PDA

View Full Version : Positional Value vs BPA


philfree
02-18-2009, 11:50 PM
This seems to be a topic that is coming up in all the draft threads. I think most understand PV and some seem to want to live by it and others would rather draft the BPA. I can see this as a conflict that real GMs have to deal with. So I guess the question is, When does BPA supercede PV? Or does it ever?

PhilFree:arrow:
~don't forget that I want a QB with the #3 pick~

Mecca
02-18-2009, 11:54 PM
They aren't separate of each other, the best player available is decided by the value of his position.

philfree
02-19-2009, 12:01 AM
They aren't separate of each other, the best player available is decided by the value of his position.

That may be true some or even alot but I don't think it always is.


PhilFree:arrow:

keg in kc
02-19-2009, 12:01 AM
And different positions will have different values to different teams based on scheme and their roster.

Draft boards are pretty complicated organisms.

Rain Man
02-19-2009, 12:08 AM
I think it's a multiplicative thing.

If you rate a player at 100 and his position value is 90, his score is 9,000.

If you rate a player a 96 and his position value is 93, his score is 8,928.

It's probably not an exact multiplicative thing, though. I think the player rating should get more weight. I'd rather draft a star at a slightly less important position than a solid player at a slightly more important position. I think the key to winning a Super Bowl is to have some players who are legitimate stars and playmakers.

philfree
02-19-2009, 12:23 AM
I think it's a multiplicative thing.

If you rate a player at 100 and his position value is 90, his score is 9,000.

If you rate a player a 96 and his position value is 93, his score is 8,928.

It's probably not an exact multiplicative thing, though. I think the player rating should get more weight. I'd rather draft a star at a slightly less important position than a solid player at a slightly more important position. I think the key to winning a Super Bowl is to have some players who are legitimate stars and playmakers.

Interesting way of looking at with the numbers. There has to be a point where the player with positional values skill level is lesser to a point that a player who has less positional value but a higher skill level has more overall value.

PhilFree:arrow:

keg in kc
02-19-2009, 01:35 AM
The problem with your analogy, Kevin, is that legitimate stars and playmakers are usually the guys that play positions that are considered premium in the draft and free agency. Not that there aren't guys who make a difference at fullback or punter, but they won't ever be as valuable as even middling skill position players, in the same way that a guard or center, no matter how good, will never be as valuable as a good left tackle. Because that position is so key.

To match your example, Tony Gonzalez would be an example of a non-star position player with star quality, but I'm not sure I can think of a whole lot more. In the confines of your discussion, I'd say that's probably what it would take to tempt me into drafting a less important position over a more traditionally valuable one: a hall of fame calibre career potential. And even then, there are people who would argue that Tony wasn't even worth it because of the position he plays, and they'd have been better off getting a slightly weaker player at a position of greater impact.

Other than that, playmaking ability usually is what gets people drafted high...

Chiefnj2
02-19-2009, 08:49 AM
Ideally it would be great to hit on a top positional value pick (QB, LOT, DE/OLB, DT) in the top of the draft. However, I think it is more important to hit on your pick than it is to get positional value correct. Positional value has to be a factor, but if you miss on a top 10 pick it sets you back a lot further. Go back a year and ask yourself which teams are better - Raiders with Russell and the Bucs with Admans (DE), or if you took two guys from non top positions like Peterson and Willis?

philfree
02-19-2009, 02:21 PM
If we go by positional value and we don't get a QB we should take DE Everett Brown.

PhilFree:arrow:

Mecca
02-19-2009, 06:33 PM
If we go by positional value and we don't get a QB we should take DE Everett Brown.

PhilFree:arrow:

He's not worth the pick...

Rain Man
02-19-2009, 06:43 PM
The problem with your analogy, Kevin, is that legitimate stars and playmakers are usually the guys that play positions that are considered premium in the draft and free agency. Not that there aren't guys who make a difference at fullback or punter, but they won't ever be as valuable as even middling skill position players, in the same way that a guard or center, no matter how good, will never be as valuable as a good left tackle. Because that position is so key.

To match your example, Tony Gonzalez would be an example of a non-star position player with star quality, but I'm not sure I can think of a whole lot more. In the confines of your discussion, I'd say that's probably what it would take to tempt me into drafting a less important position over a more traditionally valuable one: a hall of fame calibre career potential. And even then, there are people who would argue that Tony wasn't even worth it because of the position he plays, and they'd have been better off getting a slightly weaker player at a position of greater impact.

Other than that, playmaking ability usually is what gets people drafted high...


But couldn't you compensate by just weighting those positions down? At some point Tony G. is worth more than a QB if the QB isn't strong.

