Chase Stuart
Footballguy
http://www.pro-football-reference.com/blog/wordpress/?p=242
In the comments to The Dungy Index, Pat requested a different sort of coaching index. Today we’re going to analyze post-season coaching records from another angle, by predicting how many post-season wins a coach or team should have based on their number of regular season victories*.
To start with, we need to break the modern era into three sections. From 1970 to 1977**, the NFL had an eight team playoff field, and a fourteen game schedule. In 1978 the NFL expanded the regular season to 16 games and the post-season tournament to ten teams. In 1982 and 1987, the NFL did not have 16 game regular seasons, so I’ve omitted the strike seasons from the data***. In 1990, the NFL again expanded the playoff field by two teams, and since then 12 teams have made the playoffs each year. In 2002 the NFL realignment changed the ordering of the playoff teams, but I’ve decided to ignore that small change for the purpose of this post. The added specificity isn’t very probative, and unnecessarily reduces our sample size.
*The normal formula for victories is Wins + (Ties/2), since the NFL counts a
tie as half a win. Here, I'm just using actual wins, which essentially equates a tie to
a loss.
**From 1970 to 1975, the home team was chosen randomly, and not by best record.
This should be kept in mind when analyzing those coaching performances.
***Yes, I know Redskins fans have a legitimate beef with this, since I'm erasing
two of Joe Gibbs' Super Bowls. My apologies, and I'll address this later.
The Early Years: 1970-1977
RW is the number of regular season wins, PW is the number of playoff wins, #TM is the number of teams that made the playoffs with that number of regular season wins, and AvgPW is the average number of playoff wins for teams with that number of regular season wins.
RW PW #TM AvgPW 8 1 3 0.33 9 3 8 0.3810 18 27 0.6711 13 14 0.9312 15 10 1.5013 3 1 3.0014 3 1 3.00So what coaches exceeded expectations the most? Below is the list of coaches from the early period, with both their actual and expected (based on regular season record) number of wins.
The Middle Years: 1978-1989
And the coaches:
The Later Years: 1990-2006
Note: The Colts and Bears have each been given a half win for their Super Bowl berth.How do the most recent coaches look?
Coach ActW ExpW DiffBill Belichick 13 7 +5.6Jimmy Johnson 9 5 +3.8Marv Levy 10 6 +3.7John Fox 5 2 +2.9Mike Holmgren 12 9 +2.8Joe Gibbs 6 3 +2.8Bill Cowher 12 9 +2.5Barry Switzer 5 3 +1.8Ted Marchibroda 2 0 +1.6Andy Reid 8 7 +1.2Brian Billick 5 4 +1.2Bill Callahan 2 1 +1.0Jon Gruden 5 4 +0.7Mike Tice 1 0 +0.7Bill Parcells 6 5 +0.7Vince Tobin 1 0 +0.6Sam Wyche 1 0 +0.6Sean Payton 1 1 +0.5Jim Haslett 1 1 +0.5Norv Turner 1 1 +0.5Jerry Glanville 1 1 +0.5Dan Reeves 5 5 +0.4Herman Edwards 2 2 +0.1Dick Vermeil 3 3 +0.1Pete Carroll 1 1 0.0Lovie Smith 2.5 3 0.0Ray Rhodes 1 1 0.0Jim Mora Jr. 1 1 0.0Rich Kotite 1 1 0.0Mike Shanahan 8 8 -0.1Bobby Ross 3 3 -0.3Bruce Coslet 0 0 -0.3Tony Dungy 8.5 9 -0.3Art Shell 2 2 -0.3Jeff Fisher 5 5 -0.3Don Shula 3 3 -0.3Dom Capers 1 1 -0.3Jim Fassel 2 2 -0.4Butch Davis 0 0 -0.4Lindy Infante 0 0 -0.4June Jones 0 0 -0.4Eric Mangini 0 1 -0.5Buddy Ryan 0 1 -0.5Dave Wannstedt 2 3 -0.5Chan Gailey 0 1 -0.8George Seifert 7 8 -0.8Mike Martz 3 4 -0.9Marvin Lewis 0 1 -1.0Mike Ditka 1 2 -1.1Jack Del Rio 0 1 -1.3Dennis Green 4 5 -1.4Dick Jauron 0 1 -1.5Tom Coughlin 4 6 -1.6Steve Mariucci 3 5 -1.7Mike Sherman 2 4 -1.7Wayne Fontes 1 3 -1.8Wade Phillips 0 2 -2.0Jack Pardee 1 3 -2.3Jim Mora 0 5 -4.6Marty Schottenheimer 3 10 -6.5Now, several coaches span the three eras, so we need to compile a career list. To avoid staring at 100 coaches, only coaches with at least 4 expected playoff wins are included:
Now let me try and appease our Joe Gibbs fans. The 1982 Redskins went 8-1, which is close to 14-2. So we might want to give him 2.17 expected wins in 1982, except there was an additional round of the playoffs that year. That extra game was against the 4-5 Detroit Lions, so I don’t want to give Gibbs too much credit here. Let’s give him 3 expected wins with an 8-1 Redskins team, and since he won four playoff games that year, he should have one extra win in the “Diff” column.The 1987 Redskins went 11-4, but won all three games with replacement players. We should probably put them at 8-4, which we’ll make equivalent to 0.71 expected wins. Since Gibbs won three games, that’s an extra 2.3 wins for him. So while Gibbs is listed at +2.6, it’s probably fairer to place him at 5.9, which is right behind Chuck Noll for tops on the list.
