VBD Baseline (1 Viewer)

[freshly_shorn]

In reading this thread & others over the years, it seems that different people want VBD to do different things for them.

I guess what I'm getting after is what exactly is value & what determines it? If #1 above is true, then there has to be a "true" value for every player in every league; meaning that there has to be a "right" baseline for that league to determine that value; meaning that all other baselines are wrong. Setting a baseline so you get a list that looks good to you may not necessarily be anywhere near what the list would look like if you were able to actually determine each player's "correct" value.

Value is simply a way to compare players from different positions, hence the advent of VBD. Baselines are an arbitrary way to create a quantifiable value. You have to set a hard point to create a 'value', hence a baseline.

 

[Hear-the-Footsteps]

In reading this thread & others over the years, it seems that different people want VBD to do different things for them.

Some appear to want it to match their past drafts. So they tweak the baselines until they basically get an ADP for their league based on what's happened in former drafts. I really don't see the point in this & think it's redundant. It's also self-fulfilling as you're forcing your baselines to give you a ranking that matches that ADP.

Others want something that "looks" right for a draft: "there's no way a defense should have enough value to show up as a second-round choice; I'll move the baseline until they're a 7th". This may be regardless of whether their particular league in fact does put incredible value on D's. - it just doesn't "look right", so I'm adjusting the baseline. Keep in mind that VBD doesn't purport to be a draft guide, but a tool to show where value might lie.

And still others may want it for "season-long value". Which, to me, is the whole point of using projections for a season as the generator of fantasy points.

If we're to assume that the theory behind VBD is sound (that all players have value & leaving aside the fallability of projections), then I think we need to ask the following questions:

1. Does league set-up affect player values? In other words, are player values specific to a particular league & will they be different in another league with different parameters? If so, why?

2. How much do starting requirements affect player values? Why?

3. How much do total roster requirements affect player value? Why?

4. If, as some have suggested, there is no "right" baseline then it follows that players either a) have no value, or b) player value doesn't matter. All you're doing is manipulating numbers that mean nothing, since there's no "right" then there's no "wrong" either. Is that correct?

I guess what I'm getting after is what exactly is value & what determines it? If #1 above is true, then there has to be a "true" value for every player in every league; meaning that there has to be a "right" baseline for that league to determine that value; meaning that all other baselines are wrong. Setting a baseline so you get a list that looks good to you may not necessarily be anywhere near what the list would look like if you were able to actually determine each player's "correct" value.

Are any of the assumptions I've made in my last paragraph wrong?

I know there are some very analytical minds that post here - both members & staff. We've had dozens of these threads over the years & it's always disappointed me a little that more staff didn't get involved (specifically, but not limited to, Joe). Maurile used to hop in every now & again, but he's about the only one I can recall. So what usually happens is these threads devolve into advice on how to tweak baselines so that everyone gets a list that looks right.

I couldn't agree more - especially with the bolded part about the staff.Look I think the staff here is some of the smartest around with respect to FF. And I think it is BY FAR the best FF site around. I really do.

But I specifically sent PMs to 3 staff members (because of how much I value and respect their opinions) with links to 2 different baseline threads asking them if they get a chance to jump in them and give some feedback.

No PM response. And none of them jumped into either of the threads I linked. Oh well.

I think it would be beyond frustrating to fully learn an app like DD - from the weekly weights, to SOS, to how to stack your team for certain weeks because of opponents in your division, etc etc - just to have a cheatsheet that is less than optimal because the baselines used weren't so solid.

Again, well thought out post. Solid analysis and thoughts. And I agree about the staff involvement in jumping in a real strategy thread.

 

[CalBear]

I guess what I'm getting after is what exactly is value & what determines it? If #1 above is true, then there has to be a "true" value for every player in every league; meaning that there has to be a "right" baseline for that league to determine that value; meaning that all other baselines are wrong. Setting a baseline so you get a list that looks good to you may not necessarily be anywhere near what the list would look like if you were able to actually determine each player's "correct" value.

Are any of the assumptions I've made in my last paragraph wrong?

At the end of the season, you can measure value by VBD over a worst-starter baseline. Based on worst starter for 1/2/2 leagues, Peyton Manning had 120 VBD points in 2006, which meant he was less valuable than Stephen Jackson but more valuable than Frank Gore to have on your roster. (Leaving aside the issue of what you had to pay at the beginning of the season to get Manning vs. Gore). In 2005, Manning had just 34 VBD points, which meant he wasn't particularly valuable; you'd have been better off with Warrick Dunn. At the beginning of the season, the question is much more difficult. VBD is based on the specific number of fantasy points projected for each player, but it's impossible to predict exactly how many fantasy points a player will score. Really, it would be better to represent each player's fantasy scoring as a probability curve; Manning has a very tight curve (you can be pretty confident he will be close to his projected numbers), while Gore has a very broad curve (he could compete with LT, or he could be the next Barlow). Risk and upside are not factors accounted for by VBD numbers.

Also, backups have value, but their value changes based on the rest of your lineup. If you start your draft WR-WR-QB, you will almost certainly be picking RBs who are below a worst-starter baseline; because you need to start them, their value on your team is larger than their value on anyone else's team. Whereas if you started your draft by getting two 16-game starting RBs, the value of the #25 RB is fairly low to you. You can partially address this issue by using a lower baseline (median backup, or worst backup), but changing that baseline will have effects beyond just comparing value of backups.

