IPB
 

Welcome Guest ( Log In | Register )

Buy Eagles tickets or search for cheap sports tickets including all playoff games and super bowl tickets @ LetsGetTickets.com
> The detailed analysis of 4th and 1, 4th and 8
nephillymike
post Sep 30 2017, 09:07 PM
Post #1


Token Axis Member


Group: Members
Posts: 14352
Joined: 23-April 04



Using the two sources from the previous thread, here are the main statistics and probabilities that apply:

http://www.advancedfootballanalytics.com/i...d-epa-explained

http://blog.minitab.com/blog/the-statistic...ing-on-4th-down

Avg net points starting a drive at your own 12 yard line = 0.0 points

Avg net points starting a drive at your opponents 9 yard line = 4.5 points

Avg net points starting a drive at your opponents 34 yard line = 2.9 points

Avg net points starting on your opponents 45 yard line = 2.2 points

Avg net points for a drive starting at your own 42 = 1.3 points

% chance of getting 1st down on 4th and 1 non goal line situation = 70%

% chance of getting 1st down on 4th and 8 non goal line situation = 32%

Expected start of drive when D. Jones punts from around the opp 42 = 12 yd line

The Eagles Dilemma: Go for it 4th and 8 from the NYG 42 or punt it:

Expected points from getting the 1st down = 32% x 2.9 Exp pts from 34 = 0.93 pts
Expected points from getting stopped at the 42 = 68% x (1.3) = (0.88pts)
Net expected net points of going for it = +0.05 negligible
Net expected pts by NYG after punt to the 12 = 0.0

Total net expected points from going for it vs punting = +0.05 - 0.0 = +0.05. Very close to break even, thus why the 42 is right on the edge of the go for it/punt line on the chart.

The model says go for it 4th and 1 from anywhere. How can this be true?

The easiest case to disprove should be the 4th and 1 from your own 9.

If we go for it and get the first after an avg gain of 3 yds to have 1st down from the 12, then what is the average points we can expect if successful = 0.0

If we fail to get it and they get the ball at the 9, 1st and 10, what is the average points the other team should get = 4.5 exp pts.

When we apply the %'S to these outcomes you get:

70% x 0.0 = 0.0
30% x (4.5) = (1.35)
Net exp pts from going for it = (1.35)

Now compare this to punting:

If we punt from the 9 instead, on a average net of 36 yards, they get the ball on the 45 and based on the study, a drive starting there generates (2.2) expected points.

If we punt, the other team is expected to score 2.2 points. If we go for it, the other team is expected to get 1.35 points.

Going for it saves us 2.20-1.35 = 0.85 points.

Why does this seem so counterintuitive? I said previously that the chart was wrong and I intended to crunch some numbers and show how wrong it was! But I was wrrr. Why?

You can see that without factoring in the punt, going for it at the 9 is a loser proposition, netting 1.35 points for the other team. That matches the gut feel. Of course it's a loser. Who the hell goes for it 4th and 1 on the 9?

However, people don't think about the opportunity we give the opposition by punting to them. We are giving them field position where teams have historically scored 2.2 points. THAT is the part people over look. We save giving them an embarrassing 1.35 pts expectation by going for it, only to turn around and give them an acceptable 2.2 points as a result of following sound football coaching principles!

I find this fascinating.

Sure it needs to be tweaked to suit team's strength, momentum, time left etc. Sure you may be able to find other studies with slightly different success rates on 4th down and drive stats based on different time period and that may tweak it slightly. But the big picture is what the base chart shows is sound math and logic for a neutral situation. These are basic in that I make basic assumptions;

we will get 3 yards on the 4th and 1 if we make it.
If stopped at 4th and 8, we won't get any of the 8 yards on our failed attempt
Nobody will return our punt from our 9 for a TD.

These are conservative assumptions, and if I used more realistic assumptions, the decisions to go for it 4th and 1 or 8 would look better.

I was shocked in what I found in this analysis. I had intended to crow about how I was right and the chart was wrong.
Go to the top of the page
 
+Quote Post
4 Pages V   1 2 3 > »   
Start new topic
Replies (1 - 14)
Zero
post Oct 1 2017, 05:11 AM
Post #2


Soldier #1654 of the Eagle Empire


Group: Members
Posts: 13785
Joined: 24-April 04



I didn't follow the links, I also didn't see where it analyzes the potential for the other team scoring if you don't succeed.
Go to the top of the page
 
+Quote Post
nephillymike
post Oct 1 2017, 06:34 AM
Post #3


Token Axis Member


Group: Members
Posts: 14352
Joined: 23-April 04



QUOTE (Zero @ Oct 1 2017, 05:11 AM) *
I didn't follow the links, I also didn't see where it analyzes the potential for the other team scoring if you don't succeed.


