What's new
Fantasy Football - Footballguys Forums

This is a sample guest message. Register a free account today to become a member! Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!

AVBD: Solution to a problem of standard VBD (1 Viewer)

Unless you feel 100% sure that your top ranked RB actually will finish the season #1, doing that you are going to end up over projecting the top guys compared to your actual beliefs.

It may look aesthetically pleasing in that it looks like an end of season positional list. But if you think there's a range of spots your top guy could actually finish, then the points of the #1 guy isn't the best expression of your beliefs on him.
Sure, there are a cluster of guys that I think will compete for WR1, but at the end of the day I need to put them in some order. And once that order is established, I need to assign a value to each spot. I don't see how the average performance of the WR1 for the guy I consider to have the best odds of finishing WR1 is a bad estimate.

I would say that over time this method is way more accurate than any method that assigns values without regard to the historical performance of the position. Will Dodds be right by projecting his RB1 to score more points than the average RB1 scores each year? Maybe. But if he does that year after year, he's overvaluing the RB1 position.

 
Unless you feel 100% sure that your top ranked RB actually will finish the season #1, doing that you are going to end up over projecting the top guys compared to your actual beliefs.

It may look aesthetically pleasing in that it looks like an end of season positional list. But if you think there's a range of spots your top guy could actually finish, then the points of the #1 guy isn't the best expression of your beliefs on him.
Are you 100% sure your top WR will finish the season #1? Your top QB?

If not, to project the top RB to score 230 and the top WR to score 220 points, when the top WR will likely score about 240 and the top RB will likely score over 300 (5-year averages), you're undervaluing the RB. In 2014, five RBs scored more than what Dodds is projecting for the #1 RB, so even if you think your guy is just likely to be in the top 5, he should be projected higher than he is.

 
I don't typically have time to do my own projections these days, so I use FBG's. And there are structural issues with them. I don't think it's correct to project someone to duplicate Murray's 2014, but I also think it's wrong to have the #1 RB projected at 1600/12 total, which is what Dodds has right now.
IMHO, Dodds is doing it right.

Dodds is projecting Le'Veon Bell to be the #1 RB, and he's projecting Le'Veon Bell to get 1600/12. But he is not projecting the #1 RB to get 1600/12.

Those two things are completely different.

The projection for Bell is stated as a point: 1600/12. But the projection isn't really a point; it's a range. It's an X% chance of 1900/16, a Y% chance of 1600/12, a Z% chance of 1300/8, and so on.

Same for all the other players.

When we do projections for Le'Veon Bell, we're giving you a point roughly in the middle of his realistic range (something like the average of each point within the range weighted by its probability, at least in theory).

We don't actually publish projections for RB1, but if we did, it would not simply be a point toward the middle of of Le'Veon Bell's range. If Bell ends up finishing as RB1, it will be because he ended up toward the top of his range, not toward the middle of it. If Jamaal Charles or Matt Forte or even Andre Ellington ends up as RB1, they will also likely have ended up significantly above the middle of Le'Veon Bell's range (and obviously above the middle of their own ranges as well).

In evaluating Le'Veon Bell's value, what matters is not how the end-of-season RB1 usually performs. What matters is how Le'Veon Bell should reasonably be expected to perform. And since Bell has a substantially less-than-100% chance of being RB1, the projection for Le'Veon Bell will necessarily be substantially lower than the (unpublished) projection for RB1.

 
Last edited by a moderator:
I don't typically have time to do my own projections these days, so I use FBG's. And there are structural issues with them. I don't think it's correct to project someone to duplicate Murray's 2014, but I also think it's wrong to have the #1 RB projected at 1600/12 total, which is what Dodds has right now.
IMHO, Dodds is doing it right.

Dodds is projecting Le'Veon Bell to be the #1 RB, and he's projecting Le'Veon Bell to get 1600/12. But he is not projecting the #1 RB to get 1600/12.

Those two things are completely different.

The projection for Bell is stated as a point: 1600/12. But the projection isn't really a point; it's a range. It's an X% chance of 1900/16, a Y% chance of 1600/12, a Z% chance of 1300/8, and so on.

