November 17, 2010

Gregory Campbell & Referees


After reading a little bit about the mess Colin Campbell put himself in. I couldn't help but wonder a few things:
- Did Dean Warren know Colin was out to get him when he refereed Campbell's son's game?
- How many other times has Colin asked Stephen to discuss a call about his son with the referee who called it?
- Do the referees think that these calls could impact their employment/advancement?
- Are there other player's that result in similar concerns from Colin?

With that said, I compiled a quick list of the number of games that referees were involved in with Gregory Campbell (GP) and the number of calls against Gregory Campbell in those games. (ERR = 2 standard deviations of the Call%).

There's a little more variation in these results than would be expected, especially since these results should be less variable (due to the fact the NHL uses 2 refs and they are randomly matched). I don't really want to say more than that, but interestingly, Stephen has 0 calls against Greg. Kerry and Dean have the most and they're gone (although I don't think that Kerry's retirement had anything to do with Colin's son).

In the AHL Gregory averaged 1.37 Penalty minutes per game, in the NHL he's averaged 0.92. Although I believe it is common to have fewer penalties in the NHL compared to the AHL.

Update: Sorry, I just loved this exchange on TSN:
Duthie: But, do you believe the emails to former referee-in-chief Stephen Walkom were inappropriate?
Campbell: No they weren't inappropriate...
...
Campbell: Well, it is inappropriate.

November 2, 2010

New Jersey

Updated Prediction Model
I've done quite a bit of work on my prediction algorithm's and I am extremely happy with the results: With less than dozen games per team so far, I have a pretty good idea of the probabilities of a team making the playoffs. Although my model is designed to predict individual games and not outcomes over the course of a season it seems to do a reasonable job at doing so anyway.

Results
I used my algorithm to review the 2009 - 2010 results and grouped teams into 3 categories:
Top 10, Middle 10 and Bottom 10 after 154 games were played.

Top 10: 8.4 teams in this group were expected to be in the playoffs, 8 made it (Dallas & NYR didn't make it)
Middle 10: 6.1 were expected to make the playoffs in this group. 5 actually did.
Bottom 10: 1.5 were expected to make the playoffs in this group. 3 made it: Detroit 35%, Boston 32%, Nashville 7%.

Overall these results suggest that my model does not differ significantly from the true results (although my model is based on 4 seasons, which include last season)

My model says they have a 3% chance of making the playoffs
And on that note I can say with a bit of confidence that: New Jersey is probably not going to be in the playoffs (namely, they are statistically out of the playoffs at this point). Right now their win percentage is 23%. They're Pythagorean percentage sits at 18% (GF^2/(GF^2 + GA^2). I should mention, they have been a little unlucky in the neighborhood of -15 GF. Including these goals in their totals would only improve their winning percentage to 41%. Now with the injury to Parise they have an up hill battle for the rest of the season.

October 12, 2010

Kovalchuk Mess 2

I don't understand why New Jersey desperately wanted Kovalchuk so bad, but it looks like their benefits may be offset by cap issues caused by Kovalchuk.

For those who don't know:
...the Devils iced a roster of 15 skaters and two goaltenders in a 3-1 loss to Pittsburgh [Monday October 11], many are wondering if the NHL should step in and slap the team for violating the CBA; if what the Devils did with their roster because of the salary cap can be deemed good for the game. [Yahoo]
I'm not one to care what sort of things the NHL does to punish the New Jersey devils - anything monetary is just part of doing business for them and the team has already lost enough: the game.

However, I do want to comment on what effect these sorts of things have on the probability of winning. I did a study a couple years ago that showed that whether a player got more or less ice time in a game their absolute number of points received dropped (if they got less ice time it's because they didn't have enough time to get the same number of points, and if they got more time it was because their scoring rate fell due to fatigue).

For the intent of this explanation I will assume the simpler situation: points are constant for each player with respect to number of minutes played. Which means if you increase a players ice time their scoring totals for that game on average are constant.

How does this affect New Jersey?
Normally when a player is injured a replacement is used. In this case New Jersey couldn't afford the replacement. So not only did New Jersey lose the benefits of having Volchenkov, Pierre-Luc & Rolston they also couldn't fit marginal players (who also get points) to contribute at least a little.

Cap Cost:
A "marginal" forward will generally contribute about 0.3 Points/game and a defenseman about 0.17 points per game. So the salary cap cost them 0.77 points or about 0.3 goals (equal to about a $55,000 fine...).

Actual Cost:
The loss of Volchenkov [0.22 points/game], Rolston [0.5 Points/game] & Pierre-Luc [ 0.1 points/game] for a total of 0.82 points per game (not that much different due to Pierre-Luc being more of a fighter than player and Volchenkov's inability to score).