See the part about if we fail to get the first from the nine, what are the expected points we are to give up is 4.5 x the chance we don't succeed of 30% = 1.35 points expected to give up if we fail to get the 1st on our nine yard line.

For the 4th and 8, see the expected points we are to give up if they get the ball at their 42, which is 1.3 points x the chance we don't get the 1st down on 4th and 8 = 0.88 points expected to give up if we fail to get the 1st on their 42.

These two are the potential of the other team scoring if we fail in each scenario.

As far as the links, one analyses the chance of getting a first down at various 4th down distances and the other lists the expected net points from drives starting at various places on the field. The drives is NFL data,T he 4th down probability is Big Ten college data. I couldn't find NFL 4th down data but it does match the 32% mentioned by Z Pedey. I suspect That since NFL scores less points, the % may be slightly less.
Go to the top of the page
 
+Quote Post
Zero
post Oct 1 2017, 06:54 AM
Post #4


Soldier #1654 of the Eagle Empire


Group: Members
Posts: 13785
Joined: 24-April 04



QUOTE (nephillymike @ Oct 1 2017, 07:34 AM) *
See the part about if we fail to get the first from the nine, what are the expected points we are to give up is 4.5 x the chance we don't succeed of 30% = 1.35 points expected to give up if we fail to get the 1st on our nine yard line.
This sounds like BS to me. From the 9 teams will only get an average of 1.35 points? Not buying it.
Go to the top of the page
 
+Quote Post
Rick
post Oct 1 2017, 08:18 AM
Post #5


Hall of Famer


Group: Members
Posts: 1486
Joined: 15-September 10



Excellent analysis of the numbers. That's the info I've read previously. The problem people have with the numbers is the same problem I see people have in business all of the time. With the numbers people are looking at each individual drive vs. the big picture of the numbers over time. The key here is the numbers are looking at net points, not points per drive, meaning, with ALL things factored in, historically, a team who has gone for it in a particular situation nets x number of points. This means, in some situations, a team may make it and may not score, may score, etc. or NOT make it and not score or the other team may score. The bottom line is, if you're netting out a positive number of points--over time--it makes sense to go for it in certain situations.

Obviously, this doesn't mean you hold to those numbers every time in every situation. If you're up 5 points with 30 seconds left in the game, you probably shouldn't go for it at 4th and 3 on your 30 yard line. That's where the human has to come in and figure out whether it makes sense.

The numbers don't lie...
Go to the top of the page
 
+Quote Post
nephillymike
post Oct 1 2017, 09:30 AM
Post #6


Token Axis Member


Group: Members
Posts: 14352
Joined: 23-April 04



QUOTE (Zero @ Oct 1 2017, 06:54 AM) *
This sounds like BS to me. From the 9 teams will only get an average of 1.35 points? Not buying it.

No Z.

They get an average of 4.5 points from the 9.

However, because we have a 70% chance of getting the 1st, there is only a 30% chance that happens so you multiply them together to get an expected points of 1.35.

It's probability and expected outcomes math.

Go to the top of the page
 
+Quote Post
Zero
post Oct 1 2017, 09:36 AM
Post #7


Soldier #1654 of the Eagle Empire


Group: Members
Posts: 13785
Joined: 24-April 04



QUOTE (Rick @ Oct 1 2017, 09:18 AM) *
Excellent analysis of the numbers. That's the info I've read previously. The problem people have with the numbers is the same problem I see people have in business all of the time. With the numbers people are looking at each individual drive vs. the big picture of the numbers over time. The key here is the numbers are looking at net points, not points per drive, meaning, with ALL things factored in, historically, a team who has gone for it in a particular situation nets x number of points. This means, in some situations, a team may make it and may not score, may score, etc. or NOT make it and not score or the other team may score. The bottom line is, if you're netting out a positive number of points--over time--it makes sense to go for it in certain situations.

Obviously, this doesn't mean you hold to those numbers every time in every situation. If you're up 5 points with 30 seconds left in the game, you probably shouldn't go for it at 4th and 3 on your 30 yard line. That's where the human has to come in and figure out whether it makes sense.

The numbers don't lie...

I'm still not buying it. If this is based on attempts, how often do teams go for 4th and 1 on their own 9, or for 4th and 8 at the 50 even? Numbers don't lie, but where the numbers come from is relevant to the validity of the results.
Go to the top of the page
 
+Quote Post
nephillymike
post Oct 1 2017, 09:52 AM
Post #8


Token Axis Member


Group: Members
Posts: 14352
Joined: 23-April 04



QUOTE (Zero @ Oct 1 2017, 09:36 AM) *
I'm still not buying it. If this is based on attempts, how often do teams go for 4th and 1 on their own 9, or for 4th and 8 at the 50 even? Numbers don't lie, but where the numbers come from is relevant to the validity of the results.