The QB, WR, and RB curves have different shapes, and changes to your VBD baselines will affect relative positional value based on those curves. Unfortunately images are turned off on this bulletin board, but see these two graphs:

Fantasy point curves

Normalized fantasy point curves

The first graph is the 3-year average for points at each position, for players ranked #1 to #30. The second is based on the same data, normalized so that #12 in each series is equal in value. (I think the second graph shows the curves better).

What you'll see from those graphs is that different parts of the curves behave differently. The biggest QB dropoff is in the first three or four QBS, and the curve is pretty much linear after that. WRs have a fairly flat curve until about #10, then it drops off for a bit, and then is linear with the least dropoff from about #14 onwards. RBs drop precipitously from #1 to #10, then are linear with WRs until about #22, then they start to tail off.

Where you place your baselines along those curves makes a difference. For example, if you used #10 as the baseline for all three positions, Tomlinson's value (and all the rest of the top RBs) would be off the charts, because the RB curve is so steep in that section. The more typical #24 baseline is at the end of a fairly shallow part of the curve, so the value isn't quite as skewed to RBs. It looks like if you drop the baselines further than that, you'll probably increase RB value relative to WRs, because the WR curve is flatter (due to positional scarcity).

 

[Uruk-Hai]

I don't think so. To say "no single baseline is perfect for every league" does not automatically mean "all baselines are equal in any league and therefore worthless." Why not? There may be consensus that some baselines are more sound than others. I haven't meant anyone that prefers "lowest roster spot" over "lowest starter," though if they did I'd love to hear the logic behind it. I would also argue that while there may be a lot of custom tweaking for aesthetic/drafting purposes, worst starter (in at least one position) is a driver for most of them.

I'm going to use this as a jumping-back-in-point, mainly because this may be the item I worded the most poorly & caused some misunderstanding in what I'm asking/commenting on. Also, the questions I asked in my prior post were somewhat Devil's-Advocate type, but I asked them because sometimes we accept things that we "know" & forgot how we got to know them in the first place. Anyway, I'm speaking more here of pre-draft scenarios than post-season since that's where we are on the calendar. Also, let's assume all projections (wherever we get them) are perfect because that's a whole other can of worms I don't want to open here.

In any particular league, a player has a certain (projected) value, which is represented by a number. How do we arrive at that number? Well, in VBD we do it by setting a baseline. But how do we know that the baseline we choose gives us the correct (projected) value for that player? Because by moving that baseline around, the value number attached to that player changes & it affects his placement among other players (other positions, who have the same "problems" with their own baseline adjustments). The player's value amongst his own position may stay constant, but his value opposite other positions changes by adjusting baselines.

Yet it seems to me that that can't be right. Hines Ward has a (projected) value against Ahman Green given my league's parameters. It is what it is - it's either greater, less, or exactly the same but it's something exact, a set number that tells me how much more (or less) valuable Ward is in comparison to Green for my league's purposes.

For the sake of argument, let's say we know what the apprpriate baseline that gives me "true" value for my league. Yet I can manipulate the baseline to give me something else - sure my overall list may look "more right" but now I don't have Ward's correct value. What I'm saying is that there has to be one set way to measure what Ward is going to do that reflects his true value to me in my league - anything else gives me an incorrect measure (value #).

It may be that, as some have said, we use a different baseline for each position & whether we're talking about starters or bench players. Hell, it may be different for each individual player. But it seems to me that there can only be one (for each league, position, and/or player) that's right. Just whilly-nilly changing them doesn't seem right to me. Bad analogy maybe, but it's like saying: "I don't want a $1 bill to be 1/5 the value of a $5 bill anymore, I'll make it 1/3 because that seems more right to me". That doesn't change the fact that when you go out into the real world, it's still only going to buy you $1 worth of stuff. You can also say: "Hines Ward should have more value than Ahman Green" and adjust the baseline to make it so. But he may not really be more valuable & all you've done is fool yourself.

There's only one way to compare Ward to Green & be right (again, for any particular league).

What is it & why?

 

[Uruk-Hai]

In reading this thread & others over the years, it seems that different people want VBD to do different things for them.

I guess what I'm getting after is what exactly is value & what determines it? If #1 above is true, then there has to be a "true" value for every player in every league; meaning that there has to be a "right" baseline for that league to determine that value; meaning that all other baselines are wrong. Setting a baseline so you get a list that looks good to you may not necessarily be anywhere near what the list would look like if you were able to actually determine each player's "correct" value.

Value is simply a way to compare players from different positions, hence the advent of VBD. Baselines are an arbitrary way to create a quantifiable value. You have to set a hard point to create a 'value', hence a baseline.

Ok, so how do you know that the arbitrary baseline you choose gives you an accurate representation of value across positions? How do you know that you're not skewing value all to hell in a way that doesn't represent what really happens in your league?

 

[Uruk-Hai]

I guess what I'm getting after is what exactly is value & what determines it? If #1 above is true, then there has to be a "true" value for every player in every league; meaning that there has to be a "right" baseline for that league to determine that value; meaning that all other baselines are wrong. Setting a baseline so you get a list that looks good to you may not necessarily be anywhere near what the list would look like if you were able to actually determine each player's "correct" value.

Are any of the assumptions I've made in my last paragraph wrong?

