immediately assume it's law. Even though we see some nice data here I think we still need to remain skeptical.
I have an idea. How about someone write up a statistics article about confounding variables, correlation coefficients, cause and effect, sample sizes, type 1/2 errors, etc... and the various caveats that come with statistics as a reminder to readers that what we often see shows direction for a certain sample (sometimes strong, sometimes weak) but isn't necessarily law? There are many variables that I feel may influence numbers that cannot or are not tracked.
For instance, a team focusing primarily on dump and chase strategies might potentially have better numbers in this regard. There may be external factors (player skill, size, speed, coaching, etc..) that may affect these numbers. I've seen teams pass-in or carry-in incredibly well and some that are so bad that they need to rely on dump and chase and hard work along the boards.
This is good work. I would be interested in seeing more data - specifically from other teams / players.
@mrthemike
Well, I agree that more data will be better. That's why the conclusion of my article read "finally, while Geoff's data has been extremely helpful in breaking down the Flyers' performance this year, our conclusions will be a lot stronger when we have multiple teams to compare. If you're interested in joining this project, let me know"
As for your suggestion, I think most of our readers are uninterested in lessons on statistical analysis; they just want to know the hockey-relevant conclusions and trust the writer to make clear what specific caveats might be attached to that particular investigation. I try hard to show everything I need to demonstrate how strong (or weak) the conclusions are while keeping the article readable and accessible for a general audience.
If you read through what I've written in this series, you'll find that I've shown and discussed correlation coefficients, reliability, p-values, co-linearity, and causality wherever I think it's an important consideration. I've emphasized in certain places that we have only a hint at something and it needs further investigation. If there's a place where I've overlooked an important consideration, please bring it up so I can investigate further.
In this particular article, I've presented most of the conclusions pretty strongly because I think the evidence is pretty strong. We have nearly 5000 zone entries by the Flyers and nearly 5000 by their opponents from this year, plus last year's playoffs, plus this year's playoffs, plus a handful of games not involving the Flyers, and across the board the shots per entry with possession and shots per entry without possession always fall into a pretty narrow range.
The best any Eastern conference (which means 4-6 games) team did against the Flyers this year on the dump-and-chase was Toronto at 0.30 shots per dump-in. The worst any Eastern conference team did against the Flyers at carrying it in was 0.48 shots per carry-in. As I showed in the article, the best any Flyer did individually on dump-ins was 0.31 shots and the worst any Flyer did on carry-ins was 0.52 shots. You're right that the specific numbers will move a little bit as we get more data, but I'm very very comfortable saying that dump-and-chase is much less successful than carrying the puck in.
I agree that more data will be helpful, and that -- as I tried to make clear in those specific cases -- some of the things suggested by our current data set can't yet be taken as firm conclusions. But I object strongly to the implication of your broad opening statement "we do need to make sure we don't make broad statements about this stuff"; I've tried pretty hard to limit my broad statements to places where the data seems to support a broad statement.
@mrthemike
" and the various caveats that come with statistics as a reminder to readers that what we often see shows direction for a certain sample (sometimes strong, sometimes weak) but isn't necessarily law?"
Finally, while Geoff's data has been extremely helpful in breaking down the Flyers' performance this year, our conclusions will be a lot stronger when we have multiple teams to compare. If you're interested in joining this project, let me know (bsh.erict -at- gmail or @BSH_EricT) and I'll forward the templates we've used so you can track games next year or make a summer project of going through last year's games on Game Center Live.
The idea that there was no data for the "failed zone entry" was exactly what I was looking for (I even went and read all the stuff on this page http://www.broadstreethockey.com/2011/10/18/2497456/zone-entry-article-archive !)
Glad to see I was not crazy for missing it.
Intuitively, I think you are right, that teams are too conservative in using a dump and chase strategy. I just think the justification is going to be in this gap that SCM identified. How often do teams fail to get the puck into the offensive zone when attempting to carry versus dumping/tipping the puck in? How often on those failed attempts do teams get possession back? Until you show traditionalists that their perception of those events does not match the data, there will be friction.
I would also note, this guy (http://www.acthomas.ca/papers/act-jqas-2-1.pdf) agrees with you too.
Good work.
@Eric T.
Thanks for the summary. I guess I am less concerned about your specific analysis but how some readers might misconstrue. This is a larger data set then I had initially thought. Thanks for the clarification!
Any chance we can see the second table with goals per carry/pass and goals per dump/deflect instead of shots per carry/pass and shots per dump/deflect. I suspect the sample size is too small to draw any conclusions, but I am curious.
@Tach
Yup, agreed, it's definitely a missing piece. Unfortunately, it really is a tough thing to track because the "failed entry" line is so fuzzy.
