Some thoughts on ratings

[CoachMac]

Naples39 put out another couple great roms and it got me thinking about ratings and back at making some 93 roms of my own. Plus Trudatman is always busy trying to build ratings. Slapshot had a great template he used for ratings that was purely stat based and only really included the overall.

So I got thinking how to improve the rating system for these 20 year old video games.

Physical attributes that I think EA got mostly right in 93/94.

Speed
Agility

Shot Power

Stickhandling

Behavior they got right.

Checking

Shooting Bias (Roughness)

Aggression

Fighting

Weight

Means to much in the game and did you know almost every player gained 8 to 16 pounds from 93 to 94

Endurance

Means nothing when playing with Line Changes OFF as most people do.

Here are the 4 ratings that need improvement.

Defensive Awareness is based on a players +/-. This is bunk as many of the best defensive players had poor or avg +/- from playing against other teams top lines. This attribute should be a team attribute with bonus points given to Selke award type players (Dirk Graham, Guy Carbonneau, Federov Etc...). Players on teams with low GAA should have basically high DA and players on teams with poor GAA should have low DA.

Offensive Awareness Should be based on Points Scored and it sort of is but to many players have high awareness.

These 2 attributes should have more range than they do especially in 94.

Then players like Gretzky would stand out more as awareness is what made him great.

Shot Acc is based on Shooting %, this is horrible. Many goons had high shooting %'s because they only put it on goal twice and one went in. This should be GOAL SCORING and based strictly on that.

Pass Acc same as above though I don't know what it is based on. This should be ASSIST.

[trudatman]

I have to disagree on the shooting percentage and passing skill ratings. the goon with a great shot is still a nearly useless goon. his great shot matters not from the bench or the defensive zone. I'm a stats guy.

[kingraph]

I think that is what Mac was saying...a few goons have high shot accuracy, but that is incorrect.

Coming up with ratings for hundreds of individual players is one of the hardest tasks of making a new ROM. I always wonder how EA comes up with their ratings. Assume they have a large team dedicated to doing just that.

[trudatman]

I think I understand what he means, but I don't think I understand what you mean. how is it incorrect? a percentage is a percentage. if a goon scores on 19 out of 100 shots, that's 19% and I'd give him a great shooting rating. he's still going to have trouble scoring from the bench/pressbox/fourth line. if he can't skate and/or possess the puck well enough to get near the net, his great career/season shooting percentage means not much. if he's rated AGI 35 SPE 55 OFF 25 DEF 65 POW 65 CHE 85 ACC 100 CON 35 END 35 BIA 35 PAS 35 REP 100, he isn't going to lead the team in scoring even if you put him on the first line right wing unless you only attempt shots with that one goon.

I doubt EA has too many people working stats, as you really only need one or two people to fully understand and implement the formulae. if I didn't have other commitments and it was my full-time job, I could rate the entire league in a week of five eight-hour shifts, easily. I really like my methods for creating stats, but I understand the issues with overrated scrubs. if it didn't make the process at least twice as hard, I'd implement a quality of opposition component to my overall ratings.

as of right now, I use stats and fan feedback to come up with most of the ratings and plus/minus divided by total time on ice to generate overall ratings. if a player breaks a threshold of a certain number of minutes played in a season (about 25 players do), I give them an overall boost of about one. if they reach another level (maybe five do), I give them another overall point. I'll likely tweak that in the future to give most first-liners slightly bigger boosts and Chara/Karlsson/Kesler types of minute eaters even more, but I don't like the idea of punishing a dude who happens to play on a fourth line just because he rarely faces Toews/Stamkos; he is facing NHL players and if his team is scoring twenty more goals than they are letting in, per season, when he's out there, he's getting a pretty good rating from me, scrubby goon or not.

I like that people want to further explore how ratings are created. as much as I believe in my system and would be willing to teach it, I understand that people don't like seeing Adam McQuaid rated better than Ovechkin in overall ratings. the thing is, when you take control of my Ovechkin, he has the skills needed to score much more frequently than McQuaid. when left to his own artificial intelligence devices, he'll skate lazily and defend poorly. I insist that it works, but I get why people don't like seeing jacked-up scrubs and muted (-looking) stars. when you are controlling Marty St. Louis, he's going to flow a lot better than these "overrated" goons.

great conversation. I love this stuff. a dude around here wanted me to teach him my formulae, but he didn't want to talk on the phone to learn it, so it didn't happen. the gist is in the posts, but the details are available verbally to anybody willing to listen. there is a lot going on when ROMs are made. I don't like the shortcut of first-liners and 85, second 65, third 55, fourth 45. the personality of the player is lost. my guys very much seem to play like their real counterparts. I'd love to read more about how others handle this stuff.

thanks, guys.

[jackandjose]

Goons aren't out there to shoot. Most of there shots are probably high percentage like rebounds I'm guessing, so there shooting percentage will be inflated a bit.

[smozoma]

For shooting accuracy, I'd combine their shooting percentage with their total number of shots-per-game, and maybe also factor in their shot power. The idea being that guys who shoot a lot AND score a lot are probably very accurate (or have a quick release or whatever), while guys who don't take a lot of shots and score a lot probably just take high-percentage shots. Then factor in shot power because if a guy has a hard shot he doesn't need to be quite as accurate.

[trudatman]

agreed. I would do that if this was a job and not a hobby.

[smozoma]

Ideally, shooting distance would be factored in, too, so guys like Andreychuk who just bang it in don't get a high accuracy. Some interesting shot data here:

http://www.behindthenet.ca/nhl_shot_statistics.php?ds=13&s=9&f1=2013_s&f2=5v5&f7=40-&c=0+1+3+5+4+7+8+9+10+11+12+13

http://somekindofninja.com/nhl/index.php?season=Regular&year=2013-2014&shots=For&team=&ice_player_name=&withPlayer=On+Ice&player_name=Phil+Kessel&goalie_name=&event=Shots+and+Goals&game=Home+and+Away&strength=All&time=Regulation&search=Search

I wish they gave separate shooting distance averages for goals and saved shots though.