|
Sour stats from Elias
Cold, Hard Football Facts for August 23, 2006
By the Cold, Hard Football Facts staff
In the orchard of gridiron analysis, many "pundits" simply pick the low-hanging fruit. It's always easier to chew on existing numbers, blindly swallowing their validity and regurgitating stale arguments accordingly. After all, most of the NFL's historical data is readily available in the pigskin produce aisles, such as those stocked by the Elias Sports Bureau.
Occasionally, though, the time is ripe to challenge even the most revered institutions. It's at these times that "pundits" and public alike should snack instead on the sweet, juicy fruit of the Cold, Hard Football Facts. After all, truth is always in season at our gridiron greenhouse.
Consider, for example, the winning percentages of the game's greatest coaches.
If you were to rely on the Elias Sports Bureau, the football fieldhand that harvests the official data of the NFL, you'd have an incorrect list of the game's winningest coaches. 
Philly fans, for example, might be shocked to find that Andy Reid has a better winning percentage than Green Bay's Earl "Curly" Lambeau, a man of such renown that the NFL's most famous stadium carries his name. Elias inaccurately reports that Lambeau (shown here in statue form) has a better winning percentage than Reid.
Denver fans might be shocked to find that Mike Shanahan has a better winning percentage than Cleveland's Paul Brown, the legend who built not one but two NFL franchises, including the greatest dynasty in pro football history. Elias inaccurately reports that Brown has a better winning percentage than Shanahan.
Miami and Chicago fans might be shocked to find that Don Shula has a better winning percentage than George Halas. Elias inaccurately reports that Halas has a better winning percentage than Shula.
These are not obscure stats that might be overlooked and forgotten like a moldy lemon in the back of the pigskin crisper bin. Lambeau, Brown, Shula and Halas are among the greatest and most recognizable names in the history of the game.
Elias even miscalculates the winning percentage of the man for whom the Super Bowl trophy is named. Elias reports that Vince Lombardi won 73.9 percent of his games. The truth, according to the Cold, Hard Football Facts, is that Lombardi won 72.8 percent of his games.
The power of the peach-pit abacus
The Elias Sports Bureau touts itself as "the world's foremost sports statisticians and historians." It boasts a reputable staff of stats enthusiasts and the most sophisticated number-crunching computers known to sporting man.
Yet even Elias is no match for the omnipotent power of the Cold, Hard Football Facts and our proprietary peach-pit abacus. (Hey, even the 225 Club needs a little fruit from time to time, no matter how vile it tastes.)
We recently reviewed the records of the winningest coaches in NFL history because, well, that's what we do for fun. We began by studying the list that appears in the 2006 NFL Record & Fact Book, which is provided to the league by Elias. Some of the numbers looked a little rotten, so we decided to squeeze them ourselves on the peach-pit abacus.
In many instances, our numbers contradicted those provided by Elias. It turns out our numbers are correct. Those provided by Elias are wrong. (The lists of winningest coaches provided by Elias and by the Cold, Hard Football Facts can be found below. A comparative list can be found here.)
It was a shocking find for the Cold, Hard Football Facts crew. After all, the NFL Record & Fact Book is our Old Farmer's Almanac, the one book we turn to and trust more than any other. Uncovering its flaws felt like walking in on your wife as she serviced the immigrant farmhands.
Comparing apples, oranges and bananas
The problem begins with Elias's use of three different formulas to calculate the winning percentage of coaches.
The problem continues with Elias's failure to simply recalculate winning percentages as it changes the formula – an effort that would take all of five seconds with something as simple as Microsoft Excel, let alone sophisticated statistical databases.
The problem is compounded by the slightly embarrassing discovery that Elias fails to properly round off the winning percentage of the game's most famous coach.
The key, though, is Elias's stubborn insistence on comparing coaches based upon three different formulas. Naturally, you end up with three different sets of results.
Formula 1
Formula 1 completely dismisses ties when compiling a coach's winning percentage. If Coach X posted a 7-3-2 record, his winning percentage is .700 (7/10).
It makes sense to a degree. Baseball uses this kind of formula when it disregards walks in the calculation of a batting average.
Elias uses Formula 1 for all NFL games played through the end of 1971.
Formula 2
Formula 2 includes all ties in a coach's winning percentage. Each tie is calculated as a half-win and a half-loss.
If Coach Y posted a 7-3-2 record, his winning percentage is not .700, as it was in Formula 1. Instead, it's .667 (8/12).
Statistically speaking, Formula 2 makes far more sense than simply dismissing ties. It counts all the games in a coach's history, and that's the reason why Elias and the NFL made the proper decision of adopting it on May 24, 1972.
Formula 3
Instead of just recalculating the winning percentages of all coaches using the new formula, Elias chose to do something that seems infinitely more difficult.
Enter Formula 3, which blends Formula 1 and Formula 2 for a coach whose career spanned the 1971 and 1972 seasons. For these individuals, all the ties on their résumé through 1971 are dismissed from their coaching record. All the ties from 1972 and beyond are counted in their record.
If Coach Z posted a 7-3-2 record with one tie in 1970 (which is dismissed) and another in 1974 (which is counted), his winning percentage is .682 (7.5/11). Again, you can see how a different formula can have a big impact on a coach's career percentage.
According to Elias, three coaches with the identical record will have three different winning percentages, solely because Elias insists on using three different formulas to calculate something as simple as won-loss percentage.
In this example:
- Coach X went 7-3-2. He posted a .700 winning percentage because his career ended before 1972 (7/10 = .700).
- Coach Y went 7-3-2. He posted a .667 winning percentage because his career began in 1972 or later (8/12 = .667).
