I think this is an interesting perspective to look at the season from. I created a power ranking of all the teams starting from week 1 and modified it to adjust for wins/losses from week to week. Week 17 rankings (strongest to weakest) 1. Broncos 1113 2. Texans 1104 3. Falcons 1091 4. 49ers 1077 5. Colts 1073 6. Patriots 1072 7. Seahawks 1069 8. Cowboys 1058 9. Bears 1045.4 10. Redskins 1044.7 11. Bengals 1041 12. Panthers 1027 13. Chargers 1026 14. Ravens 1023 (ranking moved due to math error by me in week 17) 15. Vikings 1019 16. Packers 1019 17. Giants 1009 18. Rams 990 19. Saints 985 20. Steelers 979 21. Dolphans 974 22. Buccaneers 968 23. Bills 946 24. Jets 942 25. Browns 929 26. Titans 912 27. Cardinals 911 28. Lions 899 29. Raiders 889 30. Jaguars 883 31. Eagles 870 32. Chiefs 851
Yeah I agree the Ravens look out of place but its how the math worked out. The formula used is called ELO and its an accepted ranking system. Its not necessarily saying the Ravens are the 25th most talented but more so based on SOS vrs wins/losses they ranked 25th.
So, Houston gets it's *** kicked by Green Bay, New England and Minnesota, but finishes ahead of them?
I agree with Broncos at #1 but Cowboys at 8 and Panthers at 12 is just plain crazy talk. I see what you did there but this list is not 100% accurate at all due to the wild placings of some teams.
If I were to objectively look at the 31 other teams and ask myself if we could beat them in a best of 7 type series (since anybody can beat anybody on any given day) I'm not sure there are 10 teams i could say we could. I should start a thread on this. Naaaa.
Accepted by whom? All these rating systems are BS. It all comes now to wins and losses and that is all. You can take statistics and use them anyway you want to to prove your own conclusions. Each team played 16 games this year and the 12 teams with the best won-lost record made the playoffs. It is as simple as that. It appears to me that there are some people on here who would rather explain a teams performance through charts and graphs, and they fail to realize that all that really matters is whether the team , won or lost during the season. Any rating system that shows the Ravens as the 25th rated team in the NFL is obviously a flawed rating system.
Actually you could get some good mileage out of a system that showed that the team's performance was diminished by factors that vary randomly and may very well improve significantly next year, such as turnovers. What that would show is that there's a core of ability within the team that belies its record.
Too many variables. I do think a weighting system for all 4 units could be established. For example, ST Td's scored/Td's allowed/FG %/FG% 50 yds plus/Net Punting Avg/Avg Field position created Even then the two games ST scored a Td, we lost, the games we allowed a ST Td we split (Bills/Seahawks) I say 4 b/c the GM/Scouting/Coaching unit is as important as the 3 units on the field. Carp's missed FG statistically were negligible, the impact however was huge on W's and L's.
Too many variables IMO. A lot of sports throw players out there and let their talent work for them. Scheme, playcalls, etc., play a much larger role in Am. Football that most other sports. Crazy rule changes, subjective refereeing (more so than other sports), inconsistency between stadiums, crowd noise...I really could go on. There's also the fact that the NFL is so entrenched in philosophies. There are objective strategies in sports like ***. Football, Baseball, and Basketball. In Football, what happens is just what the coach thinks is best. Machines with a lot of moving parts tend to break more easily.
If you can model the economy, weather, migration patterns, etc you can model this. If anything there are less variables then those things I named. This has a defined set of rules.
see above. They model the economy already. Which has crazy amount of variables. They can model football.
could you explain more about the system you used. i.e. what other factor other then wins and losses you used
As an econ student, we model the economy but no economist accepts them as practically useful. We can develop IS-LM curves until the cows come home, but we only use them for theoretical problems and accept that they are very limited.
Problem is, there are physical interactions that fluctuate on a play by play basis You cannot model for example a int for a Td, a offense may run 60 plays in a game however that 1 play can literally lose the game.
Sure, which is why "We can't model" is inaccurate. We can model, it's a matter of the view and how accurate is it. Disagree about them being very limited. They should be looked at in their context and not as the complete story. No I can predict the probability of say schaub making a bad pass based on past passes he's thrown. Econ, weather, etc all also deal with physical interactions in some way. Anything dealing with humans deals with physical interactions in some way as we are all physical entities. The question is accuracy, not can it be done.