If TE is worth 60 positional points and QB is worth 100 positional points, and if it was a simple multiplicative deal, you would do a fictional draft of a young Tony Gonzalez against QBs as follows:

Tony G = 100 value points x 50 position points = 5000

Joe Montana = 100 value points x 100 position points = 10,000
Drew Brees = 90 value points x 100 position points = 9,000
Chad Pennington = 70 value points x 100 position points = 7,000
Alex Smith = 40 value points x 100 position points = 4,000
Ryan Leaf = 1 value point x 100 position points = 100

So you would take Tony G if Montana, Brees, and Pennington are gone.

I actually think there should be some other factor in there, though, and this is just a simple example. I'd do some sort of shallow exponent of value points because I think the value difference between a 100 and a 90 is bigger than the value difference between a 50 and a 40. So Tony G., being a 100, might get enough add value to pass a Chad Pennington at a value of 70, but not enough to pass a Drew Brees at 90.

That's my theory, anyway.

philfree
02-19-2009, 07:20 PM
He's not worth the pick...

Sure he is because he has positional value and he also plays a position of need. Those things combined give him tremendous value I would think. Why isn't he worth the pick?

PhilFree:arrow:

'Hamas' Jenkins
02-19-2009, 07:49 PM
Sure he is because he has positional value and he also plays a position of need. Those things combined give him tremendous value I would think. Why isn't he worth the pick?

PhilFree:arrow:

God fuck.

'Hamas' Jenkins
02-19-2009, 07:50 PM
But couldn't you compensate by just weighting those positions down? At some point Tony G. is worth more than a QB if the QB isn't strong.

If TE is worth 60 positional points and QB is worth 100 positional points, and if it was a simple multiplicative deal, you would do a fictional draft of a young Tony Gonzalez against QBs as follows:

Tony G = 100 value points x 50 position points = 5000

Joe Montana = 100 value points x 100 position points = 10,000
Drew Brees = 90 value points x 100 position points = 9,000
Chad Pennington = 70 value points x 100 position points = 7,000
Alex Smith = 40 value points x 100 position points = 4,000
Ryan Leaf = 1 value point x 100 position points = 100

So you would take Tony G if Montana, Brees, and Pennington are gone.

I actually think there should be some other factor in there, though, and this is just a simple example. I'd do some sort of shallow exponent of value points because I think the value difference between a 100 and a 90 is bigger than the value difference between a 50 and a 40. So Tony G., being a 100, might get enough add value to pass a Chad Pennington at a value of 70, but not enough to pass a Drew Brees at 90.

That's my theory, anyway.

With the way you cook up comparisons using bullshit numbers, I really worry about the accounting practices of your company.

philfree
02-19-2009, 07:52 PM
God ****.


Why the outrage?


PhilFree:arrow:

HemiEd
02-19-2009, 07:59 PM
With the way you cook up comparisons using bullshit numbers, I really worry about the accounting practices of your company.
I would guess you have more pressing things to worry about.

'Hamas' Jenkins
02-19-2009, 08:02 PM
Why the outrage?


PhilFree:arrow:

It's not outrage, it's despair.

He's not worth the pick because he's not good enough to warrant a top five selection.

It's the same reason why Siavii wasn't worth his pick. Yeah, he played a position of need that had good value where we were picking, but he sucked.

philfree
02-19-2009, 09:46 PM
It's not outrage, it's despair.

He's not worth the pick because he's not good enough to warrant a top five selection.

It's the same reason why Siavii wasn't worth his pick. Yeah, he played a position of need that had good value where we were picking, but he sucked.

So positional value isn't the be all end all of the NFL draft? Hmmm.....I think I see.


PhilFree:arrow:

Mecca
02-19-2009, 09:49 PM
Are you trying to play dumb?

philfree
02-19-2009, 09:54 PM
Are you trying to play dumb?

Nope I'm just trying to figure out at what point BPA overcomes PV.


PhilFree:arrow:

Rain Man
02-19-2009, 11:19 PM
With the way you cook up comparisons using bullshit numbers, I really worry about the accounting practices of your company.


The numbers are hypothetical to discuss a concept. Do you not like the concept, which is strong and valid and impossible to disagree with, or do you just not like the hypothetical numbers?

Anyong Bluth
02-20-2009, 03:00 AM
Except on GIANT problem.....

BPA vs Positional value AREN'T the only things you have to look at.

I mean, Jesus, what about draft # value, or should I say compensation cost.

Because the rookie contracts are completely ****ed up in terms of compensation for a player that hasn't played 1 down in the NFL, certain positions are cap suicide to realistically pick X player as certain positions. What's even worse is that the money is so high at the front and tails off really quickly. The $$ should be curtailed in and given to veteran players that have proven it on the field, and give the guys after the early rounds a better rookie contract.

Point being. This debate is pretty absurd b/c you have to look at other factors that will significantly impact that players value, and the same player should have different value totals from team to team. That still doesn't mean you should reach.

I still say that the argument is strong for using a 3rd pick on a QB.

