Examining GVT per Million: Has Brian Burke Signed Efficient Player Contracts?
by Nirbo on Sep 1, 2010 7:03 PM EDT
Email
PrintEditor's Note: Here's a good first FanPost for anyone wondering what they should be striving for. Nirbo analyses Brian Burke's roster using GVT per million and gives us an idea of what we might be able to expect in a best-case scenario.
This is a pretty simple exercise for my first FanPost. Been a lurker and occasional commenter for a while, and I hope this adds at least a limited amount to the preseason discussion.
While I love many of the names and attached abilities that he has brought in, I haven't been sold on Burke's cap-management strategy. It seems he has overpaid on a significant chunk of contracts, particularly in free agency. I wanted to see if the stats backed this up.
I'm going to use GVT vs. $M to assess players whose current contracts are Burke's responsibility: players signed or traded for under his watch. I will exclude Kadri, Mueller and others who don't have historical NHL data to go on, though I'm sure we can have fun with eGVT at some point.
Here is Tom Awad's original series on how GVT is calculated for those who are interested and haven't seen it already.
Before I start, just a couple of notes and caveats.
-
GVT assumes an average level of competition. It also normalizes against the league average rather than team average for most of its calculations, so it (understandably) penalizes players on bad teams.
-
Because it was designed for comparison between eras, GVT relies on stats that have been kept for a long time. This means that offense (goals and assists) is probably better captured than defense (shots allowed and plus-minus).
-
A player is effectively penalized for getting more ice time. 20 goals are more valuable if they come in 15 minutes a game than 20.
Given these factors, we can expect a player such as Francois Beauchemin, +.083 in qualcomp and 10th in the league in ice time on the 29th-ranked team, to suffer under this measure.
I'm going to come right out and say that given the number of past injuries and short career histories in this group, there is an issue with taking a look at raw GVT, but that's what I'm going to do for now. I've included three calculations of GVT: last year's GVT, an average of the last 3 years, and the best year of the last 3 to represent the "hope" scenario, which I don't think is unreasonable to examine in light of the expected improvement on last year's horror show. I don't care to look at the doomsday scenario.
Here's a quick chart, then, of the players Burke has signed and acquired to play for the 2010-2011 Maple Leafs.
| Player | Cap ($M) | GVT | GVT/$M | ||||||
| 09-10 | 3Y best | 3Y avg | 09-10 | 3Y best | 3Y avg | ||||
| D | |||||||||
| Beauchemin | $3.80 | 3.8 | 3.8 | 2.7 | 1.0 | 1.0 | 0.7 | ||
| Gunnarsson | $0.80 | 5.4 | 5.4 | 6.8 | 6.8 | ||||
| Lebda | $1.45 | 1 | 5.8 | 3.5 | 0.7 | 4.0 | 2.4 | ||
| Komisarek | $4.50 | -0.8 | 4.1 | 2 | -0.2 | 0.9 | 0.4 | ||
| Phaneuf | $6.50 | 8.4 | 17.8 | 11.6 | 1.3 | 2.7 | 1.8 | ||
| F | |||||||||
| Armstrong | $3.00 | 4.9 | 7.7 | 5.93 | 1.6 | 2.6 | 2.0 | ||
| Bozak | $3.73 | 3.9 | 3.9 | 1.0 | 1.0 | ||||
| Brown | $0.54 | -1 | -0.7 | -1.1 | -1.9 | -1.3 | -2.0 | ||
| Grabovski | $2.90 | 5.7 | 7.9 | 4.67 | 2.0 | 2.7 | 1.6 | ||
| Hanson | $0.65 | -0.5 | 0.3 | -0.8 | 0.5 | ||||
| Kessel | $5.40 | 11.7 | 15.1 | 10.3 | 2.2 | 2.8 | 1.9 | ||
| Kulemin | $2.35 | 7.8 | 7.8 | 3.3 | 3.3 | ||||
| MacArthur | $1.10 | 1.8 | 4.2 | 2.4 | 1.6 | 3.8 | 2.2 | ||
| Mitchell | $0.73 | 2.9 | 3.1 | 4.0 | 4.3 | ||||
| Sjostrom | $0.75 | 0.2 | 3.8 | 1.5 | 0.3 | 5.1 | 2.0 | ||
| Versteeg | $3.08 | 10.1 | 12.9 | 3.3 | 4.2 | ||||
| G | |||||||||
| Giguere | $6.00 | -2.2 | 26.4 | 7.3 | -0.4 | 4.4 | 1.2 | ||
| Gustavsson | $1.35 | -4.4 | -4.4 | -3.3 | -3.3 | ||||
I've read that league average GVT per team is by definition 120, so I'll use a criterion of 120/$59.4 = 2.02 GVT/million, rounded to 2. Any GVT/$ above 2 has been highlighted in bold. This is a crude calculation but it's also a fast one, which suits my purposes.
