Over at Broadstreet Eric T. and Geoff Detweiler have been collecting and analyzing zone-entry data for all Flyers games this year. They track each entry into the offensive zone, recording who is on the ice, how many shots and scoring chances there were before the puck left the zone and whether the puck was carried in, dumped in, passed in etc. Here is a post archiving some of their previous work. I think it is an excellent idea and the early data look promising.
In recent articles, which can be found here and here, Eric uses early data to make the argument that carrying the puck into the offensive zone is better than dumping it in and teams, or at least the Flyers they've got data on, should be more aggressive, carrying it in more often. He does put in a couple caveats that the fourth line should be more inclined to dump the puck in and top 3 lines should perhaps be more cautious late with the lead. While intuitively I think he's correct, the extra pressure you put on your opponents with the puck seems more valuable than the risk of a bad turnover, I don't think the results he cites tell us anything about whether or not teams should try to carry it in more often in marginal situations. I have a different interpretation of the data.
Their results: teams do better when they carry the puck in.
One thing there is no doubt about whatsoever is that in their dataset teams get substantially better outcomes when they carry the puck in than when they dump it or even pass it in. (Before you ask, they exclude situations where a team dumps the puck and makes a line change with little to no effort to go after the puck) As an example, when the puck was carried in the team doing so generated 0.57 shots before the puck was sent out of the zone. The similar number is only 0.22 when the team dumps it in.
On the face of it, it seems reasonable to think that this means carrying the puck in is smarter and that teams should be doing it more often. However, this ignores the circumstances. Most of the time when a player can easily carry the puck across the blue line into the zone it is both correct to do so and what he does. These situations tend to overwhelmingly favor the attacking team. In extreme cases you have breakaways and odd-man rushes. In general the defense will not be very well set up - if they were the offensive team would not be able to waltz into the attacking zone without risking losing the puck.
Now think about times where it would be very difficult for the player to cross the blue line with possession of the puck. The defense is set up, putting pressure on the puck handler. He is likely to be facing a very good defenseman. He might even have multiple defenders perfectly executing a trap. In these situations, dumping the puck is the correct move and usually what is done.
Looking at the two together, when the situation is favorable to the team about to enter the offensive zone they tend to carry it in. When it is unfavorable, they will usually dump it. Let's flip that around - when teams carry the puck in the conditions are usually very good for the offensive team and when they dump it in they are usually bad. I think it's pretty clear that the circumstances would drive the numbers in exactly the way they appear. From the numbers alone it's not clear whether or not the teams in question attempt to carry the puck into the offensive zone too rarely, too often or about the right frequency. Intuitively I agree with Eric's conclusion, but I don't think the data provide any evidence for it.
Let's shift the focus.
In my view Eric focused too much on the decision the puck carrier makes and not enough on what is happening on the ice when, and just before, he makes it. I'm not saying this to be negative, in fact it's quite the opposite. He, Geoff and Broadstreet in general write some of my favorite hockey stuff, and that's saying something since I'm a Pens fan and the Flyers are my least favorite team. I love the idea of looking at zone entries and, perhaps paradoxically, my interpretation puts more value in these metrics than his does.
Let's take a different view of some of the data from the three recent articles, including comments:
- the top line (Hartnell, Giroux and Jagr) carried the puck in 3.1 times as often as they dumped it in. For second and third liners this drops to 2.2 times as often and for the fourth liners it's all the way down to 1.4 times. Better players tend to carry the puck in more compared to dumping it.
- when the puck is carried in the results are better than for any other type of entry going by shots/entry, chances/entry, goals/entry and how often the next play is in the defensive zone.
- the team carrying the puck in gets the next shot off 69.8% of the time, just above passing it in (68.6%) and well above both deflecting (62.4%) and dumping it in (56.3%)
- carrying the puck into the zone is substantially more advantageous than getting a faceoff in the offensive zone, going by any of the above metrics.
- when the Flyers have a lead of 2+ in the first two periods or any lead in the third, so their opponents are taking risks, 58% of their zone entries are carried in or passed in where they maintained control. When trailing, their opponents are more defensive minded, this figure drops to 48%. When the game is close and the opponents are more balanced it is in the between at 55%.
What does all this say to you? To me it screams out that the ability to carry the puck, or pass it in with control, is a fantastic proxy for winning the neutral-zone and transition-game battle! Giroux isn't good because he carries the puck in, he carries the puck in because his strong play has given him the opportunity. The more the conditions dictate the decision, the better controlled entries measure how well teams and players are doing in the neutral zone. It is precisely because as I see it the conditions drive the entry decisions that I think it's such a good thing to track.
Conclusion and Suggestions
While Geoff and Eric only have 22 games so far this season, a preliminary look at the data indicates that zone entries, especially those where the puck is carried in, may tell us a lot about who is excelling in the neutral zone, on transitions from offense to defense and vice versa. This is quite promising and I have a few quick suggestions for ways they could use these stats.
Firstly, I would put it in percentages as we do for Corsi. In other words carried-in entries for divided by carried-in entries for and against. I would also consider weighting them differently to come up with one overall number. People often ask about high-value shots and other notions related to shot quality. In this case, there appear to be high-value entries and low-value entries. Another thing to look at is how often the opponents carry it in with different defenseman on the ice.
Showing posts with label Advanced Hockey Stats. Show all posts
Showing posts with label Advanced Hockey Stats. Show all posts
Wednesday, December 14, 2011
Sunday, October 16, 2011
Part III: The Aftermath of the Mike Richards and Jeff Carter Deals
Earlier this summer, I wrote extensively on the deals that sent Mike Richards and Jeff Carter from dry island Philadelphia to L. A. and Columbus respectively, promising a trilogy of sorts. After looking at what the Flyers gained in both the trades and free agency, the final step is to evaluate what the Flyers lost in those deals. While this post is certainly long overdue, the aftermath of last night’s 3-2 Kings victory over the Flyers in their only meeting this season seems like the perfect remaining opportunity to bring closure to this saga.
