In my first post here at MLHS, I wrote an introduction to advanced statistics which generated some good discussion in the comments. A number of the comments were with respect to WOWY analysis and some more specifically with respect to Kessel and Bozak. I will delve into that more here in my second post.
As explained in my introduction to advanced statistics post, WOWY is short for “With Or Without You” and looks at how the team performs when two players are on the ice together and when they are each playing apart from each other. So, for Phil Kessel and Tyler Bozak, we would look at how the team performs in three situations:
- Both Kessel and Bozak are on the ice together
- Only Kessel is on the ice
- Only Bozak is on the ice
This table shows how the Leafs have performed in these three situations according to CF% (corsi for percentage, or percentage of shot attempts for and against that the Leafs took), GF% (goals for percentage, or percentage of goals for and against that the Leafs scored) and GF20 (goals for per 20 minutes of play). I included GF20 individually because I wanted to show how Bozak impacts the teams offensive production more specifically when Bozak is on the ice.
Kessel + Bozak WOWYA WOWY comparison of Kessel and Bozak
|Kessel with Bozak||CF%||46.5%||48.8%||46.9%||44.6%||46.7%|
|Kessel without Bozak||CF%||47.1%||50.7%||43.0%||43.6%||46.3%|
|Bozak without Kessel||CF%||48.2%||39.2%||25.5%||28.5%||40.2%|
|Kessel with Bozak||GF%||42.5%||47.7%||51.7%||56.2%||49.7%|
|Kessel without Bozak||GF%||56.3%||48.9%||64.3%||46.2%||49.2%|
|Bozak without Kessel||GF%||22.7%||38.9%||75.0%||25.0%||32.1%|
|Kessel with Bozak||GF20||0.73||0.93||0.97||1.19||0.95|
|Kessel without Bozak||GF20||0.90||1.15||1.40||0.72||0.98|
|Bozak without Kessel||GF20||0.37||0.88||1.00||0.67||0.62|
Tables are good, but charts are often easier to quickly draw conclusions from. Here are WOWY charts for CF% and GF%:
Let’s start with the first chart looking at CF%, which can act as a proxy for puck possession but specifically measures which team is getting the most shot attempts. Anything over 50% means your team is getting more shot attempts than they are giving up; that is generally a good thing. As you can see, the lines for Kessel with Bozak and Kessel without Bozak are almost identical, indicating that Bozak had very little impact on Kessel’s overall statistics. Conversely, when Bozak is apart from Kessel, Bozak’s CF% stats tank, meaning Kessel (along with Lupul/JVR) has a significant positive impact on Bozak’s statistics. This would indicate that Kessel is not dependent on Bozak whereas Bozak is definitely more dependent on Kessel.
When looking at GF%, though, we see that until last season Kessel had a better GF% apart from Bozak than with Bozak. This reversed last season, when Kessel was better with Bozak than without. Over the past 4 years combined, Kessel’s GF% with Bozak and without Bozak is almost identical, which is consistent with what we saw with CF%.
Remember that sample sizes are significantly smaller when dealing with goals than shot attempts. There is much more fluctuation and randomness in these statistics, so we are more looking for trends than absolute numbers. This small sample size is a big reason for Bozak’s good GF% away from Kessel in 2012-13. In that season, Bozak only played just under 60 minutes of ice time apart from Kessel and in that ice time he was on the ice for 3 goals for and 1 goal against for a GF% of 75%. That isn’t anywhere close to a large enough sample size to draw any conclusions from. The overall trend we see in both the CF% and GF% charts are that Bozak has little or no effect on Kessel’s overall on-ice stats.
I wanted to look at GF20 specifically to see if Bozak is having an effect offensively or defensively on Kessel’s statistics. Here is how Bozak and Kessel’s GF20 WOWY chart looks.
For three straight seasons (2010-11, 2011-12 and 2012-13), Kessel had a higher goals for rate without Bozak than with Bozak but that reversed last season when Kessel’s had a better GF20 with Bozak than without. One season does not make a trend, though, and until last season the offensive production of the Leafs was significantly worse when Kessel was with Bozak than when Kessel was not with Bozak. Did something change last season? Has Bozak figured something out that has reversed the trend? We’ll have to see what next season brings. I suspect it was just an anomaly, but I’ll leave it up to Bozak to prove otherwise.
