Creating A New Metric: Points Versus Threshold

I preface this by stating that the hard work was many years prior to penning this post by the guys at Behind The Net and Hockey Prospectus.  I also directly leverage some additional work performed by Glen Miller of SNY Rangers Blog.  All credit goes to them.  The stat I created was simply an easier way for me to judge a player’s worth in terms of goals, points, and wins.

Since I started reading River Ave Blues many years ago, I started to “distance” myself from counting stats (goals, assists, etc) and tried to learn metrics to better quantify a player’s real worth to a team.  However, as readily available as metrics were to baseball fans, it was equally unavailable to hockey fans.  The guys at Behind The Net have done a great job at keeping track of a lot of advanced stats like GVT (Goals Versus Threshold), which was created by the guys at Hockey Prospectus, but it has been tough to really combine everything into one stat, like WAR in baseball, to really give you an idea about how useful a player it.

Then along comes Glen Miller of SNY Rangers Blog, who took GVT to a whole new level, and made it into a quantifiable statistic that we can use to see how many wins a player really brings to a team.  First, Miller does a great job at truly defining GVT (here), and gives readers a way to really compare it to the Rangers:

…GVT…represents a player’s value above a “replacement” or “threshold” player in a term (runs for baseball or goals for hockey) fans are familiar with. A “threshold” or “replacement level” player would be defined as the top AHL/minor league player in the organization or the highest ranking free agent on the market in-season (not a prospect) and carries a GVT or RAR value of zero. Basically the guy a club would go get in the event one of their regular players went down to injury or something to that effect. For Rangers fans think Chad Kolarik, Andre Deveaux or Kris Newbury.

In essence, you really find out a player’s worth to a team when he is replaced by someone like Newbury or Kolarik (like we saw last year).  Miller then takes the GVT of each Ranger from last year (courtesy of the guys at BTN), and with the assistance of Tom Awad (hockey metrics guru), created what he is calling Wins Above Threshold, or WAT.

Essentially WAT is just GVT/6, where 6 is the number of goals equivalent to a win in the standings.  You can check out the full Rangers list of WAT by checking out Miller’s article, but suffice it to say, the results really shouldn’t surprise you.

Taking Miller’s work –with Awad’s help– a step (very simple step) further.  Calculating wins in the standings is great, but sometimes I get the feeling that points could be better.  Following up on Awad’s comment to Miller that 6 goals = 1 win, and 3 goals = 1 point, I am just going to use that information to create Points Versus Threshold, or PVT.  PVT gives you the number of points in the standings that a player is worth when compared to a replacement player.

Just a note, you can also use these numbers and compare them to any other player on last year’s roster to see the difference.  For example, subtracting Erik Christensen’s PVT (or WAT) from Brad Richards PVT (or WAT), and you get the difference in points or wins that the addition of Richards should bring.

I’m going to follow that up by saying this really isn’t my work, it’s combining what Miller and Awad did already into something that I personally find to be a little easier to track.  Hockey is based on points, not wins (see: shootout loss), so points may be a little bit easier to judge.  After the jump, you check out the calculations for this year’s Rangers roster based on last year’s GVT numbers.

Player GVT WAT PVT
Henrik Lundqvist

29.60

4.93

9.87

Brad Richards

17.50

2.92

5.83

Marian Gaborik

10.20

1.70

3.40

Dan Girardi

9.90

1.65

3.30

Brandon Dubinsky

9.60

1.60

3.20

Marc Staal

8.70

1.45

2.90

Brian Boyle

8.00

1.33

2.67

Ryan Callahan

7.80

1.30

2.60

Artem Anisimov

7.50

1.25

2.50

Mike Sauer

7.10

1.18

2.37

Derek Stepan

7.00

1.17

2.33

Brandon Prust

6.30

1.05

2.10

Martin Biron

6.30

1.05

2.10

Mats Zuccarello

5.50

0.92

1.83

Erik Christensen

5.40

0.90

1.80

Ryan McDonagh

4.60

0.77

1.53

Ruslan Fedotenko

4.60

0.77

1.53

Wojtek Wolski

3.50

0.58

1.17

Sean Avery

2.50

0.42

0.83

Mike Rupp

1.40

0.23

0.47

Michael Del Zotto

0.7

0.12

0.23

Steve Eminger

0.50

0.08

0.17

The table is sorted in order, from greatest PVT to least.  To use the example I used above, if you were to take Christensen’s PVT (1.80) and subtract it from Richards’ PVT (5.83), and the addition of Richards should bring about 4 extra points to the Rangers.  It may not seem like much, but that’s two full spots in the playoff picture.

I want to end this post by again thanking Glen Miller, Tom Awad, and the guys at both BTN and Hockey Prospectus.  This metric I made here is a product of their hard work and dedication to creating metrics for hockey.

34 Responses to “Creating A New Metric: Points Versus Threshold”

  1. Matt J says:

    Sadly the average fan doesn’t use metrics like this. The only stats they look at is goals assists points ppg wins losses sv% and GAA (which is a dumb stat and just tells how bad the team is or how good it is). This metric is nice but people will still Take the above mentioned stats very seriously.

  2. Dave says:

    The goal is to introduce these and gradually get them to be used. Metrics in baseball didn’t take til recently.

  3. Section 121 says:

    Again, Dan Girardi ranks in with flying colors – best D man on the team, very nice.