At the end of the season, you can measure value by VBD over a worst-starter baseline. Based on worst starter for 1/2/2 leagues, Peyton Manning had 120 VBD points in 2006, which meant he was less valuable than Stephen Jackson but more valuable than Frank Gore to have on your roster. (Leaving aside the issue of what you had to pay at the beginning of the season to get Manning vs. Gore). In 2005, Manning had just 34 VBD points, which meant he wasn't particularly valuable; you'd have been better off with Warrick Dunn. At the beginning of the season, the question is much more difficult. VBD is based on the specific number of fantasy points projected for each player, but it's impossible to predict exactly how many fantasy points a player will score. Really, it would be better to represent each player's fantasy scoring as a probability curve; Manning has a very tight curve (you can be pretty confident he will be close to his projected numbers), while Gore has a very broad curve (he could compete with LT, or he could be the next Barlow). Risk and upside are not factors accounted for by VBD numbers.

Also, backups have value, but their value changes based on the rest of your lineup. If you start your draft WR-WR-QB, you will almost certainly be picking RBs who are below a worst-starter baseline; because you need to start them, their value on your team is larger than their value on anyone else's team. Whereas if you started your draft by getting two 16-game starting RBs, the value of the #25 RB is fairly low to you. You can partially address this issue by using a lower baseline (median backup, or worst backup), but changing that baseline will have effects beyond just comparing value of backups.

The QB, WR, and RB curves have different shapes, and changes to your VBD baselines will affect relative positional value based on those curves. Unfortunately images are turned off on this bulletin board, but see these two graphs:

Fantasy point curves

Normalized fantasy point curves

The first graph is the 3-year average for points at each position, for players ranked #1 to #30. The second is based on the same data, normalized so that #12 in each series is equal in value. (I think the second graph shows the curves better).

What you'll see from those graphs is that different parts of the curves behave differently. The biggest QB dropoff is in the first three or four QBS, and the curve is pretty much linear after that. WRs have a fairly flat curve until about #10, then it drops off for a bit, and then is linear with the least dropoff from about #14 onwards. RBs drop precipitously from #1 to #10, then are linear with WRs until about #22, then they start to tail off.

Where you place your baselines along those curves makes a difference. For example, if you used #10 as the baseline for all three positions, Tomlinson's value (and all the rest of the top RBs) would be off the charts, because the RB curve is so steep in that section. The more typical #24 baseline is at the end of a fairly shallow part of the curve, so the value isn't quite as skewed to RBs. It looks like if you drop the baselines further than that, you'll probably increase RB value relative to WRs, because the WR curve is flatter (due to positional scarcity).

As you say, end-of-season value is much easier to quantify.And I like your thoughts on curves (I know others here espouse this too). But VBD lives on a hard #. We've got to find it somehow, whether we do 'em by position, player, or continually changing the way we derive it as each draft pick is made. I don't care if someone thinks every NFL player should be measured against a different baseline and someone else states that it should be a combined last-starter/last-drafted. I'd like to know why they choose the baseline they do.

 

[Gamma1210]

In reading this thread & others over the years, it seems that different people want VBD to do different things for them.

I guess what I'm getting after is what exactly is value & what determines it? If #1 above is true, then there has to be a "true" value for every player in every league; meaning that there has to be a "right" baseline for that league to determine that value; meaning that all other baselines are wrong. Setting a baseline so you get a list that looks good to you may not necessarily be anywhere near what the list would look like if you were able to actually determine each player's "correct" value.

Value is simply a way to compare players from different positions, hence the advent of VBD. Baselines are an arbitrary way to create a quantifiable value. You have to set a hard point to create a 'value', hence a baseline.

Ok, so how do you know that the arbitrary baseline you choose gives you an accurate representation of value across positions? How do you know that you're not skewing value all to hell in a way that doesn't represent what really happens in your league?

I was just about to post a similar point: I have an upcoming draft for a start 2 QB league so the draft doesn't flow like the typical league. I know that at the end of 3 rounds that at least 7-8 QBs will be gone. I go to set my baseline based on this fact, when in fact i am just subjectively setting the baseline to make the draft look how I think it should which in essense destroys the point of VBD. I do think there is a correct baseline but it may be impossible to prove.

 

[Uruk-Hai]

In reading this thread & others over the years, it seems that different people want VBD to do different things for them.

I guess what I'm getting after is what exactly is value & what determines it? If #1 above is true, then there has to be a "true" value for every player in every league; meaning that there has to be a "right" baseline for that league to determine that value; meaning that all other baselines are wrong. Setting a baseline so you get a list that looks good to you may not necessarily be anywhere near what the list would look like if you were able to actually determine each player's "correct" value.

Value is simply a way to compare players from different positions, hence the advent of VBD. Baselines are an arbitrary way to create a quantifiable value. You have to set a hard point to create a 'value', hence a baseline.

Ok, so how do you know that the arbitrary baseline you choose gives you an accurate representation of value across positions? How do you know that you're not skewing value all to hell in a way that doesn't represent what really happens in your league?

I was just about to post a similar point: I have an upcoming draft for a start 2 QB league so the draft doesn't flow like the typical league. I know that at the end of 3 rounds that at least 7-8 QBs will be gone. I go to set my baseline based on this fact, when in fact i am just subjectively setting the baseline to make the draft look how I think it should which in essense destroys the point of VBD. I do think there is a correct baseline but it may be impossible to prove.

That's what I think you're doing - mirroring your league's ADP with fancier numbers that just justify the ADP. What I think VBD should do, at least in part, is show you where the real value is regardless of how your league drafts - in fact, that should be (one of) the advantages of it, IMO.

 

[mcoffman]

 

[Rounders]

In reading this thread & others over the years, it seems that different people want VBD to do different things for them.

