Victory Weighting

NFL overtime, done right

Baseball resolves ties by playing full extra innings. In basketball, tied teams play extra time in 5-minute parcels. Even in soccer and hockey, they get it partly right, with fixed overtime periods preceding those dreaded shootouts. In none of these sports does the game stop at the first score.

But in the NFL, overtime consists of a sudden-death period that ends the first time somebody scores. Sometimes, the result of an NFL overtime has some fairness. When both teams are allowed to possess the ball at least once, they can reasonably claim that the OT period tested all of one team against all of the other. (That's why OT works in those other sports.) Unfortunately, many NFL overtimes end after only one possession. When that happens, overtime tests only one of half of the winning team against half the losing team. Yet, the winning team still gets full credit for the win. That's wrong.

[As of May 2010, sudden-death rules have changed for playoff games. During the playoff the first team to score six points wins. That's an improvement, but it doesn't apply in the regular season. It's still wrong.]

College football teams use a different scheme, giving each team in a tie game an additional possession at the same place (usually the near 25-yard line). The team that scores more points in its possession wins the game. In case of tie, the teams get another possession and try again. That's more fair than sudden-death, but if each team gets more than a couple of chances, it can occasionally take a while.

Let's review: NFL overtime doesn't work fairly, but the best fix, the college OT scheme, takes too long for the NFL to implement. The good news is, I have a solution. Enter Victory Weighting. The basic idea is this:
The winner of an NFL game should get more credit if they do it in regulation.
If overtime is required, the winners shouldn't get full credit for the win. On the other hand, the loser of an overtime should get a little credit for forcing overtime in the first place. To enforce this principle, we just add another column to the standings. In addition to a won-loss record, each team receives a Strength score. Every game played adds up to four points to a team's Strength, based on how the team fared:
  • Regulation win: +4 Strength
  • Overtime win: +3 Strength
  • Tie: +2 Strength
  • Overtime loss: +1 Strength
  • Regulation loss: 0 Strength
Instead of winning percentage, teams are ranked based on their Strength.  If two or more teams have the same Strength, their overall won-loss records serve as the first tiebreaker.  Here's an example from the 2009-10 season:
NFC East RecordStrengthPct.
Philadelphia 11-5-044 0.688
Dallas 11-5-043 0.688
NY Giants 8-8-031 0.500
Washington 4-12-017 0.250
In the official standings, Dallas and Philadelphia tied for the best overall record in the division. Because the Cowboys then won the head-to-head tiebreaker, they became NFC East champions and hosted the Eagles in the wild-card round.

Under Victory Weighting, Philadelphia would have hosted that playoff game. As the team with the highest Strength, the Eagles would have won the NFC East outright, with no need for a tiebreaker. The difference is that while all the Eagles' games ended in regulation, Dallas needed overtime to beat Kansas City early in the season. As a result, the Cowboys earned only 3 Strength points from that win.  In general, Victory Weighting imposes minor changes on the standings, but those could get interesting.

Strength of Schedule and Strength of Victory

This system completely redefines two statistics that the NFL uses for tiebreakers.  For any given team,
  • its Strength of Schedule (Ss) is the sum of its opponents' Strengths over all completed games.  It increases monotonically; that is, it starts at 0 for all teams at the beginning of the season, and increases every week when the team does not defeat a 0-Strength team.
  • its Strength of Victory (Sv) is the sum of its opponents weighted Strengths over all completed games.  Each completed game adds a number of points to a team's Sv equal to the product of (a) the Strength it earned and (b) its opponent's Strength. It starts at 0 for all teams, and increases after every game that does not result in a regulation loss.  
For example, a team that has just defeated an opponent with Strength 20 in regulation would add 4 to its own Strength, 20 to its Strength of Schedule, and 80 (4*20) to its Strength of Victory.  If the team lost the same game in overtime, it would add 1 to its Strength, 20 to its Strength of Schedule, and 20 (1*20) to its Strength of Victory.

By definition, the average Strength of Schedule of all teams at the end of the season is 512.  A team whose final Ss exceeds 532 can claim to have played a tough schedule, while one whose Strength of Schedule finishes below 492 has had a weak one.

Victory Weighting Q&A

Q:  Wouldn't VW reward teams for going for overtime instead of a regulation win?
A:  Maybe.  Win, lose or draw, overtime does add at least one one point to a team's Strength.  Later in the season, playoff contenders can use that to their advantage.  If that extra Strength point helps a team clinch a playoff spot (or just eliminate a rival), then the coach's obvious decision would be to go for the tie late in the fourth quarter, rather than the win.  A team that's behind in the standings, however, must get every Strength point it can muster, so it would be better off avoiding overtime.

Q:  Overtime games between contenders could really screw up playoff races, couldn't they?
A:  Yes -- and they would become more interesting as a result.  If two teams are contending for playoff position, the impact of OT essentially doubles.  The loser of such a game isn't just gaining a Strength point for itself; it's stealing that point from the winner.

Q:  What happens to the tiebreakers?
A:  In one respect, nothing.  Victory Weighting isn't meant to address how to break ties.  The change actually resides in the definition of the word "record."  In Victory Weighting, a team's record has two components:
  1. its Strength
  2. its winning percentage (the traditional definition of "record")
If two or more teams have the same Strength, the one with the best winning percentage has the best record.  At any tiebreaking stage that depends on total or partial "records," the appropriate Strengths are compared, then (if necessary) winning percentages.

That's it.  The tiebreakers themselves (head-to-head, division records, common opponents, etc.) stay the same, and so does the order in which they apply.

Q:  Don't playoff scenarios get more complicated?
A:  At first glance, yes.  Instead of three possible results (win, loss, tie), Victory Weighting offers five for each game (regulation win, overtime win, tie, overtime loss, regulation loss).  For a two-team race, the number of distinct scenarios for each week increases from 9 to 25; for three teams, from 27 to 125; and so on.

That sounds ominous, but remember three things:
  • Overtime itself occurs in only about 3% of NFL games, and ties happen almost literally once a decade.
  • If a team needs a certain result just to avoid elimination, then from its perspective, the scenarios reduce from 9 down to 4 (or from 27 to 8, or so on).
  • Victory Weighting produces fewer ties in the standings.
In other words, we're adding a lot of scenarios, but few of them are likely to occur.  Furthermore, the scenarios themselves are simpler, since there will be fewer ties to break.  You can still use the traditional win-loss-tie model to compute the scenarios, but they won't be exact anymore.  They will, however, provide some very good guesses.

Last Updated:  3 September 2012

No comments: