Weekly game probabilities are available now at the nytimes.com Fifth Down. This week I conduct a little experiment about our perceptions of team strength and reality.
- Home
- predictions
- Weekly Game Probabilities - Divison Round
Subscribe to:
Post Comments (Atom)
For anybody who wants to know how these convert to spreads, they are:
Steelers -7.5
Packers -6.5
Bears -10.5
Patriots -7.5
On my own model I have the lines as:
Steelers -5.5
Packers +0.5
Bears -10.5
Patriots -10
My model has Falcons -∞
The bookies seem to have
BAL .39, PIT .61
GB .43, ATL .57
SEA .18, CHI .82
NYJ .22, NE .78
I just wonder what it is people THINK Atlanta have to make them favourites. They don't pass the ball as well as Green Bay, they don't defend the pass as well as Green Bay, they're as good as even with Green Bay in the running game. They are better than them at not turning the ball over, but is that really enough to overcome the offensive AND defensive deficiencies?
We'll just have to see, but the difference between the consensus view and the statistical view makes this game so much more intriguing.
The Falcons are better or even in every aspect of the game except WR depth.
They play a different style of game that is balanced and not prolific statistically except in scoring. 5th in both offense and defense.
They have a better running game Turner, Snelling, Mughelli.
They have a better TE.
The better receiver ... Roddy White.
The better O'line.
GB has better corners.
Atl has better safeties.
The rest is pretty much a was give or take a slight edge here and there.
When you play a ball control game ... You are also keeping the ball away from yourself. Which hurts your stats. The key to the Falcons success however is that they can usually finish their drives.
I just wonder what it is people THINK Atlanta have to make them favourites....
"13 wins, that's all that counts", the Atlanta people tell me. That and they already beat GB at home (by 3) once this year. And the great majority of people still take winning close games as a sign of strength rather than weakness.
What Atlanta people have been telling you "13 wins, that's all that counts"?
To back up Gitelson's comment, this is what people think Atlanta have on Green Bay:
http://www.nfl.com/news/story/09000d5d81da2266/article/preparation-home-dome-make-the-difference-for-falcons
Also, here's a really great article that I think is very relevant to this conversation, as it acknowledges and offers an objective explanation for why the Falcons are outperforming their efficiency stats:
http://sports.yahoo.com/nfl/news?slug=jc-preparedmikesmith011111
I think the contrast between Smith and McCarthy on record in close games is striking. Smith is objectively a better game manager than McCarthy. Just ask Packers fans:
http://www.totalpackers.com/2010/12/20/whats-with-mike-mccarthy-and-clock-management/
Smith pays attention to every minute detail, and when you're looking at close games that are going to come down to a field goal one way or another, it makes sense that that will often be the difference between victory and defeat.
Before you say it, I understand there's probably some luck that factors into the differences, but there's a lot of coaching that factors in as well, and it would be foolish to suggest otherwise.
What Atlanta people have been telling you "13 wins, that's all that counts"?
Well, just in the last few hours, comments under Brian's NY Times piece...
"Given their respective win-loss records, I would say that the statistics aren't any measure"
"I can think of 13 reasons why Atlanta isn't overrated. The only statistics of any consequence are wins." Etc.
David - I know you said it was an objective view, but I just can't take seriously anyone who tries to infer trends from such small samples.
"Smith has compiled a 13-7 (including the playoffs) mark in games decided by six points or less. [..] well ahead of Packers coach Mike McCarthy, who at 11-16 in such games ranks last of the eight."
Neither of thse are statistically significant, and then he goes on to say that McCarthy's Packers are 0-2 against Smith's Falcons, as somehow proof that Smth has the edge of McCarthy.
Of course there are some coaching factors in there (in the GB-ATL game earlier in the year, Smith went for the TD on 4th and goal from the 1, when most coaches would have kicked the FG to take the lead) but I seriously doubt McCarthy is as bad in close gams, nor Smith as good, as the author tries to suggest.
Jim, fair enough. There are undoubtedly tons of Atlanta fans out there who would say "13-3, that's all that counts." I was thinking within the confines of this site, where the majority of comments, with a few exceptions, don't look solely at the record.
Ian, it's an article, not a statistical study. 20 games from each coach is enough to at least point to a potential trend, and if you were to go through those games (the benefit of the small sample size), you could see things Smith did to help his team win games (like the one you mentioned and the Bears game mentioned in the article) and probably even more noticeably, the mistakes McCarthy made to help his team lose.