So I'd encourage those traditionalists to use their own eye test to fill in this gap. When I watch games focusing on zone entries, I tend not to think there are all that many turnovers in the brief moment between when the offensive player commits to trying to carry the puck in and when he crosses the blue line. There aren't all that many that get broken up in that time frame, and the offense regains possession on a substantial fraction of those.
So while we're definitely missing a piece of the puzzle in this data, my eye doesn't tell me it's a huge missing piece, just a minor correction.
@David Johnson
Yeah, the sample size will be way too small for that -- we have nearly 10k zone entries, but by the time you break it down to a single player's individual dump-ins, you're looking at around 100-200 entries generating 25-50 shots. It won't be too much trouble to produce the data if you want me to, but it's going to be Rorschach data (what you see in it says more about you than the players).
@Eric T.
I am not so much interested in how individual players do but see if there are any trends between "top six" and "bottom six" guys. Are the Giroux's and Jagr's better at converting on the rush than Couturier, Talbot, etc. As a believer in shot quality (i.e. some players/playing styles can sustain elevated on-ice shooting percentages) I was wondering if we could see those effects in this data.
@David Johnson
OK. Here's what I have, though I'm not dividing it top 6/bottom 6 because that's not really how the Flyers worked this year...
For entries with possession:
Player shots goals
Giroux 169 10
Jagr 163 12
Hartnell 114 14
1st line: 36 goals on 446 shots, 8.1%
Briere 112 8
Simmonds 111 10
Schenn 65 2
Voracek 154 11
Read 113 8
van Riemsdyk 59 3
2nd lines: 42 goals on 614 shots, 6.8%
Couturier 81 7
Talbot 41 2
Rinaldo 45 1
4th line: 10 goals on 167 shots, 6.0%
Overall (including entries by defensemen and part-timers): 106 goals on 1463 shots, 7.245%
For entries without possession:
Giroux 38 3
Jagr 23 1
Hartnell 33 2
1st line: 6 goals on 94 shots, 6.4%
Briere 15 3
Simmonds 36 2
Schenn 19 1
Voracek 29 0
Read 25 0
van Riemsdyk 22 2
2nd lines: 8 goals on 146 shots, 5.5%
Couturier 42 5
Talbot 35 2
Rinaldo 20 1
4th line: 8 goals on 97 shots, 8.2%
Overall (including entries by defensemen and part-timers): 35 goals on 483 shots, 7.246%
@Eric T.
Hmm. It munched my formatting. Hope you don't find that too unreadable.
Edit: I figured out how to unmunch it.
@Eric T.
Thanks Eric. Certainly not statistically conclusive, but pretty much as I would expect. The skill guys (first line and to a lesser extent the second lines) get higher conversion rates on the rush and the majority of their goals on the rush. I look forward to seeing data from other teams to see if the same trends continue.
@David Johnson
Yeah, I agree it's not meaningful at this data size. If I add in just simple square-root-of-N noise, it looks like this:
With possession...
1st line shoots 8.1% +/- 1.3%
2nd lines shoot 6.8% +/- 1.1%
4th line shoots 6.0% +/- 1.9%
Dumping it in...
1st line shoots 6.4% +/- 2.6%
2nd line shoots 5.5% +/- 1.9%
4th line shoots 8.2% +/- 2.9%
It might prove to be real eventually, but we need more data.
Just a note, in baseball, it's been proven that stealing bases is generally an unfavourable tactic. Unless the player steals successfully at a very high clip (something around 80%), the risk of losing an out outweighs the potential gains from the steal.
Other than that, this research is awesome. Being a minor hockey coach myself, I'm definitely going to place a big focus on playing possession hockey this coming season in light of this data.
@Anthony D.
Yeah, that was the point -- playing conservatively (hoping to maximize the chance of getting one run instead of going for a big inning) actually costs the team runs in most cases. Just like in football (where conservatively kicking costs points relative to going for it on fourth down in most cases) or tennis (where conservatively hitting a soft second serve costs points in most cases).
Eric, read both these articles and think its fascinating stuff! Sounds like a potentially game-changing insight if the conclusions hold up with more data. Well done!
One thing I'm left wondering, how many times does an attempted carry-in result in an odd-man rush the other direction? Theoritically, you dump and chase more often because a carry-in across the blueline would be more apt to result in a turn over. If turn-overs from those attempts result in the carrier getting caught up ice, you could get a higher occurence of resulting odd-man rushes.
Given how significantly more dangerous those are, I'd be very curious to see: does the shot gain from a carry over (.56 to .29?) get completely offset by the potential odd-man rush coming back (.80 x however frequent that is)?
That might temper the conclusions a bit and could explain why teams don't push the carry-in more, particularly if you've got bottom sixers out against too sixers.
