Warrior General Games

Originally posted by Sean1995
Then obviously you didn't read any of my previous posts. Okay, long thing short, stobie's data can be misleading and doing it on one QB can make things more accurate. That's why I did it.

They key words in your statement are: can and can. [Emphasis above is my edit.]

You've entered the world of forecasting errors be using a small sample. Are we going to start getting into t-tests and Type I vs. Type II errors?

How do you propose we normalize for all of the multi-variable co-factors that are implicit when analyzing just one QB?

Originally posted by Xars
Your numbers are virtually identical to mine.

Your fail rate for non-hurries is 12.00% while mine was 11.90%.

Your fail rate for hurries is 17.34% while mine is 17.33%.

We are talking about the exact same statistics (virtually) but you keep looking at success rate rather than fail rate.

The success rate change from 88.00% to 82.66% doesn't seem like a lot, but the fail rate change of 12.00% to 17.34% is.

Offenses are supposed to move the ball. There is supposed to be success.

If a Passing Offense gets the ball at their 20, then they have to make 8 successful first downs to score a TD. Since there are some extra yards gained on completions, lets assume 6 is needed.

With a success rate of 88% (or fail rate of 12%), there's 77.44% chance of getting 2 first downs. There's a 68.15% chance of getting 3 first downs. There's a 59.97% chance of getting 4; 52.77% chance of 5 and 46.44% chance of 6.

Now let's factor in hurries. With a success rate of 82.66%, there's a 68.33% chance of getting 2 first downs; 56.48% chance of 3; 46.69% chance of 4; 38.59% chance of 5 and 31.90% chance of 6.

Now lets assume there are 16 possessions by the Offense in a game. If they all start on the 20 yard line, then the expected TD rate is 46.44% x 16 = 7.43 TDs for normal and 31.90% x 16 = 5.104 TDs for hurried throws.

That's a 31.3% reduction in Touchdowns.

Now each drive isn't 80 yards, but then again neither is each catch always 10+ yards. There are also sacks which yield negative yards and significantly reduce the chance of getting a first down, particularly third down sacks which are pretty much a guarantee of a fail.

So do hurries matter as they are currently in the game? Yes. Do they matter enough? That depends on the individual's judgement.





Are you actually being serious with your arguments? I really can't tell. You keep talking about all non-hurried situations vs all hurried situations. But it's almost impossible to force hurries more than 5 times in a row, you realize that right? You are dramatizing the number difference with your weird exponential logic. We are not talking about whether forcing 50 consecutive hurries makes any difference from forcing none. It's about what one hurry does to one pass. Don't care about what 50 hurries do to result of a game, but you are just emphasizing that.

And yeah, success rate decreasing by 5% means the same as failure rate increasing by 5%, what's the difference?? Why do you keep on focusing on a portion of a portion being higher than another portion of the other portion?

Originally posted by Sean1995
All I see from my calculations is that if Dline is extremely lucky to force consecutive hurries, it still only forces three-and-out 5.34 times more out of 100 first downs (not drives).

You are not going to oppose to this statement though, right? I really hope not.

Are you even being serious right now?

Originally posted by Corndog
Are you even being serious right now?

I really am.

Originally posted by Sean1995
You are not going to oppose to this statement though, right? I really hope not.


I didn't evaluate it. Then again, why is this not significant? Isn't an increase of 5.34 times out of 100 significant? I think so. Why do you not think so? Because you're eyeballing the number, rather than calculating it's effect?


To your earlier comment, Yes, I'm serious and I'm not using weird logic.

Originally posted by Sean1995
And yeah, success rate decreasing by 5% means the same as failure rate increasing by 5%, what's the difference?? Why do you keep on focusing on a portion of a portion being higher than another portion of the other portion?

Depending on the starting point of success and fail, there's either no effective difference or a huge difference.

If the start point is a coin flip (50/50), then modifying the success/fail rate by 5% of the total outcomes changes the mix to 55/45.

But if the starting point is a 1% chance (1 out of 100), then a 5% change in success rates raises it to a 6% success rate.

Think of insurance on your car or house. The insurance company charges 1% of your house's value to protect against fire, hurricanes, etc. On a $200,000 house, that's $2,000 per year in premium.

Now if the loss (risk/fail) rate goes to 6%, then the premium they have to charge is $12,000 per year or 6x more.

Looking at the value of the house minus the premium change is to compare $198,000 to $188,000 and that doesn't seem like much. Upping the premium charge from $2,000 to $12,000 is a huge change.

So yes, perspective matters.

Originally posted by Xars

I didn't evaluate it. Then again, why is this not significant? Isn't an increase of 5.34 times out of 100 significant? I think so. Why do you not think so? Because you're eyeballing the number, rather than calculating it's effect?


Originally posted by Sean1995
All I see from my calculations is that if Dline is extremely lucky to force consecutive hurries, it still only forces three-and-out 5.34 times more out of 100 first downs (not drives).

If every single hurry can make a difference between a new first down to a three-and out 5.34% of the time, I would not be complaining. But it's assumed that three consecutive hurries are forced on QB. If there is a 30% chance that QB is hurried, this happening is 2.7%. We are talking about how much hurries can hurt offense at its best 2.7%. Why are you not getting this?

Originally posted by Xars
Depending on the starting point of success and fail, there's either no effective difference or a huge difference.

If the start point is a coin flip (50/50), then modifying the success/fail rate by 5% of the total outcomes changes the mix to 55/45.

