Putting the Pats 2007 hype aside, the two games those teams played that year would lead me to believe the Giants were a lot closer to the Pats as team than the Pats 18-0 record says. Manning was playing mistake free football and the Giants were a much more physical team with a D-line the Pats couldn't handle. The Giants could have won the first game they played as well, I do not think the Super Bowl was that much of a fluke.
I'm not sure if you made those rankings based on teh results of your latest madden season or what. But you just wasted everyone's time with this thread.
Do any of you guys bashing these rankings realize that he didn't just rank them on opinion? He has a formula that he uses that takes various stats and ranks the teams. If you think it's a waste of time, by all means come up with something better. Personally I enjoy experiments like this.
colts that low is a joke. and who cares about there shitty run game. you dont need a run game when you have a guy named peyton manning as your quarterback.
and minn is in a huge downward spin not even close to being number 1. they will be lucky to get out of the wildcard round. you cant just base rankings off stats. you have to go by the coaching, players, and how they have performed the last few weeks
I think that the Giants just matched up better against the '07 Pats*, but they weren't necessarily better.
I get that people enjoy bashing the crap out of others on this board, its kinda the charm, but I do factor in net wins to the calculation I do. I just don't think that winning in the past has nearly as much to do with future ability to win as does your ability to score or produce or prevent the production of yardage.
Far be it for me to criticize someone who is trying to use statistics to understand an underlying process, but I see two big issues with the ratings (both of which you freely acknowledge) that account for a lot of the negative comments. First, your rating is based on how balanced teams are, downgrading teams that are very strong in one area but mediocre in another because the theory is that an unbalanced team should be easier to stop. You're certainly entitled to believe that, but it is only a theory, and not one that I think is particularly well-supported. I'm not going to go through this in detail, but let's just look at a very quick test based on Super Bowl teams. Here are the league offense and defense scoring rankings, respectively, for the SB teams for the last 9 years: Pittsburgh (20, 1) and Arizona (3, 28); New England (1, 4) and the Giants (14, 17); Indianapolis (2, 23) and Chicago (2, 3); Pittsburgh (9, 3) and Seattle (1, 7); New England (4, 2) and Philadelphia (8, 2); New England (12, 1) and Carolina (15, 10); Oakland (2, 5) and Tampa Bay (18, 1); New England (6, 6) and St. Louis (1, 7); Baltimore (14, 1) and the Giants (15, 5). I could keep going back, but you see the point. I see no evidence to suggest that a team that is balanced between offense and defense is any more or less likely to get to the Super Bowl than one that is very strong on only one side of the ball. I'm not going to go back and look in more detail at things like rushing versus passing offense or defense, but I suspect that the same pattern will emerge - sometimes balanced teams win and sometimes they lose, and sometimes teams that are particularly strong in only a few aspects of the game win and sometimes they lose. If you're going to tilt your ratings in favor of balance, you have to expect that people who don't agree with that theory are going to think that the ratings aren't sensible, and it would be good if you could justify that theory. The second point, which is probably even more serious, is that you count a game in Week 1 the same as you count one in Week 16, and that's just silly. I can't see any reason why you do this, since adjusting for when a game is played is easy to do; wherever you use a simple average over games, just use a weighted average where the weights decrease as you go back in time. This is completely standard in all forecasting applications, since everyone knows that recent data are generally more informative for prediction than are older data. Take a look at the Wikipedia page for Exponential smoothing to see a good discussion of what I'm talking about.
Vikings at #1 is freaking nuts. They have lost 3 of their last 4 and have lost their identity. They might not get out of the Wild Card round.
Normally, I give some kind of sarcastic remark, as most replies here imo dictate, be it in the name of fun or just to be obnoxious. That being said........ This is one of the best, most well thought out replies I have ever seen to an OP that the vast majority of the board vehemently disagrees with. Seriously, well done! I can not add anything other than that I totally agree.