Thanks! Greg
Hello,
Did you happen to track the zone entries for your parent club (Stanley Cup Champion Los Angeles Kings)?
I'd bet the difference between pre and post Terry Murray Kings zone entries was large. But I don't think it was just a matter of the players deciding or being coached to carry the puck in, they just were very often not in a position to carry the puck in the zone under Murray (lots of players back in the neutral zone to defend/trap/breakup plays). Quite frustrating to watch failed break-in after failed break-in.
What do you think contributes to successful zone entries?
@Greg
Yup, this is a common concern. I need to update this study for the full end-of-year data set, but in http://www.broadstreethockey.com/2011/12/13/2630489/does-aggressive-equal-risky I looked at it and found no evidence that the carry-in results in more chances the other way.
Of course, since we don't include failed attempts (where the person turns it over before crossing the blue line), we're probably missing some plays where a turnover just before the blue line leads to the play heading back the other way. My hunch is that there aren't enough turnovers in that split-second between committing to carrying it in and reaching the blue line to make a significant impact on these results, but I don't have the data to prove it right now.
With that being said, I'm pretty sure odd-man rushes are an insignificant component. There were only 290 of them all year, out of 9680 even strength entries (and the Flyers played a pretty aggressive style; this number might be high compared to the rest of the league). Undoubtedly some significant fraction of those came on turnovers in the offensive zone (after the zone entry -- which means it's already included in the study I linked above), so we're looking at something like 1% of all entries being an odd-man rush following a neutral zone turnover. Odd-man rushes are more dangerous, but not so dangerous that they can have a significant impact if they only occur at that frequency.
@Bob Knob
We didn't do the Kings, except in the one game they played against the Flyers.
I'm not enough of a tactician for you to want my opinion about how a successful zone entry comes about. There are people much more qualified than me to break that down.
firstly, AMAZING research. everyone is awed, or ought to be anyway.
i'd really like to see data for more teams, because it seems like line-matching strategies would make a major difference in whether a given trio is coached to favor dump-ins or carry-ins. the carry-in-always method seems to make lots of sense in power vs. power situations, when the team with possession can be reasonably certain that the defending players are of more or less equivalent skill. but do dump-ins still make more sense for designated defensive lines? i.e. what effect does dumping the puck in have on rates of shots against, comparative to trying to carry it? it might be that the best defense is a good offense (as charron says), but are bottom-six players in the nhl really good enough to play that way against top-six competition?
this may, of course, be nonsense. i do apologize if it's stupid in some way i have not yet noticed.
@Anthony D.
Just a note, in baseball, it's been proven that stealing bases is generally an unfavourable tactic. Unless the player steals successfully at a very high clip (something around 80%), the risk of losing an out outweighs the potential gains from the steal.
That rate is actually 70%.
@Derek Zona
Well depends upon what the situation is with the game.
Best thing Ive read this week. Keep up the great work!
That wasn't my point lol. Stats show that in baseball, playing "small-ball" is likely to cost you more runs over the long haul, and that just letting your guys hit is a more favorable strategy. Ironically, in baseball culture, a coach will oftentimes get criticized for just letting his players hit instead of going for steals, calling for sacrifice bunts, etc. You could say that a coach is showing more guts by letting his players determine the game's outcome instead of micro-managing. Regardless, the point about being agressive makes a lot of sense in hockey. Reminds me of the 03-04 Lightning, that had "Safe is Death" as a team motto, and that commonly used a 2-1-2 forecheck. A far cry from Tort's team now...
@Anthony D
Yeah, and what I'm saying is that small-ball is the conservative play. It minimizes variance: reduces the chance of scoring zero runs or a lot of runs, hoping to get precisely one run. The aggressive play, the gambling play, is to let the players hit and take the chance that you'll get zero runs in the hopes of getting several.
Stealing bases may often be framed as an aggressive play, but if it minimizes variance, it's better described as conservative.
Another example of numbers showing what every single NHL hockey coach already knows.
There's a reason why most team have 3 or even 4 players lining up along the blue line and standing ground. They want to force the attacking team to dump the puck by not giving them any openings to carry it in.
Frank
Excellent article! As per your math, would it not make sense for a third line (checking line) playing tough opposition to dump and chase? i.e. factoring their chance of gaining zone entry with the puck, getting a shot on net, and their shooting percentage vs. losing possession and the reverse happening at the other end of the rink vs a superior (Giroux) line...
FYI, I don't think I am more qualified than you or anything (I have no idea how much knowledge you have of the game, I'm just some guy who played).
I think the question of what leads to a good entry is probably a few steps down the line in what to record so it can be analyzed, and it's a good question anyways. Was just wondering if possibly you or anyone had some different insights on what contributes to a good zone entry.