I don't know if you looked at the first blog I posted to, but all it took was a simple Google search for "McCarthy game management" and clicking on the first link. If you look at the bottom of that article, you'll see "The Packers are bad in close games; here's why", linked to here:
http://www.totalpackers.com/2010/12/24/the-packers-are-bad-in-close-games-heres-why/
It basically says the antithesis of everything the Yahoo! article and I have been saying about the Falcons. There are very clear reasons for why the Falcons are winning games and the Packers are losing games, and they're not things that you would expect to change overnight.
I've said about the model before that it can be right for the majority of teams but wrong for some. On average, teams are probably good in some of these "coaching" statistics and bad in others and it balances out to the point it doesn't have a major impact on a team's record. There's clear evidence to suggest, however, that McCarthy has consistently proven poor in game management areas and Smith has consistently shown attention to detail to be one of his greatest strengths. Combining statistics with actually watching football games, you can very justifiably infer that coaching has had an effect on these two teams' fortune over the last 3 years.
Jim, fair enough. There are undoubtedly plenty of Falcons fans that would say "13 wins, that's all that counts". I was thinking within the confines of this site where the majority of comments, with very few exceptions, look beyond record.
Ian, it's an article, not a statistical study. I think more than 20 games from each coach is enough to point to a potential trend. When that surface analysis is built upon by actually looking back at some of those games (the advantage of a small sample), you can see examples of where Smith has won his team games(like the one you mentioned and the Bears game mentioned in the article), and where McCarthy has cost his team.
I don't know if you looked at the blog post I linked to, but all it took was a simple Google search "mccarthy game management" and clicking on the first link. If you go to the bottom of the article, you see a link to "The Packers are bad in close games. Here's why." I have included a link to that article here:
http://www.totalpackers.com/2010/12/24/the-packers-are-bad-in-close-games-heres-why/
It basically says the antithesis of everything the Yahoo! article and I have been saying about the Falcons.
I have said before that the model can be right about the majority of teams but wrong about some. It is likely that for the majority of teams these "coaching" statistics balance out and do not have a major effect on record. It seems exceedingly likely, however, that Smith is exceptionally strong in these areas and McCarthy is exceptionally poor.
By combining statistics with actually watching football games, there is plenty of evidence to infer that McCarthy's failure at managing games and Smith's outstanding attention to detail have had a major effect on their teams' fortunes over the last 3 years.
"I think more than 20 games from each coach is enough to point to a potential trend."
This is a quantitative question. For 20 binomial trials with success probability of 0.5, the variation (as captured by the 5-95 confidence band, which is often used for instance in political poll margin of errors) is +-20%. For 27 such trials it is +-17%.
FWIW the 20-80% confidence bands are +-10% and +-9%.
David - I read the blog post but again, that's lots of small sample inferences, which are all based on the following argument - "The Packers are bas in close games, the Packers are bad in this stat, therefore sorting the stat will resolve their close game problems".
For instance, the point about the Packers poor 4th down conversion percentage. Yes, it's something to be concerned about, but when only 3 of the top 10 teams at 4th down conversions made the playoffs, it's doesn't exactly suggest that it's a stat one needs to excel in.
Like I said, on balance a lot of these stats probably even out for most teams. The difference for the current discussion is that the Packers are bad in these areas across the board and on the flip side the Falcons are good across the board. It immediately jumps out in reading the article that all of the categories cited as evidence for why the Packers are bad in close games are the exact same categories the Falcons excel in. Yes, it's small sample inferences, but as you said it's A LOT of small sample inferences, the combination of which becomes indicative of coaching ability.
Also very important to note is that the article does not say "sorting the stat will resolve their close game problems". It says the Packers are bad in close games because they are bad in these stats and leaves it at that. Aside from the general pessimistic nature of the article, this is because again these are not areas that the teams can just improve in overnight. For a few teams, two of which are the Falcons and Packers, the combination of those stat categories is actually indicative of coaching ability and therefore does hold some predictive value.
As Brian once said: Bad Coaching can ruin your team...
I wouldn´t say McCarthy is a bad coach but certainly a bad play caller, thus costing GB many games. His timid calls and bad decisions are in a league with Norv the Smurf and Andy Reid. All 3 never learn and therefore will always fail. Nobody needs big sample sizes for that to prove.
GB is the better team by FAR, but Coach Smith is the better coach. So it will be a tight game.
Brian: Bring in a coaching factor and your Model will be perfect :-)
Go Bears, Karl from Germany
The fact that Vegas has Atlanta as less than a 3 point favorite at home despite having the best record in the NFC means that Vegas clearly agrees that Atlanta has been fortunate this year.
