re: Predictability of LSU on Offense since 2008 - Why the box is stacked

There's an old expression: "It's easier to throw mud into the water than to take it out."

You've certainly muddied up the water. I'll try to make things crystal clear.

quote:

---

My model made no assumptions on wins/losses. An analysis deals with a hypothesis, ONE. Mine was that less miles is predictable, and the data bore that out.

---

First, based on your post history, it's clear you had a "hypothesis" to make Miles look bad. It's very clear that you bash Miles at any opportunity. But you know what, that's okay. You're free to bash Miles and to conjure up any hypothesis of your choosing.

The problem is you went data mining to find some data, any data, that might support your "hypothesis." That's why you created YPC/YPA when you tested to see if LSU was right to increase the number of rushing plays.

Second, the data did not bear out your hypothesis that Miles is predictable. The data does not show that an opponent can predict on any given play when LSU will run or when LSU will pass. The data merely shows that LSU's rushing attempts have tended to increase from 2008 to 2014.

Third, your conclusion is faulty because you specifically fail to consider confounding variables: "My model made no assumptions on wins/losses."

Your "conclusion" that Les is predictable is based on an increase in rushing attempts over several years. However, an equally possible and more plausible explanation for the trend is that LSU's rushing attempts increased because LSU was winning more games and needed to pass less.

quote:

---

Go read my OP again. I SPECIFICALLY REFERENCED YPC:YPA. I made no distinction on YPC, because that wasnt the analysis.

---

Oh I get that YPC alone was not part of your analysis. And I think everyone understands why you wanted to include YPA as part of the denominator in your analysis.

We know that YPC have increased over the years.



YPC are a true indication of rushing success, and as you said: "You would expect [LSU to rush more] if we were running the ball well over this time." So you did not want to use YPC alone because it fricks up your analysis.

When you put YPA into the denominator, it looks like rushing success diminishes over time but only because YPA has tended to increases over time.

YPA
2008: 6.7
2009: 7.0
2010: 6.7
2011: 7.7
2012: 7.3
2013: 10.0
2014: 10.3

Look at that! Only two years did YPA decrease. That's damn near a straight line going up. Of course, becuase it's in the denominator, it makes the number smaller so your trend is down.

That's why you wanted to include 2014. YPA for 2014 is higher than any year on your chart.

In short, you needed to rushing success to look bad ("You would expect this to be true if we were running the ball well over this time. Funny enough, however, the yardages tell a different story: "), so you threw YPA into the denominator to "show" that the "yardages tell a different story."

quote:

---

Moreover, I used the ratio because we are looking at the hypothesis of predictability, NOT yards per play. In that binomial outcome there are two options and BOTH need to be represented in a probabilistic framework for the analysis to make sense.

---

I hope you didn't think anyone would be blinded by your bullshite.

Your second graph was not used to show predictability. You clearly stated in the OP why you presented your second graph:

quote:

---

You would expect [LSU to rush more] if we were running the ball well over this time. Funny enough, however, the yardages tell a different story:

---

Thus, the second graph was supposed to show that LSU was not "running the ball well over this time."

And THAT'S why you threw YPA into the denominator. You wanted to make it look like LSU was NOT rushing well over this time. And yet, as we've seen, YPC increased over this time:





It's not just my choice to use YPC to show rushng success. Anyone who was being honest about measuring rushing success would use that metric.

Putting YPA into the denominator was your attempt to use data mining to skew the analysis.

As HC, and with Gundy as OC, he had a receiver catch 7 TDs IN ONE GAME. Seems impossible to believe given the current climate.

quote:

---

Obviously Gundy's spread attack in the wide open Big 12 has shattered any records previously held, but I'm not sure Miles was at the bottom of the barrel as an OC. I can't find any stats other than that blurb on Miles old OSU bio, so it's hard to say what that means.

---

If you look at his offenses, they were extremely mediocre. 1996 was okay, but look at passing in those years. In 1996 (his best year with regards to passing), they averaged 167ish ypg. The other two years were at the bottom of the barrel in all of division 1A.

He wasn't good. It's known by most rationale people he wasn't a good OC, but people think he's some kind of good offensive coach. Everywhere he's supposed to have a strength has suffered since 2007 with the exception of 2 years (2011/2013).

Our TEs have progressively gotten more and more useless. Our OL has actually regressed. Our offense has regressed.

2 decent offenses in 5 years, not including this one yet. That isn't a good sign for him.

quote:

---

Second, the data did not bear out your hypothesis that Miles is predictable.

