Valuation of players and draft picks

[torontogrudlies]

The rough formula for valuation of draft picks, which I found during a Google search:

Code: Select all

value=(-0.855*LN(picknumber)+4.9663) *100


This results in the #1 overall pick being valued at 496.63, #2 at 437.37, 21st pick at 236.32, 33rd pick 197.68, etc.

I liked this one better than the old one which began at about 3000, because there was no real logic used to create that one. For this one, a "career approximate value" statistic was used.... basically, this estimates the value of a particular player over the course of his career. The #1 picks analyzed thus averaged a value of about 500, #2 was about 437, and so on.

Good point was raised in the original thread... if the #1 pick is worth 500, an elite QB should be much more than 300 as originally speculated.


http://harvardsportsanalysis.wordpress. ... aft-picks/

[torontogrudlies]

The coaches also have ratings for each position, we could perhaps figure that in?

The coaches have a rating for each position but that doesn't tell us what qualities they are looking for at each position, unless you're talking about adding additional values as part of Tuesday Morning. We could make the Coach and GM values be averaged and have the weighting change based on how much approval rating they have. That way a lame duck coach has less control over what happens to the team while Bill Belicheck has near total control.

I was looking at the OVR formulas and unless we want to mess with the PRDH and PRDL values and change the starting values we can't change the "total weight" as listed in the spreadsheet. So if we want to make PBK and CAT weighted higher we have to decrease something accordingly, so when we design this we have to come up with traits which are valued as mutually exclusive. So the more pass happy a coach is the more CAT and PBK factor into it and the less STR and BTK do.

We should decide if we want this to be based only on pass run percentage or some other created stat that is more specific to each position. Pass Run percentage is the easiest and works pretty well for most things. But it isn't as nuanced as individual position ratings could be. You should decide soon if you think run/pass is good enough or if we want to try to make something more complex.

Here's another thing I thought of very early on (months ago) but didn't really think of again till just now.

We not only need to consider these ratings and setting them to value, we also need to simulate how much of this information is "known." A scout for any team can likely dig up a list of everyone's 40 time, or how much a particular lineman bench presses, but what about the third-year QB who only took snaps in three games? Not everybody is going to know that he's got a 94 throwing accuracy. His own team is going to know significantly more, and teams with excellent scouting will have some information but there will be question marks. In fact, this was one of the reasons I designed a GM rating for "Risk"...

Another guideline I wanted to use: I want to avoid using random number generation/chance. When we do use it, it should be done with a seed, so that the same series of "random" numbers will result. Otherwise, the trade which Team A balked at in week 2, they're jumping at in week 3, although nothing has really changed.

[Drizzt_13]

Don't worry about what the total numbers come to now, you can always multiply the output of the evaluation formula by a constant so that it has the appropriate value compared to draft picks. All we should be worried about now is ratio of values not the actual values. If a 99QB is only 1.5 times as valuable as an 85 QB then we have a problem no matter what the actual values are. But if a 99 QB is worth 3 or 4 times as much than it's okay even if we only value them and 100 and 25 points respectively, we can just multiply them both by 6 or something to fit with the draft value chart.

A quick comment about draft studies, one thing you have to consider is that teams picking towards the top tend to be at the top because they are bad at finding and developing talent, while consistently good teams tend to pick towards the bottom. Think of what the value of a quarterback like Sanchez would be if he was on a team with a line like the bears rather then a team with one of the best o-lines and a solid defense. Also one thing that has changed recently is that top picks are no longer paid ridiculous sums of money, which changes a lot of old studies which found that mid round picks were better than early ones because if you made a mistake at the top you lost a lot of cap space for the next 5-6 years.

I agree that the 3000 chart is completely irrational but I feel we should keep something close to that because a lot of teams still really overvalue top picks. Look at how much griffin went for, look at the fact that no ones willing to give up a first round pick for Mike Wallace, even though he's a young proven player and most draft prospects probably won't be that good. I think we should overvalue draftpicks because a lot of teams clearly value them more than they're worth, it's an issue of being closer to NFL reality than game balance.