Beginning with my familiar approach, let’s take a look at both Richards and Carter’s average ice time from last season per nhl.com:
Unsurprisingly, we see that both guys gave the Flyers a good chunk of minutes in all situations. With the exception of Carter’s reduced role on the PK thanks to the emergence of Darroll Powe, Flyers head coach Peter Laviolette was not afraid to send out either player when he felt he needed a boost in any particular area of the ice. In order to give these minutes their proper context, we will begin by looking into where both players stacked up amongst Flyer forwards in point production, once again thanks to nhl.com:
As we can see, both players seemed to match their top-6 ice time with top-6 scoring numbers both at even strength and on the power play. If we take a look at a few more key statistics according to Behind the Net and Time on Ice, it will become quite apparent why Richards and Carter are so good at what they do:
Breaking these numbers down, beginning with Richards, his negative Corsi score is perhaps the first thing that stands about his totals. However, if we judge his performance according to Eric T.’s Balanced Corsi, we see that according to his zone start he is actually around 3 shots better per 60 minutes than we might expect. His balanced zone shift is also a little higher than we might expect, and if we couple this data with his extremely impressive 53.6% Fenwick with the score tied, there is a lot here to suggest that Richards is carrying the water at even strength.
Moving to Carter, his totals are just as impressive. Carter actually was put in tougher defensive spots than Richards, and his Corsi ON score is a little more than 4 shots higher per 60 minutes. His Balanced Corsi is around 7 shots higher than what we might expect from a player put in similar situations, and his BZS is around 3 percent to the good. His Fenwick score, though lower than Richards still suggests that he was also doing a major part driving the play forward for the Flyers considering his zone starts.
What is even more impressive is that the above analysis doesn’t even take into account the elephant in the room: quality of competition. Below is a chart of the toughest CorsiRelQoC scores of every player listed as a Center on Behind the Net last season, minimum 20 games played:
Both Richards and Carter show up in the conversation with guys who are playing against some of the toughest players in the league. Though they may not score upwards of 80 points per season, both players are certainly producing at elite levels considering the players that they are expected to face night-in and night-out.
What is more, thanks to JaredL we are able to take a look at how the Flyers performed during the past two seasons with and without either Richards or Carter on the ice:
Unsurprisingly, these numbers fall in line with everything else we’ve seen – they were able to send the play in the right direction while eating the majority of the team’s tough-minute assignments. Jared was also kind enough to provide data that looks into how some of the Flyers’ other key players performed in situations both with and excluding one of Richards or Carter on the ice during the same time-frame:
Once again, we see that no matter the situation, each player was better with one of either Richards or Carter on the ice except for James van Reimsdyk whose data has a noticeable discrepancy in quality of competition. In order for the Flyers to remain one of the premier Stanley Cup contenders in the Eastern Conference, it is looking more and more like the big line of JVR, Claude Giroux and Jaromir Jagr is going to be asked to carry the mail against top-tier competition in the absence of Richards and Carter. These numbers seem to suggest that it is certainly possible, but we will have to wait until each plays an adequate number of contests before we can finally say whether Paul Holmgren’s plan will pay off in the long run. So far, the Flyers are off to an excellent start, but Giroux & Co. will have to keep up their play in the absence of what was one of the league's most formidable one-two punches up front.
Beginning with my familiar approach, let’s take a look at both Richards and Carter’s average ice time from last season per nhl.com:
| Player | Games Played | ES TOI/Game | Team Rank | PP TOI/Game | Team Rank | SH TOI/Game | Team Rank | Total TOI/G | Team Rank |
| Richards | 81 | 13:47 | 4 | 2:56 | 3 | 2:08 | 3 | 18:52 | 2 |
| Carter | 80 | 14:39 | 2 | 2:56 | 4 | 0:39 | 6 | 18:14 | 4 |
Unsurprisingly, we see that both guys gave the Flyers a good chunk of minutes in all situations. With the exception of Carter’s reduced role on the PK thanks to the emergence of Darroll Powe, Flyers head coach Peter Laviolette was not afraid to send out either player when he felt he needed a boost in any particular area of the ice. In order to give these minutes their proper context, we will begin by looking into where both players stacked up amongst Flyer forwards in point production, once again thanks to nhl.com:
| Player | ES G (Team Rank) | ES A (Team Rank) | ES Pts (Team Rank) | PP G (Team Rank) | PP A (Team Rank) | PP Pts (Team Rank) |
| Richards | 15 (T-5) | 24 (4) | 39 (6) | 5 (T-4) | 16 (1) | 21 (1) |
| Carter | 28 (T-1) | 21 (5) | 49 (3) | 8 (1) | 9 (T-3) | 17 (3) |
As we can see, both players seemed to match their top-6 ice time with top-6 scoring numbers both at even strength and on the power play. If we take a look at a few more key statistics according to Behind the Net and Time on Ice, it will become quite apparent why Richards and Carter are so good at what they do:
| Player | Corsi ON | CorsiRel | Score-Tied Fenwick % | CorsiRelQoC | SF/60 | Zone Start % | Zone Finish % |
| Richards | -1.23 | 1.1 | 53.6 | 0.752 | 30.6 | 46.8 | 50.1 |
| Carter | 3.34 | 7.8 | 50.5 | 0.896 | 29.2 | 43.8 | 51.9 |
Breaking these numbers down, beginning with Richards, his negative Corsi score is perhaps the first thing that stands about his totals. However, if we judge his performance according to Eric T.’s Balanced Corsi, we see that according to his zone start he is actually around 3 shots better per 60 minutes than we might expect. His balanced zone shift is also a little higher than we might expect, and if we couple this data with his extremely impressive 53.6% Fenwick with the score tied, there is a lot here to suggest that Richards is carrying the water at even strength.