Presumably goals against will go up a little with fewer skaters, lets assume it is half the effect of offense (conservative estimate) or about 0.15 goals.

Let's use New Jersey's scoring numbers from last year for demonstration purposes: 2.7 GF/gp & 2.3 GA/gp and won 48 games (59%). The new numbers would be 2.4 GF and 2.45 GA for an expected winning percentage of 49%. So in effect New Jersey goes from being an excellent team to being below average.

Also, I think these estimates are conservative - effects could be much higher. Study this is based on was assuming small changes to icetime, not 20% increases.

Diving Update

As per a request in a comment I am posting an update re: diving. My original post can be found here.

- The Y-axis is the number of diving calls in the previous 100 games - ordered by game #
- The X-axis is the "season" year using the year of the last game (so we're in the NHL season 2011 right now) - The first dot represents # of diving calls in the 2007-2008 season between games 0 & 100.

The graph does a good job summarizing what has changed since 2007: Diving calls have disappeared. One thing that may be happening is that the NHL could be calling them something else on the scoresheet: "unsportsmanlike conduct" instead of "diving" - which I'm going to watch for. There have been 151 Diving calls in 3690 games (1 diving call every 25 games).

Also, when I previously reported these numbers I also reported how often the referee also called a penalty against the player who "caused" the dive. In 2006 & 2007 it was close to 80% (77% and 82% respectively), since then it has jumped to 90%, so of the 151 penalties 15 were independent of any other penalty.

Does anyone believe there are actually fewer players diving though?

July 21, 2010

Clark vs. Perron

When I saw the story about MacArthur my first question was: who? I try to keep on top of all major NHL contributors and that one I had never heard of. So the obvious next question was: why on earth is a no name NHL contributor get $2.4 million from an arbitrator? Then came a story shortly after Perron signed for $2.15M, which is a name I certainly do know.

MacArthur (25):
G: 16
A: 19
Salary: $2.5M

Perron (22):
G: 20 (+25%)
A: 27 (+40%)
Salary: $2.15M

First off Atlanta did the right thing to abandon a player that is worth about $1.5M. MacArthur shows no signs of improvement. His PP stats are terrible. David Backes is getting paid $2.5 for an extra 12 assists compared to MacArthur. Callahan is making $2.3 for similar numbers. Fehr (#18 overall) is making $2.1. Disappointing Bernier is getting $2M. Stafford (much more potential) is making $1.9. Steen $1.7. Ladd $1.6M etc.

Presumably Atlanta wanted to sign him this year. It's not like they really needed his services at the deadline. That's what makes this arbitration award so bad: Atlanta traded a 3rd & 4th rounder for nothing because the arbitrators were generous.

Of course the Thrashers could have shot themselves in the foot by not providing good alternatives to what MacArthur's agent provides. There are plenty of players in this category (no shortage of supply).

July 20, 2010

Kovalchuk Mess

New Jersey finally pushed the CBA rule makers to their limits. By signing a contract that is almost guaranteed to not be completed the NHL did the right thing to step in and say no.

Of course that made people ask the simple question:

Why now?
The answer to this is probably that Kovalchuk's contract is well over the 40 mark, a point where very few players continue to play (especially snipers who typically retire early: Bure, Naslund, Sundin).

Also, the NHL is aware Kovalchuk wanted $10M/year. If he plays 10-11 years and then retires, he'll have that.

A Better Way?
Of course there is a much easier way of dealing with these contracts. That is to institute a "maximum salary cap hit" for long term SPC. Any contract year's cap over 4 years would be subject to these maximums (that is to say a 36 year old signing a 4 year contract would not be impacted at all, but a 5 year contract would, but would only apply to the last year). Say $4M for players over 35, $2M for players over 37 and $1M for players over 39. However the sum of the contract would still be the same, so the difference would be applied to the years there there is no maximum:
Proposed Cap Schedule Current Cap Schedule
$9.56 $6.00
$9.56 $6.00
$9.56 $6.00
$9.56 $6.00
$9.56 $6.00
$9.56 $6.00
$9.56 $6.00
$9.56 $6.00
$9.56 $6.00
$4.00 $6.00
$4.00 $6.00
$2.00 $6.00
$2.00 $6.00
$1.00 $6.00
$1.00 $6.00
$1.00 $6.00
$1.00 $6.00
=$102 =$102


I'm sure others could see how it would affect players like Luongo and Hossa. But at least it would create a consistent system as opposed to this ad-hoc system full of surprises. All of the sudden these long terms contracts are worthless!

What will the NHL accept?
Obviously the Devils will go back the drawing board, however, I'm not sure how much cap space they can give up. I'm sure N.J and the NHL will take a lot about this over the next few days and a slightly modified contract will be approved (and this whole process will turn out to be a joke).