Brian, read your Fifth down post. As usual I liked it, but I must say shame on you for implying that the Falcons, Packers and Seahawks did not play meaningful games in week 17, when in fact all 3 of them did.
The Falcons-Panthers game was only 'meaningful' to one of the teams on the field.
GB is a very good team. They had the "double-whammy" schedule this year (vs the AFC East and the NFC East) and still had a 10-6 record. So they are a very good team. And now they appear to have a legitimate running game.
When ATL and GB met earlier in the season, the final score was a home-team victory 20-17, which is the final score expected of two evenly matched teams.
I'm an ATL fan, but I have to admit that I wouldn't mind seeing GB@CHI and PIT/BAL@NE on championship Sunday in the cold and (hopefully) snow.
I just re-read my previous post and realized that I had assumed a Pats victory over the Jets. I apologize to any Jersey/B fans I might have offended. A game in the meadowlands would be just as compelling as one in Foxboro.
And sorry Seahawks fans, but a game in the Pacific NW just wouldn't be as compelling.
David: First let me say I am completely agnostic about Atlanta -- I'm an AFC guy since way back to AFL days and have no dog in the Falcons are great-or-fake fight. If the Falcs make the SB, good for you.
But beyond that, the articles you cite about quality of coaching are "just so stories" made up afterward in light of whatever came before. They have evidential worth of zip nada. I can quote you pages of stories about how the "Mangenius" personally won an array of games for the Jets and returned has-been Pennington to Comeback Player of the Year, etc. etc. Next season ... ooops. The same writers pilloried him after going 4-12. I can quote "Parcells is a bum who will never succeed" stories from after his first 3-13 year. And on and on.
Here is the evidential reality re coaching and close, one-score games: Vince Lombardi's Packers, exactly 50%. Bill Walsh's 49ers, 43%. But Vince Tobin has the *all-time best* record in 20 or more close games, going 15-5, 75%! (though unfortunately only 13-38, 25% in all his rest). Dick Jauron won 8 close games for the Bears one year, was named Coach of the Year, praised to the sky for doing what even the great Halas and Butkus could never do, got a new contract and then a new job in Buffalo -- and had losing records in *all* his other nine seasons. (Showing just how *expensive* it can be for GMs to believe in the close game fallacy)
What does this evidence tell you? That Smith knows the secret to winning close games that Tobin knew, and which Jauron knew for one season then forgot, but which Walsh and Lombardi could never figure out?
"Smith pays attention to every minute detail, and when you're looking at close games that are going to come down to a field goal one way or another, it makes sense that that will often be the difference between victory and defeat."
No -- it doesn't make any sense at all. Let's assume for argument's sake that Smith pays more attention to detail than Belichick, who doesn't win so many close games (and more than did Walsh, Lombardi, etc.).
That doesn't just apply in the last minute -- superior coaching makes the team better all game through. That means a team that formerly played close games against mediocre teams and got whomped by good teams now whomps mediocre teams and plays its close games against good teams. It's gotten better, thanks to its coach!
But its close games are still determined overwhelmingly by luck. Why? Because ... obviously ... the closer the game the smaller is the amount of random chance tht can change its outcome, regardless of the quality of the coaching, good *or* bad.
It should be really hard to deny that simple logic. But when tempted to, say: Vince Tobin.
You can't say the Packers are an objectively better team. Yes, statistics are important and good skill indicators, but they are not everything. The Packers may be better, but not objectively.
Jim- "That doesn't just apply in the last minute -- superior coaching makes the team better all game through. That means a team that formerly played close games against mediocre teams and got whomped by good teams now whomps mediocre teams and plays its close games against good teams. It's gotten better, thanks to its coach!"
I agree with everything in that paragraph (though probably not the magnitude you imply). That is actually what the Falcons did though. Look at the schedule: http://espn.go.com/nfl/team/schedule/_/name/atl/year/2010/atlanta-falcons
The Falcons are probably the most predictable team in the NFL if you value them at the level most people have and not the level the efficiency model does.
And yes, it still does make sense that attention to detail can make the difference in close games. In blowouts, it doesn't matter if you know how to perfectly execute the two minute drill. In close games it does. How often do you hear analysts say "this game is going to come down to who has the ball last"? If you effectively manage the game, you can actually make it so that you are that team that has the ball last.
As far as your citing of other coaches, forgive me for being young and not much of a football historian, so I don't claim to know what Walsh and Lombardi's strengths and weaknesses are, nor do I claim to know anything about Vince Tobin (which was of course your point). From my limited understanding though, Lombardi was a master motivator and Walsh was a master strategist. It is of course true that "good coaching" is a combination of many skills, not just one, and from the numbers you cite it could certainly be the case that they excelled in many areas but were only average on game management. Of course on this one I'm to some extent talking out of my ass.