Moving to Carter, his totals are just as impressive. Carter actually was put in tougher defensive spots than Richards, and his Corsi ON score is a little more than 4 shots higher per 60 minutes. His Balanced Corsi is around 7 shots higher than what we might expect from a player put in similar situations, and his BZS is around 3 percent to the good. His Fenwick score, though lower than Richards still suggests that he was also doing a major part driving the play forward for the Flyers considering his zone starts.
What is even more impressive is that the above analysis doesn’t even take into account the elephant in the room: quality of competition. Below is a chart of the toughest CorsiRelQoC scores of every player listed as a Center on Behind the Net last season, minimum 20 games played:
| Rank | Player | Team | CorsiRelQoC |
| 1 | BRANDONDUBINSKY | NYR | 1.436 |
| 2 | ARTEMANISIMOV | NYR | 1.412 |
| 3 | HENRIKZETTERBERG | DET | 1.383 |
| 4 | DAVEBOLLAND | CHI | 1.353 |
| 5 | PAVELDATSYUK | DET | 1.175 |
| 6 | BRIANROLSTON | N.J | 1.084 |
| 7 | JORDANSTAAL | PIT | 1.037 |
| 8 | PATRICEBERGERON | BOS | 1.026 |
| 9 | OLLIJOKINEN | CGY | 1.006 |
| 10 | STEPHENWEISS | FLA | 1.004 |
| 11 | PATRICKMARLEAU | S.J | 0.998 |
| 12 | NATETHOMPSON | T.B | 0.973 |
| 13 | MARCUSJOHANSSON | WSH | 0.953 |
| 14 | BROOKSLAICH | WSH | 0.95 |
| 15 | BRADMARCHAND | BOS | 0.921 |
| 16 | DAVIDBACKES | STL | 0.907 |
| 17 | JEFFCARTER | PHI | 0.896 |
| 18 | TOMASPLEKANEC | MTL | 0.895 |
| 19 | SAKUKOIVU | ANA | 0.879 |
| 20 | DARRYLBOYCE | TOR | 0.857 |
| 21 | MARTINHANZAL | PHX | 0.837 |
| 22 | JERREDSMITHSON | NSH | 0.83 |
| 23 | STEVEOTT | DAL | 0.809 |
| 24 | DAVIDLEGWAND | NSH | 0.805 |
| 25 | PAULSTASTNY | COL | 0.802 |
| 26 | MIKERIBEIRO | DAL | 0.8 |
| 27 | MARTYREASONER | FLA | 0.796 |
| 28 | BRENDANMORRISON | CGY | 0.771 |
| 29 | MIKERICHARDS | PHI | 0.752 |
| 30 | JORDANCARON | BOS | 0.745 |
Both Richards and Carter show up in the conversation with guys who are playing against some of the toughest players in the league. Though they may not score upwards of 80 points per season, both players are certainly producing at elite levels considering the players that they are expected to face night-in and night-out.
What is more, thanks to JaredL we are able to take a look at how the Flyers performed during the past two seasons with and without either Richards or Carter on the ice:
| Player On-Ice | Corsi/60 | Time (mins) | Corsi QoC |
| Both | 3.398 | 211.85 | 0.646 |
| Richards | 1.286 | 2053.55 | 0.51 |
| Carter | 4.953 | 1950.467 | 0.386 |
| Neither | -2.322 | 3514.417 | -0.214 |
Unsurprisingly, these numbers fall in line with everything else we’ve seen – they were able to send the play in the right direction while eating the majority of the team’s tough-minute assignments. Jared was also kind enough to provide data that looks into how some of the Flyers’ other key players performed in situations both with and excluding one of Richards or Carter on the ice during the same time-frame:
| Player On-Ice | With | Corsi/60 | Time (mins) | Corsi QoC |
| Giroux | Either | 4.698 | 1379.317 | 0.839 |
| Giroux | Neither | 4.204 | 784.9 | -0.456 |
| Briere | Either | 8.912 | 895.383 | 0.069 |
| Briere | Neither | -0.775 | 1238.717 | 0.224 |
| Hartnell | Either | 2.674 | 987.267 | -0.576 |
| Hartnell | Neither | -1.230 | 1121.883 | 0.329 |
| van Riemsdyk | Either | 2.194 | 1148.467 | 0.528 |
| van Riemsdyk | Neither | 4.426 | 704.967 | -0.169 |
Once again, we see that no matter the situation, each player was better with one of either Richards or Carter on the ice except for James van Reimsdyk whose data has a noticeable discrepancy in quality of competition. In order for the Flyers to remain one of the premier Stanley Cup contenders in the Eastern Conference, it is looking more and more like the big line of JVR, Claude Giroux and Jaromir Jagr is going to be asked to carry the mail against top-tier competition in the absence of Richards and Carter. These numbers seem to suggest that it is certainly possible, but we will have to wait until each plays an adequate number of contests before we can finally say whether Paul Holmgren’s plan will pay off in the long run. So far, the Flyers are off to an excellent start, but Giroux & Co. will have to keep up their play in the absence of what was one of the league's most formidable one-two punches up front.
Friday, July 15, 2011
Yeah, But: QualComp
"Going by these stats, X is better than Y." "Yeah, but…"
In the Yeah, But series, I will be taking a closer look at the stats lurking in the background. We don't care about them per se, but they provide context and help us compare and rate players. My initial plan was to start the series with a discussion of shooting percentage, but after reading this article at arcticicehockey by Dirk Hoag this seems like a good time to take a deeper look at QualComp.
