Professor Frink Wants YOU!
by tom slick on Nov 4, 2009 3:31 PM EST
Email
Print
via i38.tinypic.com
Actually, I need your opinions, my good people.
Despite having been happy with my old rating algorithm for some years, I'm not quite as happy these days. I've discovered that when you publish something for all to see, you have a lot less margin for error. As much as I like the simplicity of my spreadsheet, I think I can do a little better.
My main concern is that the existing method overly penalizes winning teams like the Saints, who are outstanding on one side of the ball but mediocre on the other. To rank them in 7th place is IMO something that should probably be addressed. Last year, for example, the Steelers won the Super Bowl with a subpar offense and a great defense. So far this year, 3 of the 4 winningest teams have a below average defense (or offense). If this trend is the new reality in the NFL, I'd rather change than stubbornly stick with something that's worked in the past, but not quite as well now.
At the risk of getting too cute with a simple process, I'm thinking of adding a couple of new variables to the formula. Winning percentage is one measure of a team's quality that is undeniably relevant, as is Point Differential. Rather than weighing offense and defense at 50% apiece, my new mix would be:
Offense (curved, as before), 25%
Defense (curved), 25%
Point Differential per Game (curved), 25%
Winning Percentage, 25%.
Here are what the Week 8 rankings would look like using the new algorithm:
(I omitted Win% from the table due to space and readabilty considerations, but it does figure into the final ratings.)
Though hardly perfect by any means, this version seems to "smell" somewhat better, at least to me.
Since I publish this for your amusement, kind readers, I would appreciate YOUR feedback on this momentous issue. Do the results seem more realistic this way? Does the added complexity strike anyone as a negative? If so, is that negative outweighed by better results?
Like Linus waiting for the Great Pumpkin, I look forward to any constructive remarks youse guys have to offer. Thanks, everyone!
This FanPost was written by one of The Falcoholic's talented readers. It does not necessarily reflect the views of The Falcoholic.
31 comments
|
0 recs |
Do you like this story?
Comments
I say
What makes your Power Rankings interesting is when a few teams stick out from popular perception. I like the “simplicity” of how it is now… But if you do decide on changing it, definitely scrap Win %. Power Rankings from sportswriters are usually lined up pretty closely with Win % – we don’t need another one of those. Especially early, a team may have a great Win % but are barely beating bad teams. Change the formula too much, and it won’t be able to find these possible over-rated teams (Broncos?) or possible under-rated teams (Packers, Eagles, Ravens, etc).
"Statistics are like bikinis. What they reveal is suggestive, but what they conceal is vital." - Aaron Levenstein
Agreed
If you were to change anything at all, it would be what I suggested initially and something you’ve recently posted, and that’s point differential. Simply add it to the mix and you should still be able to see those teams that have a good record but are barely winning (overrated), as well as those who might not have a stellar record but are barely losing (underrated). Keep winning % out of it entirely because it can be deceiving.
"This is America, if we can’t self-righteously look down on others and blame them for our faults, the commies win."-----Cormican on Bleeding Green Nation w/r/t fans overreacting to the Eagles signing of Michael Vick
Thank you both
for your perspectives, I do appreciate it. I agree with you that Point Differential (PD) is a much better component to add to the mix, relative to Win%. PD has a decades-long history of correlating very highly to Win% and IMO is the best single “simple” stat out there.
The core problem is just how to balance out the current bias against “one-sided” teams like the Saints, Vikings & Broncos (all O, no D or vice versa). Though that bias has a certain intuitive appeal, I’m increasingly convinced that it’s excessive. Fixing it by adding only PD to the mix hasn’t yet yielded results I’m comfortable with (though the whole thing is too much like reading ink blots or tea leaves).
Though I’m not totally on board with using Win% either, it has its pluses. It’s a halfway decent, simple indicator of team quality. Provided it’s not weighted too heavily, I don’t think it will screw up the ratings to the point that they begin to look like one of those Sportswriter Poll Composite Power Rankings that march in virtual lockstep with the W-L records. You’d have to weigh Win% by at least 50-60% to start doing that. I’d rather avoid using it if at all possible, though.
Thanks again, guys!
i agree about win% being weighted too heavily
most of the time. although, i’m a bit curious about it’s impact at a low weight win% rate of 10% or so. overall, i haven’t missed it though and i appreciate the simplicity your system.
Some people think football is a matter of life and death. I don't like that attitude. I can assure them it is much more serious than that. - Bill Shankly
by armchair quarterback on Nov 4, 2009 8:06 PM EST up reply actions
Hmm
I’m not saying that you are, but it reads like you are trying to force the numbers to spit out a predetermined result. The best thing about your formula thus far has been the simple fact that you haven’t tried to force it into any set result, instead just accepting the end data as it is. If you keep tweaking the formula just to arrive at a result that you like or feel comfortable with, then really you’re just turning it into an entirely subjective thing.