Same for all the other players.

When we do projections for Le'Veon Bell, we're giving you a point roughly in the middle of his realistic range (something like the average of each point within the range weighted by its probability, at least in theory).

We don't actually publish projections for RB1, but if we did, it would not simply be a point toward the middle of of Le'Veon Bell's range. If Bell ends up finishing as RB1, it will be because he ended up toward the top of his range, not toward the middle of it. If Jamaal Charles or Matt Forte or even Andre Ellington ends up as RB1, they will also likely have ended up significantly above the middle of Le'Veon Bell's range (and obviously above the middle of their own ranges as well).

In evaluating Le'Veon Bell's value, what matters is not how the end-of-season RB1 usually performs. What matters is how Le'Veon Bell should reasonably be expected to perform. And since Bell has a substantially less-than-100% chance of being RB1, the projection for Le'Veon Bell will necessarily be substantially lower than the (unpublished) projection for RB1.
Demaryius Thomas is substantially less than 100% likely to be WR1, yet he's projected within 20 points of Le'Veon Bell, as are Jordy Nelson and Dez Bryant (all within 3 points right now). And WR36 is projected for 120 points (about 10 too many), and RB24 is projected for 140 points (about 10 too many). So by Dodds' current rankings, and standard VBD principles, Thomas, Nelson, and Bryant have about the same value as Le'Veon Bell, and more than guys like Jamaal Charles and Marshawn Lynch.

I don't think that's a good estimate of value.

Dodds currently has no RB projected to rush for more than 1208 yards (Charles). In the past five years, at least 5 RBs have rushed for more than that yardage total every year. In 2012 it was 10 guys. That's not an exceptional performance, it's what should be expected of top RBs.

To put it another way, Nelson and Thomas scored 229 and 228 points in 2014. They're projected to score 217 and 215 points in 2015; just a marginal decline. Bell and Lynch scored 287 and 269 points, and they're projected to score 238 and 225 points, a drop of over 40 points each. And no one is taking their place on the rankings. So if you're using projections to choose first and second-round players (via either VBD or AVBD), you're in trouble, because the projections are lying to you.

 
Dodds currently has no RB projected to rush for more than 1208 yards (Charles). In the past five years, at least 5 RBs have rushed for more than that yardage total every year.
Are you once again mistaking Dodds' Le'Veon Bell/Jamaal Charles/et al. projections for Dodds' (unpublished) RB1-RB5 projections? If not, what's the point of that comparison?

Bell and Lynch scored 287 and 269 points, and they're projected to score 238 and 225 points, a drop of over 40 points each.
I can think of a pretty good reason to expect a drop-off for Bell this season. Three of them, actually.

And no one is taking their place on the rankings.
Somebody is obviously taking their place in the rankings.

If you mean that nobody is taking their place in the projections -- which projections? If you're talking about the unpublished projections for RB1, how can you tell? If you're talking about the published projections for Le'Veon Bell, obviously nobody's going to completely take his place, but DeAngelo Williams is partially taking his place.

 
Last edited by a moderator:
Dodds currently has no RB projected to rush for more than 1208 yards (Charles). In the past five years, at least 5 RBs have rushed for more than that yardage total every year.
Are you once again mistaking Dodds' Le'Veon Bell projections for Dodds' (unpublished) RB1-RB5 projections? If not, what's the point of that comparison?

Bell and Lynch scored 287 and 269 points, and they're projected to score 238 and 225 points, a drop of over 40 points each.
I can think of a pretty good reason to expect a drop-off for Bell this season. Three of them, actually.

And no one is taking their place on the rankings.
Somebody is obviously taking their place on the rankings.

If you mean that nobody is taking their place in the projections -- which projections? If you're talking about unpublished the projections for RB1, how can you tell? If you're talking about the published projections for Le'Veon Bell, obviously nobody's going to completely take his place, but DeAngelo Williams is partially taking his place.
We're talking about VBD, which means that the numbers you put in for expected points matter.

