NHL/AHL/European Player Rating Formulas

[AlbinoAssassin]

I have noticed that the EA Sports NHL player ratings seem very inaccurate and biased, as I am sure many other people have too. I have spent months researching an unbiased and accurate way to rate hockey players from the NHL and other professional hockey leagues, and I think it is now complete enough to share with this site.

In my rating system, players are rated using mathematical formulas, which means that there is zero bias towards individual players. The basic stats of each player (Games Played, Points, Penalty Minutes and Plus/Minus) are the 4 big factors that go into rating players.

Lets start with the basic formulas for forwards, defense, and goalies.

Forwards

Points ÷ Games Played
Multiply by 50
Add 50
Add (Plus/Minus ÷ 10)
Add 2 if player has 50-99 PIM (to give tougher players a better representation)
Add 5 if player has 100+ PIM (ditto)
Subtract 0.1 for every game missed (takes away some ability from players who are often injured, as that can wear them down in the long run)

Defense

Points ÷ Games Played
Multiply by 60
Add 55
Add (Plus/Minus ÷ 10)
Add 2 if player has 50-99 PIM
Add 5 if player has 100+ PIM
Subtract 0.1 for every game missed

Goalies

Goalie is a lot different in terms of rating than the forward and defense. The main number is based on save percentage. This is a list that shows what sv% matches what number

.913 - 78
.912 - 77
.911 - 76
----------
.910 - 75
----------
.909 - 74
.908 - 74
.907 - 73
.906 - 73

As you can see, for every save percentage above .910, the number increases by 1, while for every two save percentages below .910, the number decreases by 1

After you have your sv% number, next is Goals Against Average:

3.25-3.49: subtract 2
3.00-3.24: subtract 1
---------------------
2.75-2.99: add 1
2.50-2.74: add 2

For every .25 of a goal against average that is 3.00 or over, you subtract 1 from the number you got from the sv%, and for every .25 of a goal against average below 3.00, you add 1 to the number

Then, it gets easier from there

Add ((Amount of wins - amount of losses) divided by 10)
Add 1 for each shutout

Finally, we factor in Games Played for goalies. A goalie who plays less than 80% of his team's games (65 games in the NHL without rounding up) loses 0.1 from his rating for the amount of games under 65 that he played, and vice versa if he played more than 65. This is to put backup goalies behind starter goalies, and it also accounts for possible injuries.



How past seasons are weighted

The past two seasons are weighted instead of just the last one, as if a player had an unexpectedly bad or good season, it seems illogical to think that the player will play at that exact pace the next season. The most recent season (2012-13) is worth 70% of the rating and the season before (2011-12) is worth 30%.

What other leagues are worth

I have researched what other leagues are worth using these formulas, by studying the success of players who play in one of the leagues about to be mentioned below one year, and then play in another one of them the next year. For all of these leagues, all of the results I have seem to fit these league standards very well.

Players in other leagues still follow the same formulas as the ones seen earlier in this post, except that when it is all said and done, you must multiply their ratings by these numbers to get their NHL standard rating:

AHL: 0.7
KHL: 0.7
DEL, Swiss-A, SEL, SM-Liiga, and Czech League: 0.65
Slovakian Extraliga: 0.6
ECHL: 0.5
The second division leagues for the leagues that are worth 0.65: 0.45

If a player plays in more than 1 of these leagues in one season, there is a separate rating for each team played with in that season, and then the percentage of games played with that team compared to the other teams is the percentage of the player's rating with that team shown in his total rating for that season. (example: if a player plays 10 games in the NHL and 40 in the AHL, 80% of his rating is from his AHL performance and 20% comes from his NHL performance

The league where the player played the most games that season is the league where the number of games he played in total that season is counted (for example, if a player plays 23 games in the NHL and 52 in the AHL, he is considered an AHL player. since he played 75 games total that season, and an AHL season is 80 games, he would be considered to have played 75 games out of 80)

