Jump to content

Random Packer News & Notes


Leader

Recommended Posts

1 hour ago, AlNFL19 said:

@AlexGreen#20

Okay, I fixed it with the correct number for Cooper (actually 7 average AV in Years 3-4, not 4).

The model is based on a simple idea that might not make any sense. What is does is projects a 

  • Projected Approximate Value
  • Bust Chance (Chance of Ending up with <5.0 AV)
  • Pro Bowl Chance (Chance of Hitting or Exceeding 10.0 AV)

The first two are built off of the second one.

This is the formula for Projected Approximate Value:

AV = ((0.32*K)+(0.38*M)+(0.42*O)+(0.44*Q)+(0.30*S)+(0.34*U)+(0.31*W)+(0.12*Y))/(0.32+0.38+0.42+0.44+0.30+0.34+0.31+0.12)

K = -0.042948*(Draft Position) + 6.31038

M = 0.116655*(Receiving Yards / Team Attempts Per Game) + 0.340037

O = 11.61997*(Touchdowns / Team Attempts Per Game) + 0.822505

Q = 2.369886*(Receptions / Team Attempts Per Game) - 0.964705

S = 0.003136*(Receiving Yards) + 0.907168

U = 0.278903*(Touchdowns) + 1.596048

W = 0.044089*(Receptions) + 1.059664

Y = 3.036771*(Weight/Height) - 4.0404245

All this is is a collection of the lines of best fit between the given statistic and Approximate Value. The final AV calculation takes all of the line of best fit data and finds a weighted average. The numbers that each line of best fit data point (K, M, O, etc.) are weighted by are their respective correlation coefficients. The final correlation coefficient from projection to actual AV is 0.47. 

This is the formula for Bust Chance:

Bust Chance = -22.363786*(Projected AV) + 150.260163

(Obviously with a lower bound of zero and an upper bound of 100).

This is the line of best fit between the projected AV and a column of data that is all either 0s or 100s. For each player that was a "bust", this number was a 100. For each non-bust, it was a zero.

Pro Bowl Chance = 5.122483*(Projected AV) - 12.712742

See above, but this time in terms of "Pro Bowl" designations.

 

The model was built on this (2006-2015) data, so I guess calling it a "track record" or whatever was misleading. I'm sorry. Anyway, that's the basic idea, and hopefully it'll remain somewhat useful. Obviously, it won't fit the future data as closely, but the hope is that it remains at a decently high number (and at least higher than draft position).

So you picked some stats, ran a regression on each of the stats, and then ran a weighted average on your compilation. When you compare 5 years of data to the next sequence of years, are you seeing noticeable diversions from the estimated outcomes?

Link to comment
Share on other sites

47 minutes ago, AlexGreen#20 said:

So you picked some stats, ran a regression on each of the stats, and then ran a weighted average on your compilation. When you compare 5 years of data to the next sequence of years, are you seeing noticeable diversions from the estimated outcomes?

To answer your question, I redid the model based on only 2006-2010 and here were the results:

The five-year model only correlated to out-of-sample AV at 0.35 (at 0.33 with draft position removed). However, that was still a step up from draft position, which correlated at -0.28 on out-of-sample data. Clearly, these results aren't great and could obviously get better, but I think it's a start (I consider being a bit ahead of draft position a win). Hopefully the added five years of data from 2011 to 2015 will help to level off the model and make it more accurate going forward.

Link to comment
Share on other sites

1 hour ago, AlNFL19 said:

To answer your question, I redid the model based on only 2006-2010 and here were the results:

The five-year model only correlated to out-of-sample AV at 0.35 (at 0.33 with draft position removed). However, that was still a step up from draft position, which correlated at -0.28 on out-of-sample data. Clearly, these results aren't great and could obviously get better, but I think it's a start (I consider being a bit ahead of draft position a win). Hopefully the added five years of data from 2011 to 2015 will help to level off the model and make it more accurate going forward.

How are you running the correlation when comparing draft position?

Link to comment
Share on other sites

1 minute ago, AlNFL19 said:

Just straight up finding the correlation of draft position to NFL AV (x=draft pos, y=AV).

Isn't that going to lead to a different correlation measurement than what you're doing here? I'm not sure you're comparing apples and oranges here.