If you've ever argued with someone about whether one player is better than another, you've probably had a conversation that went something like:
Pens fan: Letang had a better year than Lidström. Look at how much better his +/- and Corsi stats are!
Wings fan: Yeah, but Lidström faced much tougher opponents!
Here are their Corsi/60 and Corsi QualComp stats, courtesy of Behindthenet:
Roughly speaking, Lidström put up decent numbers against the toughest competition in the league, while Letang did very well against opponents that were a little above average. How much should we boost Lidström's numbers to compensate?
The obvious way to figure this out would be to list the Corsi QoC and Corsi rate for each player (maybe above some time-played threshold) and run a regression. Unfortunately, that won't work very well and will drastically underestimate the importance of QualComp.
Say you are watching your favorite team play on home ice. It's tied in the second period and, with both teams at full strength, you see the opposition put their best players on the ice. Let's stop there. What should you expect in the next minute or so?
The first thing is that the opponents will be good, so you shouldn't expect much. That's what I'll call the competition effect - the better the opposition, the worse your expected results will be. This is what we would like to measure. On the other hand, your coach will see who is on the ice and will probably put out one of the top lines and a good defense pairing. This means you should expect good results. I'll call this the matchup effect - the better the opposition, the better your players on the ice, which will raise the expected results.
Which one will win out depends on how important the competition effect is and, importantly, how much the coaches focus on the matchups. In the regular season, they put some importance on them, but they have to take the long haul into account. In the playoffs, matchups get a lot of focus by everyone from casual fans to bloggers to on-air analysts to the coaches themselves. Barring serious injuries like Crosby's concussion, coaches do close to everything they can to win the game they're in. A top line facing a bottom line is rare, especially on face offs that don't follow icing. In the playoffs we would expect the matchup effect to dominate the competition effect. More on that in a bit.
To see how this all works, let's oversimplify things and say that Corsi rates are simply additive (subtractive?) - if your line is +5 and you face my line, which is +2, then your Corsi rate will be 5 - 2 = 3. Here's a graphical representation a possible matchup between two teams:
You can see the competition effect within each line - the slope of the line connecting the points is -1, which comes from the (over)simplifying assumption that Corsi rates are additive. You can see the matchup effect by noting that as you move from left to right from the opponent's fourth line to their first, you will tend to also move up toward where your first line is because of matchups.
Let's look at a real-world example - the first-round series between the Vancouver Canucks and the Chicago Blackhawks. There are several things about this series that make it a good one to demonstrate the matchup effect. It's in the playoffs, when coaches go out of their way to avoid randomizing matchups. The series went seven games, with a couple overtimes for good measure, which provides more data. The sample size is in a sweet spot where it's small enough for me to be able to break it down but big enough for there to be evidence of the effects I'm trying to demonstrate. The teams also fit the bill nicely - both feature two very strong lines with a significant dropoff. As usual, I will restrict attention to 5-on-5 play where both goalies were in net.
Overall, Vancouver dominated this series 5-on-5. Chicago had no answer, apart from Crawford in goal, for either of the Canucks top lines. Meanwhile, the Blackhawks' best struggled. Looking at the matchups, the only place where Chicago had an edge was when the bottom lines faced each other. The Blackhawks show the competition effect quite well - those players, including their best, that faced the best on the Canucks had awful stats while those playing mainly against the bottom lines did better. On the other hand, the Canucks were an example of the matchup effect swamping the competition effect. The players that faced the toughest opposition were their top players, and they did extremely well. The less skilled players struggled, even though they played far worse opposition.
Here are three plots giving the QualComp and Corsi rates for all players with at least 15 minutes of ice time against both the top and bottom two lines (combined) of the other team. For QualComp, I took the weighted average of the regular-season Corsi from BTN.
It doesn't look like much of a relationship and this is confirmed in the regression, with similar results to what Dirk Hoag discussed:
Instead of averaging everything out, let's look at within-player results. That is, after all, the information we want to know - e.g. if Lidstrom played against competition similar to Letang's, what would we expect his Corsi to be? To do this, we need to create some kind of split in the data. Vic Ferrari compared the first half of the season and the second. I will instead look at performance when facing the opponent's top two lines compared to the bottom two. Keep in mind that I'm only trying to demonstrate the matchup effect. Trying to determine the importance for the league based on a 7-game sample between two teams would obviously not work all that well.
Here is a similar scatterplot to the above, but this time with the data split:
Running the same regression on the pooled data gives us:
Splitting the data up so we capture the change in Corsi for each player gives us a regression indicating that the effect is over three times as large, is statistically significant (before it wasn't even close) and an R^2 over 65 times larger.
When looking at correlations, you have to be careful to think about how coaching decisions affect everything. Good players tend to play with other good players and against other good players. This influences basically all the "yeah, but…" stats including quality of competition and teammates, zone starts and special teams. I'll explore those more in future articles. To measure effects like Qual Comp it is important to use a method that captures changes for a player or group like what I've done here, WOWY or the method Vic Ferrari used in his article on Qual Comp.
In a future article I will try using a similar methodology on a larger scale to figure out how large the effect actually is.
In the Yeah, But series, I will be taking a closer look at the stats lurking in the background. We don't care about them per se, but they provide context and help us compare and rate players. My initial plan was to start the series with a discussion of shooting percentage, but after reading this article at arcticicehockey by Dirk Hoag this seems like a good time to take a deeper look at QualComp.
If you've ever argued with someone about whether one player is better than another, you've probably had a conversation that went something like:
Pens fan: Letang had a better year than Lidström. Look at how much better his +/- and Corsi stats are!
Wings fan: Yeah, but Lidström faced much tougher opponents!