Lets do an example now:

Forward: Nazem Kadri

Year Team League GP G A P PIM +/-
2011-12 Maple Leafs NHL 21 5 2 7 8 2
2011-12 Toronto Marlies AHL 48 18 22 40 39 2
2012-13 Toronto Marlies AHL 27 8 18 26 26 5
2012-13 Maple Leafs NHL 48 18 26 44 23 15

2011-12 Maple Leafs
Points (7) ÷ Games Played (21) = 0.33
Multiply by 50 = 16.66
Add 50 = 66.66
Add (Plus/Minus ÷ 10) = 66.86
Percentage of games played in 2011-12: 30%
Total: 20.05 (rounded down to nearest hundredth)

2011-12 Marlies

Points (40) ÷ Games Played (48) = 0.83
Multiply by 50 =41.66
Add 50 = 91.66
Add (Plus/Minus ÷ 10) = 91.86
Subtract 0.1 for every game missed (takes away some ability from players who are often injured, as that can wear them down in the long run)
(Games played: 69 of 80) = 90.56
Multiply by 0.7 (League=AHL) = 63.39
Percentage of games played in 2011-12: 70%
Total: 44.37

2011-12 rating: 64.42

Percentage of 2013 final rating: 30%

2012-13 Marlies:

Points (26) ÷ Games Played (27) = 0.96
Multiply by 50 = 48.15
Add 50 = 98.15
Add (Plus/Minus ÷ 10) = 98.65
Multiply by 0.7 (League=AHL) = 69.05
Percentage of games played in 2012-13: 36%
Total: 24.85

2012-13 Maple Leafs

Points (44) ÷ Games Played (48) = 0.92
Multiply by 50 = 45.83
Add 50 = 95.83
Add (Plus/Minus ÷ 10) = 97.33
Percentage of games played in 2011-12: 64%
Total: 62.29 (rounded down to nearest hundredth)

2012-13 total: 87.14 (70%) = 60.99*
2011-12 total: 64.42 (30%) = 19.32

Final Rating: 80 (rounded down to nearest whole number)

*since Kadri played most of his season in the NHL and did not miss a game in the shortened season, nothing was taken away from his rating due to games missed

But what does 80 mean?

NHL Rating Guideline:

Forwards:
85-99: first line
75-84: second line
65-74: 3rd line
60-64: 4th line
50-59: Healthy Scratch/Minors

Defense:
80-99: First Pair
70-79: Second Pair
65-69: Third Pair
50-64: Healthy Scratch/Minors

Goalies:
80-99: Starter
65-79: Backup
50-64: Minors

Notes:
- for the 2012-13 season, I do not count NHL player's European stats. This is just a personal preference, it is OK to do otherwise
- Any player that has a rating over 99 is just rounded down to 99. These players are very rare anyways

Click to read more...

 

[AlbinoAssassin]

Also, I forgot to mention that I am currently developing a CHL rating formula, it is harder to get an accurate rating for CHL players as they are younger and the league's rules and style of play are a lot different.

 

[hatterson]

I guess I don't really see the value here. You're just reporting a slightly modified PPG. It completely leaves off the defensive aspect of the game.


Compare Bergeron:
32/42*50 + 50 + 2.4 - .6 = 89
64/81*50 + 50 + 3.6 - .1 = 93

89*.7+93*.3 = 90.2

to Joffrey Lupul:
18/16*50 + 50 + .8 - 3.2 = 103.85
67/66*50 + 50 + .1 - 1.6 = 99.26

103.85*.7+99.26*.3 = 102.47 -> 99

or Phil Kessel:
52/48*50 + 50 -.3 = 103.866
82/82*50 + 50 -1 = 99

103.866*.7 + 99*.3 = 102.41 -> 99

I love both Lupul and Kessel, but an overall player metric that leaves them significantly ahead of Bergeron seems to be significantly flawed.

