TABMathletics concludes its work involved with the Featured Factors of Regression the same way we did last year. Perhaps we’ll do it this way again next year as well. Why? Because we found a solid way to determine basic big-picture regression when it comes to win-loss record. Last year’s discovery of the regression based off of Expected Win Differential (EWD) truly helped see where certain teams were in the pecking order of NFL contenders and pretenders. Thus, we think we can use EWD again in 2014 without problem.
Look at the intro to Expected Win Differential to gain a full understanding behind this method of madness. Bottom line, though, it worked in 2013. Last year’s study projected possible regression in win-loss record for 11 teams. The results were as followed:
- Detroit Lions (+3.85 EWD): Improved from 4-12 to 7-9
- Atlanta Falcons (-3.05 EWD): Declined from 13-3 to 4-12
- Houston Texans (-2.89 EWD): Declined from 12-4 to 2-14
- Indianapolis Colts (-2.74 EWD): Owned 11-5 record both seasons
- Kansas City Chiefs (+2.58 EWD): Improved from 2-14 to 11-5
- Washington Redskins (-2.37 EWD): Declined from 10-6 to 3-13
- Philadelphia Eagles (+2.05 EWD): Improved from 4-12 to 10-6
- Buffalo Bills (+1.84 EWD): Owned 6-10 record both seasons
- New England Patriots (-1.72 EWD): Owned 12-4 record both seasons
- Jacksonville Jaguars (+1.60 EWD): Improved from 2-14 to 4-12
- Pittsburgh Steelers (+1.51 EWD): Owned 8-8 record both seasons
Seven of the 11 teams regressed to some degree in win-loss record as expected. The other four teams maintained their record as is. Generally speaking, this is what we should come to expect from the EWD regression projections. The EWD-implied regression won’t always happen, because win-loss record in any given NFL season is already tenuous, but teams won’t stray too much from expectation.
These cases of regression sufficiently shows how our EWD adjustments worked. But if that’s not convincing enough, take a look at the year-to-year correlations. The correlation between each team’s 2012 true wins and 2013 true wins was 0.1779, while the correlation between each team’s 2012 EWD-adjusted wins and 2013 true wins was 0.3649. Our little system allowed the year-to-year correlation to become more than twice as strong than win-loss record itself. For now, we’ll consider this to be a winning formula.
Now let’s focus on 2014. To calculate the Expected Win Differential and determine which teams may be in line for win-loss record regression, we must take three steps. The first step involves calculating the Pythagorean Win Differential (PWD) of each team (Table 1). Remember that the Pythagorean win formula used here is slightly different that what was originally constructed, as a result of the league’s uptick in scoring. The results of this step can give you an idea which teams are most at risk for regression.
Table 1: Pythagorean Win Differential
| Team | PythW | Record | Differential | Team | PythW | Record | Differential | |
|---|---|---|---|---|---|---|---|---|
| BAL | 7.09 | 8-8 | -0.91 wins | CHI | 7.32 | 8-8 | -0.68 wins | |
| CIN | 11.12 | 11-5 | +0.12 wins | DET | 8.47 | 7-9 | +1.47 wins | |
| CLE | 5.44 | 4-12 | +1.44 wins | GB | 7.75 | 8-7-1 | -0.75 wins | |
| PIT | 8.23 | 8-8 | +0.23 wins | MIN | 6.07 | 5-10-1 | +0.57 wins | |
| BUF | 6.72 | 6-10 | +0.72 wins | DAL | 8.15 | 8-8 | +0.15 wins | |
| MIA | 7.47 | 8-8 | -0.53 wins | NYG | 5.54 | 7-9 | -1.46 wins | |
| NE | 10.53 | 12-4 | -1.47 wins | PHI | 9.39 | 10-6 | -0.61 wins | |