Some quick notes, though I'm sure folk can draw their own conclusions from this data:
Last season was thoroughly inefficient all around for the players who Burke has assembled for '10-11. Only five out of eighteen have a GVT/$M greater than 2. Same goes for the last few years: among the players with three years of service to draw upon (I omitted the others from the 3-year average column), only three of eleven hit the threshold.
What immediately stands out is that whatever his strengths, Burke does not seem to have the measure of how much to pay in free agency: none of Francois Beauchemin, Mike Komisarek, Brett Lebda, Colby Armstrong or Clarke MacArthur hit the efficiency threshold in 09-10.
Burke does seem to love reclamation projects and potential, however. When we consider only the best year in the last three for each of these players, the ratio jumps to twelve out of eighteen: nothing to crow about, but representative of exciting possibility nonetheless. Even the reviled Colby Armstrong contract could be a relative bargain if he returns to his 08-09 level of productivity, and Beauchemin and Komisarek bring qualities which aren't best measured by GVT. Phaneuf really will have to return to the player he was in his first three seasons to justify $6.5 million, however.
Phil Kessel, Kris Versteeg and Nikolai Kulemin look like a strong core to build the forward group from, and Bozak would certainly be added if we adjusted for games played or bonuses achieved rather than promised. These players are all young and signed to reasonable contracts given their abilities. Our core forward group has promise and is efficient enough.
Ultimately, the free-agent strategy which Burke has pursued seems like a misjudgment. However, the group his team has assembled is young and actually includes a surprising number of players who have put it all together in the past. With some pruning of the Fingers and stingier choices in the future, this could quickly become an efficient, and hence winning, roster.
Looking forward to any comments, especially any corrections or addenda as I'm a student of this stuff, not a master. Is GVT/$ a reasonable way to look at salary efficiency, particularly with GVS also available? Do any players jump out as having drastically different values than the past year, or past 3 years, indicate?
PensionPlanPuppets.com is a fan community that allows members to post their own thoughts and opinions on the Toronto Maple Leafs and hockey in general. These views and thoughts may not be shared by the editor of PensionPlanPuppets.com.
42 comments
|
3 recs |
Do you like this story?
Comments
Great first post
From reading about GVT and what you have assembled here, my big concern is lack of history.
It’s really difficult to gauge how cost efficient this team will be because the peaks and valleys of players’ careers tend to curve out over time and by having so many young and/or unproven players tends to limit the effectiveness of this measure.
When we examine the bias that GVT has against so-called ‘non productive’ players (ie defensive forwards, shut down D-men) it kind of throws the projection out of whack.
You really need some guys who are good defensively – a necessary evil as it relates to GVT.
There are two schools of thought on what players are worth.
First it is comparing the cost of the player vs what value they bring to the team in terms of statistics. After calculating ‘inflation’ (which is basically directly related to how the salary cap fluctuates) you try and cost compare what a player potentially brings to your team in comparison with similar players at a similar stages of their careers.
The second is economics: If you feel your team needs player type A to be successful, then what is the minimum amount of assets you need to spend to acquire said player.
Obviously these are overly simplified, but basically this it what it boils down to.
I tend to ascribe somewhat to both methods. When you notice a deficiency in your team you need to find a way to address it – the debate comes down to if you could have gotten that player for less assets or if you could have gotten a similar player for less cost.
Burke seems to have paid a premium for certain players (either through trade or cap space) but it simply comes down to this: if he didn’t we wouldn’t have Kessel, Komisarek(don’t all cheer just yet) Beauchemin and a few others. If you want to make significant changes, you need to pay the price to do so. It comes down to availability.
To be honest, for this (or any team) to be successful, they need to be greater than the sum of their parts.
Here’s hoping, and holy shit start the season already!
"We can't solve problems by using the same kind of thinking we used when we created them."
Albert Einstein
by Say *plan the parade one more time*... on Sep 1, 2010 8:47 PM EDT reply actions
This is exactly the kind of analysis I was hoping to see at some point (and do myself, if I had time). The lesson I took from Moneyball is that there are huge inefficiencies in the team management of professional sports teams, and that you can gain a huge advantage by properly analyzing the (right) numbers. I believe the same is true of hockey, except that the stats collection may not be mature enough to support this kind of analysis. However, we have to start somewhere and this post is a perfect place.