I just realized based on your name you're probably running an alternate league in which this isn't the case so you might not want to do that, in which case I don't know what I'm talking about.

Here's another thing I thought of very early on (months ago) but didn't really think of again till just now.

We not only need to consider these ratings and setting them to value, we also need to simulate how much of this information is "known." A scout for any team can likely dig up a list of everyone's 40 time, or how much a particular lineman bench presses, but what about the third-year QB who only took snaps in three games? Not everybody is going to know that he's got a 94 throwing accuracy. His own team is going to know significantly more, and teams with excellent scouting will have some information but there will be question marks. In fact, this was one of the reasons I designed a GM rating for "Risk"...

Another guideline I wanted to use: I want to avoid using random number generation/chance. When we do use it, it should be done with a seed, so that the same series of "random" numbers will result. Otherwise, the trade which Team A balked at in week 2, they're jumping at in week 3, although nothing has really changed.

I don't know if this is feasible but could we have GM's do their won evaluation of each players OVR, which could be modified by a combination random factor, a number based off of the number of games the player has started/played in (madden tracks these) and the coaches and coordinators skill at the relevant position. I don't know how to make it consistent from week to week, I'm also worried that if we try to combine this with a different OVR formula for each GM, then each GM effectively generates a separate roster in which all the players have a slightly different "perceived" rating. I don't know enough to know if this is doable without limiting which players each GM evaluates.

[torontogrudlies]

Don't worry about what the total numbers come to now, you can always multiply the output of the evaluation formula by a constant so that it has the appropriate value compared to draft picks. All we should be worried about now is ratio of values not the actual values. If a 99QB is only 1.5 times as valuable as an 85 QB then we have a problem no matter what the actual values are. But if a 99 QB is worth 3 or 4 times as much than it's okay even if we only value them and 100 and 25 points respectively, we can just multiply them both by 6 or something to fit with the draft value chart.

Sounds reasonable. Keep in mind I'm also supplementing this formula with other factors which can get subtracted off, reasons why the player isn't a good fit for a particular team. The objective of that part is that a really low or negative number means the team will look to trade the player.

A quick comment about draft studies, one thing you have to consider is that teams picking towards the top tend to be at the top because they are bad at finding and developing talent, while consistently good teams tend to pick towards the bottom. Think of what the value of a quarterback like Sanchez would be if he was on a team with a line like the bears rather then a team with one of the best o-lines and a solid defense. Also one thing that has changed recently is that top picks are no longer paid ridiculous sums of money, which changes a lot of old studies which found that mid round picks were better than early ones because if you made a mistake at the top you lost a lot of cap space for the next 5-6 years.

I agree that the 3000 chart is completely irrational but I feel we should keep something close to that because a lot of teams still really overvalue top picks. Look at how much griffin went for, look at the fact that no ones willing to give up a first round pick for Mike Wallace, even though he's a young proven player and most draft prospects probably won't be that good. I think we should overvalue draftpicks because a lot of teams clearly value them more than they're worth, it's an issue of being closer to NFL reality than game balance.

With regard to worth, so far I'm using two different concepts of it when it comes to the player valuation. First, the value a team comes up with for a player, and then, the common value which is the average of what all 32 teams come up with. Couldn't something similar be true when it comes to draft picks?

Example: maybe the #1 overall pick is worth 496 like in that formula I led off with here. But, my team is in dire need of a franchise QB and a good RB, and the GM's philosophy favors an influx of fresh, young players over veterans. And, maybe, the GM's attribute for negotiation just isn't that strong. So, my team ends up valuing that pick at 800. This means my team would give up 800 points worth of other draft picks, players or combinations thereof. (I think.)