When you discount all the RB performances pretty significantly, and don't discount the WR performances to the same amount, VBD (or AVBD) is going to spit out numbers which don't make sense. And frankly it's pretty wimpy for Dodds to take the position that he knows that at least 5 RBs will rush for 1200 yards, but he's only going to project one to do it.

Your RB projections are generally better than Dodds' in this matter; not that you necessarily are more correct in terms of who is going to score what, but that your projection curve is\ more likely to fit reality. Your WR projections, not so much. Why would you project only 3 RBs to rush for more than 1200 yards, and none to rush for more than 1500 yards, but project three WRs to go for over 1500 yards receiving, including one over 1800 yards? Do you really think Julio Jones going for 1800 yards receiving is more likely than one of Le'Veon Bell, Marshawn Lynch, or Jamaal Charles going for 1300 yards rushing? Really?

 
Why would you project only 3 RBs to rush for more than 1200 yards...
That's not what I'm projecting. My best over/under for how many RBs will go over 1200 yards is unpublished.

If we were to roll 100 six-sided dice, I would project about 17 of them to land on six.

But you would take a look at my projections for each die and ask, "How can you project zero of them to land on six? They're all projected to land on 3.5? WTF?"

It is perfectly rational to project that the top 17 dice will all land on six without projecting any individual die to land on six (or even four or five). Similarly, there's nothing inherently illogical about projecting each individual running back to score fewer fantasy points than the end-of-season RB5 is expected to score.

Do you really think Julio Jones going for 1800 yards receiving is more likely than one of Le'Veon Bell, Marshawn Lynch, or Jamaal Charles going for 1300 yards rushing? Really?
My Julio Jones projection is not very good right now. First draft is always hard. Sorry. Ignore that one for now.

ETA: Even so, you can't infer that last statement from my projections. I didn't publish the odds that "one of Le'Veon Bell, Marshawn Lynch, or Jamaal Charles" would go for 1300 yards. The fact that each is projected to rush for fewer than 1300 yards means that you can reasonably infer that I think each one has a less than 50% chance to get there. But suppose I think each has about a 40% chance to get there. That would mean about a 78% chance that at least one of them will get there -- and there's no reason to think I'm assigning Jones a 78%+ chance of getting 1800 yards receiving.

 
Last edited by a moderator:
And frankly it's pretty wimpy for Dodds to take the position that he knows that at least 5 RBs will rush for 1200 yards, but he's only going to project one to do it.
Here's a more accurate way to look at it. He's not projecting only one to do it. He's projecting at least five to do it.

Which five?

DeMarco Murray is projected for 1215 yards, so count that as about a 52% chance, or about 0.52 players with a 1200-yard season.

Jamaal Charles is 0.51 players.

Eddie Lacy is 0.47 players.

Marshawn Lynch is 0.46 players.

Yadda, yadda, yadda down the list, for an eventual total of at least five players.

 
Last edited by a moderator:
How would you handle this in leagues with flex positions? Especially multiple flex spots? Don't a good portion of H2H redrafts use flex spots there days?

Because then, in theory, you're comparing your player to the entire pool of flex-able players - not just those at his position. And yet, most leagues have a mandatory minimum number of starters at each spot. Now, assume that all drafters are picking "best value" (instead of "I'll fill all my RB/WR spots first" or "I'm drafting all Packers"). How do you determine that value?

My long-time (almost 30 years) local league is far from typical, I think, in either set-up, scoring, or drafting habits. But I'll use it as an example of what I mean.

We're mandated to start 1QB, 1RB, 2WR, 1TE, 1K, 1D. We then have 2 flex spots that can be any position with a cap of 2 total starters at every position except WR (where you can start up to 4). Total roster is double the mandatory starters, plus two at any position.

How would you compare players in this scenario? It's a problem I've always struggled with from a VBD standpoint.
I use a historical method of determining how many flex spots to allocate to each position. If the starting lineup requirements are QRRWWWFF in a 12 team league, I know there will be at least 24 starting RBs and 36 starting WRs. Then I'll see how many RB and WRs would have made up the 24 flex spots in the previous year. If it was 14 and 10, then I'll just set my RB baseline at RB38 and WR baseline at 46. Sometimes I'll move the baseline down even lower to include the top reserves for things like injuries, bye weeks, etc.
Right, and that's what I would often do to set a hard baseline for standard VBD drafting.

