Stopping the New Romo Myth

[TheDude]

See my follow up post above. I'm agreeing that stats aren't everything but they can be highly correlative and predictive. One must look at the data then use that powerful computer between your ears to interpret the data. But it would be a mistake to dismiss it as not being predictive over time. There are going to be outliers no matter for the most part with stats and football. Exceptions shouldn't be used to dismiss the statistics though.

Exceptions have to be evaluated to compute the underlying hypothesis...too many exceptions can change equations and coeficients. But I digress.

Stats need large sample sizes of controlled variables. Multi year stats may get some significance (Defend pass and be better at passing) but when you look at one game, winning the historical stat and losing the game is little consolation.

In reality, and maybe some one has done the analysis, Brian Burke or someone, the formulae should is likely extensive = A*passing *** + B* def pass + C*Turnover diff(adjusted for inconsequentials like End of half hail marys) + D*scoring/per drive + E*Penalties (adj maybe by down and yardage) + F* Large Plays +G +H. Otherwise, passing and defending the pass or turnovers or other variables might have embedded auto correlation.

Still the GB game likely plays out differently if the last possession before halftime 3rd and 1 is converted. And that decision had cascading impacts doesnt show up in stats and was influenced arguably by a factor outside of the field of play. Just too many intangibles.

Anyway, Im kind of over the topic, its just the stats in this thread (Hypothetical Rogers rating hold ROmo static, # of drives/scoring pro rata for season vs game, etc.) aren't explaining idiosyncrasies of that game. So if you can point to this game with an issue, the Denver 51-48 game with an issue, the Az game a couple years ago, the Sea GB Fail Mary, etc., then you end up in a morass of minutiae that is great fodder for topics and good discussion, but not starts to erode perceived significance.

 

[TheDude]

You can't simply assume that both teams would get shut out on 5 more possessions because "each game is unique."

In this game, the teams had fewer than normal possessions, and scored fewer points than normal. It's not really about understanding how to do a calculation, but about understanding that there may be a connection between those two events.

I never alluded that there would be a shut out with 5 more possessions or score 10 more TDs (5 each). Each game is unique

If GB had a higher score/drive in the season, they by default would be expected to win a heads up game because, outside of an onsides kick, Turnover for TD, etc, possessions in an individual game basically are equal. +/- the end of half start of 3rd quarter potential. You have to make each drive count in the playoffs. the reverse happened to GB the next week when it only got 6 points off of 2 turnovers in side the Seattle 23.

So if equal drives end up projecting a final score of 5-4, 16-15, 19-16, etc. It will always favor the historical leader. Dallas didnt blow the end of half because it had to maintain the lower scoring/per drive.

 

[percyhoward]

It doesn't make a whole lot of sense to run 800 hypotheticals for what the Dallas defense didnt do to lower Rogers to his average (i.e. they played poorly or whatever) when leaving Romo (rating at 143) static.

If I'd been responding to a point about the performance of Green Bay's defense, then it would have made sense to bring Romo's rating into it. I was responding to a comment about the Dallas defense though.

Dallas had opportunities to win this game with the defense playing exactly as it did. They could have done more, but 21 points wasn't enough

That much is obvious, but 21 points on 8 possessions is not the same as 21 points on 10 or 11 possessions, and that is an important point that's being missed. Had the Cowboys not been able to run the ball at will, or had fewer than 4 incompletions, or had the Dallas defense not allowed so many 3rd-down conversions, both teams probably would have had more possessions and scored more points anyway.

These are reasons each team had 8 possessions. They are not reasons to laud the performance of the Dallas defense, or to criticize that of the offense. In fact, the opposite is true.

 

[ufcrules1]

I can see why when one is so entrenched it is difficult to pivot even with the facts.

lol @ facts

 

[TheDude]

If I'd been responding to a point about the performance of Green Bay's defense, then it would have made sense to bring Romo's rating into it. I was responding to a comment about the Dallas defense though.


That much is obvious, but 21 points on 8 possessions is not the same as 21 points on 10 or 11 possessions, and that is an important point that's being missed. Had the Cowboys not been able to run the ball at will, or had fewer than 4 incompletions, or had the Dallas defense not allowed so many 3rd-down conversions, both teams probably would have had more possessions and scored more points anyway.

These are reasons each team had 8 possessions. They are not reasons to laud the performance of the Dallas defense, or to criticize that of the offense. In fact, the opposite is true.

So we lost as the stats predicted.

If I'd been responding to a point about the performance of Green Bay's defense, then it would have made sense to bring Romo's rating into it. I was responding to a comment about the Dallas defense though.