Here are the things that come to mind:
- Just how reliable is GVT as a predictor of future GVT?
- What is the effect of competition? You allude to it in your post but I’m not sure we have enough stats to do a more stringent analysis of it. If we sign a cheap guy with high GVT/$ and stick him on the first line, there’s obviously no way he’ll keep up his GVT, but it would be nice to predict approximately how much will his GVT fall.
- As a thought experiment (I’d prefer to do this with a real experiment but I don’t possess the NHL team with which to try it), consider a team of exactly 20 Grabovski-quality players (extend to G and D too) each signed to a $3M contract with a GVT of 6. The total GVT is 120, so this is by definition an average team; we would expect this team to be evenly competitive with any other team that can be assembled with 120 GVT. Really, this is a sanity check to make sure that the whole premise is valid. I guess my question from this point is – is GVT (and GVT/$) usable in isolation for a player, or does it depend heavily on the model and system (and team) on which the player recorded their stats?
Back to your question of whether GVT the number to use: my feeling is that it’s well short of capturing everything we need, but it’s the best we’ve got right now. Anyway, thanks for your post and I’m looking forward to where this discussion will lead.
Well, first I think a team of Grabovski’s would be hilarious – there would be a pineapple shortage and injured Habs everywhere.
I do see an issue with your thought experiment – most professional sports teams (and I think hockey more so than most) thrives best on variety. Over all, I think it is better to have guys that do some things very well and they are used in situations which play to this strength.
My belief is a team specialists, used properly, will out perform a team of clones with equal GVT.
"We can't solve problems by using the same kind of thinking we used when we created them."
Albert Einstein
by Say *plan the parade one more time*... on Sep 1, 2010 8:56 PM EDT up reply actions
I’m not talking about using Grabovski clones, I’m talking about using Grabovski-quality players. We still have a shutdown pairing and a checking line and so forth, they just aren’t the best at it.
In the end, I still agree with you; I don’t think this team would be able to beat an average team built more traditionally. A normal first line should be able to roll this team and their other lines should be able to hold. This leads me to believe that there’s a fair bit more work to be done before we can just plug in the GVT.
So yeah, basically I’m providing no additional value by stating “there’s more work to be done”. :(
Thank you both for the thoughtful replies, first off.
I think you are touching on an important point here: when folk speak about players who bring “intangibles” rather than goal-scoring to the table, they aren’t claiming that these players are more valuable than goal scorers.
The point instead is that a team requires a mix, both to defeat opponents’ exploitation of particular weaknesses and to make sure that you have five-man units that can work together – the passer, the shooter, the screener, the pincher and the responsible one is but one example of how this sometimes works, though it’s more complicated than NHL10 would have us believe.
It would be very difficult to put together stats which show what kind of ‘mix’ wins hockey games, given all the factors and noise involved. Inevitably three “identical” players on the same line would develop different roles and seem to be good at different things, so it’s hard to take a player out of their context and predict what they’d do in a hypothetical situation like that.
great first post. i’m a sucker for charts, numbers, etc.
i’m uneasy using GVT as a proxy to measure the ‘success’ of primarily shut-down type defensemen though. if you sort the GVT list at Behind the Net and isolate defensemen, the best GVT Dmen are all offensive types – the top 12 all have 40 pts or more, and 8 of them have 49 pts or more. komi and beauch can’t compete with those guys. even a guy like Volchenkov – who’s kind of a beast even though he was a stinky Sen – only had a GVT of 4.7. At a salary of $4.3M, he’s ‘inefficient’ based on your 2.02 definition. i don’t think that’s fair, or accurate.
further, if you look at a guy like mark fraser (who???) from New Jersey, he had a GVT of 2.2 at a salary of $500k for a GVT per mill of 4.4. duncan keith had the best GVT at 22.4… which divided by his new salary of $5.5M (cap hit) gives a GVT per mill of 4.07. Is Mark Fraser a better D than duncan keith? a better deal?
i’ve said this a few times here, but the idea of GVT per million, or points per million, is flawed as well. it inherently penalizes players that don’t score, and players with a high salary. a guy scoring 30 points at $1.5 million is scoring 20 pts per million. A player earning $6M has to score 120 points to be equivalent… if he ‘only’ scores 90 points, he’s less efficient. the reality is there’s a lot fewer players that can score 90 points that can score 30 points, so they get paid a premium. and 1 90 point player is more important to a team (generally) than 3 30 pt players.
anyways, i’m rambling. great post, and good discussion starter.
Yesterday is dead, but not my memory.
agreed
its an interesting idea for a metric, but GVT/$ can’t exactly capture the sliding scale.