But based on your OP, are you saying that you'd compare, say, RB13 against RBs 14 down to 38 AND WRs down to 46? If so, which WR is the first "below" the RB13?

 
How would you handle this in leagues with flex positions? Especially multiple flex spots? Don't a good portion of H2H redrafts use flex spots there days?

Because then, in theory, you're comparing your player to the entire pool of flex-able players - not just those at his position. And yet, most leagues have a mandatory minimum number of starters at each spot. Now, assume that all drafters are picking "best value" (instead of "I'll fill all my RB/WR spots first" or "I'm drafting all Packers"). How do you determine that value?

My long-time (almost 30 years) local league is far from typical, I think, in either set-up, scoring, or drafting habits. But I'll use it as an example of what I mean.

We're mandated to start 1QB, 1RB, 2WR, 1TE, 1K, 1D. We then have 2 flex spots that can be any position with a cap of 2 total starters at every position except WR (where you can start up to 4). Total roster is double the mandatory starters, plus two at any position.

How would you compare players in this scenario? It's a problem I've always struggled with from a VBD standpoint.
I use a historical method of determining how many flex spots to allocate to each position. If the starting lineup requirements are QRRWWWFF in a 12 team league, I know there will be at least 24 starting RBs and 36 starting WRs. Then I'll see how many RB and WRs would have made up the 24 flex spots in the previous year. If it was 14 and 10, then I'll just set my RB baseline at RB38 and WR baseline at 46. Sometimes I'll move the baseline down even lower to include the top reserves for things like injuries, bye weeks, etc.
Right, and that's what I would often do to set a hard baseline for standard VBD drafting.

But based on your OP, are you saying that you'd compare, say, RB13 against RBs 14 down to 38 AND WRs down to 46? If so, which WR is the first "below" the RB13?
Yes. In that example, I'd calculate the AVBD for RB1 - 38 and WR 1 - 46. By dividing the gross results by the number of players at each position - 1, I normalize each player's score. They can now be compared across positions. To answer your second question, it will depend on the order of the results from the first steps. It could be WR10 or WR20 depending on the AVBD value of RBs vs WRs. Adjusting the baseline player to accommodate flex spots or injury risk doesn't change the overall method.

 
I don't typically have time to do my own projections these days, so I use FBG's. And there are structural issues with them. I don't think it's correct to project someone to duplicate Murray's 2014, but I also think it's wrong to have the #1 RB projected at 1600/12 total, which is what Dodds has right now.
IMHO, Dodds is doing it right.

Dodds is projecting Le'Veon Bell to be the #1 RB, and he's projecting Le'Veon Bell to get 1600/12. But he is not projecting the #1 RB to get 1600/12.

Those two things are completely different.

The projection for Bell is stated as a point: 1600/12. But the projection isn't really a point; it's a range. It's an X% chance of 1900/16, a Y% chance of 1600/12, a Z% chance of 1300/8, and so on.

Same for all the other players.

When we do projections for Le'Veon Bell, we're giving you a point roughly in the middle of his realistic range (something like the average of each point within the range weighted by its probability, at least in theory).

We don't actually publish projections for RB1, but if we did, it would not simply be a point toward the middle of of Le'Veon Bell's range. If Bell ends up finishing as RB1, it will be because he ended up toward the top of his range, not toward the middle of it. If Jamaal Charles or Matt Forte or even Andre Ellington ends up as RB1, they will also likely have ended up significantly above the middle of Le'Veon Bell's range (and obviously above the middle of their own ranges as well).