General Thoughts
One problem here is sample size. Only two teams in the middle era made the playoffs with 8 wins, and neither won a game. Only two teams in the early era won thirteen or more games, and both won the Super Bowl. So John Madden and Don Shula get no credit for winning a Super Bowl since they were the only teams with those number of wins in the era, while Marty Schottenheimer and Bud Grant catch a break since the other coach also lost his playoff game with 8 wins. Note: this may be the only time I’ve ever seen it written that Schottenheimer or Grant caught a break.
Still, the data only has two real blips, when teams with more regular season wins are expected to have fewer post-season wins than teams with fewer regular season wins. Teams with ten wins in the middle years didn’t fare very well, and to the extent that a coach had several ten win seasons over that period, he’s probably not punished enough for his post-season losses (and he’s overcompensated for his post-season wins). The other blip is in the recent era, where only two teams had fifteen wins, and both teams went one and done. Based on the expected wins trend for 10, 11, 12, 13 and 14 win teams, I think we’d be better off expecting 2.00 wins from teams that win 15 games than 1.00 wins. So you might want to subtract one win in the “Diff” column from Dennis Green and Bill Cowher.
If you do that, you might notice that before last year, Bill Cowher would have 1.5 fewer actual post-season wins than we’d expect. Of course, last year he won four games in the post-season en route to a Super Bowl victory. That’s not very surprising, though, because I don’t think there’s much value to any of the above data. Playoff numbers are overvalued, because the results mean so much more to fans, and the sample sizes are incredibly small. As a result, grandiose statements that aren’t statistically significant, are made. I don’t believe Bill Cowher changed anything in going from a “bad” post-season coach to tying the (Joe Gibbs) record for most post-season wins in 2005.
Which brings me to my last point. In addition to thinking the data doesn’t have much predictive value, I’m not sure it has any value at all. We might look at Marty Schottenheimer’s -7.2 rating and conclude that he’s a much better regular season coach than playoff coach. We could say that Chuck Noll was a much better playoff coach than regular season coach, as evidenced by his +6.1 rating. And the data would support that. But we could just as strongly argue that Marty Schottenheimer gets the most out of average players in the regular season, but in the playoffs, coaching talent is much less important than player talent. Conversely, one could state that Chuck Noll couldn’t reach his players, was a terrible coach, but because of how talented his team was, they always won in the post-season.
You might like the former theory better than the latter, but let’s be clear: the data equally support both theories. To suggest that coaching ability matters more than player ability in the playoffs, or vice versa, is purely speculative, and not proven by anything presented here.
In the comments to The Dungy Index, Pat requested a different sort of coaching index. Today we’re going to analyze post-season coaching records from another angle, by predicting how many post-season wins a coach or team should have based on their number of regular season victories*.
To start with, we need to break the modern era into three sections. From 1970 to 1977**, the NFL had an eight team playoff field, and a fourteen game schedule. In 1978 the NFL expanded the regular season to 16 games and the post-season tournament to ten teams. In 1982 and 1987, the NFL did not have 16 game regular seasons, so I’ve omitted the strike seasons from the data***. In 1990, the NFL again expanded the playoff field by two teams, and since then 12 teams have made the playoffs each year. In 2002 the NFL realignment changed the ordering of the playoff teams, but I’ve decided to ignore that small change for the purpose of this post. The added specificity isn’t very probative, and unnecessarily reduces our sample size.
*The normal formula for victories is Wins + (Ties/2), since the NFL counts a
tie as half a win. Here, I'm just using actual wins, which essentially equates a tie to
a loss.
**From 1970 to 1975, the home team was chosen randomly, and not by best record.
This should be kept in mind when analyzing those coaching performances.
***Yes, I know Redskins fans have a legitimate beef with this, since I'm erasing
two of Joe Gibbs' Super Bowls. My apologies, and I'll address this later.
The Early Years: 1970-1977
RW is the number of regular season wins, PW is the number of playoff wins, #TM is the number of teams that made the playoffs with that number of regular season wins, and AvgPW is the average number of playoff wins for teams with that number of regular season wins.