I also dont think that guys sub-$1m should have a multiplier-effect to their GVT; a guy making $500k should not get a doubling factor on his GVT/$…
by Death_By_Leafs on Sep 1, 2010 9:32 PM EDT up reply actions
Thanks.
I agree completely with the idea that GVT measures scorers much better than it does shutdown types. When I look at the careers of a Michael Peca or Bobby Holik, I take askance at the idea that significantly more of their “value” comes from the area labelled “offense” rather than “defense”.
But I’m OK with that; no stat is going to incorporate all aspects of hockey into one, certainly not an individual stat in any case. I’d rather spend time understanding what, exactly, GVT does measure, rather than searching for the mythical acme stat.
What I love about GVT is the simple logic behind it: dividing up the goals that happen, and doing so using only stats that are readily available in NHL game logs, in this case going back to 1944 (other than ice times which are estimated where not available). It may not be the finest-boned stat, but it’s a huge step up from points and plus/minus.
To your point about Mark Fraser, I think we’re missing a fundamental assumption; that under this analysis, teams are assumed to spend to the cap. If the most efficient team comes in under cap, just replace someone with a higher absolute GVT (with lower GVT/$) to get a higher GVT team until you run out of cap space. The ultimate number that you’re trying to maximize is the team GVT (or team GVT/$ – it’s the same thing) subject to a list of players with given GVT and salary, and the team salary floor and cap.
I absolutely agree that the high-salary players are penalized from a GVT/$-efficiency point of view, but it’s an inefficiency that must be tolerated to eat up salary space to get to the salary cap. Mark Fraser is a better deal than Duncan Keith if you have other ways to get to the salary cap that are better than 4.07 GVT/$.
Great post
I do think you need to account for the fact that GVT is versus threshold. So if you just do GVT/$, you’re dividing a threshold figure by a raw one. I’d do GVT/(dollar above threshold), assigning a value to a threshold contract.
I did something like that here.
Hey, a good point. Of note, the post you’re linking to was one of a number of quality analyses on the site which I used as inspiration here.
I certainly considered using (Cap-Min Contract) for the value; however, as much as that might be the more “correct” way to do it, it inflates low-contract efficiency even further than already occurs. Christian Hanson, for example, gets ten times the value that Dion Phaneuf does out of their respective stats under this measure; subtracting the threshold, that number jumps to 40. A minimum contract would get infinite “efficiency”.
I know that GVS (linked in the fanpost) is available and is a more formal way of expressing what you suggest, but I wanted to confine this to stats already in the discussion; I’d rather really get to know the available stats and be versed in the limitations, rather than skim over a whole bunch of others.
It really wasn’t intended to be a Final Word sort of analysis, just something interesting to start discussion.
Mission accomplished, for sure.
I’m interested in figuring out Gabe Desjardins’ GVS better too.
You’re right about the “infinite GVT/threshold million” problem. A flaw in my approach for sure. But I think we need to compare apples ot apples somehow. Maybe the answer is in calculating your GVT/million threshold (currently 2), you should do 120 divided by 59.4 minus some hypothetical “threshold” salary (I used something but $17.5MM, but I’m happy to see a more rigourous than my “$750K per player sounds about right” spitballing. Then you’ll come up with a GVT/million threshold of 3, which sounds better. For example, Armstrong’s best GVT is 7.7, so 2.6 GVT/$1MM—just a little below the “efficient” threshold, which is what I’d expect for a UFA.
Anyway, very good food for though.
by The '67 Sound on Sep 2, 2010 8:54 AM EDT up reply actions
Tracking 3 season averages
can be misleading as hell, and GVT seems to me to be a lot about context.
GVT for offense is purely a measure of his assists and goals relative to his ice time, in comparison to a “threshold” player. On Defense it is a weighted plus/minus statistic that allocates proportion to D and F on the basis of their ice time proportion.
The defensive aspect is the one that I feel is misleading, and I think looking at a team OR player in isolation without the context of other tools can be amazingly misleading.
The idea that John Mitchell is a more valuable player than Fredrik Sjostrom is a bit ridiculous to me… particularly from a defensive perspective.
I’m not discounting GVT as a valuable statistic, but to reify it and adjust it to a cost basis just skews things even further from the point of the original stat in my mind.
What you’re getting out of GVT is the player’s value above or below a REPLACEMENT level player for the EXACT same situational ice time (from a +/- or point perspective). Unfortunately you’re then assigning an “average” cost? Players don’t all play in exactly the same situations, so reverting back to a mean value is completely unreasonable in this situation.