I do a bit more than look at offense vs defense. I believe you know that but just posted overall offense vs defense numbers because they were quicker to type out. Let me address how the system works more specifically. I simply look at 13 statistics, 5 each on offense and defense, 2 on special teams and net wins. The offensive stats are points scored, yards gained rushing and passing, 3rd down completion % (this seems to correlate well with passing yards), giveaways (Ints+fumbles ignoring who recovers the ball). The defensive stats are points against (all points not just against the defense), yards allowed rushing and passing, sacks and takeaways (INTs plus forced fumbles). On special teams I look at total return yards (kickoff and punting) and net punting yardage. I scale down sacks, turnovers and special teams. I take each stat, and find the z-score for each team. Then simply add them. This has the effect of finding the most balanced high performing teams. I do nothing special to adjust for balance other than the simple addition and selection of stats to evaluate. Since we're having fun with Indy, lets look at their last super bowl year. They were 23rd in points allowed, 21st in total yards allowed but 2nd against the pass. Matched up with a Manning passing offense that puts points on the board quickly they would do well. The 2007 Colts had a passing attack ranked 2nd while having the 18th ranked ground game. I believe that looking at all aspects of the game would have rated that team quite well as they excelled at multiple aspects of the game. This year Indy has the number 1 passing attack but the last place ground game. Indy is currently rushing the ball for 84.6 yards per game, 31.4 yards less than league average, 82 ypg less than the Jets, more than 1.5 standard deviations under average. This year Indy is not significantly above average in any defensive stat other than points allowed. They average roughly average against the run, the pass, takeaways and sacks. I'm sure I'm missing something with those stats but they do tell a picture. Indy is currently a good, perhaps very good, but not a dominant team. They are significantly weaker than the 2007 version who ranked #2 against the pass and 14 places higher rushing the ball. I've been comparing Indy to the 80s Broncos with Elway. Lets look at their rankings a bit: P = passing rank, R = Rushing rank, DYA = defensive points allowed 87: P=3 _ R=12 _ DYA=9 - Side note, this was a really good team. 88: P=6 _ R=19 _ DYA=22 90: P=7 _ R=14 _ DYA=20 This year's Colts: 09: P=1 _ R=32 _ DYA=23 - If Manning weren't awesome this team would be struggling to get in the playoffs. Another weakness my system may have just looking at stats is that it doesn't count the fact that the Colts barely tried against the Jets. That one game may well have pulled their averages down so much that they slipped significantly in my rankings. Unfortunately I haven't had time the last few weeks to keep up with my system. The last rating period I have is week 12. In week 12 I had Indy 4th after Minn, NO, and NE. Every one of their stats are relatively lower since then except special teams and passing defense which are slightly better. The problem I have with changing the way I do this is that I use NFL.com for my stats and I just look at the team stats page. Next year I'll probably try to do a game by game type of data collection method. That will allow me to weight games in a method similar to what you are asking. One fear I have is that NFL.com's game by game stats don't match with the team stats. When I've done spot checks in the past I've found mismatches in the numbers. Interestingly adding this feature would probably drop Indy down even more than I have. Minnesota would probably be lower and SD higher. Keeping track of individual games will also allow me to do opponent adjustments as rushing for 300 against the Bills isn't quite the same as rushing for 300 vs the Packers. My system treats those the same right now as well What I hang my hat on is that NFL schedules are roughly equal over the course of a season, give or take. I know this isn't strictly true and would like to correct for that as well. This really gets to the point of posting this stuff. Suggestions on how I can improve. ------------ I keep defending my Colts slotting. The only other criticism I've seen against the actual rankings is that SD is too low and Minn too high. I think those may be fixed with an adjustment such as Jeff pointed out. The rest of the posts are basically that I'm an idiot and don't watch football but don't suggest where I'm wrong. I can't really work with that.
I would say that you need to be weighting the variables differently. Every study in every sport finds that nothing predicts winning percentage better than points scored and points given up; once you have those two, there's not a lot left. By having three additional offensive measures over points scored and two over points allowed that you're weighting equally, you're effectively downweighting the most important thing (points). I would say that you need to weight your 13 z-scores to reflect how strongly they relate to overall quality, based on historical data. You could do that a couple of different ways. One is some form of factor analysis, where you find the linear combination of variables that accounts for the most common variability. An even better way would be based on some sort of regression analysis, trying to find the linear combination of variables that best predicts winning percentage or individual game victories (some form of logistic regression would be sensible for that). These are the kinds of things that are done in all of the sabermetrics stuff. You could also do it informally, of course, but I definitely feel that if you find some sort of sensible weighting of variables, you'll find better results. If you then use weekly data, and use some weighted combination of overall (season) score with the previous week's results, you could effectively address the two main criticisms that I can see.