1st- you're already in the realm of contract numbers for what you're going to have to be paying your QB.

2nd- Obviously you have to think the guy is worthy of taking at that value. Everyone's shitty personal opinion aside, the top 2 QB's in this year's draft have been talked by enough people that it's not going to be some shocker if the Chief's took a QB with their pick. This isn't a,

DV: "Oh shit, they just took our guy Olshanky!" <tissue> </tissue>
<tissue> </tissue>
<tissue> CP: "<mutter> Quick! Look up the roster and see who played next to him"</mutter></tissue>
<tissue><mutter> </mutter></tissue>
<tissue><mutter> Krumrie: "Sa-vi, Sea-vee, See-a-vee??? I don't know, but he's Samoan"</mutter></tissue>
<tissue><mutter> </mutter></tissue>
<tissue><mutter> Gunther: "Hold up, I'm on the phone with Goodcent's ordering the party sub- Penny club fine with everyone?" </mutter></tissue>
<tissue><mutter> </mutter></tissue>
<tissue><mutter> DV: "No Mayo on mine"</mutter></tissue>
<tissue><mutter> </mutter></tissue>
<tissue><mutter> Gunther: "K, alright. So he's Samoan. Tough mother****ers. I can coach him up"</mutter></tissue>
<tissue><mutter> </mutter></tissue>
<tissue><mutter> Krumrie: "Agreed, put this kid through a full season and a couple of my slap drills and he'll be tank!"</mutter></tissue>
<tissue><mutter> </mutter></tissue>
<tissue><mutter> DV: "Well, I'm sold, call in the pick. Alright, whose got the red glitter puff pen?"</mutter></tissue>
<tissue><mutter> </mutter></tissue>
<tissue><mutter> CP: "I'll be done with it in a second.... Oh Jesus, stop crying, it will be 1 second, I'm putting racing stripes on my black leather trenchcoat."</mutter></tissue>
<tissue><mutter></mutter></tissue><tissue><mutter>
I still say, unless they evaluate these 2 guys and just think they're totally not gonna pan out, you benefit 2 fold taking your QB this year.

1st, his draft position and contract will be more in line with what a starting QB's cap hit will be.

2nd, it's going to take most QB's at least 2 or 3 years to fully come into his own in the NFL, and don't give me this Matt Ryan, Flacco bullshit. Great, every once in a while you're going to have some statistical outliers, not to mention, look at the 2 situations closely, and you'll see both organizations did things to help them come along and succeed. Picking up Turner, not asking your QB to do too much b/c you've got one of the top 3 D's in the league, etc, etc ....

With the NFL rule changes protecting QB's even more these days, the increased importance of the passing game in the NFL, not to mention DB's these days can't get away with shit compared to the guys even 5-10 years ago, and you're looking at a position that is even more important that if they are successful, provides you with a player that will play longer than just about any other position minus kickers, punters and a few freaks of nature. Translation: Pick a winner, and you've got even more increased value based on number of years of service.

Back to the point of the maturation of a drafted QB, and in relation to their number of years of service. If you get a QB 1st. You then have a few years to bring in younger talent to build a complete team as your QB starts hitting his prime. I'd say DT is maybe the only other position that can take close to or as long to fully adjust to the Pro level.

So, say you put off going for a QB to groom. You are essentially giving up the opportunity to acquire more talent on the cheap while he learns the ropes, and you conceivably hope that if you can spot talent properly, those other early draft guys taken after your QB should already be up to snuff and you've got a talented young nuculeus that are still playing for cheap, and afford you the ability to bring in targeted FA that are of value.

You don't even have to bother responding with the question, " What if you pick a shitty QB!?! " Answer's simple: rinse and repeat. It's about a 90% certainty that you're not going to cut that guy before 3 years- if not 4. Any QB signing bonus spread over that time, and the fact that the rest of his contract isn't gaurenteed, means you're not looking at salary cap hell just b/c you picked poorly, and would not prevent you from being able to look at taking a 1st round QB again after that 3-4 year window.

Yeah, no one wants to think of having to redo, after taking X QB 3-4 years before, but reality is that you've gotta pull the trigger at some point if you really want to ever reach elite talent and win a ring.</mutter></tissue>

'Hamas' Jenkins
02-20-2009, 03:07 AM
The numbers are hypothetical to discuss a concept. Do you not like the concept, which is strong and valid and impossible to disagree with, or do you just not like the hypothetical numbers?

It's the same shit, different day.

You create an artificial set of numbers, with no basis in reality to the point being discussed, in order to beautify your point.

It's cooking the books.

The concept of assigning a value to a position is fine.

The way in which you do it is not. It relies on revisionist history. It's intellectually dishonest and overly reductive.

You are attempting to predict an unknown future by relying on past externalities.

It's much more complicated than a multiplication table, even with a sliding player value as the player approaches 100.