I just realized based on your name you're probably running an alternate league in which this isn't the case so you might not want to do that, in which case I don't know what I'm talking about.

Yeah, but my league is intended to model professional football....I prefer the alternate universe thing because it allows for a little more creativity, and I would get obsessive-compulsive if I were playing with the actual NFL rosters and things started to stray from the reality.

Thus, the input and feedback of others will be instrumental in getting this right! My players aren't real, so if some running back in the opposite conference drives over a bunch of mailboxes and gets a DUI, I would think nothing of it. If it happened with real NFL rosters, and Matt Forte showed up in the news with such an incident, most people would immediately say "wait, something isn't realistic."

Here's another thing I thought of very early on (months ago) but didn't really think of again till just now.

We not only need to consider these ratings and setting them to value, we also need to simulate how much of this information is "known." A scout for any team can likely dig up a list of everyone's 40 time, or how much a particular lineman bench presses, but what about the third-year QB who only took snaps in three games? Not everybody is going to know that he's got a 94 throwing accuracy. His own team is going to know significantly more, and teams with excellent scouting will have some information but there will be question marks. In fact, this was one of the reasons I designed a GM rating for "Risk"...

Another guideline I wanted to use: I want to avoid using random number generation/chance. When we do use it, it should be done with a seed, so that the same series of "random" numbers will result. Otherwise, the trade which Team A balked at in week 2, they're jumping at in week 3, although nothing has really changed.

I don't know if this is feasible but could we have GM's do their won evaluation of each players OVR, which could be modified by a combination random factor, a number based off of the number of games the player has started/played in (madden tracks these) and the coaches and coordinators skill at the relevant position. I don't know how to make it consistent from week to week, I'm also worried that if we try to combine this with a different OVR formula for each GM, then each GM effectively generates a separate roster in which all the players have a slightly different "perceived" rating. I don't know enough to know if this is doable without limiting which players each GM evaluates.

Yes, I think I hear what you're saying.

Code: Select all

tempovr = (CInt(temppt * tempsco / 100) + .POVR) + CInt(tempvariance * Rnd(-1 * .PGID) + (tempvariance / 2))


This, for example, is a formula which a GM uses to get a modified OVR on a player. It combines the present OVR and the potential (max the player can reach, as the database is presently laid out) and the better the GM is at scouting, the more accurately he will see the max potential. Then again, if it's a GM with no patience who just cares about "win now!" then by design he'll see more of the "now" OVR anyway. (I believe you have a suggestion on the table to create several potential max ratings, rather than just an OVR.... similar to the FPS FBPRO model, which I like. This would make this specific formula obsolete, but my main point in posting this was to point out the Rnd part of the formula.

The " Rnd(-1 * .PGID)" will generate a random number which will be the same every time, because it uses a negative seed based upon this player ID. (This is how randomization is designed in VB) Therefore although it's a unique number, it will still be used if we re-run the same script the following week on the same player.

Maybe we should try to take part of this back to the drawing board. Like, ok, we have a quarterback.

Here are his attributes.

Which of these attributes are common sense to every GM?

Which attributes can be picked up on by someone with good scouting, even without seeing the player in significant playing time?

Which ones can't be seen very well unless he's played in some games? If he's played in 10 games, we probably know he's got a 94 throwing power, but if only in 3 games, most GMs might get only a sketchy idea of his THP attribute. How to represent this in a formula? Keeping in mind that some GMs will factor risk, and we can use "unrandom" numbers as described above.

Which attributes will that player's own team have more insight on? They see him every day at practice. However, if they're a team which isn't good at developing a QB *coughTHEBEARScough* they might not have much more of a handle on this than the next team.

[Drizzt_13]

With regard to worth, so far I'm using two different concepts of it when it comes to the player valuation. First, the value a team comes up with for a player, and then, the common value which is the average of what all 32 teams come up with. Couldn't something similar be true when it comes to draft picks?