RW PW #TM AvgPW 8 1 3 0.33 9 3 8 0.3810 18 27 0.6711 13 14 0.9312 15 10 1.5013 3 1 3.0014 3 1 3.00So what coaches exceeded expectations the most? Below is the list of coaches from the early period, with both their actual and expected (based on regular season record) number of wins.
Code:
Coach ActW ExpW DiffTom Landry 12 6 +6.0Chuck Noll 8 5 +3.2Don McCafferty 4 2 +2.4Don Shula 8 7 +1.2Dick Nolan 2 1 +0.6Red Miller 2 2 +0.5John Madden 8 8 +0.3Jack Pardee 0 0 -0.4Lou Saban 0 0 -0.4Bud Grant 7 7 -0.4Dan Devine 0 1 -0.7Hank Stram 0 1 -0.7Joe Schmidt 0 1 -0.7Chuck Fairbanks 0 1 -0.9Nick Skorich 0 1 -1.0George Allen 2 3 -1.3Don Coryell 0 2 -1.6Paul Brown 0 2 -1.9Chuck Knox 3 5 -2.0Ted Marchibroda 0 2 -2.3
Code:
RW PW #TM AvgPW 8 0 2 0.00 9 11 17 0.6510 13 27 0.4811 19 23 0.8312 25 21 1.1913 3 2 1.5014 13 6 2.1715 6 2 3.00
Code:
Coach ActW ExpW DiffBill Walsh 10 6 +3.6Tom Flores 7 4 +3.0Chuck Noll 8 5 +2.9Bum Phillips 4 2 +1.9Raymond Berry 3 2 +1.3George Seifert 3 2 +0.8Sam Wyche 2 1 +0.8Forrest Gregg 2 1 +0.8Bill Parcells 5 4 +0.5Bud Carson 1 1 +0.4Ray Perkins 1 1 +0.4Ray Malavasi 3 3 +0.3John Robinson 4 4 +0.3Chuck Knox 4 4 +0.2Dan Reeves 4 4 +0.2Jerry Glanville 1 1 -0.1John McKay 1 1 -0.1Dick Vermeil 3 3 -0.1Joe Gibbs 4 4 -0.2Joe Walton 1 1 -0.3Jerry Burns 1 1 -0.3John Mackovic 0 0 -0.5Walt Michaels 0 0 -0.5Neill Armstrong 0 0 -0.5Don Coryell 2 2 -0.5Monte Clark 0 1 -0.6Bud Grant 0 1 -0.6Marty Schottenheimer 1 2 -0.7Sam Rutigliano 0 1 -0.8Bullough/Erhardt 0 1 -0.8Marv Levy 1 2 -0.8Leeman Bennett 1 2 -0.8Red Miller 0 1 -1.0Tom Landry 5 6 -1.1Buddy Ryan 0 1 -1.3Mike Ditka 5 7 -1.8Don Shula 3 7 -3.7
Code:
RW PW #TM AvgPW8 2 7 0.299 15 34 0.4410 26 51 0.5111 45 43 1.0512 44.5 33 1.3513 33.5 23 1.4614 19 11 1.7315 2 2 1.00
Coach ActW ExpW DiffBill Belichick 13 7 +5.6Jimmy Johnson 9 5 +3.8Marv Levy 10 6 +3.7John Fox 5 2 +2.9Mike Holmgren 12 9 +2.8Joe Gibbs 6 3 +2.8Bill Cowher 12 9 +2.5Barry Switzer 5 3 +1.8Ted Marchibroda 2 0 +1.6Andy Reid 8 7 +1.2Brian Billick 5 4 +1.2Bill Callahan 2 1 +1.0Jon Gruden 5 4 +0.7Mike Tice 1 0 +0.7Bill Parcells 6 5 +0.7Vince Tobin 1 0 +0.6Sam Wyche 1 0 +0.6Sean Payton 1 1 +0.5Jim Haslett 1 1 +0.5Norv Turner 1 1 +0.5Jerry Glanville 1 1 +0.5Dan Reeves 5 5 +0.4Herman Edwards 2 2 +0.1Dick Vermeil 3 3 +0.1Pete Carroll 1 1 0.0Lovie Smith 2.5 3 0.0Ray Rhodes 1 1 0.0Jim Mora Jr. 1 1 0.0Rich Kotite 1 1 0.0Mike Shanahan 8 8 -0.1Bobby Ross 3 3 -0.3Bruce Coslet 0 0 -0.3Tony Dungy 8.5 9 -0.3Art Shell 2 2 -0.3Jeff Fisher 5 5 -0.3Don Shula 3 3 -0.3Dom Capers 1 1 -0.3Jim Fassel 2 2 -0.4Butch Davis 0 0 -0.4Lindy Infante 0 0 -0.4June Jones 0 0 -0.4Eric Mangini 0 1 -0.5Buddy Ryan 0 1 -0.5Dave Wannstedt 2 3 -0.5Chan Gailey 0 1 -0.8George Seifert 7 8 -0.8Mike Martz 3 4 -0.9Marvin Lewis 0 1 -1.0Mike Ditka 1 2 -1.1Jack Del Rio 0 1 -1.3Dennis Green 4 5 -1.4Dick Jauron 0 1 -1.5Tom Coughlin 4 6 -1.6Steve Mariucci 3 5 -1.7Mike Sherman 2 4 -1.7Wayne Fontes 1 3 -1.8Wade Phillips 0 2 -2.0Jack Pardee 1 3 -2.3Jim Mora 0 5 -4.6Marty Schottenheimer 3 10 -6.5Now, several coaches span the three eras, so we need to compile a career list. To avoid staring at 100 coaches, only coaches with at least 4 expected playoff wins are included:
Code:
Coach ActW ExpW DiffChuck Noll 16 10 +6.1Bill Belichick 13 7 +5.6Tom Landry 17 12 +4.9Jimmy Johnson 9 5 +3.8Bill Walsh 10 6 +3.6Tom Flores 7 4 +3.0Marv Levy 11 8 +2.8Mike Holmgren 12 9 +2.8Joe Gibbs 10 7 +2.6Bill Cowher 12 9 +2.5Andy Reid 8 7 +1.2Bill Parcells 11 10 +1.2Jon Gruden 5 4 +0.7Dan Reeves 9 8 +0.6John Madden 8 8 +0.3George Seifert 10 10 0.0Mike Shanahan 8 8 -0.1Dick Vermeil 6 6 -0.1Tony Dungy 8.5 9 -0.3Jeff Fisher 5 5 -0.3Bud Grant 7 8 -1.0Dennis Green 4 5 -1.4Tom Coughlin 4 6 -1.6Steve Mariucci 3 5 -1.7Chuck Knox 7 9 -1.8Don Coryell 2 4 -2.1Don Shula 14 17 -2.8Mike Ditka 6 9 -2.9Jim Mora 0 5 -4.6Marty Schottenheimer 4 11 -7.2
General Thoughts
One problem here is sample size. Only two teams in the middle era made the playoffs with 8 wins, and neither won a game. Only two teams in the early era won thirteen or more games, and both won the Super Bowl. So John Madden and Don Shula get no credit for winning a Super Bowl since they were the only teams with those number of wins in the era, while Marty Schottenheimer and Bud Grant catch a break since the other coach also lost his playoff game with 8 wins. Note: this may be the only time I’ve ever seen it written that Schottenheimer or Grant caught a break.
Still, the data only has two real blips, when teams with more regular season wins are expected to have fewer post-season wins than teams with fewer regular season wins. Teams with ten wins in the middle years didn’t fare very well, and to the extent that a coach had several ten win seasons over that period, he’s probably not punished enough for his post-season losses (and he’s overcompensated for his post-season wins). The other blip is in the recent era, where only two teams had fifteen wins, and both teams went one and done. Based on the expected wins trend for 10, 11, 12, 13 and 14 win teams, I think we’d be better off expecting 2.00 wins from teams that win 15 games than 1.00 wins. So you might want to subtract one win in the “Diff” column from Dennis Green and Bill Cowher.
If you do that, you might notice that before last year, Bill Cowher would have 1.5 fewer actual post-season wins than we’d expect. Of course, last year he won four games in the post-season en route to a Super Bowl victory. That’s not very surprising, though, because I don’t think there’s much value to any of the above data. Playoff numbers are overvalued, because the results mean so much more to fans, and the sample sizes are incredibly small. As a result, grandiose statements that aren’t statistically significant, are made. I don’t believe Bill Cowher changed anything in going from a “bad” post-season coach to tying the (Joe Gibbs) record for most post-season wins in 2005.
Which brings me to my last point. In addition to thinking the data doesn’t have much predictive value, I’m not sure it has any value at all. We might look at Marty Schottenheimer’s -7.2 rating and conclude that he’s a much better regular season coach than playoff coach. We could say that Chuck Noll was a much better playoff coach than regular season coach, as evidenced by his +6.1 rating. And the data would support that. But we could just as strongly argue that Marty Schottenheimer gets the most out of average players in the regular season, but in the playoffs, coaching talent is much less important than player talent. Conversely, one could state that Chuck Noll couldn’t reach his players, was a terrible coach, but because of how talented his team was, they always won in the post-season.
You might like the former theory better than the latter, but let’s be clear: the data equally support both theories. To suggest that coaching ability matters more than player ability in the playoffs, or vice versa, is purely speculative, and not proven by anything presented here.