I guess what I'm getting after is what exactly is value & what determines it? If #1 above is true, then there has to be a "true" value for every player in every league; meaning that there has to be a "right" baseline for that league to determine that value; meaning that all other baselines are wrong. Setting a baseline so you get a list that looks good to you may not necessarily be anywhere near what the list would look like if you were able to actually determine each player's "correct" value.

Value is simply a way to compare players from different positions, hence the advent of VBD. Baselines are an arbitrary way to create a quantifiable value. You have to set a hard point to create a 'value', hence a baseline.

Ok, so how do you know that the arbitrary baseline you choose gives you an accurate representation of value across positions? How do you know that you're not skewing value all to hell in a way that doesn't represent what really happens in your league?

I was just about to post a similar point: I have an upcoming draft for a start 2 QB league so the draft doesn't flow like the typical league. I know that at the end of 3 rounds that at least 7-8 QBs will be gone. I go to set my baseline based on this fact, when in fact i am just subjectively setting the baseline to make the draft look how I think it should which in essense destroys the point of VBD. I do think there is a correct baseline but it may be impossible to prove.

That's what I think you're doing - mirroring your league's ADP with fancier numbers that just justify the ADP. What I think VBD should do, at least in part, is show you where the real value is regardless of how your league drafts - in fact, that should be (one of) the advantages of it, IMO.

I hear you Uruk-Hai, but my understanding of why you change the configuration is so that you get a better understanding of how your league drafts.....maybe I'm missing the point....but looking back at my drafts and take the first 10 rounds (which is a 10 team league, so 100 players) I can find the # of QB/RB/WR/TE taken. So when I plug this information in as my baseline the DD is telling me who is more valuable based upon my leagues drafting tendacies as well.

Like I said maybe I'm not understanding your point and VBD, but that's what I think

 

[simsarge]

Proverbial

 

[Mr. Fusion]

 

[CalBear]

And I like your thoughts on curves (I know others here espouse this too). But VBD lives on a hard #. We've got to find it somehow, whether we do 'em by position, player, or continually changing the way we derive it as each draft pick is made. I don't care if someone thinks every NFL player should be measured against a different baseline and someone else states that it should be a combined last-starter/last-drafted. I'd like to know why they choose the baseline they do.

VBD lives on a hard number, but there isn't one correct number to use. The only thing to do is to know that different baselines will give you different results, have some concept of what those results might be, and don't get married to the VBD number.

 

[Bobcat10]

 

[Uruk-Hai]

And I like your thoughts on curves (I know others here espouse this too). But VBD lives on a hard #. We've got to find it somehow, whether we do 'em by position, player, or continually changing the way we derive it as each draft pick is made. I don't care if someone thinks every NFL player should be measured against a different baseline and someone else states that it should be a combined last-starter/last-drafted. I'd like to know why they choose the baseline they do.

VBD lives on a hard number, but there isn't one correct number to use. The only thing to do is to know that different baselines will give you different results, have some concept of what those results might be, and don't get married to the VBD number.

C'mon, Cal. If you really believe that, then you can't believe in VBD. The whole premise is to assign a value to a player that accurately reflects his value vs another's. If it doesn't matter what the number is, you may as well gut-draft. That's all many are doing by shifting baselines anyway - "now it looks like it should!". You're not getting new, accurate results - you're getting more wrong, different results. Again, Hines Ward has an exact relationship value-wise to Ahman Green in any particular league (still talking preseason with a static set of projections & league parameters). It can't be a bunch of different relationships. The value difference is one number. Changing the baselines around gives you a bunch of different value numbers for the same relationship - that cannot be right.

Look, I'm not a mathmetician but intuitively I feel that there has to be one real value for each player - not many. I mean, hell - I can make 32 kickers appear to be the most valuable players by playing with baselines. Do you think that's right? Why not?

 

[Mo_Steel]

URUK:

I am with you, I have been beating my head against a wall trying to figure out the right baseline for my league; however, I also feel that it ends up being a gut decision as I don't know.

While continuing to think on it, I am thinking that maybe it's a point total and the baseline should be set on a POINT total per position; i.e. maybe looking back at historical values of said league you can figure out what each position had to score on average to win (H2H), and then convert to an entire season, from there the true value may be at total points set above this otherwise it may be a mirage... Not sure really, but I finally settled on looking at the projections for all draftable players in my league; we have 10 teams with 30 man rosters... I found what the last player drafted on total points at each position and set my baseline there. This at least makes it easy to read as none of the VBD's end up in the negative.

I will keep searching for the perfect Baseline but also with the awareness that it may only be Value to the individual.

 

[CalBear]

Look, I'm not a mathmetician but intuitively I feel that there has to be one real value for each player - not many. I mean, hell - I can make 32 kickers appear to be the most valuable players by playing with baselines. Do you think that's right? Why not?

Your intuition is wrong. "Value" is an extremely vague concept that is impossible to pin down to one number, even in domains where the equations are much simpler. What is the value of $1? [Please avoid circularity in your response].

 

[Uruk-Hai]

URUK:

I am with you, I have been beating my head against a wall trying to figure out the right baseline for my league; however, I also feel that it ends up being a gut decision as I don't know.

While continuing to think on it, I am thinking that maybe it's a point total and the baseline should be set on a POINT total per position; i.e. maybe looking back at historical values of said league you can figure out what each position had to score on average to win (H2H), and then convert to an entire season, from there the true value may be at total points set above this otherwise it may be a mirage... Not sure really, but I finally settled on looking at the projections for all draftable players in my league; we have 10 teams with 30 man rosters... I found what the last player drafted on total points at each position and set my baseline there. This at least makes it easy to read as none of the VBD's end up in the negative.