 

[AlbinoAssassin]

I guess I don't really see the value here. You're just reporting a slightly modified PPG. It completely leaves off the defensive aspect of the game.


Compare Bergeron:
32/42*50 + 50 + 2.4 - .6 = 89
64/81*50 + 50 + 3.6 - .1 = 93

89*.7+93*.3 = 90.2

to Joffrey Lupul:
18/16*50 + 50 + .8 - 3.2 = 103.85
67/66*50 + 50 + .1 - 1.6 = 99.26

103.85*.7+99.26*.3 = 102.47 -> 99

or Phil Kessel:
52/48*50 + 50 -.3 = 103.866
82/82*50 + 50 -1 = 99

103.866*.7 + 99*.3 = 102.41 -> 99

I love both Lupul and Kessel, but an overall player metric that leaves them significantly ahead of Bergeron seems to be significantly flawed.

I understand what you mean, and though my system doesn't represent defense significantly (only plus/minus really affects it), scoring is the name of the game when it really comes down to it. Take a player like Jay McClement- not a high scorer but an excellent defensive player. Then take a player like Ilya Kovalchuk, who is not as solid defensively but scores a lot. The majority of people would put Kovalchuk above McClement in their lineup because of this. Kovalchuk is an excellent scorer, and would be used on pretty much every team's first line. McClement is a solid defensive forward, rather than an offensive star, and typically he'd be put on the 3rd or 4th line. These ratings are supposed to reflect that. Here are the ratings of both players:

Kovalchuk:

2011-12 New Jersey Devils 77 37 46 83 33 -9
2012-13 New Jersey Devils 37 11 20 31 18 -6

Rating: 93

McClement:

2011-12 Colorado Avalanche NHL 80 10 7 17 31 -8
2012-13 Toronto Maple Leafs NHL 48 8 9 17 11 0

Rating: 65

What I am trying to say is, these ratings shouldn't be interpreted as "if your rating is better, you are the better player". It can be argued that McClement is as valuable to have on a team as Kovalchuk, and having seen what McClement has done for our blue and white this year, I would agree with that statement. These ratings should be interpreted as a way to find where a player belongs in a team's line-up, based on the role he plays. Here is part of the rating guidelines list from my original post:

Forwards:
85-99: first line
75-84: second line
65-74: 3rd line
60-64: 4th line
50-59: Healthy Scratch/Minors

This would mean that Kovalchuk fits in as a 1st liner, and McClement fits in as a 3rd or 4th line player. These roles are used on these players in the NHL, and these ratings try to reflect that.

 

[hatterson]

Either way you're basically saying lines should be arranged by PPG. The adjustments for plus/minus and PIM don't really affect the overall ratings much in comparison with PPG.

Lets take an 82 game season where a guy plays all 82 games. Here's how the numbers work out.

For every +6 the player gains an extra point.
For PIM >=50 the player gains 3 points.
For PIM >=100 the player gains 5 points.

And since PPG is the only rate statistic you're using as missed games goes up it becomes even more important.

A player who has 3 games played and 3 points would still be rated as a low 90s player.

You can functionally reformat your ratings and say

1st liner: .7ppg and above
2nd liner: .5-.7 ppg
3rd liner: .3-.5 ppg
4th liner: .2-.3 ppg
Scratch: below .2ppg

 

[Sucro]

Players in other leagues still follow the same formulas as the ones seen earlier in this post, except that when it is all said and done, you must multiply their ratings by these numbers to get their NHL standard rating:

AHL: 0.7
KHL: 0.7
DEL, Swiss-A, SEL, SM-Liiga, and Czech League: 0.65
Slovakian Extraliga: 0.6
ECHL: 0.5
The second division leagues for the leagues that are worth 0.65: 0.45

I don't want to be off topic, but is pretty doubtful to rank AHL that high. It shouldn't be ranked higher than SEL and SM-liiga. Same for Ranking DEL above Slovak Extraliga.
Also, don't forget KHL's analog of AHL - VHL.

