When people post power rankings (1 Viewer)

[Chase Stuart]

As far as I can tell, there are at least four different systems you could use:

1) A purely retrodictive system, based solely on previous wins and losses. Almost no one uses this.

2) A retrodictive system based mostly on wins and losses, with appropriate judgments made for SOS, margin of victory and how the teams played.

3) A predictive system, based purely on how people think teams will do in the future.

4) A combination predictive/retrodictive system.

I think most people use #3, or a combination of #3 and #4. What I don't get is how do you deal with the following issues:

A) Peyton Manning goes down after the Colts go 10-0, and he's out for the season. Where are the Colts now ranked? Where are the Colts ranked after losing their first post-Manning game?

B) You believe Team B is better than Team A, but believe Team A is more likely to win the Super Bowl. Which should be ranked higher? Followup question: You know Team B is better than Team A, but Team A plays in a weaker conference and thus is more likely to win the SB. Which should be ranked higher?

C) How do we judge someone's rankings? Presumably some power rankings are better than others; how are they judged? I'd imagine they'd be judged by end-of season results (we can come up with a scoring system later).

D) How should SOS be factored into EOY results? Let's say I predict Team A to be better than Team B, but Maurile predicts Team B to be better than Team A. It turns out that Team A played the hardest SOS in the league and went 6-10, while Team B played the easiest SOS in the league and went 8-8. Both of us agree that Team B was better in 2007, after adjusting for SOS. Who should be considered to have "been correct"?

With the hope of answering these questions, I'm going to run a power rankings contest. But I want to be extremely specific on the rules before we start.

 

[SeniorVBDStudent]

As far as I can tell, there are at least four different systems you could use:

1) A purely retrodictive system, based solely on previous wins and losses. Almost no one uses this.

2) A retrodictive system based mostly on wins and losses, with appropriate judgments made for SOS, margin of victory and how the teams played.

3) A predictive system, based purely on how people think teams will do in the future.

4) A combination predictive/retrodictive system.

I think most people use #3, or a combination of #3 and #4. What I don't get is how do you deal with the following issues:

A) Peyton Manning goes down after the Colts go 10-0, and he's out for the season. Where are the Colts now ranked? Where are the Colts ranked after losing their first post-Manning game?

Rankings exclude injury effect. Ranking is based on the team you have today. If Manning was gone for the year, Indy would be #9-12 ish.

B) You believe Team B is better than Team A, but believe Team A is more likely to win the Super Bowl. Which should be ranked higher? Followup question: You know Team B is better than Team A, but Team A plays in a weaker conference and thus is more likely to win the SB. Which should be ranked higher?



You're predicting who will win the superbowl, so you may see 3 AFC teams ranked ahead of 1 NFC team, even though only 1 AFC team will go.

C) How do we judge someone's rankings? Presumably some power rankings are better than others; how are they judged? I'd imagine they'd be judged by end-of season results (we can come up with a scoring system later).

How teams finished in the division. Equal finishes are determined by consensus. So ultimately there is no final judgement of rankings down to that level of detail unless a specific group is judging based on specific criteria. In your case, I would poll the group to determine the "winner".

D) How should SOS be factored into EOY results? Let's say I predict Team A to be better than Team B, but Maurile predicts Team B to be better than Team A. It turns out that Team A played the hardest SOS in the league and went 6-10, while Team B played the easiest SOS in the league and went 8-8. Both of us agree that Team B was better in 2007, after adjusting for SOS. Who should be considered to have "been correct"?

Closest to picking exact record. SOS is not relevant.

With the hope of answering these questions, I'm going to run a power rankings contest. But I want to be extremely specific on the rules before we start.

 

[Chase Stuart]

Rankings exclude injury effect. Ranking is based on the team you have today. If Manning was gone for the year, Indy would be #9-12 ish.

So you're in the ballpark of record isn't important, it's the team you have today? Because in my scenario Indy is 10-0 but you're ranking them 9-12ish. I don't think that's wrong, but I want to be clear. And if a 10-0 team can be ranked outside the top 8, it seems record is a very small part of the equation, but rather, it's your subjective opinion about the team. That's predictive and not retrodictive, which I think is correct but I'm not sure is the way people actually do their rankings.

You're predicting who will win the superbowl, so you may see 3 AFC teams ranked ahead of 1 NFC team, even though only 1 AFC team will go.

Another ambiguous one, IMO.For example, let's say we decide the AFC winner has a 60% chance of beating the NFC winner. We decide that the Colts, Chargers, Patriots and Steelers are the four best teams in the NFL. We think the NFC stinks outside of Chicago. We give the Patriots a 50% chance of winning the AFC, the Colts a 30% chance, and the Chargers and Steelers 8% chances. We give the Bears a 60% chance of winning the NFC.

This gives us the following SB odds: Patriots - 30%; Chicago - 24%; Indy - 18%; SD/Pit - 5% each. In that system, we'd have the Bears ranked 2nd, because they're more likely to win the SB than the Colts. But we agree that the Colts (and even the Steelers and Chargers) are better. I'm confused by your statement, since you said we may rank 3 AFC teams ahead of 1 NFC team. It seems like you're arguing for the other way around.