Example: maybe the #1 overall pick is worth 496 like in that formula I led off with here. But, my team is in dire need of a franchise QB and a good RB, and the GM's philosophy favors an influx of fresh, young players over veterans. And, maybe, the GM's attribute for negotiation just isn't that strong. So, my team ends up valuing that pick at 800. This means my team would give up 800 points worth of other draft picks, players or combinations thereof. (I think.)

I think having the draft picks be valued reasonably and then inflated to the level at which they tended to be traded for in the NFL through other factors is a good idea. What I'm a little worried about is if a team that doesn't over value the pick gets it do we see an insane amount of trades or something? But that's an issue with the logic that will go into trades rather than the logic of player valuation, and since this is obviously becoming a really complicated project and app I think it would really help to keep all the parts of this separate in the forums so maybe we should make another thread for discussing the trade logic. Or maybe we're getting ahead of ourselves and just want to finish player evaluation.

Yes, I think I hear what you're saying.

Code: Select all

tempovr = (CInt(temppt * tempsco / 100) + .POVR) + CInt(tempvariance * Rnd(-1 * .PGID) + (tempvariance / 2))


This, for example, is a formula which a GM uses to get a modified OVR on a player. It combines the present OVR and the potential (max the player can reach, as the database is presently laid out) and the better the GM is at scouting, the more accurately he will see the max potential. Then again, if it's a GM with no patience who just cares about "win now!" then by design he'll see more of the "now" OVR anyway. (I believe you have a suggestion on the table to create several potential max ratings, rather than just an OVR.... similar to the FPS FBPRO model, which I like.

I have no idea what the FPS FBPRO model is but I would definitely be interested in looking at it. My main ideas about the OVR cap was that I didn't really like the idea of hard capping player OVR because I didn't think it made sense to say it's impossible for a player to improve beyond a certain point for all abilities. I think it makes sense for some things, like STR and SPD to say that person is going to never pass a certain point, but for things that are technique based (RBK TAK AWR) it made less sense to hard cap them. But if we still wanted a measurement of potential we could have some factor which would be part of the progression formulas which would determine how quickly they advance in these areas, and which a good GM would be able to estimate.

I also had this other idea (and this is getting way ahead of myself) that things like the 40 yard dash and other drills would show a value very close to the max, but would not be accurate to what the players actual speed was. This would represent that players don't always play as strong or fast as they are, even if you are a hugely strong DT you may not have the technique to properly apply that strength, or if you're a fast HB but you don't read your blocking well you may hesitate a little and come off as slower. It would be easy for teams to determine the 40 times and reps of a player but harder to determine the speed they actually play at, which makes a lot of sense for draft logic as teams often get blown away by athletes who never turn into football players.

These are just some random ideas, and I'm willing to discuss them now but we should probably not try to really figure out how we're going to do these until we get further along

This would make this specific formula obsolete, but my main point in posting this was to point out the Rnd part of the formula.

The " Rnd(-1 * .PGID)" will generate a random number which will be the same every time, because it uses a negative seed based upon this player ID. (This is how randomization is designed in VB) Therefore although it's a unique number, it will still be used if we re-run the same script the following week on the same player.

Seems like a good solution to me, you kind of implied we were reworking the specific formula but I thing the issue with randomness has been solved

Maybe we should try to take part of this back to the drawing board. Like, ok, we have a quarterback.

Here are his attributes.

Which of these attributes are common sense to every GM?

Which attributes can be picked up on by someone with good scouting, even without seeing the player in significant playing time?

Which ones can't be seen very well unless he's played in some games? If he's played in 10 games, we probably know he's got a 94 throwing power, but if only in 3 games, most GMs might get only a sketchy idea of his THP attribute. How to represent this in a formula? Keeping in mind that some GMs will factor risk, and we can use "unrandom" numbers as described above.