| NYJ | 5.34 | 8-8 | -2.66 wins | WSH | 4.76 | 3-13 | +1.76 wins | |
| HST | 4.14 | 2-14 | +2.14 wins | ATL | 5.87 | 4-12 | +1.87 wins | |
| IND | 9.44 | 11-5 | -1.56 wins | CAR | 11.71 | 12-4 | -0.29 wins | |
| JAX | 3.08 | 4-12 | -0.92 wins | NO | 10.84 | 11-5 | -0.16 wins | |
| TEN | 7.51 | 7-9 | +0.51 wins | TB | 5.23 | 4-12 | +1.23 wins | |
| DEN | 11.71 | 13-3 | -1.29 wins | ARZ | 9.49 | 10-6 | -0.51 wins | |
| KC | 11.12 | 11-5 | +0.12 wins | SF | 11.57 | 12-4 | -0.43 wins | |
| OAK | 4.89 | 4-12 | +0.59 wins | SEA | 12.88 | 13-3 | -0.12 wins | |
| SD | 9.23 | 9-7 | +0.23 wins | STL | 7.57 | 7-9 | +0.57 wins |
PythW: Pythagorean Wins, or (Points Scored ^ 2.4) / ((Points Scored ^ 2.4) + (Points Allowed ^ 2.4))
If you need any indication for how tenuous a team’s win-loss record can be in comparison to Pythagorean record, just look at the Houston Texans. They tied the NFL record for worst decline in a single season after going from 12-4 to 2-14. However, the team only dropped six games in Pythagorean record. This win differential marks a 40 percent decline from the true win differential.
The Texans headline the NFL with the league’s best Pythagorean Win Differential, which should make them a candidate for notable improvement. Fellow “Minus-7” teams in Atlanta (13-3 to 4-12) and Washington (10-6 to 3-13) also post strong PWD totals. On the flip side, the Jets seemed destined for a big fall in 2014. They have by and far the worst PWD from last season.
Moving forward to the next step, Table 2 shows the Turnover Win Impact Regression (TWIr) for each team heading into the 2014 season. Note that we made one adjustment to the TWIr formula, as 2013-scoring-based PWD results dictate a 40 points-per game-differential equivalency. Thus, the denominator is no longer 38 and is now 40. Check out the results.
Table 2: Turnover Win Impact Regression
| Team | Margin | Points | TWIr | Team | Margin | Points | TWIr | |
|---|---|---|---|---|---|---|---|---|
| BAL | -5 | -20 | -0.41 wins | CHI | +5 | +20 | +0.41 wins | |
| CIN | +1 | +4 | +0.08 wins | DET | -12 | -48 | -0.98 wins | |
| CLE | -8 | -32 | -0.65 wins | GB | -3 | -12 | -0.25 wins | |
| PIT | -4 | -16 | -0.33 wins | MIN | -12 | -48 | -0.98 wins | |
| BUF | +3 | +12 | +0.25 wins | DAL | +8 | +24 | +0.65 wins | |
| MIA | -2 | -6 | -0.16 wins | NYG | -15 | -60 | -1.23 wins | |
| NE | +9 | +36 | +0.74 wins | PHI | +12 | +48 | +0.98 wins | |
| NYJ | -14 | -56 | -1.14 wins | WSH | -8 | -32 | -0.65 wins | |
| HST | -20 | -80 | -1.64 wins | ATL | -7 | -28 | -0.57 wins | |
| IND | +13 | +52 | +1.06 wins | CAR | +11 | +44 | +0.90 wins | |
| JAX | -6 | -24 | -0.49 wins | NO | Even | 0 | +0 wins | |
| TEN | Even | 0 | +0 wins | TB | +10 | +40 | +0.82 wins | |
| DEN | Even | 0 | +0 wins | ARZ | -1 | -4 | -0.08 wins | |
| KC | +18 | +72 | +1.47 wins | SF | +12 | +48 | +0.98 wins | |
| OAK | -9 | -36 | -0.74 wins | SEA | +20 | +80 | +1.64 wins | |
| SD | -4 | -16 | -0.33 wins | STL | +8 | +24 | +0.65 wins |
TWIr: 0.8175 * (“Points” / 40)
The Super Bowl champions unsurprisingly posted the best turnover margin in the 2013 season, but that could very much lead to some regression in 2014. Even if the team fields an elite defense once again this season, it does not necessarily mean the Seahawks can equal its takeaway rate from last year. Meanwhile, the league-worst Texans fittingly finished with the worst turnover margin. They will see turnover regression compound with their PWD to set up a 2014 season that features potential for serious improvement.