I will keep searching for the perfect Baseline but also with the awareness that it may only be Value to the individual.

You touch on a couple of things that I was hoping this thread would move toward, but first I'd like to make sure I understand you. What do you mean "what each position had to score to win"? How would you determine this?As for negative values. they don't bother me. The relationship between the numbers does: +5 to -5 is the same as +10 to 0 for comparison purposes.

 

[Uruk-Hai]

Look, I'm not a mathmetician but intuitively I feel that there has to be one real value for each player - not many. I mean, hell - I can make 32 kickers appear to be the most valuable players by playing with baselines. Do you think that's right? Why not?

Your intuition is wrong. "Value" is an extremely vague concept that is impossible to pin down to one number, even in domains where the equations are much simpler. What is the value of $1? [Please avoid circularity in your response].

Really, so if I make all 32 kickers the most valuable players in a league, that's a valid ranking?As for value-as-vague-concept, is it vague in Joe's 2 team, 2 player example from his original VBD article? It looked pretty cut & dried to me. Are you saying that the idea can't be extrapolated out mathmatically? And yet, we're told to come up with a hard number to use VBD. So, is any old number we can come up with valid? If not, which ones are & why?eta: as for your dollar question, I can only answer that it's always 1/5 of a $5 (how's that for circularity?)

 

[freshly_shorn]

In reading this thread & others over the years, it seems that different people want VBD to do different things for them.

I guess what I'm getting after is what exactly is value & what determines it? If #1 above is true, then there has to be a "true" value for every player in every league; meaning that there has to be a "right" baseline for that league to determine that value; meaning that all other baselines are wrong. Setting a baseline so you get a list that looks good to you may not necessarily be anywhere near what the list would look like if you were able to actually determine each player's "correct" value.

Value is simply a way to compare players from different positions, hence the advent of VBD. Baselines are an arbitrary way to create a quantifiable value. You have to set a hard point to create a 'value', hence a baseline.

Ok, so how do you know that the arbitrary baseline you choose gives you an accurate representation of value across positions? How do you know that you're not skewing value all to hell in a way that doesn't represent what really happens in your league?

Because my baselines come from how my league has drafted for the past three years. I arbitrarily chose to use the round in which the last starting QB is typically taken- round 7. So, in my 12 team league, thats 84 players. I use a weighted average from the last three years by each position taken among the 84. It works well for my league.As for jumping into a league whose habits you have no idea of, you can only rely on ADP and experience when choosing a baseline. Joe and David's experience has been 100 players- I find that 84 works better for me.

 

[CalBear]

Look, I'm not a mathmetician but intuitively I feel that there has to be one real value for each player - not many. I mean, hell - I can make 32 kickers appear to be the most valuable players by playing with baselines. Do you think that's right? Why not?

Your intuition is wrong. "Value" is an extremely vague concept that is impossible to pin down to one number, even in domains where the equations are much simpler. What is the value of $1? [Please avoid circularity in your response].

Really, so if I make all 32 kickers the most valuable players in a league, that's a valid ranking?

Did I say that? No. I think it would be a stupid ranking. The fact that there's no perfect ranking doesn't mean that all rankings are equally valid.

As for value-as-vague-concept, is it vague in Joe's 2 team, 2 player example from his original VBD article? It looked pretty cut & dried to me. Are you saying that the idea can't be extrapolated out mathmatically? And yet, we're told to come up with a hard number to use VBD. So, is any old number we can come up with valid? If not, which ones are & why?eta: as for your dollar question, I can only answer that it's always 1/5 of a $5 (how's that for circularity?)

Worst starter, Joe's Secret Formula, median backup, last backup, and last player are all baselines for which it is possible to make reasonable arguments. And there are others. None of them is correct; you just have to decide which argument makes the most sense to you, just like when you're deciding whether to invest in Google or in Exxon. No one can give you The One Right Answer.

 

[freshly_shorn]

Look, I'm not a mathmetician but intuitively I feel that there has to be one real value for each player - not many. I mean, hell - I can make 32 kickers appear to be the most valuable players by playing with baselines. Do you think that's right? Why not?

Your intuition is wrong. "Value" is an extremely vague concept that is impossible to pin down to one number, even in domains where the equations are much simpler. What is the value of $1? [Please avoid circularity in your response].

It's not that it's vague- a more appropriate term IMO is 'relative'. Value, in terms of the VBD system, is relative to the point at which you choose to compare everyone, and that point can vary from year to year and from scoring-system to scoring-system.But value is also the measurement of how we as owners view players. VBD incorporates that by telling us that if a player's ADP is lower than his projected, relative 'value', we should wait to pick him at a more appropriate time.

So there are two separate concepts of value within VBD. Baselines is one, and perception is another. They have to be used together.

 

[Mo_Steel]

URUK:

I am with you, I have been beating my head against a wall trying to figure out the right baseline for my league; however, I also feel that it ends up being a gut decision as I don't know.

While continuing to think on it, I am thinking that maybe it's a point total and the baseline should be set on a POINT total per position; i.e. maybe looking back at historical values of said league you can figure out what each position had to score on average to win (H2H), and then convert to an entire season, from there the true value may be at total points set above this otherwise it may be a mirage... Not sure really, but I finally settled on looking at the projections for all draftable players in my league; we have 10 teams with 30 man rosters... I found what the last player drafted on total points at each position and set my baseline there. This at least makes it easy to read as none of the VBD's end up in the negative.