Which attributes will that player's own team have more insight on? They see him every day at practice. However, if they're a team which isn't good at developing a QB *coughTHEBEARScough* they might not have much more of a handle on this than the next team.

I think we should make another thread for scouting because that's going to be complicated. I'm sorry if I'm being too assertive in this whole project but I think it would really help if we tried to keep everything as compartmentalized as possible because every time we try to do something we come up with all these ways it could relate to everything else and then we go off on a lot of different tangents. Could we focus in this thread on figuring out a formula that takes in the ratings of a player determined by scouting and a whole bunch of other factors and produces a trade value number?

I would really like some feedback on my idea for using both pass/run and aggressiveness to modify the way in which the OVR formulas are weighted for each GM. I want to know if you think it's worth the effort and if it's feasible with regard to the programming. I'll post a spreadsheet and some example equations in a bit.

[torontogrudlies]

With regard to worth, so far I'm using two different concepts of it when it comes to the player valuation. First, the value a team comes up with for a player, and then, the common value which is the average of what all 32 teams come up with. Couldn't something similar be true when it comes to draft picks?

Example: maybe the #1 overall pick is worth 496 like in that formula I led off with here. But, my team is in dire need of a franchise QB and a good RB, and the GM's philosophy favors an influx of fresh, young players over veterans. And, maybe, the GM's attribute for negotiation just isn't that strong. So, my team ends up valuing that pick at 800. This means my team would give up 800 points worth of other draft picks, players or combinations thereof. (I think.)

I think having the draft picks be valued reasonably and then inflated to the level at which they tended to be traded for in the NFL through other factors is a good idea. What I'm a little worried about is if a team that doesn't over value the pick gets it do we see an insane amount of trades or something? But that's an issue with the logic that will go into trades rather than the logic of player valuation, and since this is obviously becoming a really complicated project and app I think it would really help to keep all the parts of this separate in the forums so maybe we should make another thread for discussing the trade logic. Or maybe we're getting ahead of ourselves and just want to finish player evaluation.

It would definitely rate its own thread... I figured this might become bigger, which is why I added a separate subforum. Dealing with the trade logic, supply/demand will be complex but will be awesome to have a working model.

Yes, I think I hear what you're saying.

Code: Select all

tempovr = (CInt(temppt * tempsco / 100) + .POVR) + CInt(tempvariance * Rnd(-1 * .PGID) + (tempvariance / 2))


This, for example, is a formula which a GM uses to get a modified OVR on a player. It combines the present OVR and the potential (max the player can reach, as the database is presently laid out) and the better the GM is at scouting, the more accurately he will see the max potential. Then again, if it's a GM with no patience who just cares about "win now!" then by design he'll see more of the "now" OVR anyway. (I believe you have a suggestion on the table to create several potential max ratings, rather than just an OVR.... similar to the FPS FBPRO model, which I like.

I have no idea what the FPS FBPRO model is but I would definitely be interested in looking at it. My main ideas about the OVR cap was that I didn't really like the idea of hard capping player OVR because I didn't think it made sense to say it's impossible for a player to improve beyond a certain point for all abilities. I think it makes sense for some things, like STR and SPD to say that person is going to never pass a certain point, but for things that are technique based (RBK TAK AWR) it made less sense to hard cap them. But if we still wanted a measurement of potential we could have some factor which would be part of the progression formulas which would determine how quickly they advance in these areas, and which a good GM would be able to estimate.

I also had this other idea (and this is getting way ahead of myself) that things like the 40 yard dash and other drills would show a value very close to the max, but would not be accurate to what the players actual speed was. This would represent that players don't always play as strong or fast as they are, even if you are a hugely strong DT you may not have the technique to properly apply that strength, or if you're a fast HB but you don't read your blocking well you may hesitate a little and come off as slower. It would be easy for teams to determine the 40 times and reps of a player but harder to determine the speed they actually play at, which makes a lot of sense for draft logic as teams often get blown away by athletes who never turn into football players.

