Somewhere, sometime (possibly when Oilers fans turned on Cory Cross or maybe Marc-Andre Bergeron, although it would make more sense if it was a forward), Vic Ferrari made a comment that stuck with me to the effect that teams don’t win and lose games because of their bottom six forwards or bottom pairing defencemen. In essence, he was saying that people are foolish to get too worked up about perceived deficiencies in that area, because the games are won and lost elsewhere.
It’s something that made a lot of sense to me intuitively and sort of stuck with me. After I wrote my post a few days ago looking at how teams did in the various game states relative to average, I decided to try and take a swing at breaking it down into top six and bottom six forwards. The advantage of doing so should be obvious: it allows you to get a better understanding of what “good” means for top and bottom sixers, as well as understand whether teams that look similar in terms of total results actually have significant differences in terms of how they achieve those results.
I’m starting out with a pretty narrow focus: I’m just looking at 5v5 results. In order to do this, I’ve defined goals as being either “Top Six” goals or “Bottom Six” goals. I defined a Top Six guy as one who finished in the top six on his team in ES ice time in at least 60% of the games at which he played. I think that this produces a pretty respectable list of guys, which can be found here.
If there were two or more Top Six forwards from the same team on the ice, I called a goal a Top Six goal for that team. Otherwise, it’s a Bottom Six goal. As should be obvious, a team can score a goal that’s a Top Six goal for it and a Bottom Six goal for the opposition. Similarly, a Top Six player can score a goal that gets credited as a Bottom Six goal - if, for example, Hemsky scored a goal while out with Stortini and Reddox, I count it as a Bottom Six goal. Here’s what the results look like:
Some brief explanation is probably required for the final two columns. They represent the top and bottom six’s ratios of GF/GA, respectively, divided by the respective league averages for top and bottom six forwards. The idea there is to generate a number where 1 is average, better than 1 is above average and worse than 1 means that the group was below average.
I think that this passes the initial sniff test. With one spectacular exception, every team’s top six forwards outperformed their bottom six forwards. If you think that coaches try to put the players on the ice who give them the best chance of winning the game the most often, this would seem to suggest that they’re pretty good at picking out those players. I’ve got some comments on this, which I’ll just proceed to get into.
*Whether coincidence or not, the top 13 Top Sixes made the playoffs. Seven of the top eight made it to the second round. It might be worth doing the same thing for previous seasons to see whether there’s anything to this effect.
*Obviously goaltending matters and is influencing where the various groups of players fall. Boston’s forwards probably look relative to a lot of teams by reason of the spectacular save percentage posted by the Bruins. With that in mind (and as another smell test), this seems to do a pretty good job of sussing out the groups of forwards who I’d consider to be the best in the NHL.
*For all the hype about the Red Wings…I’m kind of surprised that their bottom six did so much worse than league average. I don’t have the shots data broken down in this fashion but, according to timeonice.com, all of the Wings players who played any amount of time were in the black shotswise.
*Spezza/Heatley/Alfredsson did not post their typical numbers this year. I broke down the Sens Top Six number into situations in which two of the three were on the ice. The Sens were EV+ 53 and EV- 51 in that situation, leaving EV+ 18 and EV- 21 for top six situations in which two of those three players weren’t on the ice. Looking at the +/- that those players have put up in the past, they were difference makers for Ottawa. Last year, they weren’t.
Taking that a step further, I went and took a look at the timeonice EV stats. It looks to me like there were a couple of things going on. Those guys had a weird year save percentage wise - Heatley and Spezza were well below the team save percentage (.904 and .900 respectively, on a .918 team). If you buy the argument that players don’t have a lot of control over that, that strikes me as some bad luck on their part. Their shooting percentages look a hair low in comparison to their established norms as well, which will hurt. I would expect a bounceback year for that line next year.
*I don’t know what to make of St. Louis. Worst Top Six performance and best Bottom Six performance? Bizarre. That might go a long way towards explaining why they were able to surprise everyone - their performance is just completely out of step with what you would generally expect.
Looking at the Blues skaters who I identified as bottom sixers, here are some additional numbers:
Lots of great PDO’s there, and numbers that I would think are unlikely to repeat. Bergulund, Perron and Oshie look like a pretty fantastic collection of young forwards - they’re the three main culprits - but they’re probably not going to be posting .940+ ES save percentages going forward.
*Outside of 2003-04, I have a hard time imagining Tampa’s top six ever doing well in this, which is strange, given the money that the Lightning devote to those players.