You have to factor in the player's projected ceiling and floor before they played a down, as well as their positional value, the cost of their contract relative to others and your cap situation, where they will fit on your team with your scheme, and you have to apply your own calculus thereafter.

What Tony Gonzalez became is immaterial to this discussion, because none of these draft picks have become anything yet. Yeah, he became the greatest TE in history, but there was another TE, David LaFluer, picked in the same part of the draft, who was a complete bust. So, your grading must also encompass the potential risk of a given pick busting based upon their skill set and their personality and overall mental and physical health.

If it were as simple as differential calculus, it wouldn't be a problem. But it's not.

Rain Man
02-20-2009, 10:26 AM
You are attempting to predict an unknown future by relying on past externalities.



That's how you build predictive models, though. You look at the past, figure out what went right or wrong, and then attempt to predict the near past by using the information you had in the far past.

And I'm only using pro players because I know nothing about college players and no one knows their value yet.


It's much more complicated than a multiplication table, even with a sliding player value as the player approaches 100.

You have to factor in the player's projected ceiling and floor before they played a down, as well as their positional value, the cost of their contract relative to others and your cap situation, where they will fit on your team with your scheme, and you have to apply your own calculus thereafter.

What Tony Gonzalez became is immaterial to this discussion, because none of these draft picks have become anything yet. Yeah, he became the greatest TE in history, but there was another TE, David LaFluer, picked in the same part of the draft, who was a complete bust. So, your grading must also encompass the potential risk of a given pick busting based upon their skill set and their personality and overall mental and physical health.


I agree with all of this, but that's another equation that feeds into the player value. You use all of that to assign a value to a player.

If it were as simple as differential calculus, it wouldn't be a problem. But it's not.

Hey, hey, hey. You can diss me, but don't EVER diss Sir Isaac Newton.

And actually, if I was a GM I'd be building equations to evaluate potential draftees. If you use the past as a predictive tool, you can identify predictors of success and failure, and you can also offset your own conscious or subconscious biases. Now, it wouldn't necessarily be a situation where you automatically take the guy whose total score (position and value) is 87.43 over the guy who graded out at 87.42, but it would certainly help you identify the players with the best chance of success.

I would argue that your own evaluation system is essentially building an evaluation equation, but you're probably just not writing it down in mathematical form.

lazepoo
02-20-2009, 10:51 AM
That's how you build predictive models, though. You look at the past, figure out what went right or wrong, and then attempt to predict the near past by using the information you had in the far past.

And I'm only using pro players because I know nothing about college players and no one knows their value yet.




I agree with all of this, but that's another equation that feeds into the player value. You use all of that to assign a value to a player.



Hey, hey, hey. You can diss me, but don't EVER diss Sir Isaac Newton.

And actually, if I was a GM I'd be building equations to evaluate potential draftees. If you use the past as a predictive tool, you can identify predictors of success and failure, and you can also offset your own conscious or subconscious biases. Now, it wouldn't necessarily be a situation where you automatically take the guy whose total score (position and value) is 87.43 over the guy who graded out at 87.42, but it would certainly help you identify the players with the best chance of success.

I would argue that your own evaluation system is essentially building an evaluation equation, but you're probably just not writing it down in mathematical form.



This is all true, and assigning values to come up with a rating would certainly help to organize things, but isn't the bias going to come out anyway since the values for certain traits will be subjective? An analysis purely driven by stats wouldn't be very useful since a lot of positional skills aren't quantified, so that leaves a lot of room for bias to creep in and skew the final ratings.

'Hamas' Jenkins
02-20-2009, 11:47 AM
And I'm only using pro players because I know nothing about college players and no one knows their value yet.

And that's why your predictive model is doomed from the outset. Football does not lead itself to statistical analysis. It's not baseball. PECOTA projections work in simplistic sports like baseball, but there are too many variables to accurately create a model, and not enough games to properly construct a sample size.

That's why you don't see people using sabermetric analysis in short baseball series. They are useless.

Rain Man
02-20-2009, 06:00 PM
This is all true, and assigning values to come up with a rating would certainly help to organize things, but isn't the bias going to come out anyway since the values for certain traits will be subjective? An analysis purely driven by stats wouldn't be very useful since a lot of positional skills aren't quantified, so that leaves a lot of room for bias to creep in and skew the final ratings.

Yeah, you're right, but at the same time if I'm a GM loving a player and then I code in his Wonderlic and his cone times and his game statistics and his combine scores and my grade for his interview and his teammates' opinions of him and his neck length or whatever, and if that score comes up and says, "He has a 56 percent chance of success", and then I do the same for another guy that I like but don't love, and the model gives him a 73 percent chance of success, I'd probably step back and really think about why I'm so enamored with Player A, and whether he's really the best choice.