Here are their Corsi/60 and Corsi QualComp stats, courtesy of Behindthenet:
| Player | Corsi | Corsi QualComp |
| Lidström | 2.65 | 1.205 |
| Letang | 12.01 | 0.281 |
Roughly speaking, Lidström put up decent numbers against the toughest competition in the league, while Letang did very well against opponents that were a little above average. How much should we boost Lidström's numbers to compensate?
The obvious way to figure this out would be to list the Corsi QoC and Corsi rate for each player (maybe above some time-played threshold) and run a regression. Unfortunately, that won't work very well and will drastically underestimate the importance of QualComp.
Say you are watching your favorite team play on home ice. It's tied in the second period and, with both teams at full strength, you see the opposition put their best players on the ice. Let's stop there. What should you expect in the next minute or so?
The first thing is that the opponents will be good, so you shouldn't expect much. That's what I'll call the competition effect - the better the opposition, the worse your expected results will be. This is what we would like to measure. On the other hand, your coach will see who is on the ice and will probably put out one of the top lines and a good defense pairing. This means you should expect good results. I'll call this the matchup effect - the better the opposition, the better your players on the ice, which will raise the expected results.
Which one will win out depends on how important the competition effect is and, importantly, how much the coaches focus on the matchups. In the regular season, they put some importance on them, but they have to take the long haul into account. In the playoffs, matchups get a lot of focus by everyone from casual fans to bloggers to on-air analysts to the coaches themselves. Barring serious injuries like Crosby's concussion, coaches do close to everything they can to win the game they're in. A top line facing a bottom line is rare, especially on face offs that don't follow icing. In the playoffs we would expect the matchup effect to dominate the competition effect. More on that in a bit.
To see how this all works, let's oversimplify things and say that Corsi rates are simply additive (subtractive?) - if your line is +5 and you face my line, which is +2, then your Corsi rate will be 5 - 2 = 3. Here's a graphical representation a possible matchup between two teams:
You can see the competition effect within each line - the slope of the line connecting the points is -1, which comes from the (over)simplifying assumption that Corsi rates are additive. You can see the matchup effect by noting that as you move from left to right from the opponent's fourth line to their first, you will tend to also move up toward where your first line is because of matchups.
Let's look at a real-world example - the first-round series between the Vancouver Canucks and the Chicago Blackhawks. There are several things about this series that make it a good one to demonstrate the matchup effect. It's in the playoffs, when coaches go out of their way to avoid randomizing matchups. The series went seven games, with a couple overtimes for good measure, which provides more data. The sample size is in a sweet spot where it's small enough for me to be able to break it down but big enough for there to be evidence of the effects I'm trying to demonstrate. The teams also fit the bill nicely - both feature two very strong lines with a significant dropoff. As usual, I will restrict attention to 5-on-5 play where both goalies were in net.
Overall, Vancouver dominated this series 5-on-5. Chicago had no answer, apart from Crawford in goal, for either of the Canucks top lines. Meanwhile, the Blackhawks' best struggled. Looking at the matchups, the only place where Chicago had an edge was when the bottom lines faced each other. The Blackhawks show the competition effect quite well - those players, including their best, that faced the best on the Canucks had awful stats while those playing mainly against the bottom lines did better. On the other hand, the Canucks were an example of the matchup effect swamping the competition effect. The players that faced the toughest opposition were their top players, and they did extremely well. The less skilled players struggled, even though they played far worse opposition.
Here are three plots giving the QualComp and Corsi rates for all players with at least 15 minutes of ice time against both the top and bottom two lines (combined) of the other team. For QualComp, I took the weighted average of the regular-season Corsi from BTN.
It doesn't look like much of a relationship and this is confirmed in the regression, with similar results to what Dirk Hoag discussed:
| Coeff. | t | R2 | |
| Corsi QualComp | -0.54 | -0.23 | 0.0016 |
Instead of averaging everything out, let's look at within-player results. That is, after all, the information we want to know - e.g. if Lidstrom played against competition similar to Letang's, what would we expect his Corsi to be? To do this, we need to create some kind of split in the data. Vic Ferrari compared the first half of the season and the second. I will instead look at performance when facing the opponent's top two lines compared to the bottom two. Keep in mind that I'm only trying to demonstrate the matchup effect. Trying to determine the importance for the league based on a 7-game sample between two teams would obviously not work all that well.
Here is a similar scatterplot to the above, but this time with the data split:
Now that we're looking at differences for each player, we can see a much stronger relationship. It is stronger for the Chicago players but the most impressive change is on the Vancouver side. Aggregating all the minutes, there was a positive relationship between opponent strength and Corsi rate due to the matchup effect. Splitting it up we see that the Canucks players tended to be less productive against better opposition, although the effect was quite weak - this almost surely would be bigger over a more reasonable sample.
Running the same regression on the pooled data gives us:
| Coeff. | t | R2 | |
| Corsi QualComp | -1.92 | -2.87 | 0.1054 |
Splitting the data up so we capture the change in Corsi for each player gives us a regression indicating that the effect is over three times as large, is statistically significant (before it wasn't even close) and an R^2 over 65 times larger.
When looking at correlations, you have to be careful to think about how coaching decisions affect everything. Good players tend to play with other good players and against other good players. This influences basically all the "yeah, but…" stats including quality of competition and teammates, zone starts and special teams. I'll explore those more in future articles. To measure effects like Qual Comp it is important to use a method that captures changes for a player or group like what I've done here, WOWY or the method Vic Ferrari used in his article on Qual Comp.
In a future article I will try using a similar methodology on a larger scale to figure out how large the effect actually is.
Wednesday, June 29, 2011
Why Shooting Stats Are Better Than Goals
Let's say you are asked to rank the NHL teams halfway through the season. Which stats should you use to do this?