In evaluating Le'Veon Bell's value, what matters is not how the end-of-season RB1 usually performs. What matters is how Le'Veon Bell should reasonably be expected to perform. And since Bell has a substantially less-than-100% chance of being RB1, the projection for Le'Veon Bell will necessarily be substantially lower than the (unpublished) projection for RB1.
I'm confused as to what purpose these projections serve if their cumulative point totals diverge from the typical performance of the population. So I've got Le'Veon Bell, who I think has a 20% chance of 300 points, and 50% chance of 250 points, and 30% chance of 200 points, probability-weighted that works out to 245 points. If I do that math for all of the players, and Bell's probability adjusted score is ranked highest of all RBs, what point does it serve to say my RB1 is going to score 245 points when in reality that slot will score 275 points?

I understand the method by which he's ranking players, but attributing a projection value based on this method seems misleading. As Calbear pointed out, people would be using these 'projections' as inputs into a VBD model.

 
Last edited by a moderator:
This Uncle Grandpa is a lot smarter than the real Uncle Grandpa.
Good mornin'.

I'm obviously not too smart to realize how much easier this method could have been implemented. As Greg pointed out, just using the average of the positional player pool as your baseline subtractor gets you to the same conclusion in a much easier manner.

 
And frankly it's pretty wimpy for Dodds to take the position that he knows that at least 5 RBs will rush for 1200 yards, but he's only going to project one to do it.
Here's a more accurate way to look at it. He's not projecting only one to do it. He's projecting at least five to do it.

Which five?

DeMarco Murray is projected for 1215 yards, so count that as about a 52% chance, or about 0.52 players with a 1200-yard season.

Jamaal Charles is 0.51 players.

Eddie Lacy is 0.47 players.

Marshawn Lynch is 0.46 players.

Yadda, yadda, yadda down the list, for an eventual total of at least five players.
I get that argument, but it doesn't work in an discussion about VBD, because of what VBD is used for. VBD is used to determine relative value of players at different positions. For that to work at all, the assumptions made about the players need to be reasonably similar: assumptions about things like regression and probability. And I don't think the assumptions which result in Antonio Brown being projected to have 126 receptions are congruent with the assumptions which result in Marshawn Lynch getting 238 carries. And to the point around baselines, I don't think the assumptions that have Todd Gulley scoring 135 fantasy points are congruent with those that have Torrey Smith scoring 111.

When the WR projection curve looks a lot like the WR production curve, and the RB projection curve doesn't look much like the RB production curve, VBD will give you structurally inaccurate answers.

 
And frankly it's pretty wimpy for Dodds to take the position that he knows that at least 5 RBs will rush for 1200 yards, but he's only going to project one to do it.
Here's a more accurate way to look at it. He's not projecting only one to do it. He's projecting at least five to do it.

Which five?

DeMarco Murray is projected for 1215 yards, so count that as about a 52% chance, or about 0.52 players with a 1200-yard season.

Jamaal Charles is 0.51 players.

Eddie Lacy is 0.47 players.

Marshawn Lynch is 0.46 players.

Yadda, yadda, yadda down the list, for an eventual total of at least five players.
I get that argument, but it doesn't work in an discussion about VBD, because of what VBD is used for. VBD is used to determine relative value of players at different positions. For that to work at all, the assumptions made about the players need to be reasonably similar: assumptions about things like regression and probability. And I don't think the assumptions which result in Antonio Brown being projected to have 126 receptions are congruent with the assumptions which result in Marshawn Lynch getting 238 carries. And to the point around baselines, I don't think the assumptions that have Todd Gulley scoring 135 fantasy points are congruent with those that have Torrey Smith scoring 111.

When the WR projection curve looks a lot like the WR production curve, and the RB projection curve doesn't look much like the RB production curve, VBD will give you structurally inaccurate answers.
:goodposting:

I'm with CB here.

 
For that to work at all, the assumptions made about the players need to be reasonably similar: assumptions about things like regression and probability.
Sure, but that's an entirely different discussion from the one I joined the thread for, and from some of the arguments made in, e.g., posts #4, #46, and #48.

If you think our projections are bad (read: suboptimal in some systematic, correctable way) because WRs in general aren't sufficiently regressed toward the mean or whatever, you may be right, and that's certainly a discussion worth having.