These are just some random ideas, and I'm willing to discuss them now but we should probably not try to really figure out how we're going to do these until we get further along

This would make this specific formula obsolete, but my main point in posting this was to point out the Rnd part of the formula.

The " Rnd(-1 * .PGID)" will generate a random number which will be the same every time, because it uses a negative seed based upon this player ID. (This is how randomization is designed in VB) Therefore although it's a unique number, it will still be used if we re-run the same script the following week on the same player.

Seems like a good solution to me, you kind of implied we were reworking the specific formula but I thing the issue with randomness has been solved

Maybe we should try to take part of this back to the drawing board. Like, ok, we have a quarterback.

Here are his attributes.

Which of these attributes are common sense to every GM?

Which attributes can be picked up on by someone with good scouting, even without seeing the player in significant playing time?

Which ones can't be seen very well unless he's played in some games? If he's played in 10 games, we probably know he's got a 94 throwing power, but if only in 3 games, most GMs might get only a sketchy idea of his THP attribute. How to represent this in a formula? Keeping in mind that some GMs will factor risk, and we can use "unrandom" numbers as described above.

Which attributes will that player's own team have more insight on? They see him every day at practice. However, if they're a team which isn't good at developing a QB *coughTHEBEARScough* they might not have much more of a handle on this than the next team.

I think we should make another thread for scouting because that's going to be complicated. I'm sorry if I'm being too assertive in this whole project but I think it would really help if we tried to keep everything as compartmentalized as possible because every time we try to do something we come up with all these ways it could relate to everything else and then we go off on a lot of different tangents. Could we focus in this thread on figuring out a formula that takes in the ratings of a player determined by scouting and a whole bunch of other factors and produces a trade value number?

No problem... this project probably needs some direction, as I've sort of bounced around. At first, I was designing formulas for how likely a player is to get suspended, then I was working on a flowchart for how an agent is going to interface with players and teams, and negotiate for them... then it was the beginning of season logic, etc.

http://www.footballidiot.com/wiki/index ... orning_App

is a wiki area which I've been using to document stuff like the secondary attributes, although logic needs to be created for much of it.

I would really like some feedback on my idea for using both pass/run and aggressiveness to modify the way in which the OVR formulas are weighted for each GM. I want to know if you think it's worth the effort and if it's feasible with regard to the programming. I'll post a spreadsheet and some example equations in a bit.

I'm sure it would be.... maybe in testing we should pick a handful of positions and come up with formulas for them? Perhaps like, the linemen because they're going to have factors for both pass and run.... then the QBs because they're sort of the central position?

I'm also wondering if some of the detrimental factors (player too young/too old, ethical character/off the field problems, etc.) should roll into this same formula? As I have it right now, the formula is computed, and then these other factors are added (or subtracted) individually, eventually leading to the total value. The evaluation script does also return flags such as "overpaid" and "age"...the idea being that I could later use them in some sort of verbose script such as "The word is that the GM may be looking to part ways with Jones because he's overpaid"

[Drizzt_13]

Ok here is my newest proposition for these formulas

I think the core of this formula is the fact that we are using exponential growth for value, but I think using 1.22^ovr-30 is the wrong formula assuming we only modify player value by multiplying it rather than adding factors, which in my opinion makes things much easier*. It just increases way too quickly, a 99 player is 5 times the value of a 91 player but no one would give up 5 91 players for a 99 even assuming all other factors were assumed to be average.

The other things that was bothering me was that I felt exponential growth was the wrong formula for all positions. I think it works well for positions in which being really good allows you to compensate for the deficiencies of the rest of the team, pass rusher, WR, QB and HB. But I thought for other positions, where elite players will be made less relevant if the talent around them is bad, CB, FB and offensive line (and 4-3LB's, maybe). Even If you have a stud LT, but the rest of the line is bad, you'll still let pressure through. Even if you have Namdi Asumoagh they can just throw to your other corner, which was why in his last season with the raiders he was targeted "just 29 times in 441 snaps in coverage" according to PFF. Calvin Johnson can still beat double and occasionally triple teams. Skill positions are impacted by other positions but look at Larry Fitz, Matt Forte and Eli Manning if you want to see skill position players make things happen even though their O-lines are terrible. Therefore I think because it is more important to go from bad to average at these positions than it is to go from average to excellent, we need a very different function.