*Hard to overstate the degree to which the Rangers are a disaster. They’ve got Henrik Lundqvist propping up everyone’s numbers and, still, both their top and bottom sixes post bad numbers. Kevin Lowe might have actually had a pretty good summer of 2007 in that he didn’t get a lot of the guys who he was targetting.
*Finally, the Oilers. Averageish top six, averageish bottom six. Close to being a cap team. Not a lot of contracts that are likely to be outperformed in the immediate future. Guys like Gagner and Cogliano a year from needing new contracts. The Oilers are going to be a pretty significant managerial challenge because it’s going to be difficult to improve the team without, IMO, making a series of moves. Salary room is going to have to be created through the identification of guys who are disproportionately not worth their tickets. More efficient players will have to be found. This isn’t necessarily easy to do, because a series of moves are going to be necessary to do pretty much anything. Like I said when Dustin Penner was signed, they bought something closer to mediocrity sooner, at the cost of limiting the upside of the team.
We’ll see what Tambellini does but I’m not optimistic that the Oilers can be seriously improved over the course of a single summer.
Was shooting the shit with some friends earlier..
Think the Oilers, with Tambs at the helm, would target the Sedin’s? Now, I’m not saying I’d advocate this, but given that we know a shakeup is coming, and that given the list of coaches who have Vancouver backgrounds, Tambs is sticking to people he knows….
Lets assume it would cost $13,000,000 against the cap.
The Oilers package Gilbert and the 10th together to move up. How high is irrelevant, but that’s where his cap hit goes.
Moreau, Penner and Staios are dealt in salary dumps.
Nilsson is dealt for Halak.
Sedin1 - Sedin2 - Gagner (14.65)
Cogliano - Horcoff - Hemsky (10.7)
O’Sullivan - Malhotra - Pisani (6.425)
Pouliot - Brodziak - Stortini (1.8)
= 33.575
Grebeshkov (3?) - Vinvosky (8.6)
Smid (1.5?) - Souray (6.9)
2 cheap vet D + Peckham (3)
= 18.5
Halak
JDD/Journey Man Goalie
= 2
For a grand total of roughly $53,500,000.
I’m sure there’s a million holes in there that someone will drive through with a mac truck, and we’d be five kinds of fucked next season when Gagner and Cogliano look for raises and the cap goes down.
Just figured I’d float the idea out there, discussion point more than anything.
Very interesting take on things; and to think, the story we always hear about the Predators is that they’re stacked with Bottom Six guys, it’s the Top Six that’s lacking. Those Bottom Six forwards may be hard-working troopers, but they simply can’t put the puck in the net, which is kind of important.
As you seem to have the numbers to hand, Tyler, how did theses top-sixers do against the bottom-sixers in the league?
Forgive my navel-gazing here. I just need to reconcile the MTL numbers with my cocksuredness that the problems up front are mostly with the top 6-ish forwards.
Your post (which is fascinating) prompted me to look at how forwards’ EV icetime is distributed at behindthenet. Montreal forwards have had the most egalitarian distribution of 5 on 5 icetime in the league two years running. That must be why Latendresse and Lapierre make the grade as top 6 according to your method. The TOI numbers say they played less than an EV shift per game more than Lang and Koivu.
In any sort of comparison, I really think Lang and Koivu belong in the top 6 over Lapierre and Latendresse because of the quality of minutes served. To me, Lapierre and Latendresse play against bottom six forwards, therefore they are bottom six forwards. Those two also had good/lucky seasons in terms of EV GF/GA. I’m pretty sure that if this swap were made MTL’s top 6 ratio would drop to a fair bit below-average (that’s not a request, btw).
I guess I’m with Jeff, in that I’m not very compelled by Step 1 of the methodology (identifying top sixers). My two main quibbles would be:
1. It’s not close enough to what most of us would intuitively define as a Top 6 forward, which is ~ “the guys who are expected to make a difference at the offensive end”. The inclusion of McClement and Crombeen for STL is a good example, I think: they don’t do PP, and were used a lot in the Pahlsson role, so they end up ahead of Berglund et al. But no one is seriously looking at their roster composition and saying, “the Blues top 6 forwards were nowhere near good enough”, because no one would describe McClement and Crombeen as top sixers.
2. There’s not actually 6 forwards for each team (notwithstanding the impact of trades). Not sure how you can use the numbers of 5 Penguins players as representing forwards 1-6. Disregarding the Lombo/Olli swap, Conroy is forward #7. Why would his contributions go towards representing the Flames’ top 6 forwards.