Rain Man
02-20-2009, 06:05 PM
And that's why your predictive model is doomed from the outset. Football does not lead itself to statistical analysis. It's not baseball. PECOTA projections work in simplistic sports like baseball, but there are too many variables to accurately create a model, and not enough games to properly construct a sample size.

That's why you don't see people using sabermetric analysis in short baseball series. They are useless.


Beh. You can model anything and learn from it. I agree that a model in something of short duration like a baseball series will have a higher degree of error, but in that situation you're really playing the odds more than you're predicting an outcome with certainty.

All you're doing in any draft is playing the odds, too. A person's career is certainly too complex to model with certainty, but you can use predictive modeling to increase your odds.

This is why I cringe when former players become head coaches. They think like players. They don't think logically. The same with GMs to some degree. I don't think GMs optimize their use of the data that's in front of them in most cases, and rely on instincts. Instinct may work for some people, but most of us don't have that skill.

'Hamas' Jenkins
02-20-2009, 06:44 PM
Beh. You can model anything and learn from it. I agree that a model in something of short duration like a baseball series will have a higher degree of error, but in that situation you're really playing the odds more than you're predicting an outcome with certainty.

All you're doing in any draft is playing the odds, too. A person's career is certainly too complex to model with certainty, but you can use predictive modeling to increase your odds.

This is why I cringe when former players become head coaches. They think like players. They don't think logically. The same with GMs to some degree. I don't think GMs optimize their use of the data that's in front of them in most cases, and rely on instincts. Instinct may work for some people, but most of us don't have that skill.

The thing is, I don't really disagree with you. I'm a firm believer in using statistical analysis as a predictive behavior of future results. But I believe that it is much more well suited to certain sports, like baseball, because it is so easily measured statistically.

Theoretically, you could create a model that would predict the potential for success of someone of a given size with certain athletic characteristics and an average work ethic. But in football it is so much more complex.

Baseball only works once players are at the pro level and have established themselves. And it's a simple sport. The bust rate for baseball draftees is astronomically higher than football players.

Ultimately, I think that former players do make awful personnel people and coaches, because I don't think that they can divorce themselves from the huddle. You need someone who can look at it dispassionately. You're starting to see a lot more of that in the NBA now, but it too is not nearly as intricate as the NFL.

You would probably love football outsiders. But if you look at their predictive models, they are pretty heavily flawed, and it's about as statistical as you could imagine (and this is with teams of established pros rather than projecting draftees).

Rain Man
02-20-2009, 06:49 PM
The thing is...,

I've occasionally seen Football Outsiders stuff, and while I like some, I think others go a bit too far. (Weren't they the ones saying something like, "a wide receiver taken with picks 26 through 32 is safe"?)

If I was rich or retired or something, it'd be fun to pull several of us together and come up with a model to test. I don't have the time to do it right, though.

tmax63
02-22-2009, 07:50 PM
I'm totally serious here when I say that if you could come up with an equation that beats the current average "bust" rate by a few percentage points you could become a very wealthy man in a hurry. I'd bet that there is more than one team out there that would listen to it when worked up.

Rain Man
02-22-2009, 07:59 PM
I'm totally serious here when I say that if you could come up with an equation that beats the current average "bust" rate by a few percentage points you could become a very wealthy man in a hurry. I'd bet that there is more than one team out there that would listen to it when worked up.


Okay, I'm in.


What are the factors that should be taken into account in such a model?

Obvious ones would include performance in college, performance against top competitors in college, college conference/division, combine performance, and character/interview issues. The last one might be hard to get, but I bet you could get clues by looking at previous media coverage.

tmax63
02-22-2009, 10:27 PM
I would start by looking at commonalities in past known performers, both stars and busts looking back for a period of time starting 4-5 yrs ago and working back as far as you have time and relative ease of access to data then use the recent data (last 4-5 yrs) to test your equation as it develops. By starting with the outer edges of the data pool you should be able to identify the obvious early with minimal bias and effort and work towards the center where there would be more potential subjectivity but you'd have more guidelines set. In my mind I see it as debugging a computer program by setting out the basics then id'ing the anomalies and accounting/correcting them.
Other factors? 1 may be some teammate strength factor for the benefit of playing next to a future all-pro DE or having an all-world wr playing on the other side of the formation to draw the double-teams or if you are the one drawing all the doubles. a 2nd might be an injury index or something to account for durability issues

Chiefnj2
02-23-2009, 08:44 AM
Okay, I'm in.


What are the factors that should be taken into account in such a model?

Obvious ones would include performance in college, performance against top competitors in college, college conference/division, combine performance, and character/interview issues. The last one might be hard to get, but I bet you could get clues by looking at previous media coverage.

I think character/interviews and medical issues are huge, but both are unknown (for the most part) to fans and internet GMs.

