smozoma

new Player Ratings subforum!

Recommended Posts

Thanks, Halifax, for setting up this new forum section!

I guess we will evolve an organization scheme, but I'll start off by creating topics for a few of the teams, and you guys can copy the technique or maybe come up with something better.

The idea is, you rate the players on teams you are familiar with, in order to help people out with making their own ROMs. Give us the skinny on all those no-name players.

If you need some numerical help with rating players, here are some resources:

http://www.hockey-reference.com/teams/LAK/2012.html (note the "OPS" and "DPS" (offensive, and defensive, Point Shares) columns).

http://www.nhl.com/ice/playerstats.htm?fetchKey=20122ALLSASALL&sort=hits&viewName=rtssPlayerStats

http://www.nhl.com/ice/playerstats.htm?fetchKey=20122ALLSASALL&sort=avgTOIPerGame&viewName=timeOnIce

http://www.behindthenet.ca/nhl_statistics.php?ds=30&s=29&f1=2011_s&f2=5v5&f5=OTT&f7=5-&c=0+1+3+5+4+6+7+8+13+14+29+30+32+33+34+45+46+63+67#

WTF is Corsi? Fenwick? zone starts?

http://www.matchsticksandgasoline.com/section/understanding-advanced-stats

dissenting opinion on the usefulness of the advanced stats:

http://drivingplay.blogspot.ca/2011/09/on-problem-with-corsi-rel.html

Some threads about player attributes, what they mean, how they're decided..:

http://forum.nhl94.com/index.php?/topic/14307-attribute-guide/

http://forum.nhl94.com/index.php?/topic/13345-gens-player-attribute-survey/

http://forum.nhl94.com/index.php?/topic/10223-most-valuable-attribute/

http://forum.nhl94.com/index.php?/topic/12879-the-passingshooting-bias-statistic/

http://forum.nhl94.com/index.php?/topic/11928-a-few-newbie-questions/#entry93188

Share this post


Link to post
Share on other sites

thanks, guys. for the last few years I have been using plus/minus (relative to teammates) the approximate overall player ratings. for next season, I'm thinking of using "Corsi relative to the quality of the competition." am I understanding it correctly in thinking that it is a stat that shows a player's overall skill/value with a consideration of the skill/value of the players they face? it isn't just the quality of the opponents, right? it does correlate to our overall ratings, right?

http://www.behindthe...46 63 67#snip=f

if anybody would be willing to discuss this with me, I'd appreciate it. it looks like Chara, Boychuk, Marchand, Seguin and Bergeron (and Rolston!?) would be excellent and Campbell, Thornton and Paille (and McQuaid!?) would be bad. that seems right, to me. this is going to take a ton of work, but I think it would be worth it, even if not a lot of ratings change by much (most look like they won't). the ones that would change are the big minuses that face other teams' top lines, like Krejci. Krejci would go from 60 to 85, or so. that seems much more fair than the method I was using, which devalued him due to the beatings he took while facing other first-liners. anyway, again, please comment if you can.

it would work out somewhat like this:

CorsiRQC = overall

+1.25 = 100

+1.00 = 95

+0.75 = 90

+0.50 = 85

+0.25 = 80

=0.00 = 75

-0.25 = 70

-0.50 = 65

-0.75 = 60

-1.00 = 55

-1.25 = 50

-1.50 = 45

-1.75 = 40

-2.00 = 35

-2.25 = 30

Edited by trudatman

Share this post


Link to post
Share on other sites

It's an imprecise science since it's quite new, and there are so many factors. Quality of teammates has a great effect, not just quality of competition. That bergeron/seguin/marchand line, are they all around the same in skill at driving high Corsi? What about when those players are not playing together?

Here's a page that handle that: Brad Marchand Teammates with/without him Corsi

For the Leafs, the one thing that really stands out is that Grabovski is the only real plus-Corsi player. Without him on the ice, EVERY PLAYER is a minus-Corsi player. (see the right-most column). So you might try comparing the with/without columns for different players to find out if they tend to help players or hurt them.