    • Dave says:

      I’m a bit surprised that he’s almost half a point better than Staal.

      • The Suit says:

        this is just ranking offensive contributions correct?

      • RangerSmurf says:

        GVT liked Girardi last year, Staal destroyed him the year before.

        • Dave says:

          That’s why I like the stat, it really illustrates when someone has a good year.

          • RangerSmurf says:

            I like it too for a snapshot, and I think it’s pretty valuable when assessing contract worth as well.

            The one issue is that GVT tends to favor offense pretty heavily, which explains the variations btwn Staal/Girardi the last two years.

            It’s good that Girardi rated well here, but I think taking this to mean Girardi was our best defenseman loses something, especially when comparing to alot of his other metrics.

            • Dave says:

              That’s why you look at career GVT when assessing who the “best” is.

              It’s also very, very tough to really quantify the worth of a shutdown defenseman. QUALCOMP, CORSI, goals against, relative +/-, etc…all need to be taken into account there.

              • Section 121 says:

                Do you have the career GVT numbers on Staal and Girardi?

              • RangerSmurf says:

                Career is better, but the deficiencies remain, so defensive value is still short-changed.

                @Section121 –
                Staal – 4.9, 4.0, 9.0, 8.7, 27.2 in 321 games
                Girardi – 2.3, 9.2, 4.5, 6.2, 9.9 32.1 in 360 games

                Girardi has the slight edge in GVT/G, which again is in large part because he’s shown more offensive skill/production to date.

              • Dave says:

                Not at the moment, I have it somewhere on my personal laptop.

  4. Dave says:

    Also, this post was supposed to run yesterday, but in light of the news, I decided to run it today instead.

  5. The Suit says:

    Good stuff Dave. I like your take on these numbers.

    I’m still waiting for someone to combine both quantitative (advanced stats) and qualitative (observational) analysis to tell us a compelling story.

    I think there are too many people who are doing one or the other, but no one has really found a way to marry the two perspectives, which I think is key for the sport of hockey.

    Just my 2 cents…

    • Chris in MA says:

      I think you’d need to start by looking at these numbers by position… then, when two players are within X PAT points, you’d have to compare them observationally to determine which one brings more to the table.

      But even then, Im not sure theres a reliable way to marry the two pieces together.

      • Dave says:

        That’s a good idea. I think I’m going to keep track of that this season by position. It’s a simple calculation anyway.

        • Chris in MA says:

          Yeah. It’d be nice to have that around. Then, when two guys are within like 1 PAT of each other, you can start discussing which one is better to have around, etc.

          • Dave says:

            I may do it for the whole league to be honest. But it is reliant on GVT, which I don’t calculate.

            • Glen Miller says:

              Tom said they are working on getting live GVT ratings posted on the Hockey Prospectus site by October. Hopefully it works out that way because I was only able to reverse-engineer it to a point. He weighted quality of ice time and I didn’t feel right about asking what factor(s) he may have used.

    • Dave says:

      I believe that the point of the metrics is to put into perspective what a player is like on the ice for those who cannot watch the player. Observational analysis is very subjective, so it’s tough to really quantify that.

      • The Suit says:

        No I understand that, it’s just I often read advanced stats articles on ESPN where they completely twist the numbers to make their point. It’s like they do analysis backwards.

        Obviously you wouldn’t be doing that here, but I just think these things often need that extra layer.

        • Dave says:

          Twisting numbers to meet your point is human nature though. We all do it.

          I think the point of GVT/PVT/WVT is to hopefully prevent how much twisting you can do.

          • The Suit says:

            I don’t know, I think ESPN’s shit often comes off as cherry picking. They know what they want to say before they even look at the numbers. That to me just compromises their findings.

            • RangerSmurf says:

              That’s an ESPN problem more than a stats problem. Non-stats arguments can be incredibly biased as well.

            • Glen Miller says:

              The best way not to twist the numbers to meet your preconceived belief is to try not to have one when analyzing the numbers. For example; I am doing a piece on Gaborik’s season and trying to find if there were other reasons his goal-scoring dropped, maybe reasons that are likely to “fix” themselves or were not necessarily his fault. One I found which suggests he is due for a bounce back is shooting percentage.

  6. Chris in MA says:

    These are the types of things that I love seeing. Metrics are so much better than the standard statistics.

    I am not surprised that Girardi ranks that high. I still constantly see people thinking that he’s overrated, but I think he does a great job.

    I am, however, surprised that Christensen ranks higher than Fedotenko… but that is likely due to my biased against Christensen.

    • Dave says:

      I have an email out to Tom Awad about GVT, and if it includes shootout stats in the calculation of GVT.

      • Chris in MA says:

        I’d imagine that it doesnt but, let us know when you find out.

      • RangerSmurf says:

        It does. You can see pretty wild variations because of it. Zuke’s shootout value was 2.3 (through 68 games) and EC’s was 2.0.

        If you have the database (and i’m assuming you do), the shootout values are in that spreadsheet.

        • Dave says:

          There you go…shootouts are included.

          Honestly, it makes sense to have a wild variation like that, extra points in the shootout are becoming a huge necessity now.

  7. Robert V says:

    This is indeed the correct application of GVT, but shouldn’t you be calling it PVT (and WVT)?

    • Dave says:

      Thanks for the input Rob. Keep up the great work with the metrics.

      Good point on the name, was just playing off Glen’s WAT. Will change mine.