Power Ratings question

[Dingle]

Hi Guys

Never posted in this bit, but it looks ike a good place.

Especially Blogguy, I have been following the NFL ratings threads, nice work!

Question is, How do you determine the range of the ratings. I have been making my own for Basketball, and this question baffles me a little, I have a method but not sure how reliable it is. I guess you have experience with the NFL starting to lose its any given sunday tag, with the lack of parity. I assume the range has grown over this period.

Is there a certain way to calculate this?

Thanks
Dingle

[blogguy]

Wow Dingle...I figured I'd type a quick answer before hitting the sack...but I couldn't think of one!

You know...a lot of it's just feel. For me, it's kind of a mix of what the stats are saying, what the oddsmakers are saying, and what the history books are saying.

And, it's also different from sport to sport. College football isn't linear. There's a level where special teams and defensive points kick in because of the differences in talent and depth. For me, if I see a difference of 20 points or more in a college game, I multiply it by 1.5 I noticed about 20 years ago that you couldn't really create college ratings that reflect reality.

If you make the Minnesota's and Louisville's 60 better than the Temples because they love to run up the score...that spreads everything out way too much. You get teams who are a field goal different grading out 7-10 points different. If you scrunch up everything to reflect how most teams compare to each other...it understates the blowout likelihood at the extremes.

So...in college football I tend to be conservative with the numbers...but I multipy by 1.5 for any differences of 20 or more. That presents a paradox that if I get a grading of 19...it's 19. But if I get a grading of 20, it's 30. I live with the paradox because it's so darned accurate it's ridiculous. That's the level where those "bonus" points usually kick in.

My current range in the colleges is probably about 35 points from top to bottom...but most everybody's in a hunk that represents a TD worse than the best teams to 17 points worse than the best teams.

In the NFL, I think I've got it about 17 points top to bottom. That's more than it's been in the past because parity is dying.

In the NBA, I like to use the actual won-lost margins once everyone's played 10 games. Then I tweak based on common sense.

Wish I had a way to calculate it for you. I didn't realize how much I do is based on feel these days when it comes to power ratings. I just use them as a starting point rather than the end-all. I'm much more of a stat guy than a ratings guy anyway. Maybe if I put more weight on them I'd have a fancier formula. It's really just a starting point. So much of what I found to work has more to do with situational things than power ratings (matchups, styles of play, tendencies to be conservative or bully, etc...).

Hope that helps. I'm not always a fan of how the oddsmakers do things. But, when in doubt, using their range won't put you too far off. You can usually deduce their range by looking at what happens when top teams play bottom teams. Last week it looked like they had the NFL best about 10 points better than the worst on neutral fields...which turned out to be way too low in Oakland/Baltimore, Houston/Indy, Tenn/San Diego, and Cleveland/Cincy. I think the numbers from Brett or I in that other thread are closer to reality for now. Soon some powerhouses will pull back in the blow off games though, and won't win by the margins they could.

Always something you have to deal with. That's why numbers should just be a starting point to me. A part of a messy complicated, picture...lol

Thanks for your kind words...maybe Brett and others will chime in here if they still visit the TANK every so often. Good luck!

[ovrund]

great anwer, bg....I learn something from you every time a read your name...thanks

[Cincy]

An easy starting point for the power rating range in pro sports is to simply take the team with the best point differential per game and subtract the team with the worst point differential per game (assuming each team has played a significant number of games).

This will give you a decent approximation of the power rating range from top to bottom in most pro sports. This approach doesn't work when teams have significantly different strength of schedules (so it won't work for college sports).

As an example from the recently completed NFL season you can lookup the team standings on ESPN and see San Diego had the best point differential of 492-303 = 189/16 games = +11.8 point dif per game.

Oakland had the worst point differential of 168-332 = -164/16 games = -10.2

So the power rating range is approximately 11.8 - (-10.2) = 22 points

In the NFL, teams have a tendency not to run up the score. So if OAK played at SD on a neutral field I would scale the 22 power rating difference back to something like 18 points when determining a pointspread.

You can also use a power rating website like Sagarin.com to get the approximate range for power ratings. Sagarin had the best team with a 32 power rating and the worst team with a 8 power rating so this has a power rating range of 24.

Also, it is important not to base you power rating range on only a few games. For example, it is likely at least one NFL team will be 3-0 and have a very high point differential per game of say +17 points while another team may be 0-3 with a -15 point dif per game. You can't assume the range is 32 because too few games were played.