But if the starting point is a 1% chance (1 out of 100), then a 5% change in success rates raises it to a 6% success rate.

Think of insurance on your car or house. The insurance company charges 1% of your house's value to protect against fire, hurricanes, etc. On a $200,000 house, that's $2,000 per year in premium.

Now if the loss (risk/fail) rate goes to 6%, then the premium they have to charge is $12,000 per year or 6x more.

Looking at the value of the house minus the premium change is to compare $198,000 to $188,000 and that doesn't seem like much. Upping the premium charge from $2,000 to $12,000 is a huge change.

So yes, perspective matters.


I am sure my chance of dying because of ebola has increased million times these days. And if it happens, I die, so it's really something I should be worried about.... Or should I? Because I am not. My chance of surviving didn't decrease by not even a little. Is it important that the chance of dying of ebola has become million times higher than it used to be or that the absolute possibility difference of me dying or surviving didn't change a bit. A portion of a portion doesn't matter. It's the absolute difference of the portion that matters.

Originally posted by Sean1995
Then obviously you didn't read any of my previous posts. Okay, long thing short, stobie's data can be misleading and doing it on one QB can make things more accurate. That's why I did it.

I read most of your comments, to begin with. I get most of your points, but I disagree mostly on your approach interpreting the numbers given by Stobie.

I dont think you will find one empiric evaluation, where the authors have the complete data regarding a problem, not just a sample, and they leave out most of it to do a statistical research. Do a significance test, if you want to see, if the difference between no-hurried passes and nurried passes shows a statistical significant difference between each other. But just picking the data that might support your theory is definitively not the base for a objective disscusion about the effectiveness of burrying passes.

Originally posted by Evil Sports Agent
I read most of your comments, to begin with. I get most of your points, but I disagree mostly on your approach interpreting the numbers given by Stobie.

I dont think you will find one empiric evaluation, where the authors have the complete data regarding a problem, not just a sample, and they leave out most of it to do a statistical research. Do a significance test, if you want to see, if the difference between no-hurried passes and nurried passes shows a statistical significant difference between each other. But just picking the data that might support your theory is definitively not the base for a objective disscusion about the effectiveness of burrying passes.

FWIW, I didn't just pick that data that support my theory. It was the first set of data I observed. Besides, if you think that a sample size of 85 and 484 isn't large enough, you didn't learn statistics. Here is the next set of data I got from Corndog's QB. Much better this time: the QB completed 13 out of 27 hurried attempts (48.1%), and completed 64 out of 123(52.0%) in the second half of last season. Yup, CD's QB didn't throw that much last season, so the sample size is a bit smaller. The difference is still smaller than what stobie's stat is suggesting.

Originally posted by Sean1995
FWIW, I didn't just pick that data that support my theory. It was the first set of data I observed. Besides, if you think that a sample size of 85 and 484 isn't large enough, you didn't learn statistics. Here is the next set of data I got from Corndog's QB. Much better this time: the QB completed 13 out of 27 hurried attempts (48.1%), and completed 64 out of 123(52.0%) in the second half of last season. Yup, CD's QB didn't throw that much last season, so the sample size is a bit smaller. The difference is still smaller than what stobie's stat is suggesting.

LOL!

Its not about the absolut number its about the relative number. 569 out of 377427 is way to small of a sample size.

Originally posted by Evil Sports Agent
LOL!

Its not about the absolut number its about the relative number. 569 out of 377427 is way to small of a sample size.

Originally posted by Sean1995
you didn't learn statistics.

Look up the definition for sample. If 377427 is the population, and 569 is good enough as a sample.

Originally posted by Sean1995
Look up the definition for sample. If 377427 is the population, and 569 is good enough as a sample.

Depends on the significance level you use.

Outside of all your arguing, what needs to happen is this.

WR's need to catch in traffic less, be it through improved PD rolls or otherwise.
WR's need to separate easier. Greater scale of speed or more effective headfakes/pump fakes

OL need to pass protect better when built for it. Run or Balanced OL's pass protect just as well as a line built to protect the pass.

Hurries need to have a larger impact on throw accuracy
Non hurried passes need to be better just all around.

CB's need to react to errant throws better
CB's need to drive on the ball when they're facing the QB and hardly moving. Right now they almost never put themselves in a winning position.

Originally posted by Adderfist
Outside of all your arguing, what needs to happen is this.

WR's need to catch in traffic less, be it through improved PD rolls or otherwise.
WR's need to separate easier. Greater scale of speed or more effective headfakes/pump fakes

OL need to pass protect better when built for it. Run or Balanced OL's pass protect just as well as a line built to protect the pass.

Hurries need to have a larger impact on throw accuracy
Non hurried passes need to be better just all around.

CB's need to react to errant throws better
CB's need to drive on the ball when they're facing the QB and hardly moving. Right now they almost never put themselves in a winning position.

I agree mostly.

Originally posted by Galithor
When you're talking about completion % from hurries, you need to account for the fact that hurries are by and large the result of blitzing. Single coverage, especially when it's on WR3 by a FS or the TE by a SS is pretty trivial to beat at a very high %. Safeties just typically aren't quite the M2M beasts that CBs are on the outside WRs.

Forcing the QB to dump off to the TE before he reaches the depth of the Strong Safety is one of the surest ways to ensure the opponent has a big game at TE. Blitzes cause that to happen the most.


Or you can just play zone defense and run a ton of zone overloaded blitzes.