I will keep searching for the perfect Baseline but also with the awareness that it may only be Value to the individual.

You touch on a couple of things that I was hoping this thread would move toward, but first I'd like to make sure I understand you. What do you mean "what each position had to score to win"? How would you determine this?As for negative values. they don't bother me. The relationship between the numbers does: +5 to -5 is the same as +10 to 0 for comparison purposes.

Sorry about that... What I meant was look at say the playoff teams from the last couple of years, and figure the average of their top player at each position. I would attempt this in my league this year, but we just switched to Dynasty and IDP; nothing like jumping in with both feet. The majority of our owners have been playing together for the last three years, but we decided to take it to the next level since it has been a re-draft to this point. Hope that clarifies the question...So onto my thoughts: I would look at the Playoff teams depending on the number in relation to the league size; say 50% of teams or less. I would want my baselines based off at least the top 50% or better. From there, look at the rosters based on the League scoring system and get an average of the top players at each position in the system. From their you should have an idea of where to set the baselines in order to get this tiered talent; you are getting to look at eh last years totals and someone mentioned this was the only true way to see value. In theory this will also let you know by scoring where you can go on the cheap because the Point total will obviously be lower, and obviously being led by history, so it should vault top talent in that position. The you would draft as normal and if it pops Gates in the first round you have a decision to make, the baseline says he is good enough, but you still have to balance the ADP and say he will be there in round two or possibly even three. I am not this wordy so I will hang up and listen as the discussion rages on...

 

[GoBears84]

I wrote an article that FBG will publish soon that touches on this subject.

I use baselines culled from my league's past 3 years of drafts, and I arbitrarily decided to use whatever round, on average, in which the last starting QB is typically taken (12 teams), which has been round 7. So, in that group of 84 players I take a weighted average of each of the three years, giving precedence to the most recent draft.

For K and IDP, I just use the worst starter as the baseline since none are drafted in the first 7 rounds.

At the end of the day, you just need to play around with the baseline until what you see agrees with what you 'feel' it should look like. There's really no best way to do it, though I think considering your league's drafting habits can be a useful exercise in helping you determine your strategy.

There is a good discussion going on here in the Application forum on creating your own custom drop offs .As a follow-up here's a cool application for taking you DD data and creating a draft log.

Joel

 

[Uruk-Hai]

Look, I'm not a mathmetician but intuitively I feel that there has to be one real value for each player - not many. I mean, hell - I can make 32 kickers appear to be the most valuable players by playing with baselines. Do you think that's right? Why not?

Your intuition is wrong. "Value" is an extremely vague concept that is impossible to pin down to one number, even in domains where the equations are much simpler. What is the value of $1? [Please avoid circularity in your response].

Really, so if I make all 32 kickers the most valuable players in a league, that's a valid ranking?

Did I say that? No. I think it would be a stupid ranking. The fact that there's no perfect ranking doesn't mean that all rankings are equally valid.

As for value-as-vague-concept, is it vague in Joe's 2 team, 2 player example from his original VBD article? It looked pretty cut & dried to me. Are you saying that the idea can't be extrapolated out mathmatically? And yet, we're told to come up with a hard number to use VBD. So, is any old number we can come up with valid? If not, which ones are & why?eta: as for your dollar question, I can only answer that it's always 1/5 of a $5 (how's that for circularity?)

Worst starter, Joe's Secret Formula, median backup, last backup, and last player are all baselines for which it is possible to make reasonable arguments. And there are others. None of them is correct; you just have to decide which argument makes the most sense to you, just like when you're deciding whether to invest in Google or in Exxon. No one can give you The One Right Answer.

So, at the risk of being completely obtuse, you're saying that if we have Hines Ward projected to score 200 points & Ahman Green slotted for 150 - and given a particular league's roster requirement - that there's no one, precise relationship between the two, right? Anyway, to move forward: can anyone make the argument for one or some of the baselines Cal mentioned?

 

[Uruk-Hai]

It's not that it's vague- a more appropriate term IMO is 'relative'. Value, in terms of the VBD system, is relative to the point at which you choose to compare everyone, and that point can vary from year to year and from scoring-system to scoring-system.

In 2007 & in your league, what are you using for your comparison point & why?Sidenote: I hope I don't come across as belligerent here, because that's not my intention. There are a lot of people here that may not be as familiar with the concept of VBD as others, and also some who question some of the tenets associated with it. I've gotten PMs from 6 different people in the last day in regards to this thread: some are newer members & some have been around awhile. There's an awful lot of confusion in regards to VBD (apologies for any I've caused here) that the article may not address sufficiently.

 

[CalBear]

So, at the risk of being completely obtuse, you're saying that if we have Hines Ward projected to score 200 points & Ahman Green slotted for 150 - and given a particular league's roster requirement - that there's no one, precise relationship between the two, right?

That's correct. The value difference between the two is a result of external factors. If there are 20 WRs who are all projected to score around 200 points, and Ahman Green is the only RB projected to score over 100 points, Green is extremely valuable. If the situation is reversed, Ward is extremely valuable. In reality, the difference between RBs, WRs, and other positions is described by lumpy curves (see above), which makes it difficult to decide exactly how to relate the two to each other.