July 14, 2010

Northwest Expecations - 2010

I'm still tweaking these, but this is a start. I've adjusted my technique. I've pulled 5 years worth of data to generate these tables. The "young" player problem still exists (because I lack data to differentiate between a good 19 year old or a average 19 year old).

Notation:
Name [GFScore@EV, GAScore@EV]

I've separated out PP & PK now. However these values are based on the player's past performance over the last 5 years. The first 7 rows only included EV goals. there are still some bugs (eg. Johnny vs. John...), which I am working on (I have to find all of them first).

Note: the percentage besting the team's name is the percentage of salary cap used.



















The little arrows indicate whether there is a substantial change in the team's expected points from last season (more than 10 difference).

Note: These include RFA's who haven't signed yet.

Central Expectations - 2010

I'm still tweaking these, but this is a start. I've adjusted my technique. I've pulled 5 years worth of data to generate these tables. The "young" player problem still exists (because I lack data to differentiate between a good 19 year old or a average 19 year old).

Notation:
Name [GFScore@EV, GAScore@EV]

I've separated out PP & PK now. However these values are based on the player's past performance over the last 5 years. The first 7 rows only included EV goals. there are still some bugs (eg. Johnny vs. John...), which I am working on (I have to find all of them first).

Note: the percentage besting the team's name is the percentage of salary cap used.


















The little arrows indicate whether there is a substantial change in the team's expected points from last season (more than 10 difference).

Note: These include RFA's who haven't signed yet.

Pacific Expectations - 2010

I'm still tweaking these, but this is a start. I've adjusted my technique. I've pulled 5 years worth of data to generate these tables. The "young" player problem still exists (because I lack data to differentiate between a good 19 year old or a average 19 year old).

Notation:
Name [GFScore@EV, GAScore@EV]

I've separated out PP & PK now. However these values are based on the player's past performance over the last 5 years. The first 7 rows only included EV goals. there are still some bugs (eg. Johnny vs. John...), which I am working on (I have to find all of them first).

Note: the percentage besting the team's name is the percentage of salary cap used.



















The little arrows indicate whether there is a substantial change in the team's expected points from last season (more than 10 difference).

Note: These include RFA's who haven't signed yet.

Atlantic Expectations - 2010

I'm still tweaking these, but this is a start. I've adjusted my technique. I've pulled 5 years worth of data to generate these tables. The "young" player problem still exists (because I lack data to differentiate between a good 19 year old or a average 19 year old).

Notation:
Name [GFScore@EV, GAScore@EV]

I've separated out PP & PK now. However these values are based on the player's past performance over the last 5 years. The first 7 rows only included EV goals. there are still some bugs (eg. Johnny vs. John...), which I am working on (I have to find all of them first).

Note: the percentage besting the team's name is the percentage of salary cap used.



















The little arrows indicate whether there is a substantial change in the team's expected points from last season (more than 10 difference).

Note: These include RFA's who haven't signed yet.

Northeast Expectations - 2010

I'm still tweaking these, but this is a start. I've adjusted my technique. I've pulled 5 years worth of data to generate these tables. The "young" player problem still exists (because I lack data to differentiate between a good 19 year old or a average 19 year old).

Notation:
Name [GFScore@EV, GAScore@EV]

I've separated out PP & PK now. However these values are based on the player's past performance over the last 5 years. The first 7 rows only included EV goals. there are still some bugs, which I am working on (I have to find all of them first).

Note: the percentage besting the team's name is the percentage of salary cap used.



















The little arrows indicate whether there is a substantial change in the team's expected points from last season (more than 10 difference).

Note: These include RFA's who haven't signed yet.

Southeast Expecations - 2010

I'm still tweaking these, but this is a start. I've adjusted my technique. I've pulled 5 years worth of data to generate these tables. The "young" player problem still exists (because I lack data to differentiate between a good 19 year old or a average 19 year old).

Notation:
Name [GFScore@EV, GAScore@EV]

I've separated out PP & PK now. However these values are based on the player's past performance over the last 5 years. The first 7 rows only included EV goals. there are still some bugs (eg. Johnny vs. John...), which I am working on (I have to find all of them first).

Note: the percentage besting the team's name is the percentage of salary cap used.



















The little arrows indicate whether there is a substantial change in the team's expected points from last season (more than 10 difference).

Note: These include RFA's who haven't signed yet.

June 25, 2010

Goalie Statistics for 5 years

So, here is a compilation of all the saves/goals/shots etc. each goalie faced over the last 5 years (inc. playoffs). I may be missing some games, but there is a lot of data here. The "Cred" column is just adjusting the shot quality neutral save percentage based on number of shots faced (regressing to the mean).