Guys making the league minimum are pretty much the exact type of 0.0 players that other players are being compared to using GVT, and as you move away from the 0.0 marker on a per dollar basis, the GVT and Salary scales do NOT slide remotely linearly. Guys who score the most make a crap load more than the guys that score the least. That’s reality… they’re competing for a fixed amount of money, and that’s how things work in that environment. To assume there’s a linear distribution or that their SHOULD be really isn’t logical.
I get what you’re trying to do with this, I just don’t think it works.
"Success is the ability to go from one failure to another with no loss of enthusiasm."
- Sir Winston Churchill
I'm pretty sure he's talking about the Leafs.
Hey Steve, thanks for the comments. Let me first say that this was really intended to be a quick-and-dirty exercise using a stat which is in common discussion. I think any NHL stat discussion comes with heavy caveats, and GVT gets unfairly beaten on. It is certainly not a subtle and accurate picture of “ability”; but the beauty of the metric is its adherence to the data available in 65 years of NHL scorecards. It has real value, and I think I’m justified in looking at the stat as a piece in the performance puzzle, and hence in terms of efficiency.
I agree that looking at any one year, or three-year averages, depends heavily on context. I included three different numbers partially to address this, though it’s still best to know the specific history with each player. To be honest, the Leafs aren’t the best team to examine using past history, with so many very young players as well as a few significant injuries.
Your comment on the abnormal distribution is fair; the only point I’d make is that addressing the nonlinearity of the distribution takes the calculations well out of the realm of simple math, and thus away from the understanding of the non-statistician. If it has use, GVT-per-salary is a calculation that can be done in a fraction of a second with a roster sheet in front of you.
I also have some problems with GVT, particularly what’s labelled “defense” – a combination of plus/minus, shots allowed, and goalie save percentage. Weird. Maybe these things are important, but I’m not sure if I’d say they sum to reflect a player’s defensive performance at all.
I’m not sure how these caveats invalidate the exercise, though. No-one is claiming GVT measures everything; we definitely still need some nuance in the operational definitions used, but I don’t agree that taking a snapshot of ratio with salary exacerbates this problem unduly. To use your example of Sjostrom vs. Mitchell, the reason Mitchell has a higher GVT/$ for 09-10 is because he had a higher GVT. It’s not that this means he had a “better” season, but he certainly had a higher GVT, and whether we’re measuring what we think we’re measuring is a separate issue. Your prose statement is a reasonable argument, but if we were using 09-10 GVT or GVT/$ as our only evidence, it wouldn’t be supported.
You say that I am reifying GVT, but I would disagree and say that in this kind of forum, it’s sometimes better to use common numbers and discuss the caveats rather than abstracting the numbers further and further from the source data. When there are problems, they are easier to parse. We can use GVT without lionizing it.
In any case, given your strong reservations, I’d welcome a follow-up using another metric to validate the (rather general and qualified) conclusions. I may take a look at GVS myself, though I think I’m going to simply accept the inaccuracy of the linear calculation and keep the right skew in mind when looking at this stuff.
I have come to a pretty solid conclusion after going through here. You sir, are much smarter than I. Or, you just know how to talk purty.
I wouldn’t say the caveats invalidate the exercise. but I think it’s misleading to evaluate free agent defencemen signings using a stat that’s a better measure of offense than it is defence, especially when the defencemen in question aren’t really the offensive type.
that being said, i don’t disagree that the Beauch and Komi signings weren’t very successful last year, but I still think both of them have an opportunity to earn their contracts, or at least come close.
Yesterday is dead, but not my memory.
GVT’s biggest blind spot is certainly in regards to the specialists. I’m pretty sure Marc-Andre Bergeron (GVT 8.1) wasn’t the Habs’ second-best D-man last year, nor was Volchenkov (4.7 GVT) only the Sens’ 4th best. But for a lot of players it’s a very good tool as long as we keep the limitations in mind.
At a bare minimum, it’s a helluva lot more useful as a predictive tool than the traditional boxscore stats.
by The '67 Sound on Sep 2, 2010 9:01 AM EDT up reply actions
Hmmm
At a bare minimum, it’s a helluva lot more useful as a predictive tool than the traditional boxscore stats.
But all it’s really using is boxscore stats – Ice Time, Plus/Minus, Goals and Assists, and then comparing them to expected values.
If you have some idea what the expected values are, then you can get the same info just by looking over their stats in general. I think what GVT does is it allows people who AREN’T adept at compiling all the relevant data (or the stuff beyond that) into one easy to see number. Unfortunately that number is flawed.