Not really Lucky, human physical interactions vs say the velocity of dollars. Toss in injuries, variable weather, and a sabremetrics approach to the NFL won't work Thank God for it. One is far easier to predict, for example Shawne Merrimon beating Jake Long for a sack.
quick what happens to oil when a tanker physically interacts with an iceberg? What happens when frost hits the bread basket? I can keep going
No my defense is other models are use for far more complex things. I never said they weren't limited. I said they weren't very limited as you claimed. Will some be sure, will all of them, no. Question my experience all you want, if you want to have a discussion about the different models of the economy or weather or our evolution, etc I'm more then happy to.
http://en.wikipedia.org/wiki/Elo_rating_system What it does is looks at wins and losses and it awards and deucts points from your score depending on if you win or lose. It skews how many points are won or lost dependent on the rating of your opponet. In other words if 1 team is rated at 500 and the other is 1500 and the 500 team won the game the game might be worth 500 to each team. Making the 500 team score boost to 1000 and the 1500 team drop to 1000. If however the 1500 team won, which would be the expected outcome, the game may only be worth 50 points to each team. new scores would be 1550 and 450 in this example. Here is the formula if your a math geek: ExpectedScore=1/(1+10^((R_opponent-R_player)/400)) R_new=R_old+K*(Sum(Score)-Sum(ExpectedScore)) In simpler language this is comparing wins vrs losses relative to the strength of the opponets faced. The Ravens being ranked so low because maybe the games they won were worth miniumal points while the ones they lost were high value games. The opposite might be said about the Colts who are probably higher on the list then many expected. since they are playing eachother today: Ravens = 917 Colts = 1073 Ravens win = 940 +33 Ravens lose = 907 -7 Colts win = 1082 +9 Colts loss = 1050 -23 These ratings are also skewed by the fact that I did not go back 2 seasons and run this calc out, I instead started all teams out with a 1000 rating. It was too much work to do 2 seasons. 1 season required me to run that formula over 500 times and it took me hours to do it. Also it would of been more accurate but its results would of also been skewed for teams like the Colts who came in last without Luck and now are in the playoffs this year with him. Over time these anomilies minmalize themselves but the system does have flaws. The reason I ran this is because I often hear people talking about the quality of wins a team has or in Dolphan fans cases, they often talk about playing good teams tough. This system attempts to quantify that idea and then rate it numerically. I was curious also to see how it would all shake out and the results are the list I posted. Also because I had 512 calcs to do it is possible there is human error involved by me and I enetered a wrong value here or there. I did not proff check my work. I think I got it all right but its possible I made some mathamatical mistakes. Next season I am going to continue from this list and run it each week. If I made any mistakes they should self correct themselves as the weeks progress.
How does that make a difference? These things can be incorporated into models. The reality is that as long as the objective remains the same, you can create reliable models, especially on a team level. When you try to isolate individual players, it becomes more difficult, but there is robust enough data available on a team-level these days to model a lot of different events.
Limited in what sense? The largest enterprises in the world rely on economic models to predict all types of events.
I owe all of you an appology. I rechecked the Ravens score and I made a mathamatical error in week 17. The right ranking for them is 1023 ranking them much higher on the list. In my defense, it was like 3am when I was doing that one. Anyhow I am going to go through all the calcs to see if any other ones jump out at me as looking incorrect. I edited to OP to rank the Ravens propperly and I will edit the list if I find any other errors on my part. Sorry for the goof up.
Now have those high impact events happening on a 110 play x 16 game x 32 team environment each with a different outcome and variables being placed into that equation, including items of a 6% chance of occuring and "statistical modeling" for the NFL is simply not very accurate. When frost hits the farm land, it is forseeable and even a given non movable area with a given amount of crop in the ground. When you have tens of events..guess what the probability of accurately predicting outcomes will be? Take it every individual wheat plant, or droplets of oil in the tanker, then you are getting warmer.
Every entity.be it the investor or weather has an impact on the economy. I was giving you general largely seen events. Now what happened when Steve jobs got sick. Again tons of physical interactions take place. I'd the weather is perfect it has an impact.
My list predicts nothing. It is more of measuring stick of how good of a season each team had based on wins and losses and modified slightly by stretgh of opponets faced. It was not my intent to use to say anything about the future.