Now, we have to keep some perspective on these “without you” statistics for both Kessel and Bozak. When Kessel isn’t playing with Bozak he has mostly been playing with Grabovski or Kadri, both pretty solid centers so it makes it easier to maintain his “without Bozak” numbers. That said, Kessel has generally been as good or better with those other two than with Bozak over the past 4 seasons, so it makes for a difficult argument that Bozak is actually a better player than either of those guys or more deserving of #1 center role.
For Bozak, when he is not playing with Kessel (particularly the past couple seasons) he has generally been playing with 3rd line players like McClement, Kulemin, Raymond, Komarov, etc. in very defensive roles. This could have a significant negative impact on his without Kessel statistics. That said, over the past 2 seasons, Bozak’s CF% without Kessel is 27.5% and his GF% without Kessel is 37.5%. These are horrifically bad numbers and pretty difficult to justify, but could in part be due to Bozak being used to win defensive zone face offs when without Kessel. If he loses a defensive zone face off, he gets stuck on the ice for a bunch of shots against. If he wins the face off and Leafs clear the zone he heads to the bench getting no benefit of any subsequent offensive chances. This certainly would be a factor but I have serious doubts it is enough to justify a 27.5 Corsi percentage. Playing with Jay McClement a fair bit certainly doesn’t help, either, as McClement consistently put up poor possession statistics every year (that article is another example of how one can use WOWY’s in player evaluation).
I hope this provides a little more detail as to how one might conduct a WOWY analysis of a pair of players as a method of determining which players are more critical to the on-ice success of the team. In the example of Kessel and Bozak, there is little or no evidence that Bozak is integral to Kessel’s production where as there is a lot of evidence that Kessel is integral to Bozak’s production. As always, if you have any questions feel free to ask in the comments.
To better understand this analysis, please see David Johnson’s post: A Guide to WOWYs
Just had a thought reading through everyone's comments, I see a lot of posters here assuming that because bozak doesn't make kessel better that he's no good. But I'd like to see more info before coming to a judgement (full disclosure - I'm a bozak supporter - in his play/effort/talent for his pay grade, not calling him joe Thornton here but he's good value IMO)
To really get a handle on how bozak impacts the team, we need a wowy for every player with bozak, not just kessel. Can he really be faulted for not making one of the best offensive players in the world better than he already is?
Food for thought: If he doesn't slow kessel down, and makes every single player he plays with apart from kessel better, then isn't that a better representation of value?
Not trying to steal your personal time here David but I'd love to see a qualified statistician like yourself discuss this point of view.
Great article, glad to have you aboard the MLHS family David. very concise and relevant, something all analytics articles should strive for. Looking forward to reading more.
I would throw in my opinion that bozak looked like different player this year and was good to very good, I hope and expect to see that continue next year
These advanced stats are certainly interesting. It's worth relating them to other observations.
During the 2013-14 season, there was a stretch of 9 games in November and 12 games in December when Kessel played without Bozak. During that stretch, he had Kadri, Smith, and even JvR as his center. Here are Kessel's stats for that period without Bozak.
November: 3GF and 6GA for a -3 plus/minus rating over 9GP => GF%=33%
December: 3GF + 7A and 15GA for a -5 plus/minus rating over 12GP => GF%=40%
Even if Kessel's CF% didn't slip, it seems he wasn't getting quality chances. His shots were either missing / being blocked or more likely the opposing goalie could see it all the way, easily stopping Kessel.
When Bozak and Kessel are together, their shooting % might be higher, both complementing each other and GF% hovering just under 60%.
Thanks for these insights David, been really interesting to read them and begin to understand the wonderful world that is statistics.
I just have a quick question though, you have three factors that you are looking at;
Kessel With Bozak
Kessel Without Bozak
Bozak Without Kessel
Should there also be a fourth factor that is Bozak With Kessel? Or does the Kessel With Bozak look at a team stat not an individual stat so it is not needed?
Sorry if this makes no sense, you're intro post was my introduction to Hockey stats.
Kessel and Bozak having on-ice chemistry is - and always has been - a myth. Kessel makes Bozak better because he's elite. I bet everyone's numbers would be better with Kessel than without him, in the long run. Kessel will be great with anyone. I'd love to see what he could do with another player of his caliber, though. He has gotten to play with some good linemates, (Lupul and JVR), but nobody that comes anywhere close to his level. Not since Savard, at least.