Other factors are zone starts (if you start in your own end a lot, then your Corsi will probably suck).

Share this post


Link to post
Share on other sites

I can't say I'm less confused, now. do you think the system I outlined above is a good one? is there a better stat for determining overall ratings? I don't want to have to compare stats to other stats, I want to find one that works as correlating to general overall skill/effectiveness. is the Corsi relative to opponents too harsh to some defensemen? that would align the overalls well with the original game, then, wouldn't it?

edit: should I just be using Corsi relative, which would be about a -20 through 20 scale? the only difference is that it doesn't take into account how hard the opponent is, right? the one I want to use as outlined in the post above is the same, just weighted based on the difficulty of the players on the other side, right? tell me if I get it, please. wait... Corsi and Fenwick just measure whether a player being on the ice leads to more or less shots? I'm not sure I like it.

also, penalties taken per 60 is better than the way I was doing it, which was essentially per game, ignoring ice time. I wish they took that sat out another decimal (or two), but it's still going to improve my ratings if I use something like:

1.3+ = 100

1-1.2 = 85

0.8-1 = 65

0.5-0.7 = 55

0.3-0.5 = 45

0-0.2 = 35

I may use HARO+ and HARD+ to make better "awareness" skill ratings, but they'd most like be for tweaking and balancing them, so I'm not giving a defensive guy more offense and stuff like that. those rating and stick handling/puck control (and, to a lesser extent, agility) are my go tos for when I need to adjust an overall, as they aren't culled directly from a stats usage chart.

how about HART+ for overalls (I'm not all that comfortable with strictly shot-based stuff)? puts McQuaid back up and Krejci back down, but, aside from Pouliot becoming a superstar and Chara seeming mediocre, it looks pretty good:

http://stats.hockeya...utes=300&disp=1

VAN's looks perfect:

http://stats.hockeya...utes=300&disp=1

Edited by trudatman

Share this post


Link to post
Share on other sites

Oh yeah, I forgot about HARx+ ratings.

The short answer is, these are all just tools you can use to help you rate them, but it's hard to take a single stat and rate a player based on that.

Corsi stats are usually counted at 5on5, but some guys are valuable power play specialists. They might get killed 5on5, but be worth it due to the PP.

Try looking at a good team like Boston, a bad team like maybe the Islanders, and a middling team like Calgary, and see if you can find a technique that gives consistenly reasonable results.

Share this post


Link to post
Share on other sites

I am leaning toward what follows, based on HART+ (5-on-5 adjusted with regards to zone starts).

HART+ = overall:

1.25 = 100

1.20 = 95

1.15 = 90

1.10 = 85

1.05 = 80

1.00 = 75

0.95 = 70

0.90 = 65

0.85 = 60

0.80 = 55

0.75 = 50

0.70 = 45

0.65 = 40

0.60 = 35

0.55 = 30

0.50 = 25

PIT's looks perfect:

http://stats.hockeyanalysis.com/ratings.php?db=201112&sit=5v5close_f10&type=goals&teamid=24&pos=skaters&minutes=100&disp=1

BOS's looks good:

http://stats.hockeyanalysis.com/ratings.php?db=201112&sit=5v5close_f10&type=goals&teamid=3&pos=skaters&minutes=300&disp=1

VAN's looks perfect:

http://stats.hockeyanalysis.com/ratings.php?db=201112&sit=5v5close_f10&type=goals&teamid=29&pos=skaters&minutes=300&disp=1

COL's look good:

http://stats.hockeyanalysis.com/ratings.php?db=201112&sit=5v5close_f10&type=goals&teamid=9&pos=skaters&minutes=100&disp=1

EDM's look good:

http://stats.hockeyanalysis.com/ratings.php?db=201112&sit=5v5close_f10&type=goals&teamid=12&pos=skaters&minutes=100&disp=1