The only time these things need tweaking are when they aren’t right to begin with or are weighing something too heavily or not enough. Your original formula was sound enough to begin with that just adding PD should be plenty to see a more accurate data set for projecting or correlating with real-world results.
My take on your end results were that this method shows how win-loss records do not tell the whole story. You can see that while the Saints have an undefeated record, they don’t have the best defense which means they probably won’t be able to maintain that same level of success against teams that are more balanced such as the Colts or the Patriots.
The last thing I would want to see is for your formula to become watered down and convoluted just for the sake of change or to pacify a few commentors, myself included. It’s yours man, it’s completely logical, and the result is a product of sound work, so thank you, regardless of what your final changes end up being.
"This is America, if we can’t self-righteously look down on others and blame them for our faults, the commies win."-----Cormican on Bleeding Green Nation w/r/t fans overreacting to the Eagles signing of Michael Vick
I agree with what Jesse touches on here
If you tweak the formula to try to make it “line up” more with what your head is telling you (and the W-L records show), you might hurt it in other areas. That could just start a vicious cycle of you chasing your tail to make the Rankings pass the eyeball test.
"Statistics are like bikinis. What they reveal is suggestive, but what they conceal is vital." - Aaron Levenstein
You guys are my reality check. Thanks.
One of my many personality flaws is that I tend to expect perfection in areas that I have no right to do so. (Just like my dad, imagine!) My brain knows damn good and well that this algorithm is not going to be very reflective of wins, but my gut wants to see perfection, regardless of logic and intellect.
I see two choices. One is to leave things alone, accept the algorithm’s limitations, and learn to suppress that gut instinct. The other is to add variables to the formula and try to make it better (though it will never be perfect, as we know). Looking back over the years, I’ve tried the more complicated option many times, and I always end up going back to Simple.
Professor Frink appreciates your taking time out to share your collective wisdom, gentlemen. A tip o’ the old Falcons cap to you!
As an aside
And again, not that this is happening here, just that I don’t want it to, but I happen to be running into something along these lines at work. I have a bunch of really good programmers all sitting together in a room constantly writing code. Unfortunately, since they are all very experienced they aren’t very receptive to being told they aren’t coding correctly because they don’t understand the data. They review their code and say that nothing is wrong with the code, and in a sense they are correct because the functions are written properly, but what they don’t see is that they are chosing the wrong functions and therefore are getting results that don’t make sense. Then they blame the data and make changes to the data to fit their code so they can force the data to return an answer that makes more sense to them.
What they don’t see is that they are getting exactly what they asked for from their functions, just that they are asking the wrong questions. When they change the data all they are doing is forcing the results to show an answer they already have determined to be correct. It’s highly frustrating and causes a lot of rework. Makes me wonder if all programmers think their work is so infalliable.
"This is America, if we can’t self-righteously look down on others and blame them for our faults, the commies win."-----Cormican on Bleeding Green Nation w/r/t fans overreacting to the Eagles signing of Michael Vick
Following this discussion with interest...
Have to say, as a Saints fan, I think we rely too much on screwball interceptions returned for TD’s to win games or expand our leads. Is this a function of a good team or luck? Not sure, but to say we’re the 7th best team after only playing 7 games and 1 divisional competitor would not be a complete slap in my face…yet. After all we do have 8 turnovers in the last two games. We’ve just been fortunate (good enough) to also induce our opponent to be equally generous.
What about last years’ results? The Cardinals clearly surprised everyone on their way to the Super Bowl yet never cracked the top 10 of most “power polls” I saw. Where were they hovering on your list during various parts of the season?
One other question. Do you remember when Devin Hester of the Bears was routinely returning punts for TD’s? How do you account for a teams special teams in your equations or are they ignored? A modern example would be to compare say the Packers who have an off-the-street, slow, but steady Ahman Green returning kicks vs Miami who has red-hot super-fast dangerous Ted Ginn. My instinct would be that Miami should be rewarded some type of “power bonus” since Ginn was responsible for 40% of the Dolphins points this last week while teams with struggling, impotent units get deductions…
Saints Fan! GET him!
I keed, welcome. Strange how even though we all know you can’t explain everything with numbers, we keep trying anyway. I know I do.
I get the impression you’re not too familiar with my algorithm. If not, you might want to read the RATINGS COMPUTATION part of my last post (link here) to get the details on how the numbers are calculated. Though I am thinking of tweaking the math, I do want to keep things relatively simple and not get into quite the level of detail you’re describing.
