Zone entries and scoring chances

by Eric T. on Dec 14, 2011 12:58 PM EST in Philadelphia Flyers Commentary

So far, every article we've written on zone entries focused on generating shots as the assessment of how successful a given entry was. Questions about shot quality have come up on occasion, so let's look at scoring chances courtesy of Todd's scoring chance tracking.

Entry type	Shots/entry	Chances/entry	Goals/entry
Puck carried in	0.57	0.19	0.046
Puck passed in	0.51	0.19	0.036
Puck deflected in	0.25	0.082	0.016
Puck dumped in	0.22	0.073	0.013
OZ faceoff	0.34	0.14	0.024

It looks pretty clear that chances and goals move in lockstep with shots -- that no entry type (except perhaps faceoffs) is having a significant effect on shot quality. After the jump, we'll revisit yesterday's look at which plays allow the opponent to get a counterattack going.

Entry type	Next play is in def zone	Next play is scoring chance against	Next play is odd man rush against
Carry	50.9%	8.4%	2.4%
Pass	51.4%	9.1%	1.5%
Deflect	51.8%	6.7%	0.8%
Dump	58.7%	9.0%	2.8%

The differences in scoring chances and odd man rushes between dumping and carrying the puck in probably aren't significant, but they certainly don't suggest that dumping the puck in is the safer play. Two things that are worth noting:

First, as mentioned in the comments yesterday, when the fourth line carries the puck into the offensive zone, that leads to a lot of chances against. So while the top 9 should push every marginal edge, the fourth line should be a bit more cautious.

Second, a persistent concern is that dump-ins are made to look worse than they are because a dump-in is often the last resort against a defense that has closed the play off. However, the deflection plays are usually by design (a deflection to negate icing on a long dump-in to a speeding forward), so the comparison is interesting: those plays don't look any more successful than dump-ins at the offensive end, but they may actually limit dangerous counterattacks.

This leaves us with the following picture:

Carrying the puck in definitely generates more offense, and should be attempted at every opportunity when trailing.
The top 9 should usually try to carry the puck in whenever they can, and top-9 players who often carry the puck in are helping the team with their aggressive play. However, with a lead towards the end of the game there may be cause to run designed dump-in plays that give up fewer counterattacks.
The fourth line should carry the puck in when they have a clear opportunity, but should be a bit more cautious with borderline plays.

Tweet Comment 21 comments | Add comment | 5 recs |

Do you like this story?

Comments

What would you call a significant sample size for zone entries? Is it 5 games? 10 games?

by Olivier on Dec 14, 2011 1:04 PM EST reply actions

It depends what question you’re trying to answer.

Each team will have about a hundred zone entries per game, so somewhere in the 5-10 game range is enough to get a significant team total for zone entry percentage. But as you start to want to chop the data finer — looking at just when a certain player was on the ice, looking at how many shots came from each entry, separating out each type of entry — it takes a larger sample.

This work is based on 22 games of data, which is enough that I have pretty good faith in most of the team stats, and am starting to believe the individual stats. But later in the year I’ll do some reliability studies to get a more precise understanding of how long it takes a certain stat to become meaningful.

@BSH_EricT
Writer at Broad Street Hockey

by Eric T. on Dec 14, 2011 1:20 PM EST up reply actions

Questions about shot quality

DRINK!

Lightning strikes once, Hextall strikes twice!
"I think there is virtue in pissing off idiots." - Fehr and Balanced

by hintzy64 on Dec 14, 2011 1:56 PM EST reply actions

I thought we agree on no stats on Wednesday.

oh and

The fourth line should ~~carry the puck in when they have a clear opportunity~~ as it is currently constructed never see the ice

FTFY

"Call me dumb, call me stupid, whatever. I block shots."
#FireRoseman
@boknows71

by boknows71 on Dec 14, 2011 2:12 PM EST reply actions

There was no such agreement. Ben was for it. Geoff suggested Sunday for alliterative purposes. No agreement was reached.

Besides, how could the “eye island” proposal rule out a post that is based on information that arises from people watching the games and recording what they saw?

@BSH_EricT
Writer at Broad Street Hockey

by Eric T. on Dec 14, 2011 2:18 PM EST up reply actions

And how can I drink on Wednesdays if nobody says “shot quality” or “watch the game”?

Lightning strikes once, Hextall strikes twice!
"I think there is virtue in pissing off idiots." - Fehr and Balanced

by hintzy64 on Dec 14, 2011 2:29 PM EST up reply actions

Chicken and egg.

Good stuff Eric as always.