There is a story about Coughlin when he was with Jax. They had their draft board set and had Haynesworth as the top DT with Henderson #2. The night before the draft Coughlin spoke with Haynesworth and didn't like the way the interview went. He found him to be immature. The next morning he walked in and switched their spots on the Jags draft board. Jax took Henderson and for a while it looked like the good move. Henderson came in and started at a high level. Haynesworth struggled his first few years but has played a lot better the last two years.

We have no way of knowing what happens in the interviews and/or what each coach is specifically looking for.

Micjones
02-23-2009, 02:55 PM
I think the BPA approach is best employed by a team with few holes.
They come away from the Draft improving the overall quality of their team.

Teams like ours should go BPA at a position of need.
With the draftboard in mind.

Never thought it was an either or... To me it's more a both and.

CoMoChief
02-23-2009, 03:59 PM
They aren't separate of each other, the best player available is decided by the value of his position.

This is NOT true at all.

This just further proves the fact that you would take Sanchez at #3 just because he plays QB.

DaneMcCloud
02-24-2009, 02:28 AM
God fuck.

I hope that Phil is decent at golf.

DaneMcCloud
02-24-2009, 02:31 AM
This is NOT true at all.


Then prove it

philfree
02-24-2009, 09:05 AM
I hope that Phil is decent at golf.

Not to change the subject but what's this a reference too? Who Phil?


PhilFree:arrow:

Chiefnj2
02-24-2009, 09:15 AM
You have to first decide who the best players are and then adjust your board based on positional value and need.

Chiefnj2
02-24-2009, 12:23 PM
Positional value is probably a main reason why you have early draft picks that bust. Instead of just taking the best player, you are tweaking your draft board and moving certain player up because they play a "more important" position. That's probably why you have less busts at LB and OG and C.

StcChief
02-24-2009, 12:28 PM
I think it's a multiplicative thing.

If you rate a player at 100 and his position value is 90, his score is 9,000.

If you rate a player a 96 and his position value is 93, his score is 8,928.

It's probably not an exact multiplicative thing, though. I think the player rating should get more weight. I'd rather draft a star at a slightly less important position than a solid player at a slightly more important position. I think the key to winning a Super Bowl is to have some players who are legitimate stars and playmakers.exactly why Pioli is here now. I trust he knows all about this on team building.

and Freakin' King Carl is finally history. (never gets old saying it)

Kyle DeLexus
02-25-2009, 01:51 AM
Mr. Wraighn Mannnnn isn't far off from what happens according to some of what I have read on the matter, they just don't have a formula and use it as a comparitive analysis. Players are broken down and assigned a value for different skills 1-9 for example. Then each team has different needs and values of different positions that will move them up or down or completely off a board. Character has the same effect on the board with some teams, and no effect at all on other teams. Basically, you take everything and rate the players.

I believe Mr. Rain Man is on to something, but like he said it's more complicated than that and would need tweeking and fine tuning. His artificial numbers are what they are....the scouting department. No one knows what will happen with a player but your scouting department should be able to make an educated guess.

It would not be a 100% deciding factor if something like this formula was created, but it could be a useful comparitive analysis to help make a decision without bias.

Mecca
02-25-2009, 02:01 AM
Positional value is probably a main reason why you have early draft picks that bust. Instead of just taking the best player, you are tweaking your draft board and moving certain player up because they play a "more important" position. That's probably why you have less busts at LB and OG and C.

Because it's pretty easily argued that even if you have the best player in the league at those positions it's not that great of an advantage if at all.

bdeg
02-28-2009, 10:19 AM
Okay, I'm in.


What are the factors that should be taken into account in such a model?

Obvious ones would include performance in college, performance against top competitors in college, college conference/division, combine performance, and character/interview issues. The last one might be hard to get, but I bet you could get clues by looking at previous media coverage.
I was going to say this is probably best left alone, as there are so many factors you could account for there would be no way to assign any values that had any meaning. Plus drafting might just be better done by looking at an individual.

But then I got to thinking about it, and it would just be very complicated. You'd need to come up with a separate model for every position, and have like 100 factors in each model. Some of them would be the same: intelligence - maturity - drive - speed, but when you get into position-specific ratings the limits are endless.

bdeg
02-28-2009, 10:44 AM
A lot of the real work of constructing a model like this would be going back through the last ten years, grading every prospect(or maybe the top100 or so) on every factor and then grading their career so far. From that data you have to make a huge jump in trying to infer what does and does not determine success. And then build your model based on the results.

'Hamas' Jenkins
02-28-2009, 11:04 AM
A lot of the real work of constructing a model like this would be going back through the last ten years, grading every prospect(or maybe the top100 or so) on every factor and then grading their career so far. From that data you have to make a huge jump in trying to infer what does and does not determine success. And then build your model based on the results.