That much is obvious, but 21 points on 8 possessions is not the same as 21 points on 10 or 11 possessions, and that is an important point that's being missed. Had the Cowboys not been able to run the ball at will, or had fewer than 4 incompletions, or had the Dallas defense not allowed so many 3rd-down conversions, both teams probably would have had more possessions and scored more points anyway.

These are reasons each team had 8 possessions. They are not reasons to laud the performance of the Dallas defense, or to criticize that of the offense. In fact, the opposite is true.

Coming away with no points from the GB 26 from a 3rd and 1 (after converting 3 previous in the game) ended a drive that gets lost in the "they got a better than expected 21 on 8 drives" vs "That was a pivot game call/decision and lack of execution that not only ended a drive (with a blocked FG) and gave GB an extra drive where they scored 3 points."

They didnt fail to score on that drive because they were above some statistical pt/drive mean.

Points are what matter. I showed scoring under 23 points only resulted in a 16% winning conversion in the last 99 playoff games. Interestingly,half of those 16 games are accounted for by 3 teams. NYG are 4-0 (2 games each year they won the super bowl), Green Bay 2-0 (same SB year) and the NYJ during the first 2 Sanchez years. Dallas is 0-2 in those games (lost to Sea 21-20 2006, and NYG 21-17 2007). 2 additional wins were from Indy (2-1) and the Ravens (2-2) the ravens lost to the colts twice.

Any decision, mistake, etc that removes "expected points" from a given situation is more valuable on a single game basis than extrapolating historical averages and basically excusing it away.

 

[percyhoward]

So we lost as the stats predicted.

More to the point, we scored at our normal rate. Green Bay scored about 4 more points than their normal rate.

And again, that's pretending the overturn of Dez's catch was correct.

 

[5Stars]

lol @ facts

lol @ you

 

[TheDude]

More to the point, we scored at our normal rate. Green Bay scored about 4 more points than their normal rate.

And again, that's pretending the overturn of Dez's catch was correct.

or not scoring a TD or FG from the GBs 27 on 3rd and 1 when the that scenario results in a score 82% of the time historically

 

[AdamJT13]

21 points, by hook, crook or whatever, generally wont get it done in the playoffs. Since 2006 there have been 99 playoff games. Only 16 games (16.2%) has the winning team scored less than 23 points and no team in the last 2 years has won scoring less than 23.

I dont see the relevance of # of possessions, you simply have to make each drive count.

If you're judging the offensive performance, then you absolutely have to look at the number of possessions and where those possessions started. If you have 14 possessions and start four of them deep in the opponent's territory, then scoring 21 points is a subpar offensive performance -- possibly even a very bad performance. If you have eight possessions and start all of them at your own 20 or worse, then scoring 21 points is an excellent offensive performance.

Could the offense have done better? Of course. But unless the offense scores a touchdown on every possession when it is trying to score (ie., not running out the clock), then every offense could have done better.

The point is that our offense played much better than our defense did in that game. So the defense should get more of the blame.

 

[percyhoward]

or not scoring a TD or FG from the GBs 27 on 3rd and 1 when the that scenario results in a score 82% of the time historically

The number of possessions matters because it shows how Dallas' offense scored at its normal rate, and how GB scored better than its normal rate. IOW, that the Dallas offense outplayed its defense in this game.

Not sure why you're under the impression that this has anything to do with a particular play from the game that you seem to be focusing on.

 

[TheDude]

If you're judging the offensive performance, then you absolutely have to look at the number of possessions and where those possessions started. If you have 14 possessions and start four of them deep in the opponent's territory, then scoring 21 points is a subpar offensive performance -- possibly even a very bad performance. If you have eight possessions and start all of them at your own 20 or worse, then scoring 21 points is an excellent offensive performance.

Could the offense have done better? Of course. But unless the offense scores a touchdown on every possession when it is trying to score (ie., not running out the clock), then every offense could have done better.

The point is that our offense played much better than our defense did in that game. So the defense should get more of the blame.

So is your position that 1st and 10 from the opponents 20 has a significantly different scoring probability for these 3 scenarios if it was a result of :

• It is the 9th play of a drive that started on the teams 20
• It is the 3rd play of a drive that started on the opponents 40
• It was a result of a turnover

If so, then this is succumbing to the roulette history ticker problem

 

[jobberone]

So is your position that 1st and 10 from the opponents 20 has a significantly different scoring probability for these 3 scenarios if it was a result of :

• It is the 9th play of a drive that started on the teams 20
• It is the 3rd play of a drive that started on the opponents 40
• It was a result of a turnover

If so, then this is succumbing to the roulette history ticker problem

You're looking at one game. Stats are really for a larger population to be more predictive.