 

[Uruk-Hai]

So, at the risk of being completely obtuse, you're saying that if we have Hines Ward projected to score 200 points & Ahman Green slotted for 150 - and given a particular league's roster requirement - that there's no one, precise relationship between the two, right?

That's correct. The value difference between the two is a result of external factors. If there are 20 WRs who are all projected to score around 200 points, and Ahman Green is the only RB projected to score over 100 points, Green is extremely valuable. If the situation is reversed, Ward is extremely valuable. In reality, the difference between RBs, WRs, and other positions is described by lumpy curves (see above), which makes it difficult to decide exactly how to relate the two to each other.

Ok, but I was assuming that their positional rankings were already fixed; that we had projections for all players not just those two in a vacuum. Ward's projections might put him at WR #12 & Green's at RB #20 (or whatever, I don't care).It's your pick - how are you going to know who to take?

 

[CalBear]

So, at the risk of being completely obtuse, you're saying that if we have Hines Ward projected to score 200 points & Ahman Green slotted for 150 - and given a particular league's roster requirement - that there's no one, precise relationship between the two, right?

That's correct. The value difference between the two is a result of external factors. If there are 20 WRs who are all projected to score around 200 points, and Ahman Green is the only RB projected to score over 100 points, Green is extremely valuable. If the situation is reversed, Ward is extremely valuable. In reality, the difference between RBs, WRs, and other positions is described by lumpy curves (see above), which makes it difficult to decide exactly how to relate the two to each other.

Ok, but I was assuming that their positional rankings were already fixed; that we had projections for all players not just those two in a vacuum. Ward's projections might put him at WR #12 & Green's at RB #20 (or whatever, I don't care).It's your pick - how are you going to know who to take?

Again, it depends on the curves, not just the ranking. If you want an absolute number looking just at the ranking, check out AVT; I don't think most bright fantasy folks consider AVT to be as useful as VBD, but it does generate specific numbers for you.

 

[redman]

I use an average (mean) of the starting players at each position as my VBD baseline. This helps to account for deviations from the mean. It also means that around half of my starters and all of my backups have negative VBD values, but that doesn't concern me.

 

[Uruk-Hai]

I use an average (mean) of the starting players at each position as my VBD baseline. This helps to account for deviations from the mean. It also means that around half of my starters and all of my backups have negative VBD values, but that doesn't concern me.

Nor does it concern me. Why do you use starters' means as opposed to all drafted?

 

[Ramblin Wreck]

I use an average (mean) of the starting players at each position as my VBD baseline. This helps to account for deviations from the mean. It also means that around half of my starters and all of my backups have negative VBD values, but that doesn't concern me.

That would mean your 6th TE has the same value as your 12th RB assuming you start a TE and 2 RB. I don't agree with that at all.I decided to make my baselines the last good player at each position I would be willing to have as a starter. That worked out to 7 TE, 7 QB, 27 RB, and 29 WR this year. I have no idea if my numbers are good or not but I'm giving this a shot.

 

[freshly_shorn]

It's not that it's vague- a more appropriate term IMO is 'relative'. Value, in terms of the VBD system, is relative to the point at which you choose to compare everyone, and that point can vary from year to year and from scoring-system to scoring-system.

In 2007 & in your league, what are you using for your comparison point & why?Sidenote: I hope I don't come across as belligerent here, because that's not my intention. There are a lot of people here that may not be as familiar with the concept of VBD as others, and also some who question some of the tenets associated with it. I've gotten PMs from 6 different people in the last day in regards to this thread: some are newer members & some have been around awhile. There's an awful lot of confusion in regards to VBD (apologies for any I've caused here) that the article may not address sufficiently.

I already noted that- Round 7, because that's when the last starting QB is usually taken. 84 players. Then I found that on average, in the last three years, 12QB, 34 WR, 34 RB and 4 TE will be taken. My baselines are QB12, RB 34, WR 34 and TE 4. I just use worst starter for the rest of the positions. Why? Because I only care about players that are likely to be in starting lineups, and all the players taken in the first 7 round are likely to start.

 

[redman]

I use an average (mean) of the starting players at each position as my VBD baseline. This helps to account for deviations from the mean. It also means that around half of my starters and all of my backups have negative VBD values, but that doesn't concern me.

Nor does it concern me. Why do you use starters' means as opposed to all drafted?

Simple. I'm only playing my starters, so they're the basis for my analysis.

 

[wdbs]

C'mon, Cal. If you really believe that, then you can't believe in VBD. The whole premise is to assign a value to a player that accurately reflects his value vs another's. If it doesn't matter what the number is, you may as well gut-draft. That's all many are doing by shifting baselines anyway - "now it looks like it should!". You're not getting new, accurate results - you're getting more wrong, different results.

Again, Hines Ward has an exact relationship value-wise to Ahman Green in any particular league (still talking preseason with a static set of projections & league parameters). It can't be a bunch of different relationships. The value difference is one number. Changing the baselines around gives you a bunch of different value numbers for the same relationship - that cannot be right.

Look, I'm not a mathmetician but intuitively I feel that there has to be one real value for each player - not many. I mean, hell - I can make 32 kickers appear to be the most valuable players by playing with baselines. Do you think that's right? Why not?

I think your intuition is correct that there are actual values for each player relative to all of the others, as long as you discount things like who else you already have on your team, etc. The problem is that VBD is not a tool that can give you these exact values -- so the statement others made that there is no perfect baseline is likely correct.Have you read Maurile's article about calculating auction values? Although not perfect, I think his method does a little better job than VBD of getting closer to actual values. One poster above mentioned the idea of using league history to help with valuation. When you have something that you cannot solve mathematically, using available data (in this case history of actual leagues) can be used to help you approximate a solution, and this is something Maurile's method takes advantage of.