It would be very hard to grade a prospect retroactively after knowing how his career has developed and remain honest about it. With guys like Brady, you'd praise his through the roof intangibles, and with guys like Jared Allen, you'd talk about his motor...but there are guys like that in every draft that don't make it.

bdeg
02-28-2009, 11:10 AM
The model would account for that by the sheer number of guys who came out in that 10 years with those qualities but didn't make it.

bdeg
02-28-2009, 11:11 AM
I think it takes a unique combination for a player like Brady to make it, not just one thing. You couldn't lay it all on "intangibles"

bdeg
02-28-2009, 11:13 AM
One thing the model fails at is accounting for events such as Jared Allen's. He was downgraded coming out because of off-field issues. It was an issue in the pros, but the model has no way of accounting for him quitting drinking and truly exploding as a player.

bdeg
02-28-2009, 11:14 AM
I'd love to see the results as far as defensive linemen go. Is it better to take a raw guy or one that has learned some pass rush moves?

'Hamas' Jenkins
02-28-2009, 11:16 AM
I think it takes a unique combination for a player like Brady to make it, not just one thing. You couldn't lay it all on "intangibles"

But see, that's the thing. Football isn't like baseball. It's so much more complex. A pitcher never has a situation where he has to worry about the accuracy of his pitch while someone is charging the mound.

There are almost an infinite number of variables rather than a handful at most.

bdeg
02-28-2009, 11:18 AM
A pitcher has to worry about the accuracy of his pitch while keeping a runner on first.

But you do realize I made the exact same point right? read page 3

'Hamas' Jenkins
02-28-2009, 11:19 AM
I'd love to see the results as far as defensive linemen go. Is it better to take a raw guy or one that has learned some pass rush moves?

Here's your problem though:

Take a guy like Gholston. There were scouting reports that both said he had a great motor and he lacked a motor.

Well, let's say he works out. Which do you believe?
If he doesn't, which do you believe?

What if he was misused due to scheme (b/c he's an attacking player and Mangini's 3-4 is a read and react system).

bdeg
02-28-2009, 11:20 AM
You couldn't go by "scouting reports" you would have to watch film and evaluate for yourself that's the point.

And yes you'd have to be intelligent grading the career. I definitely believe it's too early to write off Gholston.

'Hamas' Jenkins
02-28-2009, 11:22 AM
You couldn't go by "scouting reports" you would have to watch film and evaluate for yourself that's the point.

And yes you'd have to be intelligent grading the career. I definitely believe it's too early to write off Gholston.

You don't think that scouts weren't watching film when they said he both had a strong motor while others say he lacked a motor?

JASONSAUTO
02-28-2009, 11:27 AM
You don't think that scouts weren't watching film when they said he both had a strong motor while others say he lacked a motor?

yep and thats whats is so fucking confusing. how do these people who actually get good film differ so much in their opinions?

bdeg
02-28-2009, 11:29 AM
So it sounds to me like he has an inconsistent motor, maybe you make a factor for that.

I think an intelligent person should be able to watch a prospect's film and make the correct judgement. One of those was unarguably inaccurate.

'Hamas' Jenkins
02-28-2009, 11:36 AM
So it sounds to me like he has an inconsistent motor, maybe you make a factor for that.

I think an intelligent person should be able to watch a prospect's film and make the correct judgement. One of those was unarguably inaccurate.

The same scouting department that picked Tom Brady thought Tim Rattay was the same quality of prospect.

For all the Pats ability to bring in great vets or make great value picks, they've wasted first rounders on Daniel Graham and Laurence Maroney, and high seconds on Chad Jackson.

Even the best scouts in the world (JJ Cowboys) picked Steve Walsh #1 overall.

And this is the problem. People go back and retroactively stress traits based upon the success that player had in the future, when other players have identical traits, but failed.

bdeg
02-28-2009, 11:39 AM
The same scouting department that picked Tom Brady thought Tim Rattay was the same quality of prospect.

For all the Pats ability to bring in great vets or make great value picks, they've wasted first rounders on Daniel Graham and Laurence Maroney, and high seconds on Chad Jackson.

Even the best scouts in the world (JJ Cowboys) picked Steve Walsh #1 overall.

And this is the problem. People go back and retroactively stress traits based upon the success that player had in the future, when other players have identical traits, but failed.
That's why it's a system that incorporates the good and the bad in order to try to determine relationships. Nothing is 100% going to make a player succeed or fail. Economics has proved that over time the numbers don't lie.

I'm not saying it would be perfect but it could bring to light key patterns.

bdeg
02-28-2009, 11:43 AM
As for players succeeding or failing when they have "identical traits," gimme a break, you think with 100+ factors per position you will ever see identical prospects?

'Hamas' Jenkins
02-28-2009, 11:45 AM
That's why it's a system that incorporates the good and the bad in order to try to determine relationships. Nothing is 100% going to make a player succeed or fail. Economics has proved that over time the numbers don't lie.

I'm not saying it would be perfect but it could bring to light key patterns.