Now with everything in place, the final step will use PWD + TWI regression to create the Expected Win Differential (EWD) for each team in the upcoming season. Teams above the regression threshold of ±1.454 wins are denoted in blue, and they will be likely to positively regress in win-loss record in 2014. Teams below the threshold are denoted in red, and they will be likely to negatively regress in 2014.
Check out the table below to see where your team stands! In total, 11 teams are outside the threshold of regression.
Table 3: Expected Win Differential
| Team | PWD | TWIr | EWD | Team | PWD | TWIr | EWD | |
|---|---|---|---|---|---|---|---|---|
| BAL | -0.91 | -0.41 | -0.50 wins | CHI | -0.68 | +0.41 | -1.09 wins | |
| CIN | +0.12 | +0.08 | +0.04 wins | DET | +1.47 | -0.98 | +2.45 wins | |
| CLE | +1.44 | -0.65 | +2.09 wins | GB | -0.75 | -0.25 | -0.50 wins | |
| PIT | +0.23 | -0.33 | +0.56 wins | MIN | +0.57 | -0.98 | +1.55 wins | |
| BUF | +0.72 | +0.25 | +0.47 wins | DAL | +0.15 | +0.65 | -0.50 wins | |
| MIA | -0.53 | -0.16 | -0.37 wins | NYG | -1.46 | -1.23 | -0.23 wins | |
| NE | -1.47 | +0.74 | -2.21 wins | PHI | -0.61 | +0.98 | -1.59 wins | |
| NYJ | -2.66 | -1.14 | -1.52 wins | WSH | +1.76 | -0.65 | +2.41 wins | |
| HST | +2.14 | -1.64 | +3.78 wins | ATL | +1.87 | -0.57 | +2.44 wins | |
| IND | -1.56 | +1.06 | -2.62 wins | CAR | -0.29 | +0.90 | -1.19 wins | |
| JAX | -0.92 | -0.49 | -0.43 wins | NO | -0.16 | 0 | -0.16 wins | |
| TEN | +0.51 | 0 | +0.51 wins | TB | +1.23 | +0.82 | +0.41 wins | |
| DEN | -1.29 | 0 | -1.29 wins | ARZ | -0.51 | -0.08 | -0.43 wins | |
| KC | +0.12 | +1.47 | -1.35 wins | SF | -0.43 | +0.98 | -1.41 wins | |
| OAK | +0.59 | -0.74 | +1.33 wins | SEA | -0.12 | +1.64 | -1.76 wins | |
| SD | +0.23 | -0.33 | +0.56 wins | STL | +0.57 | +0.65 | -0.08 wins |
EWD: PWD – TWIr; Note: The threshold for regression is ±1.454 wins.
As expected, the Texans lead the way in this study. Furthermore, bounce back years are unsurprisingly expected for Atlanta and Washington as well. Beyond that, we see three north teams stand out as they earned EWD totals above the regression threshold. News isn’t as good for four playoff teams and the Jets. Teams like the Seahawks and Patriots may be strong enough to survive some decline despite finishing below the regression threshold, but can the same be said for the Colts and Eagles?
After the strong showing from EWD in its inaugural run, this year’s results could very well help to further illustrate the validity of this stat. Let’s keep track of this year’s results and see if future adjustments are needed.