 

[inca911]

The goal of VBD in an auction setting is to tell us the true theoretical auction values for players. These values are derived from actual projections for players, and are based on the specific scoring system in use, along with the roster requirements, number of teams, and total team salary cap. All potential upside or injury risk for players is factored into the player projections. Auction values should not be artificially adjusted to attempt to predict the behavior of other owners, or to mimic what has happened in previous years during historical auctions.

Setting the correct baseline for an auction is specific to the setup of the league, and includes the following key items:

Number of Teams

Lineup and Roster Requirements

Player Scoring Projections

1. Number of Teams

With identical lineup, roster requirements, and projections, a 14 team league will have different player baselines than an 8 team league. While this may seem very intuitive, it is easily overlooked in various conversations of auction baseline. In a 14 team ppr league, there is a signicant dropoff between TE #8 at 94 total points and TE#14 at 59 total points. TE#8 is worth paying a bit of money to get in a 14 team league. Not so much in an 8 team league.

2. Lineup and Roster Requirements

A league requiring 2 QB starters will logically have a different QB baseline than a league requiring only one QB. It's simple supply and demand. If I must have 2 QBs on my roster, the baseline for them will be deeper as people must fill their rosters.

3. Player Scoring Projections

This is the second most important aspect of setting a correct baseline. Imagine an 8 team league with DST scoring as follows:

DST #1 to DST #6 all score 250 points

DST #7 to DST #12 all score 150 points.

DST #13 to DST #24 all score 50 points.

Baseline of worst starter sets the baseline at 150 points. An artifically adjusted baseline of worst starter * 1.5 (to account for backups) sets the baseline at 50 points. Clearly if you are an owner in this league you better get one of your top 6 projected DSTs or you'll have a lot of ground to makeup at another position. In this extreme example, few owners will pay anything to get DST #7, so clearly player scoring impacts where the baseline should be set.

Using an example from a 8 team, start 1 DST, with scoring projections as follows:

Baltimore 229

Chicago 222

San Diego 182

New England 181

Carolina 177

Dallas 174

Jacksonville 170

Minnesota 169

Denver 167

Miami 162

Pittsburgh 162

Philadelphia 160

My strategy is to either get a DST scoring >180 points, or to play matchups with lesser DST units. Based on player scoring flattening out at DST #6 (scores only 14 points more than DST #12), my strategy sets the baseline at DST #5. Note that DST #1 scores 52 points above the baseline, which is rougly equivalent to the difference between an RB like Addai/Westy/Parker and an RB like Caddy/TJones.

Note that in the example above my personal strategy impacts where the baseline is for a given position. This is the single most important factor in setting a baseline.

4. Setting the correct baseline is primarily dependent on the strategy of each individual owner.

An owner who decides to play a quarterback-by-committee approach will value QBs much less as he believes two QBs paired together can equal the performance of a top 5 QB. Similarly, an owner who believes in defense-by-committee will not pay much to get a top defense. These two owners will have much different baselines at those positions than an owner with a different strategy. Strategy plus player projections plus league setup determines each owner's baseline. A baseline cannot be set without looking at all of these factors and determining where you want to spend money. A 12 team start 2 QB league with a scoring system that really spreads out the QBs may have a very deep baseline at QB, as QB#24 may score significantly less than QB#22. Another 12 team start 2 QB league with scoring that bunches players together may have a baseline at QB#13. Projections and scoring systems absolutely come into play and cannot be ignored when determining a baseline. The "correct" baseline is set by deciding (at each position) when you are personally not willing to spend more than some minimum amount for a player. Once you know that point, you have your baseline for a position and can move on to determine the baseline for the next position.

Knowing the true theoretical auction cost for a player enables you to know how much value you are getting from any player. It doesn't prevent you from bidding over that number from time to time, nor does it dictate how you should build your team. The VBD concept is just one tool in a variety of options, and it's up to each owner to use the tool properly to execute a strategic plan.

 

[matuski]

 

[yrforyg]

And I like your thoughts on curves (I know others here espouse this too). But VBD lives on a hard #. We've got to find it somehow, whether we do 'em by position, player, or continually changing the way we derive it as each draft pick is made. I don't care if someone thinks every NFL player should be measured against a different baseline and someone else states that it should be a combined last-starter/last-drafted. I'd like to know why they choose the baseline they do.

VBD lives on a hard number, but there isn't one correct number to use. The only thing to do is to know that different baselines will give you different results, have some concept of what those results might be, and don't get married to the VBD number.

C'mon, Cal. If you really believe that, then you can't believe in VBD. The whole premise is to assign a value to a player that accurately reflects his value vs another's. If it doesn't matter what the number is, you may as well gut-draft. That's all many are doing by shifting baselines anyway - "now it looks like it should!". You're not getting new, accurate results - you're getting more wrong, different results. Again, Hines Ward has an exact relationship value-wise to Ahman Green in any particular league (still talking preseason with a static set of projections & league parameters). It can't be a bunch of different relationships. The value difference is one number. Changing the baselines around gives you a bunch of different value numbers for the same relationship - that cannot be right.

Look, I'm not a mathmetician but intuitively I feel that there has to be one real value for each player - not many. I mean, hell - I can make 32 kickers appear to be the most valuable players by playing with baselines. Do you think that's right? Why not?