"Call me dumb, call me stupid, whatever. I block shots."
#FireRoseman
@boknows71

by boknows71 on Dec 14, 2011 2:30 PM EST up reply actions

Confirmation Bias Thursday
Unfounded Rumor Friday
That Old Chestnut Weekend
Pet Peeve Monday
Agree to Disagree Tuesday

/s, more often than not

by flyersfaninchicago on Dec 14, 2011 2:57 PM EST up reply actions

Eye Island.

I see what you did there. Kudos.

by Phalange on Dec 14, 2011 3:48 PM EST up reply actions

I think the stats are quite interesting and the project overall seem promising and you’ve gotten some interesting results already, but I think you are overstating the case with most of these conclusions. I talked about the problem last time, but assuming the teams do the obvious in the obvious spots (carry it in on odd-man rushes and good 2-on-2 type situations, dump it when the other team is stacked up at the blue line etc) then the data might tend to look like this even if it is the case that it’s better to dump it or use some other entry method in marginal cases.

The times where you suggest dumping it in is better are pretty convincing because biases in the data go against them.

Driving Play - The Blog with Three First Lines

by JaredL on Dec 14, 2011 5:19 PM EST reply actions

There’s no question that within the existing strategy framework, carry-ins are better on average than dump-ins.

What’s questionable is whether the existing strategy is optimized, or whether being more aggressive about trying to carry it in whenever possible would improve the team’s chances. There, I admit that I am making claims that aren’t directly supported by evidence when I generally argue in favor of more aggressive plays with the puck — arguing for a change in strategy requires knowing what the marginal benefit/cost would be for each play, not what the average would be.

The disparity between average results on carry-ins versus dump-ins is so large that I suspect teams aren’t pushing the margins enough. This is consistent with a common theme in sports, where teams and fans remember and fear the spectacular failure more than they should and play overly conservatively. But you’re right that in advocating for certain strategies I’m crossing the line from “proven by the data” to “a hunch I have based on the data”.

@BSH_EricT
Writer at Broad Street Hockey

by Eric T. on Dec 14, 2011 6:23 PM EST up reply actions

Maybe I’m splitting hairs, I’m honestly not sure how far off we are. I suspect you might be right, especially for the top guys; it’s mainly that I don’t think the data tell us much at all in this regard.

The disparity between average results on carry-ins versus dump-ins is so large that I suspect teams aren’t pushing the margins enough.

I can’t be precise, but think we should expect a large disparity no matter what is supposed to happen in marginal situations. Let’s oversimplify a bit and say there are 3 situations – those where carrying it in is obviously best, those where dumping it in is obviously best and ambiguous, marginal situations. The first type are overwhelmingly more favorable than the second – if it’s obvious you should dump it in then things haven’t gone your way in the neutral zone or on-ice matchups. Based on that, assuming your heroes tend to make the obviously correct play when there is one, the stats should show a big chances/Corsi/first event edge in favor of carrying it in those spots when it’s obvious what the best play is. In marginal situations, by definition, it doesn’t matter as much what we do so it’s not going to change the results gap by a huge amount either way.

So no matter what is supposed to happen in marginal spots, which are the ones we care about, we should expect carry ins to have much better outcomes on average.

Driving Play - The Blog with Three First Lines

by JaredL on Dec 14, 2011 7:10 PM EST up reply actions

Yeah, I think we’re mostly in agreement about the facts and the way of thinking about it, but our intuition tells us different things. Which is fine.

My opinion here is probably based as much on what’s been shown in other sports as on what we know about hockey. Across lots of sports, you can show that teams or players should adopt the boom-or-bust strategy more than they do: that playing for one run in baseball is usually a mistake, that football teams should go for it on fourth down much more than they do, that tennis players should hit their second serves harder (for some players, as hard as they hit their first serve), etc. So I definitely have a preformed bias towards arguing in favor of the aggressive strategy and not dwelling on the times it fails.

@BSH_EricT
Writer at Broad Street Hockey

by Eric T. on Dec 14, 2011 7:23 PM EST up reply actions

I’m just glad the data makes intuitive sense.

Also, score effects! Could we see type of entry frequency by game state?

Tracking the Flyers scoring chances at Broad Street Hockey

by ToddtheFox on Dec 14, 2011 5:29 PM EST reply actions

Score effects seem to show up a lot more in entry fraction than in entry type.

When the Flyers trail by enough that score effects would show up (2+ any time, 1 goal in the third), 48% of their entries are with control, and those controlled entries average 0.55 shots. When it is close, 55% of their entries are with control, and those controlled entries average 0.51 shots. When they lead, 58% of their entries are with control and they average 0.58 shots.