Link to comment
Share on other sites

Just now, AlexGreen#20 said:

Isn't that going to lead to a different correlation measurement than what you're doing here? I'm not sure you're comparing apples and oranges here.

I don't think so.

I just plugged them into a correlation coefficient calculator as

(Projected AV, Real AV)

and 

(Draft Position, Real AV)

Link to comment
Share on other sites

1 minute ago, AlNFL19 said:

I don't think so.

I just plugged them into a correlation coefficient calculator as

(Projected AV, Real AV)

and 

(Draft Position, Real AV)

What's your correlation degree?

I feel like a linear correlation and a second degree correlation are going to be massively different. The way the draft tends to break down, it tends to but be linear.

Link to comment
Share on other sites

3 minutes ago, AlexGreen#20 said:

What's your correlation degree?

I feel like a linear correlation and a second degree correlation are going to be massively different. The way the draft tends to break down, it tends to but be linear.

That's a good point. It's linear, though. Probably could be improved.

Link to comment
Share on other sites

12 minutes ago, AlNFL19 said:

That's a good point. It's linear, though. Probably could be improved.

I don't think I've ever done wide receiver positionally, but I want to say the draft in general is best modeled by a 3rd degree equation. Obviously 4-6 are better but not noticeably enough to make the extra computation worth it.

  • Like 1
Link to comment
Share on other sites

If you go back to the original Waldo formula thread, we criticized and over-analyzed that as well.  So, don't be discouraged by what looks like hostility.  We are trying to understand and fine-tune the method, and hopefully make it better.  I don't recall any of the guys that pass that threshold failing in the league, for on the field reasons.

Are these the NFL or college receiving numbers? If they are NFL numbers, what do the same set of players show if we only use their college production?  Then, is there a factor that makes Greg Jennings an 11.5 and  Chad Jackson a 0?  Or are we just determining which players were the best draft value?

To compare to Waldo's formula: 

In a re-draft of 2006, Greg Jennings should have been a first round pick, that plays out in stats and player ranking across that draft class.  It is also shown in the results here.  In a 2011 re-draft, Justin Houston should have been a first round pick, that also plays out in stats and player ranking. The difference is Waldo's formula predicted it.

Is this a predictive formula, or can it be made into one?

  • Like 1
Link to comment
Share on other sites

5 hours ago, ThatJerkDave said:

If you go back to the original Waldo formula thread, we criticized and over-analyzed that as well.  So, don't be discouraged by what looks like hostility.  We are trying to understand and fine-tune the method, and hopefully make it better.  I don't recall any of the guys that pass that threshold failing in the league, for on the field reasons.

Are these the NFL or college receiving numbers? If they are NFL numbers, what do the same set of players show if we only use their college production?  Then, is there a factor that makes Greg Jennings an 11.5 and  Chad Jackson a 0?  Or are we just determining which players were the best draft value?

To compare to Waldo's formula: 

In a re-draft of 2006, Greg Jennings should have been a first round pick, that plays out in stats and player ranking across that draft class.  It is also shown in the results here.  In a 2011 re-draft, Justin Houston should have been a first round pick, that also plays out in stats and player ranking. The difference is Waldo's formula predicted it.

Is this a predictive formula, or can it be made into one?

A good predictive formula is the goal. 

Jennings is an 11.5 because that’s the average of his Approximate Value from his third and fourth seasons. That’s what all the AV numbers are. The model that gives out a projection of AV and Bust chance and all that is based on college statistics.  

Link to comment
Share on other sites

The Packers OTA are now underway.

The offence started out with the 2WR, 1 TE, 1 RB, 1FB set. (If there's a name for this please let me know as I don't have a clue).

This is interesting as that mirrors how the 49ers run their offence which I thought might happen.

We're used to seeing 3 or even 4 WRs but I theorised that we would be seeing more base 2 WR sets, the TE would essentially be the 3rd receiver, our slot guy. The 4th receiver would be our RB. The downside to this that by making the TE a receiver we would lose a blocker on the line but that is where the FB comes in. 

The 49ers have a competent TE and FB who can block and catch which helps make their offence a little more unpredictable. Unfortunately for us we have a one dimensional TE and we have a fullback with unproven blocking ability.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...