LAK's is interesting:

http://stats.hockeyanalysis.com/ratings.php?db=201112&sit=5v5close_f10&type=goals&teamid=14&pos=skaters&minutes=100&disp=1

this would give us just fourteen players at 100, so I like that, but it is putting Sobotka in that group and calling Chara average, which seems wrong:

http://stats.hockeyanalysis.com/ratings.php?db=201112&sit=5v5close_f10&type=goals&teamid=0&pos=skaters&minutes=500&disp=1

maybe doing a half HART+ and half CorsiRQC would work...

CorsiRQC = overall

+1.25 = 100

+1.00 = 95

+0.75 = 90

+0.50 = 85

+0.25 = 80

=0.00 = 75

-0.25 = 70

-0.50 = 65

-0.75 = 60

-1.00 = 55

-1.25 = 50

-1.50 = 45

-1.75 = 40

-2.00 = 35

-2.25 = 30

so, Zdeno Chara getting a 75 in one and 95 in the other would end up an 85.

Patrice Bergeron-Cleary would be (100+85)/2 = 92.5.

David Krejci would be (82+58)/2 = 70.

Edited by trudatman

Share this post


Link to post
Share on other sites

I dunno, those HART+ ratings look pretty bad. Jordan Staal one of the worst Penguins? Kunitz above Malkin? I think you're better off going with hockey-reference.com's Point Shares per GP rather than HART+.

Share this post


Link to post
Share on other sites

actually, I really like the Penguins HART+ ratings. Malkin is an offensive superstar, sure, but Kunitz is no slouch in either direction. I think Staal is WAY overrated. all of the Staals are, in my opinion. but, I'm a big plus/minus guy while most fans see flashy offense as the best thing ever. yeah, you have to score to win, but you also have to not let the other team score more than you. I think I'm going to play around with that 50/50 formula some more and see what I think of how it holds up, leaguewide.

Share this post


Link to post
Share on other sites

Only Crosby had a higher Quality of Competition than Staal, and Staal's Quality of Teammates was lower than Neal, Kunitz, Malkin, and Crosby (top 4 in order), yet he put up 50 points in 62 games, never playing with Malkin or Crosby. His Offensive zone starts was 47% (waaay less than most of the other scorers on the team, who were mostly over 60%, which greatly helps their +/-). Corsi Rel QoC, top on the team by a mile... He's pretty good!

Share this post


Link to post
Share on other sites

I see. well, then... I retract my ill-informed criticism. the other Staals still suck, though.

so, how about (2xCorsiRQC + HART+) / 3 ? that would weigh the Corsi with some HART, but not as much. are you seeing many other examples where the HART+ seems way off? I have taken a break from running numbers on my 50/50 theory, as the computer I was using for that is suffering from a severe lack of Internets (it's a humidity versus wireless connection thing), but I'm curious to know what your opinion is for skewing overall ratings. it seems to be the part most players of '94 hacks care the most about and I'd love to read about what a clearly well-informed hacking master thinks would work.

Edited by trudatman

Share this post


Link to post
Share on other sites

Leafs:

Phillipe Dupuis gets an astronomical 2.525 HART+ despite playing 8 5on5 minutes a game against the other team's worst players, and not scoring a single point in 30 games. On the plus side, he was only -2, but the goalie save percentage while he was on the ice was .966 (which you would expect when he's playing against the other team's worst players). In his defense, his offensive zone starts was 43%, but I watched almost all the Leafs games and don't remember seeing this guy once. The Leafs let him go and PIT picked him up for $600K.

Jay Rosehill ranked middle-of-the-pack. He's an enforcer.

Carl Gunnarsson ranked near the bottom, when he's actually pretty good and the only guy I really trust on D (poor man's Lidstrom.. very poor man..).