Before getting to that, we need to think carefully about what a ranking means. The best team in the NHL is the one that is the best at winning games, and so on down the line. This comes down to two things - scoring goals and preventing the opposition from doing the same. If someone says X "is the best team in the league", what they mean is that X is the best at outscoring their opponents. Similarly, if Y is the best player in the league that means that he is the best at the combination of generating goals for his team and preventing them for the other.
Success at scoring and preventing goals in hockey, like every activity, is a combination of skill and luck. For some things, e.g. roulette, luck is the dominant factor. In others, like sprinting 100m, skill overwhelmingly wins the day. Hockey falls somewhere in the middle, perhaps closer to roulette than anyone would care to admit. Getting back to ranking the teams, that means figuring out which are the strongest at the skill part. Note that I'm using skill loosely here to refer to any skills that help a team score goals and prevent them, including those like grit and mental toughness that pundits love to talk about.
There are a few ways to tease out this skill component, all of which I will use in various articles in the future. Here I will compare stats from each team in two different groups of games - each half of the season, numbered even vs. odd, etc. The idea behind this is that luck in the first half of the season and luck in the second half of the season should be completely unrelated. Sometimes your team will get lucky in the first half and unlucky in the second half of the season, but the opposite is just as likely. Think of it like two coin tosses. If you win the first coin toss then you are no more likely to win the second than if you'd lost it. In contrast with the luck factor, your team should usually be about as skilled in the second half of the season as the first. If there is no relationship, known as correlation, between luck in the first half of the season and the second any link will be due to skill.
Mostly due to the availability of data, I restrict attention to 5-on-5 situations where both goalies are on the ice. For each of the past four years, I split the season in half and look at how goals and shooting stats in the first half relate to goals in the second half. Because we care about both scoring and allowing goals, I expressed this as a percentage: goals for divided by the sum of goals for and against (GF/(GF+GA)). The same goes for shooting stats.
Here is a graph of the relationship between goal percentage in the first half of the season and the second. All data are from timeonice. See links on the right.
It looks rather weak. The numbers back that up - the correlation is just 0.13. This is not statistically significant. Even ignoring that, it's pretty clear that putting up good scoring numbers 5-on-5 with the goalies in net in the first half of the season doesn't mean much in the way of predicting performance in the second half.
The relationship between Corsi percentage in the first half of the season and goal percentage in the second half is far stronger. Corsi percentage is like goal percentage, but for all types of shots, including missed shots and blocked shots. Here is the scatterplot:
You can see a distinguishable up-and-right pattern, which indicates a stronger relationship between the two. The correlation is 0.36, which is statistically significant. Keep in mind that we're looking at how shooting ratios in the first half relate to goals in the second half.
Let's look at the best and worst teams in the first half of this last season. The New Jersey Devils were an impressively bad 10-29-2 on January 8th, with an overall goal differential of -58 (72 - 130). 5-on-5 with goalies in their goal differential was -48 (45 - 93) and goal percentage 32.6%. That is the worst goal percentage in either half for any team in any of the four seasons of data that is available at timeonice. In contrast, the Flyers looked like world beaters halfway through. Their record was 26-10-5, goal differential +30 (137-107) and goal% 5-on-5 a cool 60%. What happened in the second half? The Devils put up one of the best turnarounds in NHL history, nearly making the playoffs, and the Flyers record was mediocre. The Devils went 28-10-3, the Flyers 21-13-7. The Devils had an overall goal differential of +23 (102-79), the Flyers +6 (122-116). 5-on-5 with goalies, New Jersey had a goal differential of +23 (76-53), 58.9%, and Philly 0 (81-81), 50%.
How could the worst team in the league in the first half have a better second half than the best team by such a large margin? The answer comes down to the luck factor I discussed above. In the first half, New Jersey took 52.6% of all the 5-on-5 Corsi shots in their games. Philadelphia was actually worse, just better than even at 50.6%. Despite that, the Devils got hugely outscored and the Flyers got far more goals than their opponents. While skill may be a factor in shots going in and being saved by your own goalie, the topic of my next article, luck plays a massive role in scoring over just a half season. The Devils were clearly not getting the bounces and the Flyers were. In the second half of the season, Philadelphia's luck was about average and New Jersey actually caught the breaks.
You can see how much better Corsi stats handle luck by looking at the two teams in the graphs above. New Jersey is the red point and Philadelphia orange. You can see that the Devils are a huge outlier when you look at goals in the first and second, but not so looking at Corsi in the first half and goals in the second, though you can see that they were fortunate. The goals graph is so scattered that the Flyers don't stand out much, but you can see that they dropped off a lot by how far they are from the top of graph. On the Corsi graph they are right in the middle, so from that perspective their second-half performance should have been expected instead of surprising.
Other articles might stop there, but things get more interesting if you run a regression. Regression analysis is a tool I will use pretty frequently. It allows you to separate out different effects. In our case, we want to know how important goals in the first half are once you take Corsi into account, and vice-versa. The regression makes it very clear that Corsi% is a far, far better predictor of goal% in the second half than first-half goal%. Not only that, it appears that virtually all of the tiny amount of explanatory power you get from goal% comes from the fact that goals are a type of shot.
When the regression spits out a formula, the size of the coefficient tells you how big its effect is. When both first-half goal% and Corsi% are included, the goal% coefficient is a minuscule 0.007. For the stats nerds, the standard error is 0.087 so the p-value is an astonishing 0.936. This is about as statistically insignificant as it gets. For comparison, the coefficient for Corsi% is 0.550 (SE of 0.142, p < 0.001) which is very strongly significant. If you have a team that breaks even on goals in the first half of the season but Corsi outshoots its opponents 60-40 then they will average about 83.3 goals scored and 66.7 allowed in the second half of the season (assuming 150 total 5-on-5 goals, which is close to the league average). If instead you have a team that was even on shots but won the goal battle by that much then they will average 75.2 goals in the second half and concede 74.8.