---

Twice posted, I couldn't even get our seer to predict something as simple as which down LSU was most likely to pass on.

quote:

---

Ummm... no. As OC of OSU in 1995-1997, Miles was identical to LSU now:

1995 = 43:21.4 rushes:passes (2:1)
1996 = 48.2:25.7 rushes:passes (1.9:1)
1997 = 49.4:16.8 rushes:passes (2.9:1)

---

wow, almost identical to his time at LSU.....

I was referring to his time as head coach there

quote:

---

wow, almost identical to his time at LSU.....

---

There was a lot of issues with Miles time as HC at OSU and regards to the offense. There was supposedly a huge power struggle in the big wigs with Miles and Gundy, and so how much of the offense was Miles versus Gundy's is a bit of an issue.

Let's look at it year by year for rush:pass ratio.

2001 = 1.03:1
2002 = 0.97:1
2003 = 1.50:1
2004 = 2.95:1

So I see a trend as Les took over if anything and matches what he's done while at LSU.

Can you give me the run:pass ratio from his time as OC, and the yds/rush:yds/passattempt too?


I have a neat idea to compare the two regressions and show how he hasnt changed in 30 yrs.

quote:

---

quote:

---

First, based on your post history, it's clear you had a "hypothesis" to make Miles look bad

---

So now instead of arguing the data, you're going to argue against me as a person.

That's a logical fallacy, not even worth responding to, and strike #1.

---

Now there is some bullshite. You cut off my quote. I argued against your method, not your bias. To the contrary, I merely noted your bias and absolved you for it:

quote:

---

First, based on your post history, it's clear you had a "hypothesis" to make Miles look bad. It's very clear that you bash Miles at any opportunity. But you know what, that's okay. You're free to bash Miles and to conjure up any hypothesis of your choosing.

The problem is you went data mining to find some data, any data, that might support your "hypothesis."

---

You cut of my quote to misrepresent the truth. I see a pattern in this thread.

quote:

---

No I didnt. As i laid out, the hypohthesis is predicting a run versus a pass play.

YPC doesnt adress the other side of the binary result. No matter how many times you want to change the correct variable to YPC, it will still be wrong because that wasnt the point of the test.

Strike #2.

---

You didn't use YPC/YPA to show predictability. You used YPC/YPA to show that the increase in rushing plays wasn't justified by the results of those rushing plays:

quote:

---

You would expect this to be true if we were running the ball well over this time. Funny enough, however, the yardages tell a different story:

---

In point of fact, the YPC/YPA chart does not, in any way, shape, or form, address predictability. It's completely irrelevant to predictability. It's merely your attempt to show that the increase in rushing attempts was not justified: "You would expect this to be true if we were running the ball well over this time. Funny enough, however, the yardages tell a different story:"

quote:

---

quote:

---

Third, your conclusion is faulty because you specifically fail to consider confounding variables: "My model made no assumptions on wins/losses."

---

Not really. In fact, the model assumes that a similar output from Miles is expected in terms of wins and losses. His win/loss record isn’t a random walk relationship so having to account for it in the marginal distributional probabilities is meaningless.

---

Of course, your model assumes a similar output in terms of wins and losses. You didn't account for wins and losses. And that's okay, but then you have to consider when forming your conclusion. You never addressed them in forming your conclusion.

Of course, Miles win/loss record is not a random walk relationship. To the contrary, it increases over the time period, just like the rushing attempts increase over the time period. You completely failed to consider it as a confounding factor. Indeed, the highest rush:pass ratio occurs in 2011, LSU's best year in that time period.

quote:

---

I wasnt modeling rushing success. Thanks for proving that no matter how many times it is literally spelled out for you, you STILL cant comprehend it.

---

Clearly, you were not modeling rushing success. You modeled YPC/YPA.

And yes, I know your OP was an attempt to show that Miles is predictable, and that is precisely what you tried to prove in your first graph.

However, you were trying to prove with your second graph that the yardages "showed" that LSU's rushing efforts did not justify the increase in rushing attempts:

quote:

---

You would expect this to be true if we were running the ball well over this time. Funny enough, however, the yardages tell a different story:

---

quote:

---

That was 6 called passes in a row, in the THIRD quarter. Miles actually did make an adjustment in the game, 11 called pass plays to two running plays in the 3rd quarter - LSU ran the ball more in the 4th than the 3rd.

---

And there in lies the problem. Miles is predictable. He'll try to force things once behind, but wait until the 3rd quarter to make adjustments, then blame execution.

It's predictable when you are two TDs down to start passing.

It doesn't go against what the premise of the initial graphs in this topic state. There is little to no creativity in the playcalling. We pass when it is dictated, and only at the last second. If LSU mixes it up more, they will have a better chance for success.