But if you think our projections are bad because the distances between RBs (specified by name) in our preseason projections don't look sufficiently enough like the distances between RBs (unspecified until after the fact) going by their typical end-of-season statistics, I think that argument is demonstrably, unequivocally wrong in principle.

 
Last edited by a moderator:
CalBear, were you around for the endless discussion in 2002-2003 about AVT?
Yep. I'm not a huge fan of AVT, used in isolation. Players are individuals and each player's performance is based on a lot of different factors. But obviously what I'm saying about fitting projection and production curves is getting at the same thing that AVT was trying to.

What I'm saying is not that the RB1 should necessarily be projected to score the same as the 5-year average of RB1s. What I'm saying is that the degree to which you discount performance at one position needs to be congruent to the degree to which you discount performance at the other positions, or else VBD is going to give you structurally inaccurate results.

I wish we could embed images here, but here's a spreadsheet with charts to illustrate my point:

https://docs.google.com/spreadsheets/d/1CFCuA4K7fvsYWHpAmo-NpYmUuq5lOg1Ca5RAl4ftT5Y/edit?usp=sharing

Both you and Dodds project WRs very close to the historical curve for WR production, and both you and Dodds appear to significantly under-project the top 5-7 RBs relative to the historical curve for RB production. You're not making the same assumptions for both groups, and that's going to result in the top 5-7 RBs being undervalued by VBD relative to WRs.

It's not just that you're projecting Lynch to get 300 fewer yards and 4 fewer TDs than last year; it's that you're also discounting Lacy, Charles, and Forte, and you're not discounting in the same way for A.Brown, Bryant, J.Jones, D.Thomas.

 
I agree with using some of the historical data to evaluate positional expectations. However, why are both of you using 5-year averages? Based on the rule changes, I would think a 3-year average is a better estimation point than 5-years. The league has been moving to a more WR friendly type system. So using a 5-year average tends to slightly devalue WR and overvalue RB in the middle rounds until you have more of the 5-year data under the modified rule structure.

I guess I am thinking a 5-year average may put you at a slight disadvantage in your projections. :shrug:

 
What I'm saying is not that the RB1 should necessarily be projected to score the same as the 5-year average of RB1s. What I'm saying is that the degree to which you discount performance at one position needs to be congruent to the degree to which you discount performance at the other positions, or else VBD is going to give you structurally inaccurate results.

I wish we could embed images here, but here's a spreadsheet with charts to illustrate my point:

https://docs.google.com/spreadsheets/d/1CFCuA4K7fvsYWHpAmo-NpYmUuq5lOg1Ca5RAl4ftT5Y/edit?usp=sharing

Both you and Dodds project WRs very close to the historical curve for WR production, and both you and Dodds appear to significantly under-project the top 5-7 RBs relative to the historical curve for RB production. You're not making the same assumptions for both groups, and that's going to result in the top 5-7 RBs being undervalued by VBD relative to WRs.

It's not just that you're projecting Lynch to get 300 fewer yards and 4 fewer TDs than last year; it's that you're also discounting Lacy, Charles, and Forte, and you're not discounting in the same way for A.Brown, Bryant, J.Jones, D.Thomas.
I think you might be heading towards a true conclusion, or a conclusion that has a decent chance of being true -- namely, that I'm overprojecting the top WRs in comparison to the top RBs.

But I think the arguments you're providing in support of that possibly true conclusion -- here, look at some AVT-style numbers -- are particularly unlikely to persuade me.

It's simply not true that RBs and WRs should necessarily be regressed the same amount from their AVT-style numbers. It's less obvious with WRs and RBs, so let's consider WRs and Team Defenses instead to make the general point. In certain scoring systems, fantasy defensive production is nearly random. The end-of-year distance between DT1 and DT10 and DT20 might be significant. But as with the dice example I keep using, the rationally estimated distance between the Seahawks (ranked #1), the Bears (ranked #10), and the Falcons (ranked #20), as estimated during the preseason, might well be miniscule.