We know that +/- isn’t a solid predictor of offensive or defensive play… we’ve moved away from using those numbers in comparison of player quality over the past year or two towards things like Corsi and Rel Corsi. I haven’t taken the time to do it, but I would argue that using Rel Corsi as a predictor of “defensive” ability might be more useful than adjusted +/- statistics in the long run. It might also improve the numbers.
The specialists issue is also something that needs to be addressed, but I’m not particularly sure how it can be accomplished other than changing the ice time adjustments Awad makes for 5 on 4 or 4 on 5 situations.
"Success is the ability to go from one failure to another with no loss of enthusiasm."
- Sir Winston Churchill
I'm pretty sure he's talking about the Leafs.
by Steve Burtch on Sep 2, 2010 12:37 PM EDT up reply actions
Just to push the envelope a bit, what it sounds like you are suggesting is that one should be able to look at a chart of goals, assists, plus-minus, shots allowed, goalie save percentage when on the ice, PP ice time, ES ice time, and SH ice time, and keep in mind the yearly league norms for per-minute production in each of these categories as well as the ratio of goals to assists on the particular team. I don’t imagine this is actually what you’re proposing, but GVT is very complicated and saying that “looking over the stats” gets you the same result is disingenuous.
GVT may not be all-encompassing, but it’s in a niche where there simply isn’t much competition. The challenge I’d pose to those who wish to throw GVT out completely is to create a metric which improves on its accuracy to what we want it to measure while still maintaining its major advantages:
1) Using only data that has been available in NHL scoresheets for a long time
2) Expressing several common strengths in terms of a win- or goal-related number
3) Adjusting for the strength of the league and the value of opportunity in a sensible way.
In either case, I think more serious discussions along these lines might move to behindthenet.ca or similar – your main problem doesn’t appear to be with expressing GVT in terms of salary, but whether GVT has enough value in the first place.
Hmm
no it’s really both.
Salary doesn’t increase linearly in correlation to counting stats, but GVT seems to generally based on production using counting stats.
The fact that there isn’t a linear relationship means that to try to create one and use it as a determinant of “efficiency” is misleading.
High end production is extremely rare, and those players are paid for their rarity… not just their production per point. I don’t think you can logically compensate for high end production through exchanging one player for another… but I suppose arguments around whether a team full of semi-decent players trumps a team with one or two superstars surrounded by barely above average players will continue for a while.
I don’t know how to work around this, because I haven’t devoted the energy to it. I also think work needs to be done on the calculation of GVT, largely because of it’s defensive flaws. I know I’m not proposing a replacement but I think one could be formulated that made use of relative corsi values. I’d need to spend more time looking at it.
I’m also curious about the distribution of GVT in relation to the distribution of salary… if they line up to a large extent, then I think it’s got more value than I’m assigning, but if they don’t… there may be some flaws in comparison using both values.
"Success is the ability to go from one failure to another with no loss of enthusiasm."
- Sir Winston Churchill
I'm pretty sure he's talking about the Leafs.
Even Corsi fails when looking at specialists; although pure checking lines aren’t that common nowadays, a guy like Mike Richards or Sami Pahlsson probably gets undervalued even by Corsi.
by red army line on Sep 9, 2010 3:23 PM EDT up reply actions
We really do need to start playing some games soon. This is getting out of hand. Can some just highlight the part that says Grabovsky suck?
by Johnny Bower's Pokecheck on Sep 2, 2010 2:54 AM EDT via mobile up reply actions
Grabovski's
2.0 would (using the logic suggested by Nirbo) indicate that he is being paid exactly what he is worth. That doesn’t mean he sucks, it just means he isn’t outperforming his contract.
"Success is the ability to go from one failure to another with no loss of enthusiasm."
- Sir Winston Churchill
I'm pretty sure he's talking about the Leafs.
by Steve Burtch on Sep 2, 2010 12:38 PM EDT up reply actions
Don’t bother. He cares nothing for logic. Just gut.
Pension Plan Puppets: A Toronto Maple Leafs blog and a group therapy session.
Like reading thoughts confined to 140 characters? I'm on Twitter too.
by PPP on Sep 2, 2010 2:46 PM EDT up reply actions
By the way, I should point out that rather than adding a half-page of qualifications and reservations about GVT and about the crudity of the method, I assumed (rightly, it turns out) that I’d be better off leaving the bulk of criticisms to the multiple-paragraph comments which would inevitably follow. So I’m glad for that.
Well, we do have to start somewhere and if nothing else, you’ve started a discussion that at least a few of us are very interested in. The discussion is important in laying the groundwork and the implicit assumptions in what we’re trying to do.
For me the ultimate goal is this: list all the players in the league with their GVT (or a better number that we’re still seeking), a salary and a position. Set the salary floor and ceiling as constraints. Set 2G, 6D, 12F as constraints. Throw the whole thing into an integer programming solver (I think Excel has one) and see what comes out. Later on, we could refine the model by adding some role in addition to position (or expanding the positions); maybe a checking forward or a shutdown D and then expanding the constraints to include them (ideally, we wouldn’t need to expand the positions but instead properly credit these other roles in our GVT-esque number). It would be really interesting what an optimal solution would give us. The math geek in me would then put it into a Simplex or other non-integer solver and see if the optimum involves something like 2.5 Malkins and 0.76 Duncan Keiths.
Of course, this depends very much on how reliably we can use GVT. At this moment, it’s not the all-encompassing number that we need to be able to do this but it’s a start. It may be worth throwing it into Excel just to see what happens. Is there somewhere we can get a list of “all” the players in the league with some GVT value (perhaps normalized by minutes in some way?) and salary? I’ll give this a try if I have time and I can get my hands on the data.
Go to
Puck Prospectus and download the all-time GVT spread sheet… or just go to behindthenet.ca, and use the numbers from last season.
"Success is the ability to go from one failure to another with no loss of enthusiasm."
- Sir Winston Churchill
I'm pretty sure he's talking about the Leafs.
by Steve Burtch on Sep 2, 2010 12:41 PM EDT up reply actions
That spread sheet is all kinds of awesome.
"That’s why stats are so important – anecdotal evidence just doesn’t cut it when you’re talking about history." - Bower Power
by The '67 Sound on Sep 2, 2010 1:48 PM EDT up reply actions
The spreadsheet is terrific. I do wish there was a version that split years which were spent between teams, pretty difficult to look at just the Leafs over any period of time when the spreadsheet includes the stats from the transition year as a single line. I know I’m being picky.
That’s what I was planning to use, and I know I can find salary info at capgeek. I was just hoping someone had already combined the two since I think it’ll take me quite a while to combine it myself.
I’ll try a small proof-of-concept with a smaller data set and report back… someday.
just drop it into access and spit it out in 2 seconds.
by Death_By_Leafs on Sep 3, 2010 1:22 AM EDT up reply actions
I haven't used access in years
teaching doesn’t require me to use databases as much as engineering did.
My skills aren’t “up to date”… someone else can play with it by all means though.
"Success is the ability to go from one failure to another with no loss of enthusiasm."
- Sir Winston Churchill
I'm pretty sure he's talking about the Leafs.
An interesting note
on the subject of GVT’s defensive measure being problematic: looking at just 09-10, just 08-09, and the whole population of individual player-seasons, OGVT and DGVT are correlated at about r=.35.
Not huge, but it raises concerns. +/- should not be used in a number labelled “defense”.
agreed
although the correlation could stem from a similarity in “overall” ability. The best players may well be the best players overall at BOTH ends of the ice. i.e. Duncan Keith is good at both, Nik Lidstrom is good at both, etc.
But yeah generally speaking that doesn’t make a lot of sense.
I think Corsi makes more sense than +/-… but you’d have to come up with some sort of factor that takes into account Corsi Rel QoC, Corsi Rel QoT, and perhaps zone shift? blocked shots? hits? I dunno how we determine defensive value…. I think zone shift has some value.
"Success is the ability to go from one failure to another with no loss of enthusiasm."
- Sir Winston Churchill
I'm pretty sure he's talking about the Leafs.
I agree that a moderate correlation like that could easily be due to a number of factors other than conflation. Didn’t really examine further.
This thread is getting stale, but just to respond to what you’re proposing here, Corsi and GVT seem to be capturing different things. Because Corsi takes some interpretation, I think I could take or leave its inclusion in a portmanteau metric.
Corsi has a weaker correlation to NHL standings points than does +/-, for obvious reasons – meaning that it captures something specific, but not something that is necessarily synonymous with “success”.
Just as an example, if we compared Jason Blake and Phil Kessel last year using the Corsi numbers, we might assume that Blake was better at driving the flow of play. Corsi is most useful in the context of other numbers such as zonestarts, qualcomp and shooting percentage.
On the other hand, OGVT (and I’ll confine to that, because I don’t intend to defend DGVT) is a really neat way of combining goals, assists, and opportunity in a formula which compensates for league scoring.
Just different things. I don’t think it’s reasonable to use “success” as a definition for either, though, and that seems to be where we get into trouble with these new hockey stats, using them to say that one player is “better” than another as if we were discussing .OPS or WHIP in baseball.
Despite it's staleness
I’d like to continue the discussion…
so onward!