Further, another question arises. Let's say you think Team X is really hit or miss; they've got a 50% chance of finishing in the top 4, and a 50% chance of finishing at 15th. Another team is very likely to finish seventh. Now the first team is more likely to win the SB, but the second team is more likely to be better. Which team should be ranked higher? This seems particularly difficult to judge after the fact, as well.

Closest to picking exact record. SOS is not relevant.

I may be wrong, but I highly doubt most people look to upcoming SOS when doing power rankings. Do you think they do? I don't recall seeing something like "I've got the Seahawks as my 15th best team and the Bills as my 18th best team. Yeah, the Bills are better, but they've got a tougher schedule and the difference in SOS is greater than the difference in talent of the two teams, so I've ranked Seattle higher." It seems peculiar to do your rankings in this way.

 

[Chase Stuart]

tap

 

[Avery]

As far as I can tell, there are at least four different systems you could use:

1) A purely retrodictive system, based solely on previous wins and losses. Almost no one uses this.

2) A retrodictive system based mostly on wins and losses, with appropriate judgments made for SOS, margin of victory and how the teams played.

3) A predictive system, based purely on how people think teams will do in the future.

4) A combination predictive/retrodictive system.

I think most people use #3, or a combination of #3 and #4.

I don't think it's possible to use solely #3, "A predictive system, based purely on how people think teams will do in the future", for whether we do it with conscious intent or not, we all use past and recent success as, at least, a partial measure of future success.As an obvious example, we know Payton Manning makes the Colts better for he has proven to be a great QB in the past. Without looking at all at past success, Payton Manning is just an ugly dude with a weird name.

 

[Random]

tap

How do you determine SOS? Do you do this before or after the rankings? Since they each feed the other its kind of a circular reference.Last year I put out a few Power rankings but I believe they've since been purged.Here's how it looked in wk 6 (first week I did it)Rank Team Margin 1 Bears 0.3345 2 Colts 0.9107 3 Saints 0.0477 4 Panthers 0.4258 5 Seahawks 0.2657 6 Broncos 0.2743 7 Giants 0.1709 8 Patriots 0.1241 9 Rams 0.3489 10 Vikings 0.0012 11 Jaguars 0.1932 12 Falcons 0.1499 13 Eagles 0.7543 14 Ravens 0.0754 15 Bengals 0.0058 16 Chargers 0.2522 17 Redskins 0.2042 18 Jets 0.0138 19 Cowboys 0.0905 20 Bills 0.3182 21 49ers 0.0303 22 Chiefs 0.2626 23 Steelers 0.0828 24 Buccaneers 0.1412 25 Lions 0.1497 26 Cardinals 0.0059 27 Titans 0.2972 28 Packers 0.0606 29 Dolphins 0.1515 30 Texans 0.3939 31 Browns - 32 Raiders - And wk 13 (last week I did)Rank Team Margin 1 Colts 5.92 2 Patriots 0.16 3 Chargers 0.28 4 Ravens 0.17 5 Bears 3.56 6 Jaguars 0.10 7 Cowboys 0.10 8 Bengals 1.82 9 Broncos 0.48 10 Chiefs 0.51 11 Saints 0.45 12 Jets 0.59 13 Seahawks 1.49 14 Panthers 0.50 15 Titans 0.40 16 Eagles 0.06 17 Giants 0.46 18 Falcons 0.22 19 Dolphins 1.01 20 Steelers 0.91 21 Bills 0.53 22 Redskins 1.03 23 Browns 0.03 24 Rams 0.01 25 Vikings 0.36 26 49ers 0.08 27 Texans 1.38 28 Buccaneers 1.57 29 Packers 1.47 30 Cardinals 0.27 31 Lions 1.56 32 Raiders

 

[SeniorVBDStudent]

tap

It occurred to me after posting this, but away from a PC, that my comments were mostly ambiguous. Sorry. I'll clarify this evening...

 

[BroadwayG]

When I think power ranks, I think odds to win the Super Bowl.

 

[Chase Stuart]

As far as I can tell, there are at least four different systems you could use:

1) A purely retrodictive system, based solely on previous wins and losses. Almost no one uses this.

2) A retrodictive system based mostly on wins and losses, with appropriate judgments made for SOS, margin of victory and how the teams played.

3) A predictive system, based purely on how people think teams will do in the future.

4) A combination predictive/retrodictive system.

I think most people use #3, or a combination of #3 and #4.

I don't think it's possible to use solely #3, "A predictive system, based purely on how people think teams will do in the future", for whether we do it with conscious intent or not, we all use past and recent success as, at least, a partial measure of future success.As an obvious example, we know Payton Manning makes the Colts better for he has proven to be a great QB in the past. Without looking at all at past success, Payton Manning is just an ugly dude with a weird name.

Allow me to clarify.Team A is 5-0 when SuperStarX plays, and 0-5 when SuperStarX doesn't play. SuperStarX is now back in the lineup, after say, serving a five game suspension. If you're ranking teams based solely on how you think they'll do in the future, an argument could be made for putting Team A atop your power rankings. That's what I mean by based purely on how you think teams will do in the future (i.e., season to date record isn't relevant).

 

[Chase Stuart]

tap

How do you determine SOS? Do you do this before or after the rankings? Since they each feed the other its kind of a circular reference.

What do you mean how do I determine SOS? For what purpose? A little confused here.