Once Corsi is taken into account, goals do not at all predict future success.
Topics left for future articles:
- What about score effects?
- What about Fenwick?
- What about special teams?
- Is shooting all luck, then?
Before getting to that, we need to think carefully about what a ranking means. The best team in the NHL is the one that is the best at winning games, and so on down the line. This comes down to two things - scoring goals and preventing the opposition from doing the same. If someone says X "is the best team in the league", what they mean is that X is the best at outscoring their opponents. Similarly, if Y is the best player in the league that means that he is the best at the combination of generating goals for his team and preventing them for the other.
Success at scoring and preventing goals in hockey, like every activity, is a combination of skill and luck. For some things, e.g. roulette, luck is the dominant factor. In others, like sprinting 100m, skill overwhelmingly wins the day. Hockey falls somewhere in the middle, perhaps closer to roulette than anyone would care to admit. Getting back to ranking the teams, that means figuring out which are the strongest at the skill part. Note that I'm using skill loosely here to refer to any skills that help a team score goals and prevent them, including those like grit and mental toughness that pundits love to talk about.
There are a few ways to tease out this skill component, all of which I will use in various articles in the future. Here I will compare stats from each team in two different groups of games - each half of the season, numbered even vs. odd, etc. The idea behind this is that luck in the first half of the season and luck in the second half of the season should be completely unrelated. Sometimes your team will get lucky in the first half and unlucky in the second half of the season, but the opposite is just as likely. Think of it like two coin tosses. If you win the first coin toss then you are no more likely to win the second than if you'd lost it. In contrast with the luck factor, your team should usually be about as skilled in the second half of the season as the first. If there is no relationship, known as correlation, between luck in the first half of the season and the second any link will be due to skill.
Mostly due to the availability of data, I restrict attention to 5-on-5 situations where both goalies are on the ice. For each of the past four years, I split the season in half and look at how goals and shooting stats in the first half relate to goals in the second half. Because we care about both scoring and allowing goals, I expressed this as a percentage: goals for divided by the sum of goals for and against (GF/(GF+GA)). The same goes for shooting stats.
Here is a graph of the relationship between goal percentage in the first half of the season and the second. All data are from timeonice. See links on the right.
The relationship between Corsi percentage in the first half of the season and goal percentage in the second half is far stronger. Corsi percentage is like goal percentage, but for all types of shots, including missed shots and blocked shots. Here is the scatterplot:
You can see a distinguishable up-and-right pattern, which indicates a stronger relationship between the two. The correlation is 0.36, which is statistically significant. Keep in mind that we're looking at how shooting ratios in the first half relate to goals in the second half.
Let's look at the best and worst teams in the first half of this last season. The New Jersey Devils were an impressively bad 10-29-2 on January 8th, with an overall goal differential of -58 (72 - 130). 5-on-5 with goalies in their goal differential was -48 (45 - 93) and goal percentage 32.6%. That is the worst goal percentage in either half for any team in any of the four seasons of data that is available at timeonice. In contrast, the Flyers looked like world beaters halfway through. Their record was 26-10-5, goal differential +30 (137-107) and goal% 5-on-5 a cool 60%. What happened in the second half? The Devils put up one of the best turnarounds in NHL history, nearly making the playoffs, and the Flyers record was mediocre. The Devils went 28-10-3, the Flyers 21-13-7. The Devils had an overall goal differential of +23 (102-79), the Flyers +6 (122-116). 5-on-5 with goalies, New Jersey had a goal differential of +23 (76-53), 58.9%, and Philly 0 (81-81), 50%.
How could the worst team in the league in the first half have a better second half than the best team by such a large margin? The answer comes down to the luck factor I discussed above. In the first half, New Jersey took 52.6% of all the 5-on-5 Corsi shots in their games. Philadelphia was actually worse, just better than even at 50.6%. Despite that, the Devils got hugely outscored and the Flyers got far more goals than their opponents. While skill may be a factor in shots going in and being saved by your own goalie, the topic of my next article, luck plays a massive role in scoring over just a half season. The Devils were clearly not getting the bounces and the Flyers were. In the second half of the season, Philadelphia's luck was about average and New Jersey actually caught the breaks.
You can see how much better Corsi stats handle luck by looking at the two teams in the graphs above. New Jersey is the red point and Philadelphia orange. You can see that the Devils are a huge outlier when you look at goals in the first and second, but not so looking at Corsi in the first half and goals in the second, though you can see that they were fortunate. The goals graph is so scattered that the Flyers don't stand out much, but you can see that they dropped off a lot by how far they are from the top of graph. On the Corsi graph they are right in the middle, so from that perspective their second-half performance should have been expected instead of surprising.
Other articles might stop there, but things get more interesting if you run a regression. Regression analysis is a tool I will use pretty frequently. It allows you to separate out different effects. In our case, we want to know how important goals in the first half are once you take Corsi into account, and vice-versa. The regression makes it very clear that Corsi% is a far, far better predictor of goal% in the second half than first-half goal%. Not only that, it appears that virtually all of the tiny amount of explanatory power you get from goal% comes from the fact that goals are a type of shot.
When the regression spits out a formula, the size of the coefficient tells you how big its effect is. When both first-half goal% and Corsi% are included, the goal% coefficient is a minuscule 0.007. For the stats nerds, the standard error is 0.087 so the p-value is an astonishing 0.936. This is about as statistically insignificant as it gets. For comparison, the coefficient for Corsi% is 0.550 (SE of 0.142, p < 0.001) which is very strongly significant. If you have a team that breaks even on goals in the first half of the season but Corsi outshoots its opponents 60-40 then they will average about 83.3 goals scored and 66.7 allowed in the second half of the season (assuming 150 total 5-on-5 goals, which is close to the league average). If instead you have a team that was even on shots but won the goal battle by that much then they will average 75.2 goals in the second half and concede 74.8.