- Coach Z went 7-3-2. He posted a .682 winning percentage because his career spanned the arbitrary statistical barrier of the 1971 and 1972 seasons (7.5/11 = .682).
So there you have it. Three guys with the same record. Three guys with different winning percentages ... all because Elias, the official historian of the NFL, makes a logically indefensible argument.
Fruitful findings
As you might imagine, Elias's insistence on comparing apples, oranges and bananas has a practical misapplication for our purposes. Most notably, it creates an inaccurate list of coaching leaders.
Here are the 15 coaches with the best records in NFL history – according to the formulas utilized by the Elias Sports Bureau. (We limited ourselves to only those coaches with 100 or more career games.) This list includes the coach, his tenure, his record and the formula (1, 2 or 3) that Elias used to calculate his winning percentage.
The Elias list of the Top 15 coaches:
|
Coach (Tenure) |
Record |
Formula |
Win % |
|
John Madden (1969-78) |
103-32-7 |
3 |
.759 |
|
Vince Lombardi (1959-69) |
96-34-6 |
1 |
.739* |
|
George Allen (1966-77) |
116-47-5 |
3 |
.712 |
|
Blanton Collier (1963-70) |
76-34-2 |
1 |
.691 |
|
George Halas (1922-29; 33-42;
46-55; 58-67) |
318-148-31 |
1 |
.682 |
|
Don Shula (1963-95) |
328-156-6 |
3 |
.677 |
|
Joe Gibbs (81-92; 04-present) |
140-76 |
2 |
.648 |
|
George Seifert (1989-2001) |
114-62 |
2 |
.648 |
|
Tony Dungy (1996-present) |
102-58 |
2 |
.638 |
|
Bill Cowher (1992-present) |
141-82-1 |
2 |
.632 |
|
Curly Lambeau (1921-53) |
226-132-22 |
1 |
.631 |
|
Andy Reid (1999-present) |
70-42 |
2 |
.625 |
|
Paul Brown (1950-62; 68-75) |
166-100-6 |
3 |
.624 |
|
Mike Shanahan (1989-present) |
122-74 |
2 |
.622 |
|
Bud Grant (1967-85) |
158-96-5 |
3 |
.621 |
* Lombardi's record using Formula 1 is actually .738. Elias rounded off incorrectly. The Formula 1 calculation looks like this: (94/130 = .73846 = .738). However, as you'll see, this minor rounding error is nothing compared to the difference you get when you recalculate Lombardi's record using modern Formula 2.
Now, here's the list of all-time winningest coaches as it should look, compiled by the Cold, Hard Football Facts. We take a novel approach. We use the same modern yardstick (Formula 2) to calculate the winning percentage of each coach. Again, we limited ourselves to only those coaches with 100 or more career games.
(You can click here for a side-by-side comparison of the inaccurate Elias list and the accurate Cold, Hard Football Facts list.)
The Cold, Hard Football Facts list of the Top 15 coaches:
|
Coach (Tenure) |
Record |
Formula |
Win % |
|
John Madden (1969-78) |
103-32-7 |
2 |
.750 |
|
Vince Lombardi (1959-69) |
96-34-6 |
2 |
.728 |
|
George Allen (1966-77) |
116-47-5 |
2 |
.705 |
|
Blanton Collier (1963-70) |
76-34-2 |
2 |
.6875 |
|
Don Shula (1963-95) |
328-156-6 |
2 |
.676 |
|
George Halas (1922-29; 33-42;
46-55; 58-67) |
318-148-31 |
2 |
.671 |
|
Joe Gibbs (81-92; 04-present) |
140-76 |
2 |
.648 |
|
George Seifert (1989-2001) |
114-62 |
2 |
.648 |
|
Tony Dungy (1996-present) |
102-58 |
2 |
.638 |
|
Bill Cowher (1992-present) |
141-82-1 |
2 |
.632 |
|
Andy Reid (1999-present) |
70-42 |
2 |
.625 |
|
Curly Lambeau (1921-53) |
226-132-22 |
2 |
.624 |
|
Mike Shanahan (1989-present) |
122-74 |
2 |
.622 |
|
Paul Brown (1950-62; 68-75) |
166-100-6 |
2 |
.621 |
|
Bud Grant (1967-85) |
158-96-5 |
2 |
.620 |
One issue in particular jumps out at us. The old Formula 1 method of simply dismissing ties led to inflated winning percentages for earlier-era coaches. Essentially, modern coaches are held to a higher standard, which by definition makes for an unfair comparison. The winning percentages of coaches like Madden, Lombardi, Allen and Halas drop noticeably when measured by the more rigorous standard of modern coaches.
The response
Randall Liu, the member of NFL Communications Department who edits the NFL Record & Fact Book, was gracious when this issue was brought to him by the Cold, Hard Football Facts. He said he'd consider adding more footnotes to explain the various systems used to calculate winning percentages.
"This is how the (Record & Fact) book gets better," he said. But, he added, "we don't have any inclination of going back and changing the records."
Nobody, of course, is asking the NFL or Elias to change the records. Don Shula will always be 328-156-6. George Halas will always be 318-148-31. What needs to change is the logically indefensible system that compares coaches based upon different systems of measurement.
Why is correcting a flawed system such forbidden fruit?
Alex Stern of the Elias Sports Bureau said the service "hasn't ever addressed changing" the standards of applying different formulas to calculate the winning percentages of different coaches. He said that "statistical rules often evolve" and that you don't apply modern standards to performers from a different era.
Some standards don't change, however. A win is always a win. A loss is always a loss. And a game that ended in a tie actually happened. What needs to change is the unequal application of basic mathematical standards when measuring something as important to the NFL as the winning percentages of its most legendary figures.
Comparing apples, oranges and bananas just seems fruitless.
|