Cody Franson ranked as their 2nd-best D. He was scratched half the season and also played mostly against weak opponents.

I think I'd ignore HART+ since it appears to completely ignore quality of competition and give strange results. I like Corsi HART+ a bit better, but it makes Franson the 3rd best player on the team. If you can adjust the HART+ values based on QoC somehow, they might be ok.

I can't comment on the formula, don't want to plug the numbers into excel and stuff.

Share this post


Link to post
Share on other sites

it ain't much of a formula, it's just two C's, one H, averaged. where are you seeing Corsi Hart? I want to try half that and half CorsiRQC. really, I'm looking to get goals scored considered, not just shots.

Share this post


Link to post
Share on other sites

hmm.

how did your Leafs look to you with that one? seems alright to me. I also looked at NYR, BOS, VAN. they all looked okay. only big surprises were Chris Kelly as average and Dan Sedin above his brother.

can I get your total opinion on how to go with all of this stuff?

here's how I'd see this one working out:

CorsiHART+ = overall

1.2 = 100

1.1 = 88

1.0 = 75

0.9 = 62

0.8 = 50

Edited by trudatman

Share this post


Link to post
Share on other sites

Daniel Sedin shoots, so that'll help his Corsi when he's separated from Henrik, I guess

minimum 200 minutes

1 GRABOVSKI, MIKHAIL - -- I'd say he's as important as Kessel to this team. Great player.

2 FRANSON, CODY - -- frequent scratch, played weak opponents

3 STECKEL, DAVID -- not sure about his QoC. not a bad player, but not 3rd best. great at faceoffs.

4 KULEMIN, NIKOLAI -- had a really down year (from 30 goals to 7.. 7!) but very good defensively.

5 MACARTHUR, CLARKE -- also plays with Grabo and Kulie, so has similar stats. Good player..

6 CRABB, JOEY -- Helped by unrealistically high shooting %, but decent bottom 6 player. they let him go for nothing, though, so obviously not 6th best player.

7 KOMISAREK, MIKE -- scratched a bunch, as well. had some good stretches. some brutal mistakes.

8 KESSEL, PHIL -- 4 straight 30 goal seasons, 2 playing with total scrubs. not great defensively, but 23 PP points to help make up for his -10. should be 1st or maybe 2nd best player on the team (really wish him and grabo had chemistry..).

9 GUNNARSSON, CARL -- good player

10 LUPUL, JOFFREY -- plays well with Kessel

11 GARDINER, JAKE -- will be a good player

12 PHANEUF, DION -- overrated, but not bad. played the toughest minutes. saw a good comparison of Spezza's stats with Phaneuf on the ice vs without Phaneuf on the ice, and Phaneuf makes a big difference.

13 CONNOLLY, TIM -- invisible, but put up points with really inconsistent assignments/line pairings.

14 SCHENN, LUKE -- too many mistakes

15 LILES, JOHN-MICHAEL -- good powerplay guy

16 FRATTIN, MATT -- rookie, not good yet

17 BOZAK, TYLER -- has to be the defensive support for Kessel and Lupul. would be a good 3rd line centre.

18 LOMBARDI, MATTHEW -- this i can agree with.. not an effective player

My rankings, by position:

Forward (good):

80 - KESSEL, PHIL

75 - GRABOVSKI, MIKHAIL

72 - LUPUL, JOFFREY

68 - MACARTHUR, CLARKE

68 - KULEMIN, NIKOLAI

Forwards (ok):

65 - CONNOLLY, TIM (probably suffered from being shuffled around too much)

65 - BOZAK, TYLER

60 - STECKEL, DAVID (faceoff guy)

60 - CRABB, JOEY

Forwards (bad):

55 - FRATTIN, MATT (young rookie)

50 - BROWN, MIKE

50 - LOMBARDI, MATTHEW

Defense (good)

I found Corsi QoC worked pretty well for the Defense

75 - PHANEUF, DION (overrated, but still good)

70 - GUNNARSSON, CARL (underrated)

Defense (ok to bad)

GARDINER, JAKE (rookie)

KOMISAREK, MIKE

LILES, JOHN-MICHAEL (powerplay guy)

SCHENN, LUKE

FRANSON, CODY

In the end, every stat will have some weakness. Mixing stats like you were mentioning before might work.. i think once more people post their rankings from their favourite teams we can get a better idea, run a statistical correlation between various stats/formulas and forum user rankings.