EG = Expected Goals
G = Actual Goals
SQN = shot quality neutral save percentage
= 1- 0.0926*G /EG
SV% = real save percentage
= 1 - G / S

Note: The expected goals are adjusted for site based biased shot information.
Few things of note
- Save percentages over 0.920 are not really sustainable (8 goalies in 2010).
- I didn't realize how good Hiller is.
- Raycroft is really bad (allowed almost 100 more goals than average)

NnameCEGGSVPCTSQNCred
1Henrik Lundqvist1017510108470.9170.9220.919
2Jonas Hiller41784023230.9230.9260.918
3Tomas Vokoun95508537530.9210.9180.915
4Jaroslav Halak36773482970.9190.9210.913
5Roberto Luongo112349959290.9170.9140.911
6Craig Anderson48574524050.9170.9170.911
7Timothy Thomas83887266840.9180.9130.910
8Dominik Hasek38993633320.9150.9150.909
9Cam Ward87398368130.9070.9100.907
10Cristobal Huet65495815590.9150.9110.907
11Chris Mason66526045790.9130.9110.907
12Martin Brodeur103548718610.9170.9090.906
13James Howard24642131950.9210.9150.906
14Dan Ellis31982892750.9140.9120.905
15J.S Giguere77676936850.9120.9080.905
16Pekka Rinne32032902760.9140.9120.905
17Ilja Bryzgalov78516656670.9150.9070.904
18Miikka Kiprusoff112449549720.9140.9060.904
19Dwayne Roloson85657697720.9100.9070.904
20Kari Lehtonen66795965950.9110.9080.904
21Manny Fernandez35573133040.9150.9100.904
22Nikolai Khabibulin66516296410.9040.9060.903
23Niklas Backstrom66625435490.9180.9060.903
24Ryan Miller105168628840.9160.9050.903
25Steve Mason34513213200.9070.9080.902
26Martin Biron72026236510.9100.9030.901
27Evgeni Nabokov91877948330.9090.9030.901
28Rick Dipietro59605395580.9060.9040.901
29Josh Harding22511941900.9160.9090.901
30Jason Labarbera31632872900.9080.9060.900
31Marty Turco91328118540.9060.9030.900
32M.A Fleury96168398870.9080.9020.900
33Carey Price45393954080.9100.9040.900
34Manny Legace41123793890.9050.9050.900
35Martin Gerber51694614860.9060.9020.899
36Ray Emery51144514780.9070.9020.899
37Pascal Leclaire43633944200.9040.9010.898
38Jonathan Quick33712983130.9070.9030.898
39Ty Conklin35123003160.9100.9020.898
40Alexander Auld52454735130.9020.9000.897
41Peter Budaj51704605020.9030.8990.896
42Mathieu Garon53144735180.9030.8990.896
43Brian Elliott23342102220.9050.9020.896
44Chris Osgood50584464850.9040.8990.896
45Jose Theodore69746357000.9000.8980.896
46Antero Niittymaki59135275780.9020.8980.896
47Joey Macdonald22202122230.9000.9020.896
48Brian Boucher25862452630.8980.9000.895
49Brent Johnson33752993310.9020.8980.894
50Johan Hedberg42993864300.9000.8970.894
51Curtis Sanford23522152340.9010.8990.894
52Vesa Toskala62175586320.8980.8950.893
53Scott Clemmensen23111952160.9070.8980.893
54Ed Belfour30262793090.8980.8970.893
55Olaf Kolzig54254765430.9000.8940.892
56Mike Smith38053123540.9070.8950.892
57David Aebischer24352202450.8990.8970.892
58Curtis Joseph37683413930.8960.8930.891
59Fredrik Norrena24102102440.8990.8920.889
60Patrick Lalime31162803270.8950.8920.889
61Mikael Tellqvist26362272700.8980.8900.887
62Andrew Raycroft45534065000.8900.8860.885
63Marc Denis27792473050.8900.8860.884
64John Grahame26752373010.8870.8820.882
65Johan Holmqvist24661992680.8910.8750.876

June 8, 2010

Stanley Cup Final


CHI
PHI
#GEGS%
GEGSV%
Game 1:63.283.953.181.3
Game 2:22.396.412.891.3
Game 3:33.089.743.990.0
Game 4:33.786.252.991.9
Game 5:72.784.642.677.8
Series [3-2]2114.988.21915.386.6



Philadelphia is doing a lot better than I would have anticipated. If it wasn't for the difference in goaltending this series could easily be tied (or have gone the other way).




CHIPHIWinner
Even Strength
GF2.92.44
EGF2.832.78
GA2.352.43
EGA2.192.53
SV%89.2%90.4%
Power Play
GF6.546.64
EGF6.558.69
GA0.492.42
EGA0.60.8
SV%92.3%89.6%