You can't show that projections are suboptimal by showing that Position A is more regressed to the mean than Position B is, because appropriate regression is in part a function of uncertainty (as measured, for example, by how much turnover there is from year to year in the top ten at a given position), and there's no rule saying that uncertainty or other relevant factors are the same across all positions. There is way more yearly turnover in certain scoring systems in the top ten DT list than there is in the top ten WR list, for example, which means that DT projections should look less like AVT numbers than WR projections should. Showing that DT projections are regressed more than WR projections is not sufficient to show that the projections aren't as good as they can possibly be.

The principle is just as true when comparing WRs and RBs. It might well be the case that we're doing things wrong by regressing WRs too little or RBs too much, or both, but AVT-style reasoning is logically incapable of making that point.

 
Last edited by a moderator:
I agree with using some of the historical data to evaluate positional expectations. However, why are both of you using 5-year averages? Based on the rule changes, I would think a 3-year average is a better estimation point than 5-years. The league has been moving to a more WR friendly type system. So using a 5-year average tends to slightly devalue WR and overvalue RB in the middle rounds until you have more of the 5-year data under the modified rule structure.

I guess I am thinking a 5-year average may put you at a slight disadvantage in your projections. :shrug:
Since I have the data, I ran this, and it's not really much different.

5-year

RB1-24, 311-132 points

WR1-36, 236-110 points

3-year

RB1-24, 309-124 points

WR1-36, 236-111 points

So the RBs get a very minor downtick (within the margin of error), and the WRs are identical.

 
You can't show that projections are suboptimal by showing that Position A is more regressed to the mean than Position B is, because appropriate regression is in part a function of uncertainty (as measured, for example, by how much turnover there is from year to year in the top ten at a given position), and there's no rule saying that uncertainty or other relevant factors are the same across all positions. There is way more yearly turnover in certain scoring systems in the top ten DT list than there is in the top ten WR list, for example, which means that DT projections should look less like AVT numbers than WR projections should. Showing that DT projections are regressed more than WR projections is not sufficient to show that the projections aren't as good as they can possibly be.
OK, so let's look at this; how much turnover is there in the top 10 of the two positions?

RBs returning to the top 10 since 2005:

5 (1 instance)

4 (4 instances)

3 (4 instances)

Average: 3.7

WRs returning to the top 10 since 2005:

6 (1 instance)

4 (2 instances)

3 (6 instances)

Average: 3.6

Let's call that a wash; you can expect 3 or 4 of last year's 10 RBs and WRs to finish in the top 10.

(Incidentally, those numbers are much higher for QBs and TEs, and I think that means that top QBs and TEs are actually more valuable than their VBD numbers would suggest).

 
RBs returning to the top 10 since 2005: Average: 3.7

WRs returning to the top 10 since 2005: Average: 3.6

(Incidentally, those numbers are much higher for QBs and TEs, and I think that means that top QBs and TEs are actually more valuable than their VBD numbers would suggest).
Stud QBs and TEs are chronically undervalued by traditional VBD and its mainstream acceptance in fantasy circles.

I LOL every time Michael Fabio talks smack about taking QBs early in drafts.

 
CalBear, were you around for the endless discussion in 2002-2003 about AVT?
I can't find any of the message board discussions from 2002, but here's one from 2007.
To crudely summarize your portion of the conversation in the link, it was, "AVT sucks. Projections are better." While there was a brief explanation of the former, there was absolutely no substantiation of the latter. Have you backtested your own projections to see if you're outperforming AVT?

 
You're making this too complicated - and you shouldn't just use baseline. You should be looking at the relative value given whether a player will be available at your next pick. Strip VBD down to the next ~20 players at any given point and align with ADP. I've had much better success with that strategy.

2013: 18 leagues (112% ROI)

2014: 14 leagues (72% ROI)

 
Last edited by a moderator:
You're making this too complicated - and you shouldn't just use baseline. You should be looking at the relative value given whether a player will be available at your next pick. Strip VBD down to the next ~20 players at any given point and align with ADP. I've had much better success with that strategy.

2013: 18 leagues (112% ROI)

2014: 14 leagues (72% ROI)
No offense, but this proves absolutely nothing about the strategy.

 
Last edited by a moderator:

Users who are viewing this thread

Back
Top