The Cardinals were 18th in my 2008 EOY rankings. That team shall surely live in infamy as the most undeserving NFC Champ ever, with that wretched defense.
Just keep doing what you do
Honestly, I think people enjoy them precisely because they’re simple and yet effective. Like Selsun Blue.
Sign up for a free account today to join the discussion about all things Atlanta Falcons!
Simplicity seems to count for a lot in this thread
I hope I’m not forgetting my own screed about not expecting miracles from a very simple algorithm. Thanks.
Simplicity is a good thing at times
I’ve really just enjoyed having you guys with running weekly features. It’s a lot of fun for me to read, and it’s a testament to how the community has grown. Whether you change it or not, I appreciate the work.
Sign up for a free account today to join the discussion about all things Atlanta Falcons!
ouch
a little flakey are we???
Life is a garden. Dig it!
by Hardcore Falcon on Nov 5, 2009 10:39 AM EST reply actions
como se dice???
say what? I am sure you were not talkin to me.
Life is a garden. Dig it!
by Hardcore Falcon on Nov 5, 2009 1:01 PM EST up reply actions
Misunderstanding?
I thought you were referring to the Fanpost topic, judging by your first comment not being a reply to any other thread. If I was mistaken, my preemptive apologies.
Now that I look back, I think you meant to comment as a reply to someone else, not as a new thread. No hard feelings.
lol
yeah I meant to reply to the dandruff shampoo reference but missed it.
Life is a garden. Dig it!
by Hardcore Falcon on Nov 5, 2009 2:01 PM EST up reply actions
Realy Cool!
I like all those statistics things, I think that’s one of the coolest things in mathematics.
I also made my own “Power Ranking”, almost like yours, but with the difference that I also put the TD/game average. I think my one is realy simple and a little bit efficient, just like yours.
I forgot.
Here it is (notice that some titles are in portuguese, well, “time” = team and “divisão” = division; but that’s a kind of obvious):
Good clean fun!
Fans like yourself may enjoy the links to many different Power Rankings at sister site Blogging the Boys. Both types of rankings are included, the poll composites as well as the stat algorithms.
Thanks for the link.
I’ve enjoyed to see some of the rankings, even more the ones that only uses the math, and I compared some w/ my one. I put the Ranking in a brazilian forum about footbal, and the people liked, because we don’t see here ( = brazilian blogs/websites about football) any kind of power ranking that’s based only in the stats. tom slick, do you know where can I find an easy website to get some stats and put them into MS Excel? I was thinking about making another ranking, using also the yd/g average, in both off. and def.
What I do probably isn't the easiest way
but I’ve used NFL.com and copied my stats to the Windows clipboard. Then open Excel and do a Paste Special, then select Text (strips away the HTML code). Save the file as a .prn type then close the file. Re-open the same file as space delimited text, and you’re ready to go.
Sorry but that’s the first method I figured out that worked, and I stuck with it. It’s easy once you’ve done it a few times. Good luck!
If you are going to add anything, I would go for special teams.
Like average starting position for the offense and defense because of punts, kickoffs, field goal %, things of that nature.
Hard to do that while remaining simple,
though I do see the intuitive sense in what you’re saying. Special Teams has a lot of components that are probably really hard to boil down into a single number. Agreed that it’s not the most elegant solution for Special Teams play to be reflected indirectly in Offense and Defense ratings.
When using a limited number of variables, I personally prefer stats that have a higher correlation to Win%, such as Point Differential. You might find this chart of correlation coefficients interesting if you’re into that kind of thing.
If you’re over 40 or so, I bet you fondly remember the old Redskins Coach George Allen. He was an absolute maniac about Special Teams, his philosophy was that they were good for 3 games a year. The last study I read put that number at about one game a year, though that’s still nothing to sneeze at.
That correlation chart is awesome.
Do you know what year(s) does it covers?
"Statistics are like bikinis. What they reveal is suggestive, but what they conceal is vital." - Aaron Levenstein
If memory serves
I found it as a link in this article from Advanced NFL Stats, though I don’t see the link now. The article’s 2.5 years old, so the data’s got to be from 2006 or earlier.
I may be a little late to the game
but here’s another thought starter for your tables: mgrex03 over at StampedeBlue.com has a long running series called Finding The Winning Numbers, in which he analyses a whole boatload of stats for their ability to predict wins. Perhaps you’ll find something there that makes sense.
I already tried the TO ratio in your table but that didn’t work too well, as it ranked the Eagles ahead of the Cowboys, and that is obviously a Big No No :-)
by One.Cool.Customer on Nov 10, 2009 9:45 AM EST reply actions
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Certain photos copyright © 2012 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