So that points to carrying it in more with a lead, but without a clear trend in success on those controlled entries, but let’s check their opponents’ strategy before drawing any conclusions…

Their opponents, when trailing, have control on 52% of their entries (0.58 shots per). When it’s close, they have control on 52% of their entries (0.60 shots per). When leading, they have control on 50% of their entries (0.52 shots per).

Hmm. Maybe the Flyers manage score effects differently from other teams — recall that we saw the Flyers and their opponents trend in opposite directions in this regard when we looked at it last year in the playoffs. But really there just isn’t enough data yet — ~400 total entries when the Flyers trailed by a lot means ~200 for each team, ~100 with control. A couple percent difference isn’t significant at that sample size.

@BSH_EricT
Writer at Broad Street Hockey

by Eric T. on Dec 14, 2011 6:31 PM EST up reply actions

I feel like the Flyers score effects make intuitive sense though. Trail – just get the puck in, take lots of shots. Close – try for optimal situations. Lead – enter only if have clear possession, when most likely to generate offense.

But that could easily be me fitting the data to my biases.

Man-crushin' on Boucher since 1999 and Matt Calvert since May 2010
Broad Street Hockey - Makin' it look mean since 1967.
SB Nation Philly - Associate Editor

by Geoff Detweiler on Dec 14, 2011 7:37 PM EST up reply actions

Incidentally, I haven’t updated it at 22 games, but after 12 games I started to do a piece on score effects (before concluding there wasn’t enough data). I manually calculated the amount of time spent in big-lead/close/big-trail situations and compared the team rates. What I found was:

When leading big, the Flyers’ controlled and uncontrolled entry rates (e.g. controlled entries per 60 minutes) stayed pretty similar to when it was close
When the Flyers led big, the opponents’ controlled entry rate went down slightly and their uncontrolled entry rate went up significantly.
When trailing big, the Flyers’ controlled entry rate went WAY down. Their uncontrolled rate went up slightly.
When the Flyers trailed big, the opponents’ rates both went down a ton — by almost half.

If I were going to draw a conclusion from this, it would be that the Flyers don’t change their defensive scheme much when they are leading — they protect the blue line a little harder, forcing more dump-ins but at the expense of letting the opponent come through the neutral zone a little more easily. And that their opponents changed their system a lot more when leading, conceding almost all of their offensive threats to slow the game down, giving the Flyers fewer total entries and in particular fewer controlled entries.

I ended up deciding that the sample sizes were too small to support those conclusions. But that’s my working hypothesis.

Incidentally, score effects aren’t playing a significant role in the conclusions of this article or yesterday’s. About 75% of the entries came when the game is close, and anyway the shots per entry aren’t significantly different in the non-close situations.

@BSH_EricT
Writer at Broad Street Hockey

by Eric T. on Dec 14, 2011 8:04 PM EST up reply actions

At the risk of belaboring the point, score effects provide a good example of what I’m talking about.

When the Flyers trail by enough that score effects would show up (2+ any time, 1 goal in the third), 48% of their entries are with control,

When the Flyers trail, the opponents are going to be conservative. That means fewer odd man rushes in favor and in general you will be facing defenses that are better set up in the neutral zone and, while they might also be more conservative they will be more prepared to challenge entry at the blue line.

When they lead, 58% of their entries are with control and they average 0.58 shots.

When they lead the opponents are pressing to score and care less about defensive responsibilities. In the extreme this will lead to breakaways and odd-man rushes but more commonly you will have entries where the D isn’t particularly well set up and is scrambling to get back into position. They will more often be in weaker position to challenge entry and be more likely to take the safe option and concede the blue line trying to stop you in their zone.

One should expect more control entries and for them to be more successful with the lead. It’s similar to shooting percentage being higher shorthanded than at even strength.

There are two issues with this. This pattern doesn’t seem to hold for the opponents. Not sure what to make of that. The other problem is that the Flyers had more success carrying it in when trailing than when the game was close. This may or may not be mostly variance, but I could also see it being due to the Flyers being more aggressive when in the attacking zone because they are more desperate to score. As evidence for or against my theory, I would expect more shots/entry for other types to be higher trailing than close.

Driving Play - The Blog with Three First Lines

by JaredL on Dec 14, 2011 7:41 PM EST up reply actions

Agree with everything you’ve said here. I think it’s probably going to turn out that the lead/trail scenarios are still dominated by variance at the 200-entries-per-team level, but if the data looks the same 40 games from now, I’d read it the same “this sort of makes sense but sort of doesn’t” kind of way.

@BSH_EricT
Writer at Broad Street Hockey

by Eric T. on Dec 14, 2011 7:43 PM EST up reply actions