 

[AlbinoAssassin]

Either way you're basically saying lines should be arranged by PPG. The adjustments for plus/minus and PIM don't really affect the overall ratings much in comparison with PPG.

Lets take an 82 game season where a guy plays all 82 games. Here's how the numbers work out.

For every +6 the player gains an extra point.
For PIM >=50 the player gains 3 points.
For PIM >=100 the player gains 5 points.

And since PPG is the only rate statistic you're using as missed games goes up it becomes even more important.

A player who has 3 games played and 3 points would still be rated as a low 90s player.

You can functionally reformat your ratings and say

1st liner: .7ppg and above
2nd liner: .5-.7 ppg
3rd liner: .3-.5 ppg
4th liner: .2-.3 ppg
Scratch: below .2ppg

It is true that my rating formulas are based on Points Per Game, not entirely but to a large extent. This is because usually you see players who get more PPG get recognized more as top line players (Which is what these formulas try to show). With the other components added to the formulas, including PIM, GP and +/-, you are still given value if you are more of a defensive or a physical player; but like I said in my last reply, players like McClement, as skilled as they actually are, belong lower in the line-up.

The example of a player who plays 3 games and gets 3 points seems highly unlikely. We almost never hear of players playing only 3 games in an entire season at all professional levels (the only example I can remember of this is when Koivu battled cancer in 2001-02, and we know even as Leaf fans that he was a good player anyways), and there are conditions in the ratings that prevent a very small but successful run with an NHL team being the main source of a player's rating. As an example, I will use Jeremy Williams, a former Leafs/Marlies player who for two seasons in a row played one NHL game, and scored a goal in each.

2005-06 Toronto Marlies 55 23 33 56 77 2
2005-06 Toronto Maple Leafs 1 1 0 1 0 0
2006-07 Toronto Marlies 23 6 9 15 27 -4
2006-07 Toronto Maple Leafs 1 1 0 1 0 1

05-06 Leafs rating: 100
05-06 Marlies rating (NHL standard): 70.49
06-07 Leafs Rating: 100.1
06-07 Marlies rating (NHL standard): 53.62

Keep in mind that the games they played in one league compared to the other in a season is converted into the percentage of the season's rating.

05-06:
Percentage of games played in AHL: 98.2%
Percentage played in NHL: 1.8%

100 x 0.018 = 1.8
70.49 x 0.982 = 69.22
05-06 total: 71.02

06-07:
Percentage of games played in AHL: 95.8%
Percentage played in NHL: 4.2%

100.1 x 0.042 = 4.20
53.62 x 0.958 = 51.36
06-07 total: 55.56

(71.02 x 0.3) = 21.30
(55.56 x 0.7) = 38.89
06-07 Final rating: 60 (60.19)

I think that with the conditions set, using PPG as the basis for my formulas for forwards and defense is an accurate way to measure player ratings. But otherwise, what would you suggest as an alternative?

 

[AlbinoAssassin]

I don't want to be off topic, but is pretty doubtful to rank AHL that high. It shouldn't be ranked higher than SEL and SM-liiga. Same for Ranking DEL above Slovak Extraliga.
Also, don't forget KHL's analog of AHL - VHL.

It is impossible to tell EXACTLY how much higher the skill level is in one league compared to another. These are the best I can come up with based on the research I did using hundreds of player's stats, comparing their performance in one of these leagues one year and then their performance in another league the next using these league value numbers.

 

[Zine]

This study by Gabriel Desjardins might be of interest to you:

http://www.behindthenet.ca/projecting_to_nhl.php

 

[AlbinoAssassin]

This study by Gabriel Desjardins might be of interest to you:

http://www.behindthenet.ca/projecting_to_nhl.php

The problem I have with this study is the fact that it's containing stats from before the 2005 lockout, as scoring levels in the NHL have changed since then. This would be more useful if it was made after the lockout.

 

[tenorman85]

That is some pretty cool stuff! Thanks for sharing.