 

[TheDude]

The number of possessions matters because it shows how Dallas' offense scored at its normal rate, and how GB scored better than its normal rate. IOW, that the Dallas offense outplayed its defense in this game.

Not sure why you're under the impression that this has anything to do with a particular play from the game that you seem to be focusing on.

Because, I would hope you agree, the offense is the primary driver of points. If an offense is in a position that has resulted in a 80% scoring conversion and comes away not only with no score, but that leads to an extra possession for the other team - who scores. There is culpability there. The fact that the drive result maintained an average is irrelevant.

Dallas didnt need to score at a normal rate - it need higher. History also points to 21 points not being enough.

Your average includes this wasted drive. It is calculated based on historical drives. HIstory has no bearing on the next play

 

[TheDude]

You're looking at one game. Stats are really for a larger population to be more predictive.

No you'll see in a minute. waiting for a reply. Unless you want to answer to the affirmative on this.

Also, ever wonder why there are table limits at casinos

 

[AdamJT13]

So is your position that 1st and 10 from the opponents 20 has a significantly different scoring probability for these 3 scenarios if it was a result of :

• It is the 9th play of a drive that started on the teams 20
• It is the 3rd play of a drive that started on the opponents 40
• It was a result of a turnover

If so, then this is succumbing to the roulette history ticker problem

I never said that.

Are you saying that where you start a possession is irrelevant to the chances of scoring on that possession?

 

[TheDude]

no.

you said

If you're judging the offensive performance, then you absolutely have to look at the number of possessions and where those possessions started. If you have 14 possessions and start four of them deep in the opponent's territory, then scoring 21 points is a subpar offensive performance -- possibly even a very bad performance. If you have eight possessions and start all of them at your own 20 or worse, then scoring 21 points is an excellent offensive performance.

Could the offense have done better? Of course. But unless the offense scores a touchdown on every possession when it is trying to score (ie., not running out the clock), then every offense could have done better.

The point is that our offense played much better than our defense did in that game. So the defense should get more of the blame.

Unless I am misreading, you are saying the start of the possession is more important to predict scoring than the ending position.

 

[jobberone]

Exceptions have to be evaluated to compute the underlying hypothesis...too many exceptions can change equations and coeficients. But I digress.

Stats need large sample sizes of controlled variables. Multi year stats may get some significance (Defend pass and be better at passing) but when you look at one game, winning the historical stat and losing the game is little consolation.

In reality, and maybe some one has done the analysis, Brian Burke or someone, the formulae should is likely extensive = A*passing *** + B* def pass + C*Turnover diff(adjusted for inconsequentials like End of half hail marys) + D*scoring/per drive + E*Penalties (adj maybe by down and yardage) + F* Large Plays +G +H. Otherwise, passing and defending the pass or turnovers or other variables might have embedded auto correlation.

Still the GB game likely plays out differently if the last possession before halftime 3rd and 1 is converted. And that decision had cascading impacts doesnt show up in stats and was influenced arguably by a factor outside of the field of play. Just too many intangibles.

Anyway, Im kind of over the topic, its just the stats in this thread (Hypothetical Rogers rating hold ROmo static, # of drives/scoring pro rata for season vs game, etc.) aren't explaining idiosyncrasies of that game. So if you can point to this game with an issue, the Denver 51-48 game with an issue, the Az game a couple years ago, the Sea GB Fail Mary, etc., then you end up in a morass of minutiae that is great fodder for topics and good discussion, but not starts to erode perceived significance.

I don't disagree with you, Adam, or PH. Two are speaking about statistics and you are looking at specific variables that 'bent' the game. If you marry the two then you get a more secular view.

 

[AdamJT13]

Unless I am misreading, you are saying the start of the possession is more important to predict scoring than the ending position.

I didn't say anything about predicting. I don't do predictions.

 

[TheDude]

I didn't say anything about predicting. I don't do predictions.

Well "evaluating a drive" then same difference in this context

 

[AdamJT13]

Well "evaluating a drive" then same difference in this context

What is "more important" depends on what you are evaluating. And I'm still not sure what point you're trying to make.

If Team A starts at its own 10-yard line, drives 89 yards in six plays to get first-and-goal at the 1, it might have the same chance of scoring *from there* as Team B, which took the field at the opponent's 1 after a turnover -- but Team A earned its way into that position, while Team B started from that position. Therefore, if they both score a touchdown, Team A's possession was better than Team B's possession. Are you saying otherwise?