Basically over the 4 years Kessel's stats are the same with or without Bozak. The way some numbers guys and bloggers talk about how bad a player Bozak is, that playing with Bozak should drag him down, but the numbers show, over a 4 year period it hasn't. One thing I would like to see is how Kessel affects other player he plays with. I know a lot of people still put him down, but my thinking is that he makes everyone around him a better player.
good analysis David, thing is, is that Kessel just makes everybody better, even JVR....he is a world class player whose passes are as good as anyone in the league, and he's underrated as a play maker too....could work on his breakaway a little though :) He should forget about deking and just shoot the disc like he does on the fly...
These posts are always decent, but the next level would be to explain specifically on the ice what happens to cause this. This just shows you what, not why. This is the big thing lacking for me in advanced stats.
Nice write up. What this tells me is that we have to keep Bozak on that first line because Kessel makes him better. Which keeps his trade value higher (if they go that route).
I feel bad for saying this but I'm hoping for a small injury (Bozak) to see someone like holland can get a shot just to see if he has the potential to be up there, we have lots of options if it doesn't work out this year for centre depth.
It's very difficult to judge Bozak when he is away from Kessel. As the article stated, when he is not playing with Kessel he is probably on the PK or some kind of defensive role. Probably because of his faceoff ability. And even that sample size is probably not as large as we would like it to be. Has he ever gotten a look with the likes of Lupul, Raymond, Clarkson, Kulemin etc? If so the sample size must be miniscule.
I have heard that his play style suits Kessel's game more. Grabo would hang on the the puck for long stretches in the offensive zone which kind of mirrors Kessel's game. Carlyle has also said Kadri needs to play a North/South game instead of East/West. I'm afraid to quote Carlyle on this website but what the hell. I'm not an expert but maybe Bozak's style of play is actually the best fit for his linemates out of the current crop of centers the Leafs have. Wouldn't mind seeing Holland get a look at the 1C spot if he can improve. His style of play may fit. I think he also brings a bit of grit and sandpaper. He's not a superstar 1C but neither is Bozak. Just food for thought.
Good work David, I'm surprised by the praise you're getting in the comments section. Whenever I, or anyone else, has suggested this, everyone would lose their minds. Bozak is a very well liked person, but he's merely a placeholder on a line that does not gain anything from his presence, or loses anything from his absence. What worries me about this, is that when someone does emerge as a clear cut 1C, Bozak will struggle anywhere else in the lineup. He's not good enough offensively to lead a secondary scoring line in the top 6, and not good enough defensively for the bottom 6. The Leafs should trade him while his value has peaked.
Good stuff David....couple of questions though...
Perhaps I missed it...but what's the sample size for each category?...for example how many games total was each stat based on (kessel with, kessel without, bozak without)? Are the number of games for each stat comparable? or very different...I ask because it seems like Bozak and Kessel have been together much more than apart.....
Also are all the stats based on even strength or is it mixed bag of even,pp and short handed?
What about the impact of the competition they faced during each specific time period? not only player matchups but team matchups? (above .500 teams...below .500 teams?)...and what about goaltending quality faced?
It just seems that there could be a lot of different variables that directly affect each one of these stats....
Absolutely liked this David. I wonder how all this data collected thru stats actually impacts performance on the ice? A lot of it doesn't seem to matter. Shouldn't player interaction and player intelligence in these interactions be more important? Stats that can show good judgement or poor judgement would be way more useful to a team. If for example, you can actually prove thru stats that say Franson, makes worse judgements on the ice repeatedly compared to say Gardiner and Reilly, Wouldn't these types of stats be ideal for judgement on players? Do these exists yet? I'd like to know your perspective on these or any stats available that can help with this.
I wonder which player's advanced stats number go way up in playoffs as opposed to the regular season.
Leaf fans are the best. Regardless of what we think of Bozak, our top line was one of the best in the league. But we get all bent out of shape because it could/should be better. Meanwhile our bottom 6 was beyond brutal, and our D was pretty much an embarrassment night in and night out. Yet we continue to gripe about Bozak. There are so many places on the roster that stress me out more than JVR-Bozak-Kessel. (Note: this was not a shot at the author or the article, which was great. Just a general observation.)
or better yet, does he make the 3rd line players numbers go up when he's with them? ie. does it illustrate the fact that he may just be the ideal 3rd line center and if we had a true 1C that enhance phil is that what gets phil to be a 40 - 50 goal scorer instead of a 30 - 40 goal scorer....