Provided you know what to stress. Football outsiders has a ridiculous methodology and it's ultimately worthless as a predictive model.

PECOTA is like counting cards in black jack
Football outsiders is like try to predict the effects of a butterfly flapping its wings in China.

'Hamas' Jenkins
02-28-2009, 11:46 AM
As for players succeeding or failing when they have "identical traits," gimme a break, you think with 100+ factors per position you will ever see identical prospects?

You do realize that you are just inventing the idea of 100+ factors out of the ether, right?

bdeg
02-28-2009, 11:47 AM
I'd have to check out their model. I doubt it's that well put together. The key is this would all have to be done very intelligently which I can't really explain.

bdeg
02-28-2009, 11:48 AM
You do realize that you are just inventing the idea of 100+ factors out of the ether, right?

What's your best guess?

Or do you want me to figure out a list?

I just said it was what might be necessary to make a very good model.

It wouldn't be huge things like "route running," a receiver might have twenty factors based on how he runs routes. Most likely none of them would be weighted as heavily as something like his motivation.

Rain Man
02-28-2009, 05:33 PM
Here's your problem though:

Take a guy like Gholston. There were scouting reports that both said he had a great motor and he lacked a motor.

Well, let's say he works out. Which do you believe?
If he doesn't, which do you believe?

What if he was misused due to scheme (b/c he's an attacking player and Mangini's 3-4 is a read and react system).

It seems like a model could build in the fact that there's not a unanimous reading.

If all scouting reports say he's high-energy, he gets 100% on that attribute.

If some say he does and others say he doesn't, he gets XX% (less than 100)

If all say he's a three-toed sloth, he gets 0% on that attribute.

You can build in fuzz factors to account for differing opinions.

bdeg
02-28-2009, 07:11 PM
If you're going to go off of draft websites I would pick one which you really agree with his assessments, and just stick with that one. There's too much garbage out there to include scouting reports from a lot of sites.

bdeg
02-28-2009, 07:13 PM
It seems like a model could build in the fact that there's not a unanimous reading.

If all scouting reports say he's high-energy, he gets 100% on that attribute.

If some say he does and others say he doesn't, he gets XX% (less than 100)

If all say he's a three-toed sloth, he gets 0% on that attribute.

You can build in fuzz factors to account for differing opinions.
You wouldn't really need to if all the inputs were graded on a scale of 1-10 already, which is what I had in mind.

Let's take motivation for an example. There are 5 prospects we're looking at. One is Michael Oher, who didn't even read the book written about him. Doesn't seem very motivated. Then we've got Clay Matthews, let's say interviews show he is very motivated to live up to his family name and won't quit til he's in the hall. And 3 players who seemed motivated but didn't stand out.

This is an important trait so we assign it a high value(its all relative), let's say 8.

Oher gets a grade of 3/10
nameless all get 5
Matthews gets a 9

Oher gets 2.4 points, matthews gets 7.2, and everyone else gets 4

These would be added to each prospects' other ratings, and the resulting scores compared.

Actually, now that I think about it something like motivation should almost be multiplied into the equation. Without it, it doesn't matter what else the guy has. But it obviously can't be put in that simply. Im thinking something with the paramaters if motivation < 2 -500 (random big #, only relative to the equation. you'd have to calibrate it based on how you value motivation after your research and the total possible # of points)
3 -250
4 -100
5 - 0
6 - +100 and so on.

In order to figure out how to weight things, you would run a huge multiple regression, or possibly a set of them. With many inputs(prospects) this would tell you what predictor variables had what effects on average. The important part is nailing the predictors and relationships.

Chief Roundup
03-06-2009, 04:53 PM
There also should be a better way of rating players. That we can use as a better guideline as to where each player will evaluate out at.

KCrockaholic
03-07-2009, 12:44 AM
I havent really seen any of the post on this thread, but to answer the question, I would say BPA will overcome PV depending on what type of team you are. For example, the Lions should draft PV. The position with the most PV would be QB. QB is the best way to rebuild a team along with O-line. Because the Lions have so many holes on their team, they cant afford to take BPA in this situation because, guess what? Stafford and Sanchez are not the BPA in this draft. That title would go to Aaron Curry, Jason Smith, or possibly Michael Crabtree...None of these players is the best way to rebuild a foundation of a terrible 0-16 team. Now for a different team such as the Cardinals or Cowboys it would make more sense to go for BPA. These are teams who are not far from a serious SB run. One more big time playermaker could send this team over the edge and help them reach the SB. Positional Value would probably not help them out much unless, the BPA also happens to be great PV...Well thats my thoughts on it anyway.

KCrockaholic
03-07-2009, 12:48 AM
If you're going to go off of draft websites I would pick one which you really agree with his assessments, and just stick with that one. There's too much garbage out there to include scouting reports from a lot of sites.

Best and only site i stick with is nfldraftcountdown.com