I propose we use -(1.22^(31-(ovr-70)))+(1.1+1.1^30), though I would like to change the 1.22 part of this.This function is basically the previous function reflected across y=x. with this the difference between a 70 LG and an 80 LG is huge, but the difference between an 80 LG and a 90 LG is still big but nowhere near as large.

I could run a bunch of stuff with the variable but I was mostly wondering if I could get some feedback on this, before we make estimations for value and the like, what value should a 99 have compared to a 90, or an 80. keep in mind we may only be able to produce the result we want at one end of the spectrum using these formulas, so is it more important that ratio of value from 80 to 90 be more accurate if that means the ration of a 90 to a 99 is a little bit skewed or vice versa?

we should also just make a list of every factor which should be a part of this formula, a number of the GM values were mentioned in relation to this but I haven't heard much about it sense.



* torontogruddlies, I think you were planning on adding figure but I think it makes much more sense to have player value be multiplied by other factors. That way instead of asking if it makes sense for a 99 to be worth 80 points more than a 95 we ask ourselves if it makes sense for a 99 to be worth 4 times as much as a 95, which is just easier to make sense of at this point. It also gives us the added benefit of just scaling it to whatever draft value chart we use.

[torontogrudlies]

Maybe we should pop a season's worth of stats into the mix. If we're analyzing running backs, look at stuff like the yards per carry and see how it compares to the OVR. I think this could answer the question of how much better a 99 OVR is to a 90 OVR. And perhaps look at a few individual ratings rather than OVR... because maybe we'd want to redo the OVR concept in our code.

[Drizzt_13]

Maybe we should pop a season's worth of stats into the mix. If we're analyzing running backs, look at stuff like the yards per carry and see how it compares to the OVR. I think this could answer the question of how much better a 99 OVR is to a 90 OVR. And perhaps look at a few individual ratings rather than OVR... because maybe we'd want to redo the OVR concept in our code.

I talked earlier about changing the way OVR is calculated slightly based on team tendencies such as pass/run and aggressiveness and I think I made a spreadsheet which models it. As for stats I would rather not bring season stats, in unless we want them to represent the way stats change the way a coach views a player. We actually have an objective way of measuring how good backs are at certain parts of the game which NFL coaches do not.

You could try and bring season stats into this but that would still be affected by the lines they run behind and the defenses they face. Though if we could change a roster so that every team has the same line and defense and only the HB differs that would tell us something interesting about what the maddens sim stats do.

[torontogrudlies]

Actually, I was thinking of looking at the stats as part of our "focus group" here, to help perfect a formula to value a player.... in other words, looking at the relevant stats next to the OVR information, for the purposes of figuring out how much correlation there is, what separates an 80 OVR from a 90 OVR from a 99 OVR. No, I don't think the app itself should pull stats at this point.

To see how well a team views a player, factors such as how many games the player has played in, plus the team's expertise at evaluating talent would probably be sufficient.

Although, at one time, I did toy with the idea of formulating an "alternative OVR" which would be based on the stats which can be seen. When scouting a player on another team, it would become relevant. Although it would in many cases not be as accurate, the same is true in real life. If a QB starts just three games and does extremely well, throws 8 TDs and just one INT, this alternative OVR might be near the top of the scale. But, a good GM would see more of this QB's actual talent (i.e. his skill ratings) whereas a GM bad at scouting would mostly work off of this alternative OVR. I had tabled this idea, but if you think it might make sense, we can look further?