Points per dollar

[Fencer]

The purpose of everything that happens on a football field is either to increase your team's point total or to restrain the opposing team's -- i.e., all football value relates closely to influence on point differential. In the Gostkowski thread, I suggested that the difference between Gost and a cheap kicker might be 1 point per game. If that was close to correct -- and by no means did I do the arithmetic carefully enough to be sure it is -- it would seem that a kicker is probably a good investment.

How about a Revis-quality CB? How much does he cost the opposition in points over the course of a season. How about the WRs that various people do or don't want to see acquired?

Has anybody done a credible job of estimating stats like this?

 

[PatsFanInVa]

Quantifying would be easier if there were no interrelationships among the variables. It's so tempting with a pure points-or-no-points example on the table, like a kicker. Still, distilling the extra mileage other offenses have to go when you reliably kick touchbacks is difficult here.

How much more difficult when you get down to the impact of an offensive lineman? Or for that matter, that O-Line unit's composition, where the guy next to you is dependent on you and vice versa to do your jobs, stunt, pick up defensive wrinkles, etc., and you're all looking to the center for your line-leadership?

I'd love to see it, but the project of boiling down team play to stats seems to always be defeated by reality. There's this stat people float around, value over replacement player, where you take a guy's performance and compare with either a JAG or whoever the backup to the guy is. I don't think that tracks to points.

Simple version: how many points is Brady worth if his O-line gets him 4 seconds per throw? Through the ceiling. How many point is Brady worth where he's got two mississippis drilled into his head, then **** and release? Somewhat fewer. Damned amazing job last year given the time to throw number, but the numbers are interdependent on each other.

And last thing, I am still seeing Pats players interviewed talking about "TRUST" on the NFLN after the SB. That was the word this year from KC onward, as in "trust" the guy next to you. If the D is on the field at the end, "trust" they're taking care of business.

BB boiled down a wealth of problems to one word and forged a team (again) from a band of 53 guys who weren't doing that well. It went top down... Way back in the KC-Cincinnati comparison, or shortly thereafter, I broke down his targets... Now granted, that changes with what the D gives you, but bottom line, I kicked out the TEs and RBs, and looked at how many targets and catches for receivers. It went straight up after the "Trust" statement. What else could he do? Let the ball go for LaFell or Amendola or even Tyms a time or two. Get rid of the ball at the amount of time available in that play. No more crying over the departed O-Line star. What do we have now, and how far can we take it?

So short answer, no, have not seen that metric. And I hate to be old school like this, but it almost feels like it would be self-defeating even if you got it to work. It would be like pursuing the Randy Ratio back in Minnesota. Start making decisions with that metric, and chaos might ensue.

But with all the stats that are available, I am sure they influence decisions... the human gut sometimes pulls serious boners, and sometimes processes info much more efficiently than a stat-crunch. For example, Marshawn Lynch from the 1 yard line? 1 for 5 that season. But everybody says (now) what Carroll's gut should have told him, right?

Look at goalline stats and you'll find they had very good chances whether throwing or running. We've thrown at the 1 yard line or thereabouts plenty of times, so have they. That's why there are goalline pass plays, to punish a team if it throws 8 jumbo bodies into run contain.

Okay, just a rant on some of the too-contrived stats, this time veering off into the very pertinent ones coaches are always processing.

Everybody uses quantitative data and anecdotal "case study"/"gut" data in the NFL. The latter has much more predictive power than people admit, because all the sample sizes are low and all the variables shift so much. In some ways the sabremetrics-style approach is indispensable, but in other ways forging guys into a team with the best chance of winning and drilling on what to do when are even moreso.

Like your question, this is just a reaction to whether you can control for the necessary variables... sort of an extension of "Why isn't Jeff George the greatest QB of all time" and other "measurables" theories from combines.

 

[Fencer]

Quantifying would be easier if there were no interrelationships among the variables. It's so tempting with a pure points-or-no-points example on the table, like a kicker. Still, distilling the extra mileage other offenses have to go when you reliably kick touchbacks is difficult here.

How much more difficult when you get down to the impact of an offensive lineman? Or for that matter, that O-Line unit's composition, where the guy next to you is dependent on you and vice versa to do your jobs, stunt, pick up defensive wrinkles, etc., and you're all looking to the center for your line-leadership?

I'd love to see it, but the project of boiling down team play to stats seems to always be defeated by reality. There's this stat people float around, value over replacement player, where you take a guy's performance and compare with either a JAG or whoever the backup to the guy is. I don't think that tracks to points.

Simple version: how many points is Brady worth if his O-line gets him 4 seconds per throw? Through the ceiling. How many point is Brady worth where he's got two mississippis drilled into his head, then **** and release? Somewhat fewer. Damned amazing job last year given the time to throw number, but the numbers are interdependent on each other.

And last thing, I am still seeing Pats players interviewed talking about "TRUST" on the NFLN after the SB. That was the word this year from KC onward, as in "trust" the guy next to you. If the D is on the field at the end, "trust" they're taking care of business.

