The Puck Stops Here
Is Goals Versus Threshold Biased Toward Goalies?
by PuckStopsHere on 09/18/10 at 02:39 AM ET
Comments (0)
Yesterday, I wrote a post listing the top 20 players in the 2009/10 regular season by goals versus threshold. One of the most obvious conclusions upon looking at the list is that half of the top 20 players (ten of them) are goaltenders. Considering that teams dress 20 players per game and only two of them are goalies (and usually only one plays in any given game) this is a high number. Is this number reasonable?
The main reason this conclusion looks reasonable is that the top goaltenders lead the league in minutes played. They play 60 minutes a game and often play in 60, 70 or even more games in a season. They are clearly important to their team when they are on the ice. A goaltender can steal a game with a great performance or give it away with a poor performance. I had a good conversation about this in comments of the previous post with a commenter called Bossy Rules.
I think it is clear that in any individual game a goaltender is more likely to decide the game (either positively or negatively) than any other player on the ice. This doesn’t necessarily mean that the goaltender is the player who earns the most wins for his team. Even the best goalies in the league sometimes lose games for their team and the worst regularly played goalies sometimes steal games for their team. Is the net positive benefit of goalies so big that it is reasonable that half of the top 20 players in the league last year were goalies?
Further when we look at the goals versus threshold rankings, we see Ryan Miller led the league and was followed by Evgeni Nabokov and then Alexander Ovechkin. Does this mean that it would have been logical for Miller to have won the Hart Trophy last year and Nabokov and Ovechkin to have been the runners up? This is certainly a different conclusion from the one the Hart Trophy voters came up with. If we accept that goals versus threshold is a good model of reality, then we have to accept that a goaltender will be MVP in almost all NHL seasons. While goaltenders have won the Hart Trophy on occasion, it is a relatively rare occurence. Have the voters been wrong this consistently?
It certainly goes against my observations of the NHL last season. I picked Alexander Ovechkin as the should be Hart winner, with Sidney Crosby and Henrik Sedin as the runners up. These are the same three nominees as the hockey writers actually selected, but in a different order. Goals versus threshold picks them as 3rd, 4th and 9th respectively in the league. All of the other players ranked ahead of Sedin are goaltenders. Was Henrik Sedin really the ninth most valuable player last year? I would have picked Ryan Miller for fourth on my ballot and had Duncan Keith as a fifth place choice (Keith comes in as top defenceman in 13th place). It is clear that defence is underrated in the goals versus threshold system, as the defensive contribution of individual players is hard to measure. This also affects forwards, who have considerable defensive value, which is not a good description of any of our three Hart Trophy nominees.
Because the defensive contribution of the position players is hard to measure, it is clearly underrated. It is reasonable to believe that the goals versus threshold contribution to goaltenders on good defensive teams is overrated, as they pick up the missing value from the top defence. This is a clear systematic reason that goaltenders will be preferentially selected in place of players of defensive value. Goaltenders being overrated and defencemen being underrated is certainly consistent with the fact that ten of the top 20 players in this system are goalies and only two are defencemen.
This increases goaltender value significantly against defencemen and also somewhat against forwards. Thus the forwards on the list (the MVP nominees for example) should rise relative to the defencemen who dominate the list. Is this enough to show that the MVP nominees are correct? Ryan Miller appears 45% more valuable than Henrik Sedin in the goals versus threshold system and it is not reasonable to imagine this effect can make it up.
Are the forwards values accurate? We know that their defensive values may be underrated, but that is probably not a significant error given that these are not great defensive forwards. Their offensive value is also probably underrated. The offensive portion of goals versus threshold assumes all goals as being equal in value and all assists as being equal in value. This is not actually true. Some players create a significant portion of the goals they are involved with, while their linemates inflate their stats as a result. This is hard to measure quantitatively and not taken into account in the goals versus threshold system. Alexander Ovechkin and Sidney Crosby are clearly players who create most of the goals they are involved with and will thus be underrated. Henrik Sedin does this too, but at a lesser level, because he often plays with his brother Daniel as a linemate (however Alexandre Burrows clearly inflated his point totals being on the Sedin line). I think it is within the uncertainties of the goals versus threshold method to say that my MVP observations from last season could be consistent with anything it reveals.
Even if the goals versus threshold valuations of players is correct, would it mean that goalies should be MVP in most (if not all) seasons? I suppose that is a manner of interpretation of MVP awards. I have always argued that the leader in win shares (the player whose performance led his team to the most wins) should be MVP. That is not entirely the same as the player with the most goals versus threshold, as we have the debate of what a threshold player might do. The threshold player is intended to be a replacement level player who is believed to be available to be called up from the minors, claimed on waivers, acquired in a trade for minimal return (say a late round draft pick) at any given time.
The way a replacement level player is used will be more significant for a goaltender than for a position player (especially one on the front line). If a starting goalie who plays 60, 70 or more goals a year has to be replaced, he will be replaced by a seldom used backup goalie or minor leaguer. This player is effectively your replacement level player. The difference between a starting goalie and his replacement will be significant.
If a frontline position player is replaced, the second line player will increase his minutes played and third (and possibly fourth) line players will as well. The replacement level player will play fourth line minutes in replacing a frontline player. The remaining minutes will be an increase in minutes by existing players in the line-up. This understates the value of a position player when he is compared to his threshold level replacement.
The goals versus threshold system attempts to correct for this by using rate stats as much as possible. Players are essentially ranked by goals, assists and points per minute of ice time in the offensive portion of this system, for example, but this isn’t enough to correct for the undervaluing of replacement players. All ice time is not created equally. In most cases, the replacement player on the fourth line will play easy minutes, where he is most likely to score. The frontline player may play some easy minutes, but he also plays in many tough situations that no replacement player would ever be used. This overrates the replacement player. It makes the comparison with threshold for a position player a different comparison than that of a goaltender. Replacing a goalie is a much more significant thing than replacing a position player.
Despite the approximations of the goals versus threshold method, it generally does a good job of showing of the value of players versus their replacement. It overrates goaltenders. It underrates defensive play. It only produces approximate valuations of players. Players who are 10 or 20% apart in value in this system may in fact be equals or be ranked in the incorrect order, but for the most part it ranks players correctly. It should not be used to pick MVP awards. This is because the threshold, replacement level of a goaltender is different than the replacement level of a position player. While a goaltender often has the biggest value versus replacement, this is because he is entirely replaced by the threshold player. A position player is replaced in part by other players who are already in the lineup increasing their ice time. The replacement player plays minimal ice time in easy situations. The comparison versus threshold is not an absolute value of what a player is worth. There is a different baseline for goaltenders. It is entirely possible for a forward to have the most win shares, while a goaltender has the biggest value versus threshold. I argue this happened last season and many other seasons in NHL history.
Filed in: | The Puck Stops Here | Permalink
Comments
Be the first to comment.
Add a Comment
Please limit embedded image or media size to 575 pixels wide.
Add your own avatar by joining Kukla's Korner, or logging in and uploading one in your member control panel.
Captchas bug you? Join KK or log in and you won't have to bother.
Most Recent Blog Posts
What Petri Kontiola’s WHC Performance Tells Us
World Hockey Championships: Preliminary Round
About The Puck Stops Here
The Puck Stops Here was founded during the 2004/05 lockout as a place to rant about hockey. The original site contains over 1000 posts, some of which were also published on FoxSports.com.
Who am I? A diehard hockey fan.
Why am I blogging? I want to.
Why are you reading it? ???
Email: y2kfhl@hotmail.com