Another idea is to rank them by salary, since it gives a good idea of true value. factor in when the contract was signed (older contracts are undervalued due to salary cap inflation).

Share this post


Link to post
Share on other sites

the more I look at leaguewide CorsiHART+ (when the score is close, adjusted for zone starts), the more I like it. there may be a few surprises near the top that one may expect to find in the middle and vice versa, but there doesn't seem to be anybody near the bottom that I feel is taking an unfair hit.

http://stats.hockeyanalysis.com/ratings.php?db=201112&sit=5v5close_f10&type=corsi&teamid=0&pos=skaters&minutes=50&disp=1

Share this post


Link to post
Share on other sites

yeah, it's not bad. if you can figure out a way to incorporate QoC (and maybe QoT), it could look very good.

Anyway, enough on this from me, hopefully the ball is rolling and some more team ratings will be posted : )

Share this post


Link to post
Share on other sites

WOW! Thanks for making this subforum! I was recently wishing for something like this. Also thanks for the advanced stats links. I have been wanting to take a look at that kinda stuff!

Share this post


Link to post
Share on other sites

the more I look at leaguewide CorsiHART+ (when the score is close, adjusted for zone starts), the more I like it. there may be a few surprises near the top that one may expect to find in the middle and vice versa, but there doesn't seem to be anybody near the bottom that I feel is taking an unfair hit.

http://stats.hockeya...nutes=50&disp=1

Just curious. Why do you prefer his CorHART+ over his HART+ based on his Goals formula?

Share this post


Link to post
Share on other sites

less outliers, I guess. conforms to my prejudices about more players. not by a lot, but enough. still seeking the best way to handle overalls. I also would consider tweaking shot accuracy to account for power and shot attempt distances. there's also the fact that shooting wide isn't considered shooting. this stuff is tricky. I'm very happy to have discussions going. please chime in if/when you have something.

Share this post


Link to post
Share on other sites

Just curious. Why do you prefer his CorHART+ over his HART+ based on his Goals formula?

"Cor" is short for Corsi, which is a measure of shot attempts per game (shots + missed shots + blocked shots. "Fenwick" is Shots + missed shots, ignoring blocked shots).

The theory goes that goals have a large bit of luck in them, while shots are a more consistent statistics, they are more "repeatable." A good player will generate more shots, not just for him, but also his linemates, even if he is unlucky that year and doesn't score much. The hole in this is that some players create BETTER shots than others, not necessarily more shots. But basically, when you are working with statistics, the more samples you have, the better. There are about 10 times more shots per game than goals, so you have 10 times more stats to work with.

I also would consider tweaking shot accuracy to account for power and shot attempt distances. there's also the fact that shooting wide isn't considered shooting. this stuff is tricky. I'm very happy to have discussions going. please chime in if/when you have something.

Definitely, I would like shot accuracy to factor in missed shots and shot distance. Missed shots is easy to find, since the NHL tracks it and lists it. Shot distance is more difficult -- it's tracked I think, but not tabulated clearly. And of course a slow but accurate shot is easier to stop than a faster shot of the same accuracy.

Share this post


Link to post
Share on other sites

New interesting stat to watch for.