Head meet sand, sand meet head. tongue.gif

It measures two things independently:

How often teams get the 1st down on 4th and 1 on any place on the field that is not a goal to go situation. It also measures how many net points teams score on drives starting at the 9.

There is plenty historical data on each of those scenarios and you can put them together to get a valid analysis.
Go to the top of the page
 
+Quote Post
Zero
post Oct 1 2017, 10:07 AM
Post #9


Soldier #1654 of the Eagle Empire


Group: Members
Posts: 13785
Joined: 24-April 04



ka.gif ac6.gif
Go to the top of the page
 
+Quote Post
mcnabbulous
post Oct 1 2017, 11:22 AM
Post #10


Hall of Famer


Group: Members
Posts: 9954
Joined: 14-January 07



Cool stuff, Mikey.
Go to the top of the page
 
+Quote Post
Rick
post Oct 2 2017, 05:06 AM
Post #11


Hall of Famer


Group: Members
Posts: 1486
Joined: 15-September 10



QUOTE (nephillymike @ Oct 1 2017, 10:52 AM) *
Head meet sand, sand meet head. tongue.gif

It measures two things independently:

How often teams get the 1st down on 4th and 1 on any place on the field that is not a goal to go situation. It also measures how many net points teams score on drives starting at the 9.

There is plenty historical data on each of those scenarios and you can put them together to get a valid analysis.

Well-said. I don't understand probability math but I understand enough to know, well, it works. Just like how they can poll a relatively-small group of people about something and get an accurate idea of how a larger group feels about a subject.
Go to the top of the page
 
+Quote Post
Pila
post Oct 2 2017, 08:51 AM
Post #12


Eusébio


Group: Members
Posts: 5541
Joined: 28-April 04



QUOTE (Rick @ Oct 2 2017, 10:06 AM) *
Well-said. I don't understand probability math but I understand enough to know, well, it works. Just like how they can poll a relatively-small group of people about something and get an accurate idea of how a larger group feels about a subject.

Ever hear of a Black Swan?

There's a saying in forecasting data science. Bet your change in probability, but save your fortune for the Black Swan.


--------------------
[img]http://dreadart67.files.wordpress.com/2008/03/benfica-2008-ii.jpg[/img]
Go to the top of the page
 
+Quote Post
The Franchise
post Oct 5 2017, 03:52 PM
Post #13


Hall of Famer


Group: Members
Posts: 1266
Joined: 23-November 15



QUOTE (nephillymike @ Oct 1 2017, 07:34 AM) *
The drives is NFL data,T he 4th down probability is Big Ten college data. I couldn't find NFL 4th down data but it does match the 32% mentioned by Z Pedey.


Not to keep kicking this dead horse, but I mentioned this in another thread, and it makes his decision all the more horrifying. Also, the Chargers have the same chance of converting against the Chiefs defense as the Patriots against the Saints defense? If the answer if yes, this exercise is useless. If the answer is no, please explain why.


--------------------
"If it came down to both teams were even, talent-wise, I think the opponent's team would win if it came down to coaching. Andy Reid got out-coached in a lot of games, man, a lot of big games."
Go to the top of the page
 
+Quote Post
Rick
post Oct 5 2017, 04:31 PM
Post #14


Hall of Famer


Group: Members
Posts: 1486
Joined: 15-September 10



QUOTE (The Franchise @ Oct 5 2017, 04:52 PM) *
Not to keep kicking this dead horse, but I mentioned this in another thread, and it makes his decision all the more horrifying. Also, the Chargers have the same chance of converting against the Chiefs defense as the Patriots against the Saints defense? If the answer if yes, this exercise is useless. If the answer is no, please explain why.

Because, again, the numbers look at the PROBABILITY over time, not for one time. Do it enough times and the numbers predict the most-likely outcome but any given play can go either way.
Go to the top of the page
 
+Quote Post
The Franchise
post Oct 5 2017, 04:35 PM
Post #15


Hall of Famer


Group: Members
Posts: 1266
Joined: 23-November 15



QUOTE (Rick @ Oct 5 2017, 04:31 PM) *
Because, again, the numbers look at the PROBABILITY over time, not for one time. Do it enough times and the numbers predict the most-likely outcome but any given play can go either way.


4th down conversions in the Big Ten translate to the probability of a specific NFL team attempting one against another specific NFL team? Or any random NFL team? Do tell. rolleyes.gif


--------------------
"If it came down to both teams were even, talent-wise, I think the opponent's team would win if it came down to coaching. Andy Reid got out-coached in a lot of games, man, a lot of big games."
Go to the top of the page
 
+Quote Post
4 Pages V   1 2 3 > » 
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:
 


NFL Football Tickets    football flash games   
RSS Lo-Fi Version Time is now: 20th October 2017 - 02:04 AM