I wish I had a good suggestion for doing it better, but I don’t think I do.
It’s not close enough to what most of us would intuitively define as a Top 6 forward, which is ~ “the guys who are expected to make a difference at the offensive end”.
Although it has the limits of any objectively based determination, I think that, for most teams, it ends up being pretty close. I take your point Matt, to a certain extent, but to a certain extent, it’s our understanding of what constitutes a Top 6 forward that might need to be revisited. Crombeen was only a Blue for 11 games, so I won’t get too worked up about him but McClement was a top sixer in the sense that he played 54 games there out of 79. At a certain point, it’s our understanding of what constitutes a top 6 forward that we need to change.
Not sure how you can use the numbers of 5 Penguins players as representing forwards 1-6. Disregarding the Lombo/Olli swap, Conroy is forward #7. Why would his contributions go towards representing the Flames’ top 6 forwards.
Keep in mind, I’m identifying players who teams treated as top 6 players. So, in the case of the Pens, so long as two of them are on the ice, it’s a top 6 goal. It’s not their numbers that define things, it’s how many of them were on the ice when a goal is scored.
In Pittsburgh’s case, they probably ran a ton of guys through opportunities to play top six minutes. If nobody did it enough to stick, I’d argue that the defining characteristic of a PIT top six goal is two of those five guys on the ice, rather than arbitrarily including someone who didn’t play enough games in the top six.
Similarly, with Calgary, I assume Conroy shows up because Keenan shuffled his lines around enough and the Flames had some injuries. The effect of including seventh forward is diluted by the fact that you still need another “top sixer” on the ice with him.
In any event, this type of analysis is something that I imagine any team using would then sit down and pick away at. You’d want to understand any wrinkles that might be driving your numbers one way or another, like Jeff and you have pointed out. As a starting point though, a line in the sand, I think that it’s awfully reasonable.
kudos for the fantastic idea and execution.
i too have thought about this, but my intuition told me the opposite of what Vic said. (my intuition is likely wrong.) i always thought that a lot of teams could improve a lot by replacing their slightly-above-replacement-level players with average players. it’s also easier and cheaper than replacing your average players with elite players.
your methodology here does seem damn reasonable to me in a lot of respects. my main issue is the heavy goaltending/randomness effect. take a look at TOR. their top 6 scored friggin’ 114 goals, but they get stuck with a brutal .93 ratio due to letting in an abysmal 122 goals against. was that really their fault? (and predictive of their future results?) and on the other end look at BOS. their top 6 scores 20 less goals than TOR but gets a sick 1.55 ratio.
don’t wanna beat a dead horse here - i know you’ve already thought about this and even mentioned it. but is there any way you could run this chart again using SDF/SDA (Shots Directed For divided by Shots Directed Against) instead of GF/GA?
You know how Barry Melrose always says such lovely things like, “the right guys are showing up on the score sheet,” and “your best players have to be your best players.” I always thought Vic was just Melrose with a TI-82.
Well, it didn’t make a lot of sense to be blaming the Oilers failings on Bergeron when he was playing maybe 15 minutes a night and was better bottom pairing option than most teams had.
Scotty Bowman was the guy who popularized the ‘your best players have to be your best players’ line, by my memory. Though I imagine you can find the same quote, or something strikingly similar, from every coach in the league.
On the Oilers numbers, they had some subpar on-ice percentage years from their better players, and some good years with percentages from Stortini and MacIntyre. But by the shots proxies, by the scoring chances, and by eye … the depth lines got badly outplayed most nights.
I think that the Oilers will be a better team next year just through maturity. Not a tonne better, but better. Gagner, Cogliano, Gilbert and Grebeshkov should all be better players on the whole. Hopefully Penner and Nilsson too, though they’re getting to the age where you pretty much know what you have.
On the market as it is now, my sense is the same as Sunny. You can get some good NHL players for your roster for under a million dollars now. I mean if average joes like Hilbert, Dvorak and Viellieux had played the minutes of Nilsson, Gagner and Cogliano over the past two seasons, I have no doubt that the Oilers would have been a better team. I guess a lot of it comes down to how much you believe in the value of developing players at the NHL level.