@Sic_Transit_Gloria Yes, to really evaluate Bozak you would need to do a WOWY analysis for everyone on the team kind of like the McClement analysis that I linked to in the article. I may do this for a future post that focuses more on Bozak specifically instead of Bozak and Kessel as a tandem. I think the resulting conclusions will be more or less the same. Bozak had a better season this past season than any of his previous ones.
@Hmmm If we are supposed to exclude power plays, here's Kessel's stats.
November: 3GF (2PPG) and 6GA => GF%=14.3%
December: 3GF (1PPG) + 7A (3PPA) and 15GA => GF%=28.6%
Can someone else check the numbers? I don't want to prematurely conclude that Kessel's offensive production took a dive. If true, no wonder Kessel was glad to see Bozak back.
I mean Kessel may have good CF% but we're not sure if his lower GF% is because the goalies are stopping him or if he's being more of a passer without Bozak. Most of the time, I'd prefer Kessel do the shooting.
@DCDC Kessel with Bozak = Bozak with Kessel as it is a team stat, I believe
@StanSmith If Kessel's stats are the same with or without Bozak, doesn't that dispel the notion that Bozak has some sort of special, irreplaceable chemistry with him? Because that's a pretty common thing I keep hearing.
I'd like to see a Kessel+Bozak WOWY for GA%. Does Bozak really help that line defensively?
@Pent_House That's is the part that's coming. The massive amounts of info that will come from these new camera systems will help get to that point
@LeafsGuru he did have a 'not small' injury ..Holland did mediocre after Kadri stunk
@Yaknowwhat All data are 5v5 data so no PP or PK stats included.
Sample sizes for Bozak away from Kessel are smallish, the last 2 seasons in particular. ~90 minutes last year. ~60 minutes in 2012-13. ~160 minutes in 2011-12. ~270 minutes in 2010-11.
And yes, there are a lot of different variables to consider. Analytics at the player level is still, and maybe always will be, very difficult. The above post is just the tip of the ice berg.
@Xxxxxnew Reimer !
@Walter_White most likely off topic comments
I'd take Bozak over most anyone on that list with the exceptions of JVR, Oshie and Voracek.
@Xxxxxnew JVR is one of the biggest steals in the league
@Xxxxxnew Oshie is a bit of a steal
@Xxxxxnew I'd be curious to see the same list, but filtered to only show centers. That might give us a clearer picture.
@Walter_White amen Walter....I like Bozak...the guy is a very reliable hockey player and excellent on breakaways and shootouts.
The 2nd line was nothing to write home about either.
@Walter_White hard to imagine going into battle in the playoffs knowing your number 1 center is wearing a Justin Bieber T Shirt underneath.
@Bring the Cup Home @Sic_Transit_Gloria regaedless of ability, he is not an ideal third line centre with his 4+ M contract, unless he is a two centre playing in a third spot due to depth.
@loomx @StanSmith It seems Kessel's chemistry with Bozak exists purely on a personal or comfort level. As the article indicates, the sample size for times when they play apart is small and in those moments Bozak still plays for the same team as Kessel and they still have each other off the ice.
Given that hockey is a largely mental game and Kessel to all appearances is not a socially confident individual it is interesting to hypothesize as to how he would react both on and off the ice to Bozak not being available at all (as in traded to another team).
@Greg Fenton We missed the playoffs because no other line could pick up the slack when our top line didn't produce (which is to be expected over an 82 game season).
@Xxxxxnew I get the on-topic part, but joking around isn't allowed either? Good to know I guess...
That's true, but generally speaking, if you are a part of the leafs organization you should have seen way more than enough to answer these types of questions.
I think the real value of analytics is in making smarter trades and off season signings. Using data to assess players you don't get the opportunity to watch play every couple of nights. Then use scouting & game tape to answer any questions and/or gaps that the data presents.
@Greg Fenton I don't think anyone would expect another line to produce as much as them, but our secondary scoring was so bad last season, particularly from the bottom 6. We can't expect the top line to carry us for 82 games. They carried our offense all season and then sputtered at the end and nobody could pick up the slack. Good teams don't have that problem.