Once Corsi is taken into account, goals do not at all predict future success.
Topics left for future articles:
- What about score effects?
- What about Fenwick?
- What about special teams?
- Is shooting all luck, then?
Putting Skaters in a Context: The World of Advanced Hockey Metrics
With the world of advanced hockey metrics continually improving, we are now beginning to see hockey players evaluated in more diverse ways than ever before. Since the beginning of many a hockey fandom, a quick glance at a skater’s goals, assists and total points has been the measure that grades offensive prowess across the league’s scorers. Now, however, the emergence of a few newer (and quite frankly, better) statistics allows us to take these age-old points totals and put them in a context, showing just how valuable a player may or may not be to his team’s success. Here at Driving Play, while attempting to evaluate different players across the league we will be commonly referring to many of these newer statistics within our analysis. Below is a quick list that will attempt to make clear just what we may be referring to if an unfamiliar term happens to appear within one or more of our posts.
A Corsi Number – Similar to a +/- statistic, Corsi gives a player a (+) upon the event of his team generating either a shot on goal, a missed shot, or a blocked shot directed at the opponent’s net while he is on the ice. Similarly, a player earns a (-) if the opponent generates a shot on goal, missed shot, or a blocked shot directed at his own net. Sometimes this can be expressed as a percentage, i.e. the percentage of the total shots that are directed at the opponent’s net while a player is on the ice. Corsi can also be expressed in a “Relative Corsi” number which is the difference between a player’s on-ice Corsi score and the shot differential while he is on the bench. Relative Corsi is generally used to look at which players are having the most positive effects on shot totals relative to their teammates.
A Fenwick Number – Since many consider shot-blocking a measurable skill in the hockey world, a Fenwick number is the same as a Corsi number, except blocked shots are taken out of the equation. So, a player will earn a (+) if his team generates a shot on goal or a missed shot whilst he is on the ice, and a (-) if either event occurs for the opponent.
Quality of Competition (QUALCOMP) – The fact is, all ice time in the NHL is not created equal. Having to line-up toe-to-toe with Sidney Crosby is a much different task than Jesse Winchester, the hockey player or the musician. QUALCOMP more or less weighs the on-ice +/- (the familiar statistic measured in goals) of a player’s opponents relative to the rest of his teammates, and averages this rating across every player faced during the season. The higher the resulting rating, the better the competition a player is facing and vice versa. There is also a CorsiRelQUALCOMP number which does the same thing, except uses Relative Corsi instead of +/-.
Quality of Teammates (QUALTEAM) – Similar to QUALCOMP, QUALTEAM weighs a player’s teammates using the exact same formula as QUALCOMP. Just like QUALCOMP, a player’s QUALTEAM rating will be higher if he is playing with first-line teammates and vice versa if he is playing with fourth-line enforcers. Also similarly, CorsiRelQUALTEAM will measure a player’s teammates using Relative Corsi.
Zone Start Percentage – A zone start percentage measures the percent of the time any player starts his shift in the offensive zone. As you might expect, players with a high defensive prowess are often called upon to start in the defensive zone frequently, and vice versa is true for those players who are more inept in their own end. This particular statistic is important in that it can directly affect a player’s aforementioned Corsi or Fenwick percentage since players who are starting in the offensive zone more frequently will have an easier time generating more shots towards the opponent’s net. What’s more, players who are more immediately deployed in defensive roles will have a harder time finding shot opportunities than their counterparts who are already starting in prime offensive positions.
Score Effects – Within the ebbs and flows of a hockey game, it has been a long-believed ideal that teams will go into more of a “defensive mode” while ahead and try and get just about every shot possible on net while behind. Using Corsi and Fenwick percentages, it has been shown that teams who enjoy an advantage in the score are commonly outshot at improving rates as the game progresses and vice versa. With the score tied, the disparity in shot totals is most close to even which is why many advanced hockey statisticians choose to look at Corsi/Fenwick with the score tied at even strength to put players’ ice time on a level playing field.
Coming back to the original point regarding putting different skaters in a context, we are now able to more closely examine the situations that different players are playing in. For this reason, it is now much easier to come to a conclusion about their value to their respective teams. Before these statistics came into play, we could look at two players, Patrice Bergeron and Ville Leino for example, who had similar point totals during the regular season (57 and 53 respectively). In a vacuum, it may seem as if they are both comparable players toward Boston and Philadelphia’s total success. However, a little scratching beneath the surface reveals that Bergeron played against much tougher competition than Leino, and Leino enjoyed the luxury of skating with better teammates. Leino started in the offensive zone a walloping 62.3% of the time compared to Bergeron’s 42.7%, showing us that Leino was given far more prime scoring opportunities to begin his shifts which undoubtedly had a positive effect. Finally, Bergeron’s Corsi and Fenwick percentages with the score tied at even strength were 52.7 and 52.8% respectively, compared to Leino’s 54.9 and 53.1%. While a higher percentage of the on-ice shots were directed at the opponent’s net while Leino was on the ice, we have of course already noted that Bergeron faced tougher opponents and played with worse teammates than Leino which gave Leino an advantage in putting up better numbers in these categories. Had Leino, a notoriously subpar defensive forward (see: 2 seconds of average shorthanded time-on-ice/game in ’10-11) been given minutes similar to Bergeron’s, the point totals most certainly would not have looked anything similar. Considering the minutes they were given, Bergeron most certainly had an excellent season while Leino performed at a level around what we might expect from a forward given “softer” minutes during each game.
Subscribe to:
Posts (Atom)