TOI QualComp (Quality of Competition, based on the Time On Ice of opponents (assuming more ice time means better player))

The 'normal' QualComp stat is based on Corsi (+/- for shot attempts)

Original Results/Graphs: http://nhlnumbers.co...or-all-30-teams -- check out your team from last season to see how your players were used

Original Explanation: http://nhlnumbers.co...sed-on-ice-time

Blogspam: http://blogs.thescor...of-competition/

Blogspam: http://theleafsnatio...or-his-own-good

Takeaway:

The goal is to be able to say, when player X is on the ice, the players on the other team that he plays against are "this" good (or bad).

The assumption is that better players get more ice time.

It's a 2-dimentional graph.

Along the up/down, is the average TOI of the defensemen the player faced, so higher should mean the player is better offensively because the opponent coach puts his best D against that player.

Along the right/left, is the average TOI of the forwards the player faced, so higher (more to the right) should mean that player is better defensively because his own coach is playing that player against the opponent's better forwards.

Washington's graph is interesting to see because they mixed their lines up a lot.

For Toronto, you can see that the top 2 lines played together almost always, and oddly it's Lupul who faces the hardest average D competiton -- this is because the Toronto coaches give Kessel extra shifts against bad D. The 1st line played against hard D. The 2nd line played against average D, but faced similar forwards.

Boston is interesting because their top 2 lines play against almost identical competition, so there is no "first" line and "second" line, there is 1A and 1B. On the other hand, those 2 lines have massively different +/-.

One thing I either missed or isn't mentioned is how powerplay and penalty killing time is worked into this or not.

Share this post


Link to post
Share on other sites

http://stats.hockeya...Tp&sortdir=DESC

Using Vancouver as an example, because they're my favourite team.

So apparently a 7th defensemen is a 100?

And Tanev, a guy who played in the AHL is also a 100?

Meanwhile poor Henrik Sedin is beaten out by 4 players on the Canucks...

K...

Edler was also an Allstar.

He's a top 20 defensemen in the league.

Dude's near the bottom.

NO GOOD.

Edited by Stefan

Share this post


Link to post
Share on other sites

I can't say I like this new stat.

Share this post


Link to post
Share on other sites

I can't say I like this new stat.

In case there's any misunderstanding -- it's not a measure of how good the players are, it's a measure of how good the players they play *against* are.

It's not saying both top lines on Boston are equally good. It's saying they play against equivalent competition, so their stats can be directly compared to each other. The Bergeron line is clearly the better line (way higher +/- and more points for Bergeron and Seguin).

Share this post


Link to post
Share on other sites

I got what it is, but not why it is. your Boston lines comparison helped. I can't say I learned much or would know what to do with this stat, but I understand it.

Edited by trudatman

Share this post


Link to post
Share on other sites

Oh, when comparing things like points, +/-, and Corsi, it's important to know what kind of competition the players play against. That's the problem with that HART+ system, it has no concept of opponent skills, so it rates bad players highly because they play weaker opponents in their time on the ice.

Anyway, I thought it was cool to see how different players are used, and it's a clever usage of the simple rule-of-thumb that the better a player is, the more ice time he gets.

Share this post


Link to post
Share on other sites

I just saw the overall ratings for EA's current atrocity. seems like their ratings "guru" could use some lessons on the finer points of this art, because, for examples, Phil Kessel isn't awesome and Chris Kelly doesn't suck. I bet they are proud of themselves.

Share this post


Link to post
Share on other sites

yeah their ratings are ridiculous.

Leafs centremen ratings:

Grabovski 82

Connolly 82

Lombardi 82

Bozak 80

Steckel (i forget, 78 or 76 maybe)

Lombardi was bad last year. Like barely NHL-worthy bad. And they rated him the same as Grabovski, who is one of the better two-way centres in the league.

82 points on the Leafs is pretty awesome, even if Kessel sucks defensively :P

Share this post


Link to post
Share on other sites

Bump... people might want to start rating their favourite teams now that the season is on and new 2013 ROMs might be coming out.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Who's Online   0 Members, 0 Anonymous, 8 Guests (See full list)

    There are no registered users currently online