BB boiled down a wealth of problems to one word and forged a team (again) from a band of 53 guys who weren't doing that well. It went top down... Way back in the KC-Cincinnati comparison, or shortly thereafter, I broke down his targets... Now granted, that changes with what the D gives you, but bottom line, I kicked out the TEs and RBs, and looked at how many targets and catches for receivers. It went straight up after the "Trust" statement. What else could he do? Let the ball go for LaFell or Amendola or even Tyms a time or two. Get rid of the ball at the amount of time available in that play. No more crying over the departed O-Line star. What do we have now, and how far can we take it?

So short answer, no, have not seen that metric. And I hate to be old school like this, but it almost feels like it would be self-defeating even if you got it to work. It would be like pursuing the Randy Ratio back in Minnesota. Start making decisions with that metric, and chaos might ensue.

But with all the stats that are available, I am sure they influence decisions... the human gut sometimes pulls serious boners, and sometimes processes info much more efficiently than a stat-crunch. For example, Marshawn Lynch from the 1 yard line? 1 for 5 that season. But everybody says (now) what Carroll's gut should have told him, right?

Look at goalline stats and you'll find they had very good chances whether throwing or running. We've thrown at the 1 yard line or thereabouts plenty of times, so have they. That's why there are goalline pass plays, to punish a team if it throws 8 jumbo bodies into run contain.

Okay, just a rant on some of the too-contrived stats, this time veering off into the very pertinent ones coaches are always processing.

Everybody uses quantitative data and anecdotal "case study"/"gut" data in the NFL. The latter has much more predictive power than people admit, because all the sample sizes are low and all the variables shift so much. In some ways the sabremetrics-style approach is indispensable, but in other ways forging guys into a team with the best chance of winning and drilling on what to do when are even moreso.

Like your question, this is just a reaction to whether you can control for the necessary variables... sort of an extension of "Why isn't Jeff George the greatest QB of all time" and other "measurables" theories from combines.

You're right that it's incredibly difficult.

Still, there ARE stats as to how much success/failure/disaster on particular plays contributes -- of course only in expected value -- to points on the board. (This could be closely related to the win-probabilities that have been discussed quite a bit in connection with the Super Bowl's Q4 drives.) Those are team results all the way through, so they're not really subject to the concerns you were raising about teamwork and trust.

From there it's a matter of judging "If we have Guy X vs. the alternative, how many more plays of what magnitude do we expect?" That part, for your reasons about teamwork, fit and situations, MUST include a considerable amount of intuition or gut feel. Still, I think it's as interesting a discussion as any.

E.g., if the Pats lose Revis, how many more pass plays of what magnitude will they give up if they don't sacrifice any run defense to compensate? Assuming they do rebalance their defense, is the harm from losing Revis significantly less?

 

[PatsFanInVa]

Revis may be a great example.

In the cold light of day, we can look at performance alone and say "Asante Samuel missed an interception that would have won us a SB." You cannot say "And Revis made that interception."

But you can say (as you point out) that the throw got forced, they knew that Revis was a low-percentage target, and they'd rather not throw at Browner either. Then there is the fact that BB worked it out, decided this year he had a D he could trust as a better alternative to Brady with 30 seconds, and what empowered that belief in the red-zone D was that the secondary was sound. All of that would have to go into the point value of "Revis on field over likely replacement."

Looking deeper than just the interdependencies, how do you measure that point value at the championship/super bowl level?

Once again, this is influenced by "Fanism." We focus on the few big plays that can go either way, and in recent memory, have been going against us. Now one went for us. The problem with the Pats is that developing the point value of a guy, even if a science in the regular season, fixes a problem that we don't have (winning enough games to get into the playoffs.) Post-season sample size is vanishingly small, smaller the deeper you go into the post-season.

Sorry to be so negative - it might be a great stat to develop, even if it is more art than science... but then do these numbers change for the post-season? Yes, of course. And even the team that's won more post-season games and gone to (and won) more super bowls than any other in the modern era doesn't have enough year-over-year data to figure out what will work.

You can only figure out the best matchups, the things you want to stop, down, distance, clock, and what you can expect from personnel as a team. That was what won us this SB, what was expected of the Defense. It was a low percentage bet, but a higher percentage bet than giving Brady the ball and 30 seconds. So it came down to trusting the D to come up big, and all percentages aside, they did. Butler jumped the route (and had seen the route in practice, to hear the post-facto talk,) knew what to do, and was able to physically do it.

Sorry. Still on very non-scientific (or at least non-statistical) footing here. I suppose you can pick the opposing offenses week over week, develop points against Pats at each level of opposing offense with and without Revis on the field, but in the absence of injury, the only control you can put in is still situational: In one situation, you pull him from the field anyway -- it's the best bet. In another situation, you put Revis on the field, as the best bet. So we're comparing inherently different situations.

I don't know whether my methodological concerns sink the whole concept any more than my "intangibles" concerns, so maybe the thing to do is sketch out how it would be done... it's the offseason, I'm game.

Want to develop the formula? Or even get started? See how far we get?