And considering what they bring to the table, the value of the really high end guys, players that don’t only put up counting numbers, but really help you win (Zetterberg, Lidstrom, Iginla, Nash, Havlat, Chara, etc) … it’s not possible to pay them too much, the cap doesn’t allow it. And with the way that the price of offensive defencemen has gone recently … forwards who can play the point on the PP effectively look like good values right now too.
As you imply, a tough situation for Tambellini right now given that he has a whack of players getting paid 500k or a million too much. It would be easier to fix if Lowe had negotiated better with the veterans and then just made one enormous contract mistake, like signing Nylander.
Vic:
Imagine if Lowe had done the exact same thing, AND signed Nylander to that albatross
I’m finding those Wings numbers HIGHLY interesting.
Looking at your spreadsheet, your Top 6 guys are Zetterberg, Franzen, Datsyuk, Hossa, Filppula, Cleary. But of those, Fil scored almost a quarter of his goals on “bottom 6″ lines, and Cleary scored a third of his on “bottom 6″ lines. The Wings didn’t really roll two top lines of those 6 guys. In fact, Cleary and Fil did spend significant amounts of time on third lines. Holmstrom only played 53 games this year, but he primarily plays on a top-2 line, and Jiri Hudler spent a decent amount of time on the top two lines as well. I don’t know that I necessarily agree with your methodology for picking out the Wings top two lines here.
Holmstrom only played 53 games this year, but he primarily plays on a top-2 line
Might want to tell that to Babcock. Holmstrom is not a particularly close call. I’m missing two DET games but I have Homer at 5/15/26/1 in terms of 1/2/3/4 line ES minutes.
The Wings didn’t really roll two top lines of those 6 guys. In fact, Cleary and Fil did spend significant amounts of time on third lines.
The Wings pretty clearly have a top 4 forwards and then it’s a bit murky. Val and Cleary are pretty obviously the next two most used forwards though. I don’t know how they can blow everyone else away without playing at least a big chunk of time in a top two role.
I guess something that should come from this is the simple fact that most teams have a lot of overlap. Oilers fans loved to bitch and moan about the MacT blender. Pretty much every coach does it. Detroit is actually pretty clean.
Re: the top 6 problem…
Maybe this is the best way to proceed, and MTL and STL are the outliers. Much was mad in the local media about Carbonneau’s insisting that Koivu was signicantly less effective when he had to pile more than 10-12 minutes of ES time in any given game. That would be why Carbo so insisted on rolling four lines; that also may (I say *may*) be why getting Schneider with Lang’s money wasn’t the best idea, but hey, water under the bridge…
If another way of zoning in on the top 6 was to be devised, I would suggest something like a Crosstab of the top-9 ES TOI with % of PP TOI. That would weed out Latendresse and Lapierre and their ilk all right.
This is great stuff Tyler. I think the methodology is fine and agree with your earlier point that we might need to change how we see what are top two lines. Often, that’s judged by who’s getting time on the PP and this exercise is really about what’s happening at EV. A lot of the time the same guys get the EV and PP minutes, but not always (obviously).
As for the Oilers, I think any big improvement will need to come on special teams. The EV numbers should improve a bit with guys maturing. If they can maintain their EV performance from this past year and improve their special teams to league average then they’ve made the playoffs and that’s at least a step in the right direction. Moving toward an elite team, however, will be mighty difficult unless some of these kids improve at a rate much higher than I expect (the year after they sign extensions).
Interesting stuff.
In terms of the Blues, Murray was running McClement against the other teams best and he was just getting killed, both in terms of the counting numbers and otherwise (corsi).
But I think that you’re right in that most of it can be explained by anomalous percentages.
I do find this very interesting, but I think the methodology of player selection could use some fine tuning.
In respect to the Leafs, the fact that Alexei Ponikarovsky doesn’t make the list is a tad bizarre. The guy finished 2nd on the team in total Ice Time, but 7th on the club in ES Ice Time Per game.
I would think PP time might be a better way to figure out who the “top 6″ forwards from a production perspective are on the club… or else just do a total TOI determination rather than just ES.
Ponikarovsky’s numbers dipped as a result of the addition of Stempniak mid-season, and frankly as time wore on, Ponikarovsky got more responsibility, not less.
I find it hard to believe that Ponikarovsky wasn’t top 6 in ES ice time in 49 games this season… but in essence you’re punishing him for playing a full 82 games, while Hanson only played 5 and he makes the list.
I realize that in the end this won’t have a huge impact because a lot of his goals came with Antropov and/or Stajan on the ice, but all the production at the end of the year with Grabovski and Kulemin will go unnoticed as he and Kulemin didn’t make the cut, though I would definitely consider them to be top 6 forwards.
Why not do it on a game by game basis, and define goals scored when the top 6 players in ES ice time within a given game are on the ice vs. everyone else?
I’m curious if there’s any info available which created the thought that save% fluctuations for individual players is bad luck?
Intuitively I’d think a player could affect sv%. For example ovechkin vs Pittsburgh. He directed a lot of shots on net but he created a lot of dangerous turnovers. All scoring chances aren’t equal.
Well no, Steve, I don’t think so. For one, I’m talking about top six ES forwards. Not some other kind of top six forwards. The PP time doesn’t matter to my definition. I’m not sure why I should define it differently to capture something that isn’t the point of the study? Guys who are top six forwards in ice time because their significant PP time makes up for their insufficient ES time aren’t top six ES forwards.
I find it hard to believe that Ponikarovsky wasn’t top 6 in ES ice time in 49 games this season… but in essence you’re punishing him for playing a full 82 games, while Hanson only played 5 and he makes the list.
48 actually. Line has to be drawn somewhere though. He’s on the wrong side this year. As you note, that’s attenuated because Poni played a lot iwth Antro and Stajan.
all the production at the end of the year with Grabovski and Kulemin will go unnoticed as he and Kulemin didn’t make the cut, though I would definitely consider them to be top 6 forwards.
Kulemin wasn’t even close. 27 top six games, 45 bottom six games.
Why not do it on a game by game basis, and define goals scored when the top 6 players in ES ice time within a given game are on the ice vs. everyone else?
Because I don’t think that the single game basis is how things really work. Some games, with special teams or whatever, the “true” first line will get third line minutes. Failing to treat them as the first line for that game, in my mind, misses the point.
danny: http://www.mc79hockey.com/?p=3003
I realize Kulemin’s numbers were low for the early portion of the season, but I’m basing my opinion on his play over the final 3rd of the year.
I think some of the issue I take with this is that player movement makes a fairly major impact, as do injuries and the like.
If you have a player brought in via trade, who suddenly eats into the minutes of what used to be a top 6 guy, then he’s suddenly the 7th guy, and he’s somehow less valuable or meaningful to the team?
Also if a player is injured and plays fewer minutes because of said injury on even strength, but plays more minutes of low impact play on the PP, I would have to say I don’t think he’s not a top 6 player… he just doesn’t play top 6 ES minutes.
I understand this is your definition for the sake of your study, and you can set it up however you like, I just don’t think I see much logic in setting the bar at 60% of the games played rather than 50%… or 70%? I guess you’re just ballparking it to get a reasonable number of players, which is fine.
I also fail to see how your 60% of games played justification counters my point of using a game by game basis… aside from using it as a definition of who the top 6 ES guys are on a given team.
In a given game, the top 6 ES players in ice time are the top 6 players FOR that individual game, as selected by the coach on any given night (which you mentioned as a verification of your methodology in the original posting). The idea that “true” first line players won’t get the lions-share of ES minutes in games with a lot of PP and PK time doesn’t really wash with me. Usually extra PP and PK time eats into the minutes of the guys at the end of the bench, not top 6 minute guys at ES.
I mean if what you’re getting at is top 6 goals for/top 6 goals against vs. bottom 6 goals for/bottom 6 goals against I don’t know why an individual game determination would be any less meaningful, over the time frame of an entire season.
To me it just seems less arbitrary than the 60% determination… but that’s just my 2 cents. Take from it what you will… or won’t.
Решил добавить RSS и получать новости, мне лично понравилось, что написал автор
Статья интересная, добавлю в избранное для дальнейшего ознакомления
Все правильно говорите, мне тоже все понравилось сдесь. Бложек действительно на славу сделан.
Достаточно интересная статья. Думаю стоит добавить в избранное для дальнейшего изучения
Для более подробного и внимательного изучения добавил в избранное. Буду изучать
Автор, можно с вами познакомиться?
Не увидел нотки благодарности за этот материал в комментариях. Скажу тогда сам:
автору огромная благодарность
Мне понравился сайт. Много интересных и адекватных мнений. Благодарю автора
неплохой ресурс подсказали, очень доволен. спс.
По моему у Вас украли эту статью и поместили на другом сайте. Я её уже видела.
Тоже верно, хотя многие бы не согласились.
Я тоже воспользуюсь.
Не вижу смысла в этой писанине.