I can though have a conversation about Belichick. When you invoked his name, I was surprised because I do think of him as somebody that is great at paying attention to detail and that would probably be good in close games. Looking back at the stats, sure enough, he has the second best record to Smith of coaches left in the playoffs at 52-33, so I'm not sure why you brought him up. Obviously I understand that by the time Smith has played 85 close games his record may be about where Belichick's is, but that doesn't change anything. It just means they're both great in this aspect.
David: You are making the mistake of assuming your conclusion, then arguing from there with "just so" stories.
Your *assume* that winning close games is a skill, thus a mark of a good team, and something that can be coached by a superior coach. Then you assume the coaching skill involved is attention to detail ... thus Smith is a *markedly* superior coach at handling details compared to others ... and, wow, maybe Walsh Lombardi with their eight titles, but lack of close game wins, were relatively weak in their attention to detail (cough, ahem).
But check your starting point. If winning close games is not a skill, not a mark of team strength, then all the rest of your chain of conclusions falls apart.
The test of whether winning close games is a skill is easy: do teams that win close games (1) continue to win close games, and (2) defeat other teams in competition going forward? And the answers in the data are very, very clear: "NO" and "NO". In baseball's hundred-year data base, pro basketall, NFL football, the data are overwhelming.
In the NFL playoffs, winning close games during the regular season is predictive of failure! At every regular season W-L level (11-5, 12-4, 13-3) the teams that win the most close games during the regular season do worse in the playoffs. And the teams with worst losing records in close games during the season *dominate*!
Which is of course exactly what you'd predict if your initial assumption was: "close games are determined by *luck*".
Example: Teams A, B, and C all get into the playoffs with 11-5 records. Team A gets its record with several close game wins, Team B has no close game wins, Team C has unlucky close game losses it must overcome to reach its 11-5. Obviously Team B is stronger on the merits than lucky Team A, and Team C to overcome several unlucky losses and still reach the same record as the others must be a *powerhouse* compared to them.
Fact: In the last 15 years of playoffs, among teams with 11-5 or better season records, the seven teams with the *worst* records in close games (13-30, 38%) in the postseason went 14-4, 78% and won three Super Bowls.
Meanwhile, the top 10 teams at winning close games during the regular season, 86-12 in them, 88%, in the playoffs went 10-9. Coin flip city.
And yes, it still does make sense that attention to detail can make the difference in close games. In blowouts, it doesn't matter ...
No, it doesn't. You refuse to accept that any skill applies equally *all game through*. The Jets are a genuinely good team with a very good defense. When the Pats blew them out 45-3, you don't think that the Pats did that with great attention to detail on every play all game through? Do you really believe that there is some sort of good quality coaching that applies only in the last two minutes of close games?
Close games are determined overwhelmingly by luck, hard fact. And being that they are, you simply can't use them as evidence that any coach has *any* skill other coaches lack -- and certainly not to claim Smith has skills lacked by Lombardi and Walsh!
As to what sports writers say about it, always remember that NFL football is an entertainment industry. Popular article writers and "expert commentators" *never* say things that contradict the myths fans love which make so much money. To the contrary, they fuel the myths.
Imagine all the big names on the Sunday post-game show going: "Another three games looking like they were determined by random chance today..."
*I'd* find that entertaining! But NFL owners and networks who want ratings, and editors who want readers to attract advertising? Not so much. "Three more examples of great last second clutch play!" sells a whole lot better for them.
David: You are making the mistake of assuming your conclusion and then arguing from there with just-so stories.
You assume that winning close games is a skill, thus a mark of a good team, and something that can be coached by a superior coach. And you're assuming the coaching skill involved is attention to detail ... thus Smith is a markedly superior coach at handling details compared to others ... and thus, wow, maybe Walsh and Lombardi with their eight titles, but lack of close game wins, were relatively weak in their attention to detail. Ahem.
But check your assumption. If winning close games is not a repeatable skill, and not a mark of team strength, then all the rest of your chain of conclusions falls apart ... maybe Walsh and Lombardi while winning eight titles were the best "attention to detail" coaches of all time.
The test of this is easy: Do teams that win close games continue to do so? In competition with other teams that don't win close games, do they win rather than lose?
There's tons of data giving the answer, and it is very clear. Look it up.
falcons fans can now shut up wins dont matter as much as efficiency in predicting future games