I see the point you’re making regarding success as a definition. I also see no reason to defend DGVT in it’s current format… so we’re in agreement on both of those. I think though that the distancing that’s going on with respect to the defensive aspects of the game is sort of unreasonable. The fact that we don’t have a “good” way to account for it doesn’t mean we should stop trying to do so in the best manner possible.
I feel that Corsi is a superior measure of puck possession and overall defensive ability to +/-, and I find the Jason Blake vs. Phil Kessel argument a tad ironic because frankly I would rate Jason Blake as a much better ALL AROUND player than Phil Kessel. Kessel is certainly more valuable than Blake as a result of the combination of things like age and potential future output. From a speed, and offensive perspective though, I’m not sure why one would be immediately considered superior to the other. Defensively I would argue that Blake IS superior to Kessel. Blake’s positioning, back checking, and desire to withstand contact in high traffic areas really isn’t in question is it?
The guy works his butt off on D, and he has far more takeaways than giveaways… unlike Kessel.
Blake’s shot totals dipped drastically last year, from 300+ to 239, which is likely a large part of why his goal totals and point production decreased. He had an 8.3% shooting percentage with the Ducks to close out the year, and an 8.1% shooting percentage in 2008-09 with the Leafs. Unfortunately the combination of his first year with the Leafs, and his start to the season in 2009-10 with the Leafs has left people with the impression that he’s useless offensively due to his 5.0% shooting percentage in those portions of his play.
Last season, Blake had 70 hits, 30 giveaways, and 62 takeaways. He fired the aforementioned 239 shots on goal, and missed the net 68 times. Kessel by way of comparison had only 13 hits, 46 giveaways, and 30 takeaways. He fired 297 shots on goal, but missed the net 102 times. If you include missed shots, Kessel had a shooting percentage of 7.52% and Blake had a shooting percentage of 5.21%… that’s a difference of 2.3 goals for every 100 shot attempts.
Considering they attempted between 300 and 400 shots, that mean’s we’re talking a difference of 7 goals on the first 300 shots, which appears to be a lot, but frankly isn’t a huge swing for an offensive player in an NHL season. It also amounts to approximately 1.3 wins (according to marginal goals per win models used by Alan Ryder at hockey analytics)… Kessel replacing Blake would be worth 2 more points in the standings offensively… but unfortunately his defensive lapses could just as easily be argued to be worth LESS wins (or a point reduction).
I’m not arguing that I prefer Blake to Kessel, but I’m saying in terms of their contributions last season, it really isn’t
"Success is the ability to go from one failure to another with no loss of enthusiasm."
- Sir Winston Churchill
I'm pretty sure he's talking about the Leafs.
by Steve Burtch on Sep 13, 2010 9:21 PM EDT up reply actions
woops
cut off my own posting…
it really isn’t as big a difference as it may at first appear. Jason Blake is a solid player, and I would argue that his Corsi numbers are just one of many ways that can be demonstrated. Phil Kessel is at this point in time more than a bit one dimensional. He does not have a well rounded all around game, and a team made up of a bunch of Phil Kessels would not be worth much on the ice, largely because they’d give up possession far too often, and they wouldn’t do a good job of getting the puck back in order to put it on net in the opposition end.
"Success is the ability to go from one failure to another with no loss of enthusiasm."
- Sir Winston Churchill
I'm pretty sure he's talking about the Leafs.
by Steve Burtch on Sep 13, 2010 9:23 PM EDT up reply actions
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Recommended FanPosts
Recent FanPosts
In Association With
Latest Puck Daddy - NHL from Yahoo! Sports
- Friday’s Three Stars: Blackhawks lose seventh straight; Wings win nineteenth straight at home
- Puck Previews: Blackhawks reeling; Sabres streaking; Holmstrom’s 1000th game
- Will Teemu Selanne remain with the Anaheim Ducks through the trade deadline?
- Nikolai Zherdev takes stick to crotch, elbow drop in KHL game (Video)
- Puck Headlines: Accused coach Pravilov dead; more Tim Thomas heat; Junior team goes nuts post-shootout
Latest stories and headlines from NHL.com
- Miller with a goal and two assists to lead the Warriors over the Hitmen 4-2
- NHL Capsules: Red Wings win 19th straight at home with 2-1 shootout victory
- Red Wings beat Ducks 2-1 in shootout for 19th straight home win
- Nathan Gerbe caps Buffalo Sabres' comeback in 3-2 shootout win over Dallas Stars
- QMJHL Roundup: A look at Friday's games
SBNation.com Recent Stories
StatCounter
Certain photos copyright © 2012 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
