Toxic Differential (1 Viewer)

[Jeff Pasquino]

An excerpt from my weekly Hot Reads article:

Everyone wants to try and anoint the best team this time of year. The playoffs are just weeks away, and with so many teams that are up and down this year it is tougher than most seasons. Rankings by "experts" are all over the map - and if you grab a list from that same person from October it may not just look completely different, it might look like it is from a different season. So how can you figure out the best teams?

Sometimes stats do not lie.

At first I was going to start this article about "turnovers" and possibly "third down efficiency" - but then I found this - Toxic Differential.

I suggest that you watch it, it is pretty interesting.

The original link only has updates through Week 13, but the latest standings can be found here for the Top 10. As they lament on that site, the Bears are well off the pace with a -4 number.

Here are the Top 10 teams:

Rank/Team/Turnover Diff/Big-play Diff/Toxic Diff/W-L1/Pittsburgh/11/26/37/9-32/Philadelphia/15/15/30/8-43/San Diego/-5/32/27/6-64/NY Giants/-1/25/24/8-45/Tampa Bay/9/11/20/7-56t/Green Bay/8/6/14/8-46t/New England/14/E/14/10-28/Atlanta/10/2/12/10-29/Kansas City/7/4/11/8-410/NY Jets/4/5/9/9-3Based on that list above, all of these teams should make the playoffs - but as we know now, Green Bay is in danger after their loss to Detroit. It is interesting to see that both Kansas City and San Diego are on the list, because if they both get in then that means the Jets or the Ravens are going home. Baltimore is not on the list (based on some research they were -8 before Monday night's game against Houston), nor is Jacksonville or Indianapolis. It is not surprising that no NFC West team makes the Top 10 either.

If anyone finds a good source for tracking these numbers for all 32 teams, let me know . I bet it would be quite interesting to see and track over the next few weeks.

 

[two_dollars]

Got any tangible results where this might have been a great predictor in the past?

There's a stat I like to look at when determining play off teams too.. it's called "their record."

 

[Truman]

This seems like something which has a pretty high variance from week-to-week. Would also be interested to know about its predictive history.

 

[kinghill]

I realize this is just based on stats and not results. Will be interested to see how this plays out. Here are the first 3 things that stuck out to me...

New England 6th?

Atlanta behind Tampa even though they beat them twice?

New Orleans is missing

 

[PlasmaDogPlasma]

Teams change over the course of a season because of injuries and/or the development of young players. The stats accumulated early in the year might have very little to do with teams that are taking the field today.

Without doing the research, I'd bet the New England is far and away the leader in this category over the last 4 weeks.

 

[Das Boot]

Definitely interesting, but it seems like mostly common sense.

We've always known that turnovers are a huge predictive indicator.

Adding 20+ yard plays as an additional metric further identifies teams with high powered offenses, but it does interestingly not reward teams that have a strong running game that grinds out yards and clock - but there really aren't very many of those left anyway.

My first thought is what is the optimum yardage number that makes this the best predictor... 20 yards is probably just a round number chosen for aesthetics... maybe using 16 yard plays, or maybe 23 yard plays etc... would fine tune it even a bit more.

 

[Doug Drinen]

Definitely interesting, but it seems like mostly common sense.

We've always known that turnovers are a huge predictive indicator.

Actually I'm pretty sure they're a crummy predictive indicator. What they are is a great retrodictive indicator. That is, the team that had the better TO margin in a particular game wins that game a very large percentage of the time. But is it true that the team with the better TO margin coming into the game wins the game a large percentage of the time? I don't think so, though I admittedly don't have the numbers in front of me. I suspect this Toxic Differential has similar properties. Yes, at the end of the year, teams with good records will nearly always have good Toxic Differentials. But I'd propose the following test: in week 8, if you give me two teams with the same record, will the team with the better ToxDiff have a better record the rest of the way? I would guess not. Maybe I'll have time to check that out today or tomorrow.

 

[Das Boot]

Definitely interesting, but it seems like mostly common sense.

We've always known that turnovers are a huge predictive indicator.

Actually I'm pretty sure they're a crummy predictive indicator. What they are is a great retrodictive indicator. That is, the team that had the better TO margin in a particular game wins that game a very large percentage of the time. But is it true that the team with the better TO margin coming into the game wins the game a large percentage of the time? I don't think so, though I admittedly don't have the numbers in front of me. I suspect this Toxic Differential has similar properties. Yes, at the end of the year, teams with good records will nearly always have good Toxic Differentials. But I'd propose the following test: in week 8, if you give me two teams with the same record, will the team with the better ToxDiff have a better record the rest of the way? I would guess not. Maybe I'll have time to check that out today or tomorrow.

I get and agree with what you're saying, but by the same token, if 2 teams are playing far enough into the season where meaningful stats are established, I'd think if one team has a clearly favorable +/- ratio of turnovers, that team would statistically tend to win more often.Likewise a favorable rating in the 20+ yard +/- measurement reflects both an "explosive" offense, AND a defense that does not give up such plays - and as such I'd expect statistically a team with a well established advantage in that metric to statistically win more often vs. a team weaker in that metric.

Combine the 2 metrics and it seems logically like a sizable advantage going into any game where one team is reasonably ahead of the other in both areas.