Introducing the Market-Driven Baseline for VBD (1 Viewer)

[Instinctive]

Nice job Fresh, good effort, keep at it and don't let folks try and stop you. Feedback is always good, there is room for everyone.

Nobody's trying to stop him - just helping understand the implications. This year I've really started to understand why you rub people the wrong way...

Show me in the rules where I am not allowed to offer support for other writers as I deem fit? Hope you have a better afternoon Instinct, sorry you feel that way.

Thanks bud. Just academic conversation going on in here. I love it. I'll be using MDB to my heart's content, so I'm good! Just seeing what the masses think, and very happy to have an engaged audience. Even if they think I'm a complete dolt.

I wouldn't go that far Just trying to tighten the principle and make it even better, I think.

 

[Concept Coop]

OK, lets go to fantasy land. Let's say the 32 picks in the draft are kickers, before the MDB. So, out of the 144 picks at the MDB, 32 are kickers. Looking at DD, Gostowski in my league is projected to have 136 FP. Kicker #32, Josh Scobee, is worth 95. So, with Mr. Scobee being the baseline, Gostowski's X-number is 41. OK, so we have the 112 other picks now. Let's say 21 QBs still went. Rodgers is worth 441, QB #21- Jay Cutler- is worth 259. That's an X-number of 182. So, that proves that MDB holds up just fine in this ridiculous example. It's the projections that keep everything honest. Less valuable positions will remain so, and this method accurately conveys exactly where your league perceives value to be.

But I don't think you have explained why it matters? As the quoted poster points out - why should I pick Foster or Rodgers over the other, because of how my league drafted in the past?Besides, any good drafter uses some sort of tiering method, which reacts to how your league is actually drafted. How they are actually drafting seems more important than how they might draft, based on how they drafted last year.

 

[moleculo]

not that anyone cares, but I've developed my own secret formula for establishing baseline. I'm even willing to share it, so I suppose it's not so secret.

Instead of specifying a baseline player, I establish a baseline score. to do that, I multiply the number of starters at each position by 1.5, and then average the overall scores for that many players. that score becomes my baseline.

so, say there are 12 teams, starting 1 QB each. 12 starting QB's. So, I average the scores for the top 18 QB's and call it a day.

Why do I do average a bunch of players' scores for VBD? My VBD baseline is now not dependent on any single players projected value, and not subject to any tiers. suppose you want worst starter, but there is a tier break between QBs 9-11 and then a big drop-off before 12-15. Now, the VBD value for QB1 will be effected by the dropoff between 11 & 12, and that doesn't make sense to me. I think the VBD value should be a reflection of the entire group of players to be drafted, not one guy way down the list.

Why do I average 1.5* number of starters? No idea. It was a matter of guess and check, until I came up with a combined VBD list that looked similar in player distribution to ADP. In actuality, my averaged VBD ends up being somewhere around QB9, so I'm looking for something slightly better than worst starter, I suppose.

 

[Ministry of Pain]

not that anyone cares, but I've developed my own secret formula for establishing baseline. I'm even willing to share it, so I suppose it's not so secret.Instead of specifying a baseline player, I establish a baseline score. to do that, I multiply the number of starters at each position by 1.5, and then average the overall scores for that many players. that score becomes my baseline.so, say there are 12 teams, starting 1 QB each. 12 starting QB's. So, I average the scores for the top 18 QB's and call it a day.Why do I do average a bunch of players' scores for VBD? My VBD baseline is now not dependent on any single players projected value, and not subject to any tiers. suppose you want worst starter, but there is a tier break between QBs 9-11 and then a big drop-off before 12-15. Now, the VBD value for QB1 will be effected by the dropoff between 11 & 12, and that doesn't make sense to me. I think the VBD value should be a reflection of the entire group of players to be drafted, not one guy way down the list.Why do I average 1.5* number of starters? No idea. It was a matter of guess and check, until I came up with a combined VBD list that looked similar in player distribution to ADP. In actuality, my averaged VBD ends up being somewhere around QB9, so I'm looking for something slightly better than worst starter, I suppose.

 

[Sarnoff]

Nice job Fresh, good effort, keep at it and don't let folks try and stop you. Feedback is always good, there is room for everyone.

Nobody's trying to stop him - just helping understand the implications. This year I've really started to understand why you rub people the wrong way...

MOP just seems to fail reading comprehension. All that's going on here is people trying to understand what the OP means by asking questions. No one is trying to stop anyone.

 

[Lab]

Actually, it does capture perfectly how my league acts.

I never said it didn't, but that's not by design. It's more a coincidence generated by the fact that most of your leaguemates are probably using projections that are roughly similar to the ones you use, so for example their model is telling them that Aaron Rodgers should go in the mid-first round, and so is yours. It's not much of a thought experiment, but imagine if (for whatever reason) all your leaguemates decide this year to wait on QBs. ... QBs would be drafted later than usual, but your model would still spit out the same exact results that it is now, so it wouldn't accurately capture the way your league would act. ... most people use similar projections and most drafts follow a fairly predictable pattern. ... MDB won't tell you whether or not [Aaron Rodgers] still be there in the second round. It'll tell you (based on your projections) where he should be taken, but the fact that 21 QBs will be off the board by round 12 doesn't give you any indication where Rodgers will go among those 21. ...

I agree with this, but this is really a characteristic of static vs dynamic VBD. I think that this MDB is probably an improvement over static VBD but not as good as dynamic VBD. The problem with dynamic VBD is it's difficult to do w/o a computer - if you want to draft with a piece of paper and that's it, you are kind of stuck with static, which is where this may help.

After reading all this, my view is best captured by what Moleculo says above. I'm a VBD believer, just like most people on this thread. One of the hardest parts of VBD is selecting an effective baseline. Baselines like "top 100" or "all starters" or "all drafted" are arbitrary and often give poor VBD results, because they do not accurately reflect how real owners draft players. Using a baseline based on ADP would be an improvement, since it would more closely reflect how real owners will draft. But because different leagues draft differently (often based on different scoring systems), even using ADP as a baseline would give less-than-satisfactory VBD results. So Freshly Shorn's proposed method of creating a baseline tailored to how your particular league has drafted historically ("MPD") seems like an improvement over a simple ADP as a baseline. Its goal is to use historical results to predict how your particular league will draft in 2012, so you can predict VBD scores before the draft begins.

An even better approach might be to start with MPD's historical averages, and then tweak them based on how you sense your league's draft will go down in 2012. This tweaking would allow you to predict changes in the 2012 draft. For example, in 2012, it certainly seems the top TEs (Gronk and Graham) will be drafted in the 1st or 2nd round of many drafts. This is a big change from prior draft years. The MPD approach will fail to capture that change, and will inaccurately value Graham and Gronk in 2012. The easiest way to fix that problem is to allow for tweaking based on your predictions of how the draft will flow. I think someone above suggested combining MPD with current-year ADP info, and that seems like a promising approach.

But Moleculo raises the key question: If you're doing all this work to improve the accuracy of your VBD baseline, why not go ahead and take the next step to dynamic VBD? Dynamic VBD is specially geared to predict how the draft will flow in real time, on a round-by-round basis. It's certainly more accurate to create a continuously improving baseline with realtime data than it is to rely on old data.

In the end, it seems a little like you're trying to build a better gasoline engine (and maybe even succeeding!), while ignoring the fact that the auto market is shifting toward electric power for cars. It seems like a better solution that MPD already exists (dVBD), so why not put your efforts into creating an improved dVBD system, instead of tinkering with static VBD baselines?

(BTW, I mean this a constructive criticism and a suggestion for future work, not an attack of any sort. You're doing good work here, but I just think it's misplaced effort.)

 

[freshly_shorn]

But I don't think you have explained why it matters? As the quoted poster points out - why should I pick Foster or Rodgers over the other, because of how my league drafted in the past?Besides, any good drafter uses some sort of tiering method, which reacts to how your league is actually drafted. How they are actually drafting seems more important than how they might draft, based on how they drafted last year.

My apologies, Coop! I meant to answer all the direct questions in my last post and neglected yours.Let me explain my bias- don't know if you saw my post about the free Draft Planalyzer spreadsheet that helps you understand how your league drafts, in terms of trends and such. So, as I developed that, I began to wonder if there wasn't something in there that could apply to baselines. Like many leagues, we have flex positions and I was never satisfied on any method to accommodate for those. So I began to think that perhaps I could look at historical trends in my league to help me determine my baselines.As part of that exercise, I decided that the most valuable players are starters. So, going into Week One, the only way to measure that is through projections. Of course, my league doesn't draft all starters before drafting reserves. So, I decided that I could easily capture what my league does do by using that same point (all starters X all teams) as my baseline, then that tells me what the league values position-wise.I don't know how to make it any clearer- I like to know what my league thinks, in every way I can. I think MDB is a better method then just picking baselines by some arbitrary measure because it is linked to how my actual league behaves. It reveals, in a simple way, what the league as a whole values. I can then use that knowledge to out manuever them at their own game. Is there other ways to do it? Absolutely yes. Are they better, or worse? I'd argue some are worse, and some are just as good. I haven't been made privy to anything better.Does that answer your question? I'm not offering MDB as something better. It isn't a revolutionary way to baseline. But it is simple, and it conveys more information about an established league than any method out there that I'm aware of.

 

[freshly_shorn]

After reading all this, my view is best captured by what Moleculo says above.

I'm a VBD believer, just like most people on this thread. One of the hardest parts of VBD is selecting an effective baseline. Baselines like "top 100" or "all starters" or "all drafted" are arbitrary and often give poor VBD results, because they do not accurately reflect how real owners draft players. Using a baseline based on ADP would be an improvement, since it would more closely reflect how real owners will draft. But because different leagues draft differently (often based on different scoring systems), even using ADP as a baseline would give less-than-satisfactory VBD results. So Freshly Shorn's proposed method of creating a baseline tailored to how your particular league has drafted historically ("MPD") seems like an improvement over a simple ADP as a baseline. It's goal is to use historical results to predict how your particular league will draft in 2012, so you can predict VBD scores before the draft begins.

An even better approach might be to start with MPD's historical averages, and then tweak them based on how you sense your league's draft will go down in 2012. This tweaking would allow you to predict changes in the 2012 draft. For example, in 2012, it certainly seems the top TEs (Gronk and Graham) will be drafted in the 1st or 2nd round of many drafts. This is a big change from prior draft years. The MPD approach will fail to capture that change, and will inaccurately value Graham and Gronk in 2012. The easiest way to fix that problem is to allow for tweaking based on your predictions of how the draft will flow. I think someone above suggested combining MPD with current-year ADP info, and that seems like a promising approach.

But Moleculo raises the key question: If you're doing all this work to improve the accuracy of your VBD baseline, why not go ahead an take the next step to dynamic VBD? Dynamic VBD is specially geared to predict how the draft will flow in real time, on a round-by-round basis. It's certainly more accurate to create a continuously improving baseline with realtime data than it is to rely on old data.

In the end, it seems a little like you're trying to build a better gasoline engine (and maybe even succeeding!), while ignoring the fact that the auto market is shifting toward electric power for cars. It seems like a better solution that MPD already exists (dVBD), so why not put your efforts into creating an improved dVBD system, instead of tinkering with static VBD baselines?

(BTW, I mean this a constructive criticism and a suggestion for future work, not an attack of any sort. You're doing good work here, but I just think it's misplaced effort.)

Thank you. I should say this method is best suited for draft preparation and mocking; I never follow what any baseline tells me verbatim. Since I'm not creating software to track drafts, DVBD is out of my league to add anything to. But, I think MDP is a better starting point than any other method. Once you enter the draft, you naturally must adapt. But, the good thing is by having MDP as a starting point, you will know exactly how the league is deviating from last year (same with if you are using custom adp).

 

[Ignoratio Elenchi]

Yes, I believe that if my league drafts more RBs than QBs at a certain point of the draft, then RBs are perceived to be the more valuable position.

Right, but the goal is not to determine how valuable players are perceived to be, the goal is to determine how valuable they are.

Obviously, this whole exercise is about identifying value- finding players that are undervalued by the 'market' and picking them at the appropriate spot.

Right, but in order to determine if a player is being undervalued by the market, you need to first determine what their actual value is. Then you can look at something like ADP to see if your league is over- or under-valuing players. If you determine a players value based on what the market thinks their value is, then no player will be under- or over-valued by the market - they'll all be valued exactly where the market values them (you see how convoluted and circular this is).

I totally disagree that a players value tells you when to take them. ADP tells you more about that. VBD tells you how a player ranks against other players and ranks them accordingly. You have to decide when to take them, using all available information.

ADP tells you when other teams are likely to take players. The point of VBD is to determine when they should be taken (i.e. their rank relative to all other players in the pool). You combine these two sources of information to best determine when to draft each player, but you don't combine them ahead of time. You need to figure out their value independently of what ADP or your league tendencies say they're likely to be drafted. Otherwise, like I said, you're just going in circles.

 

[moleculo]

let's tighten up some terminology and analogies here. here's how I see things:

Projections: my appraisal of players values, comparing players within the same position.

VBD: a method to compare players across positions, a more complete appraisal.

ADP: macro-scale estimate of player value. This is essentially the sum total appraisal of the entire FF world, and may or may not reflect how your specific league will value particular players.

MBD: an estimate of how your specific league will value players.

MBD is more valuable than ADP or VBD, as it is a reflection of a specific league, which is likely to differ from the FF universe.

IMO, the ultimate drafting method would combine dVBD with MBD, essentially looking forward to how many players your specific league is likely to draft at a particular position and setting baselines accordingly.

 

[freshly_shorn]

1) Right, but the goal is not to determine how valuable players are perceived to be, the goal is to determine how valuable they are.

2) Right, but in order to determine if a player is being undervalued by the market, you need to first determine what their actual value is. Then you can look at something like ADP to see if your league is over- or under-valuing players. If you determine a players value based on what the market thinks their value is, then no player will be under- or over-valued by the market - they'll all be valued exactly where the market values them (you see how convoluted and circular this is).

3) ADP tells you when other teams are likely to take players. The point of VBD is to determine when they should be taken (i.e. their rank relative to all other players in the pool). You combine these two sources of information to best determine when to draft each player, but you don't combine them ahead of time. You need to figure out their value independently of what ADP or your league tendencies say they're likely to be drafted. Otherwise, like I said, you're just going in circles.

1) I'd love to be able to determine how valuable they are. But in the preseason it is *all* guess work.2) Again, preseason value is all perceived- no way to truly know actual value. But I'll stipulate that you are referring to your projected fantasy points for players. But, again, MDB is not determining value directly. It is setting the baseline where you measure value from. Fantasy points determine value. All MDB does is give you a lens to look at your projections in the context of what your league does by a certain point in the draft, no different from any other baseline technique. Where this is better, IMO, is that context is the actual one you will be drafting in.

3)You can't figure out a players value within the context of your league without examining how that league operates. MDB by definition operates within that context. And I don't think this goes in circles at all, pretty straight forward. Besides, circles make me dizzy.

I'm choosing to begin measuring value at a point determined by league size and number of starters. I tally the number of each position taken by that point in the draft, and I gain clear knowledge of exactly how much value/emphasis my league-mates put on a given position. My projections don't change. My overall strategy isn't likely to change. It just tells me how valuable players are likely to be given my specific league's past draft habits rather than by some nebulous criteria that may or may not reflect the reality of how my league behaves.

Keep in mind, too, that I know my league's habits don't change much based on my studies. So I happen to know, then, that MDB will work well in my league. I also can make adjustments as you would with any model, since I'm positive at least a couple of tight ends will go earlier than tight ends have gone in the past. But I doubt the overall number of tight ends taken by pick 144 will change much. Point being, none of this is in a vacuum- baselining is only one component of the whole process, as you know.

 

[freshly_shorn]

let's tighten up some terminology and analogies here. here's how I see thingsrojections: my appraisal of players values, comparing players within the same position.VBD: a method to compare players across positions, a more complete appraisal. ADP: macro-scale estimate of player value. This is essentially the sum total appraisal of the entire FF world, and may or may not reflect how your specific league will value particular players.MBD: an estimate of how your specific league will value players.MBD is more valuable than ADP or VBD, as it is a reflection of a specific league, which is likely to differ from the FF universe.IMO, the ultimate drafting method would combine dVBD with MBD, essentially looking forward to how many players your specific league is likely to draft at a particular position and setting baselines accordingly.

Love it. And, again, I agree on the MDB-dVBD thing- my method is the starting line, and I think dVBD would take over from there when we're in the draft room. Makes a ton of sense.But.. MDB isn't more valuable than VBD as it is part of VBD. It may be more valuable than other generic baselining methods, which are also part of VBD.

 

[freshly_shorn]

Ignoratio Elenchi, I am curious on how you baseline, by the way. It's not all about me! Care to elaborate on your method? I know this topic isn't about your methods but I am curious. Heck, I'm curious about how anyone does it- Moleculo shared his and that was cool, and QSPAR teased us with his method.. Please, anyone, let us know how you do it and why it works for you! And do you think MDB has a place in your strategy, for those of you interested in dual baselines?

 

[Ignoratio Elenchi]

MBD is more valuable than ADP or VBD, as it is a reflection of a specific league, which is likely to differ from the FF universe.

But I don't see how you can state this as fact. That's exactly what this whole thread has been about, and OP has so far failed to demonstrate that MDB is actually better than other common baseline methods at determining value. Just because it "incorporates your leagues tendencies" doesn't mean it's a good way to determine baselines. One last time: Knowing your league's draft tendencies IS valuable information. But no one has demonstrated that using those tendencies to determine player values is useful. It's great to know that, every year, owner X likes to take QBs early, and owner Y always fills out his starting lineup before drafting any backups, and by the end of the third round there are always 3 top TEs off the board, etc. All of that is extremely good to know - I have all that kind of data on my own long-running league and it's great. But independently of that, you need some way to determine how much players are worth. The market doesn't dictate how much better Aaron Rodgers will be than Jake Locker this year. That's something no one knows in advance, of course, but we attempt to project - not based on perceived market values, but based on statistics, league settings, etc.

Anyway, I'm bowing out of the thread because we're just going in circles now. The same questions keep being raised and they're not really being answered. I do wish OP the best with his system and his blog, and contrary to MOP'ss weird interjection I think it was a great discussion.

 

[Ignoratio Elenchi]

Ignoratio Elenchi, I am curious on how you baseline, by the way. It's not all about me! Care to elaborate on your method? I know this topic isn't about your methods but I am curious. Heck, I'm curious about how anyone does it- Moleculo shared his and that was cool, and QSPAR teased us with his method.. Please, anyone, let us know how you do it and why it works for you! And do you think MDB has a place in your strategy, for those of you interested in dual baselines?

I don't use VBD with a static baseline. I use sort of a dVBD model, where "baselines" change constantly based on how the draft is going, what my team needs, etc. So to me, this discussion is more about theory than anything practical. As another poster mentioned earlier, trying to make VBD better is kind of like trying to build a better typewriter, when everyone's already using computers.

 

[freshly_shorn]

I agree that a dVBD system is great in draft but you have to start somewhere. dVBD takes your league's behavior into account, which MDB does in computing a starting point.

Good stuff. Thanks for your thoughts and effort!

 

[The Jerk]

A Shark Pool thread several seasons ago on using the DD with customized position rank got me thinking about what freshly_shorn is calling MDB back in 2009 or so. Then -- as now -- the debate on its utility centered on the benefit provided by customizing the baseline to fit a given league vs. the possible inaccurate valuation on players for drafting your team. To me, the most useful combination would be to have two DDs, one with the league baseline to be able to best predict what the league is going to do, and one with a different baseline that best sets true value of players across different positions.

For me, I use a standard FBG baseline for my own team valuations, but to utilize the "MDB" information specific to my league -- and also because I am at heart a tier drafter and also use a second projections source, I add a simple spreadsheet for use in addition to the draft dominator.

The way I use the "MDB" information is to use the three previous drafts of my league through 80 selections (weighted 0.5 for the previous season, 0.33 for two seasons ago, 0.17 for three seasons ago) to come up with how many players project to be selected at each position by pick 80. I then create a league VBD (basically "MDB" although I calculate it slightly differently using Excel and Visual Basic) to rank the projected top 80 picks. I use this information to create essentially a league ADP (LADP) which I include on my spreadsheet. In this way, I know when given players are expected to be chosen -- although of course, what I really know is what POSITIONS should be chosen, as the actual PLAYERS do not shake out the same, especially once the first 6-8 picks are made.

So my spreadsheet looks something like this, with QBs, RBs, WRs, and TEs, left to right.

Player name, overall rank FBG, overall rank FFsource2, LADP

Last year, for example, #1 receiver projected, Calvin Johnson, 3, 12, 8

so Calvin Johnson was #3 overall in my scoring system according to FBG, #12 according to source2, and projected as the 8th pick in the draft based on league history.

In addition, I stagger the rows to represent tier breaks across positions, clearly delineating when there is a large gap within a position, at least as I see it.

The goal for me is to see simultaneously both the projected value of the player to me, and the expected place where this player is likely to be drafted in my league. This helps me know when to slightly reach for a given player, and also which player is more likely to hang around an extra round if I am choosing between two closely-valued players.

Yes, it involves a little bit of prep time, but I've learned that the prep time is valuable in and of itself as it enables you to do some analyzing you might otherwise not do until your pick is imminent.

 

[The Jerk]

I should add...

If you don't want to create a spreadsheet, you can just place "source2" and "LADP" information in the Notes of the DD.

 

[moleculo]

so I mentioned how I do my static VBD baseline above. The Jerks post above has prompted me to spill some more beans as to what I'm doing this year, because I like how he reports multiple bits of data that only he might understand.

Beginning lat year, I started with dVBD. I do it based on ADP. I found a source that reports average draft position as well as standard deviation. I use some statistics in excel and determine the probability that each position is drafted at a given slot - probability that a RB, WR, QB, and TE are drafted at 1.11, for example. Next, I sum up the probabilities for each position for the next 2 rounds, and set that as my initial baseline. hypothetically, I might believe that in the next 24 picks, 12 RB's, 8 WRs, 2 TEs and 2 QB's will be picked, for example. I do some other stuff to auto-correct when people don't draft as I think they should...if my math says that 8 WR's should have been drafted by now but only 6 have been, I will add 2 positions to the WR baseline, because at that point, WR's are due and the league will self-correct. Last year, it seemed fairly accurate, and I liked how it worked.

Now, for each player, I list the highest I expect that player could be drafted (basically adp + stdev), the average draft position, and their projected score. This lets me know at a glance when it's OK to reach for someone I think is worth reaching for, and who has been left behind, relative to ADP. I also place a (*) next to the players name if there is a lot of uncertainty in his projections - I guess I should have mentioned I use about 7 or 8 sources for projections and average them...in addition to having a broad sense of projections, I have an idea who the experts generally agree on and who they don't...I find that to be worth considering. Matt Forte's cell reads: 6. M Forte CHI * 1.8 (1.10), for example.

What I don't do is try to evaluate what my league will do. That's definitely something that I could improve on. IMO using actual historical league data would be stronger than ADP. That's why I find this MDB to be interesting.

IMO, the way that should be done is to compare your leagues tendency vs ADP each year, see how your league differs. Then, take this years ADP, add the league-specific difference, and there you go - how your league will respond given todays landscape.

 

[The Jerk]

What I don't do is try to evaluate what my league will do. That's definitely something that I could improve on. IMO using actual historical league data would be stronger than ADP. That's why I find this MDB to be interesting.IMO, the way that should be done is to compare your leagues tendency vs ADP each year, see how your league differs. Then, take this years ADP, add the league-specific difference, and there you go - how your league will respond given todays landscape.

Good stuff, moleculnce every 4-5 years, I find that the league makes a decisive change from past draft tendencies. That clearly throws a wrench into my plans, but it's not necessarily catastrophic. If you are thoroughly prepared for a draft, and I greatly expect that you are, then you have already prepared worst-case scenario/fallback players for each position. All I'm saying is that league history will outperform ADP most of the time, except for when it doesn't. (Yes, it's a joke, but it's also kind of true!!!)In truth, some leagues could see significant shakeups this year. QB numbers were so outrageous last year that I suspect more "stud RB" owners will pick QBs earlier than ever before. It may be that QBs revert to the mean, and I would expect that, but my point here is that previous year drafts may carry less weight than normal given how unusual 2011 was in terms of top QBs.moleculo,The final step in my draft preparation, once I know my exact boring old serpentine draft slot, is to know precisely which QB, RB, WR, TE, etc. should be selected by the time of my 1st, 2nd, 3rd, 4th, etc. pick, down to about round 9 or 10. This adds a dynamic aspect to the preparation like your are describing. What's great about this is that, combined with my tiering preference, I can know exactly which 2-3 RBs and WRs I will likely be deciding in round 4 several days if not a full week before my draft. I can do a lot of specific decision-making and turn it over in my head outside of the time pressure and distractions of the actual draft -- which as commissioner, I run via excel fed to a lcd projector.Even better, often one or two players from a higher tier are left over, making the decision even easier.

 

[yrforyg]

I've always wondered: why not take an empirical approach to setting the baseline? That is, given a set of projections (e.g. dodds) and adp, why not simulate a large number of drafts in draft dominator using a variety of different baselines, then compare the strengths (total projected fantasy points) of the teams (either just starters or full team, as preferred) that are drafted as a result? The baseline that produces the strongest team is the best, right? Of course (as with any vbd method), this is dependent upon the accuracy of the projections.

 

[BigDave]

Need help. Leave a link and I'll reply to yours.

This year we are starting a mandatory TE position. In yrs. past it could be included with WR's or a Flex. (eliminated the flex)

So when I do my baseline for TE's using top 10 in a 10 team league, the numbers are off the charts versus other positions.

For example, Fitzgerald at #16 overall is +42, Julio at #19 is +47. Graham at #17 overall is +81 and Gronk at #22 is +62.

Are Top TE's that much more important to have over other positions to outscore you opponents TE each week????? Seems like it forces you to take a TE early. I know Graham and Gronk then the cliff starts in TE's. Even later they are ranked higher based on vbd.

 

[Lord Fantasy]

Found this amazing thread while trying to understand VBD and how it relates to the FLEX position.

I'm in a league where my VBD calculations say that RBs are the most valuable followed by QBs, WRs and TEs.

If have to start 1 QB, 2RB, 3 WR, 1 TE and 1 RB/WR/TE flex then am I looking to draft so that I can start 3 RBs?

Or do I draft in an effort to maximize VBD across the board and then start whoever gives me the best points that week?

Thanks!

 

[cdubz]

Found this amazing thread while trying to understand VBD and how it relates to the FLEX position.

I'm in a league where my VBD calculations say that RBs are the most valuable followed by QBs, WRs and TEs.

If have to start 1 QB, 2RB, 3 WR, 1 TE and 1 RB/WR/TE flex then am I looking to draft so that I can start 3 RBs?

Or do I draft in an effort to maximize VBD across the board and then start whoever gives me the best points that week?

Thanks!

This, just because RBs on average generate the highest value over baseline doesn't mean the RBs that are available when you pick will be the best values.

That said, if that league is non-ppr I think my personal drafting habits would generate a RB at the flex over 90% of the time.

 

[GregR]

Totally missed this thread last year. IE knocked it out of the park as usual.

Lord Fantasy... VBD is good at showing you what the value of a segment of players is Typically it can do a very good job with starters, but generally does a poor job with backups regardless of baseline (because projections seldom are created to reflect value of backups).

Once you understand player value, then you can understand the order that they should be drafted. And then you can compare that to your league tendencies or ADP to determine what players are likely to represent bargains and which are likely to be taken earlier than they should.

So note, that means take the results of VBD, and then combine it with ADP (or your own customized view of what your league will do). Once you have that, now you look the results over and you decide what combination of players available at your picks are likely to give you the best final team. It might be that a QB in the 2nd round is the highest value player available at that pick. But if there is a a QB available in the 4th who provides nearly as good of a value, then your team might score more overall if you used that 2nd round pick on another position that would drop off more by the 4th round. (At which point we're very much doing a dynamic VBD sort of thing, only now we're doing it in advance of the draft rather than during it, as we try to predict in advance what decisions we'll face and have time to study the implications of our decisions.)

There is no way to build into a baseline all of the information you need. And note the above, still only really applies to the starters. Or at least to the segment of players where your projection is what you truly expect them to score, and that represents the bulk of their fantasy worth. Often though, the value of a backup is not based on him having the 50 carries for 215 yards that we have him projected at, but is based on what he would do if he became the starter. For VBD to work with such players, we would need a projection which is not what we truly expect he'll put up, but which instead includes more strongly what his value might become.

I can tell you that trying to generate auction prices is a very good test of one's use of VBD or any other method of draft prep. You'll quickly find there is that moment where someone's projected stats compared to other backups around him don't necessarily reflect his value at all. And you'll also find you have to deal better with that dividing line between starter quality players and the rest where a single baseline doesn't do a good job in helping you calculate prices.

 

[Dinsy Ejotuz]

This seems like as good a thread as any, I guess.<br /><br />There were 72 RBs that started at least one game in a 14-team league of mine last year. That number really surprised me and I did some fiddling around with what actually happened. I wanted to figure out how to value RBs in light of that.<br /><br />Cumulatively, each team's most-started RB accounted for 39% of all RB starts (11.0 games/team)<br />The 2nd most-started RB had 29% of all starts (8.3 games/team)<br />3rd = 17% (4.8 games per team)<br />4th+ = 17% (~5 games per team, but only 12 teams started more than 3 RBs in the season)<br /><br />The average PPG were as follows:<br /><br />Most started = 15.6<br />2nd most = 11.6<br />3rd = 9.4<br />4th+ = 7.4<br /><br />So my question is...<br /><br />Since you know that, on average, you're likely to need both a RB3 and a RB4 in about five games why wouldn't you set the VBD baseline for RBs at 7.4 ppg?<br /><br />Then, if you've got a RB3 that's projected to score 10.4 ppg, you'd subtract the baseline (7.4ppg) and multiply by five (the number of games you expect to need him, on average) -- so that the player had a baseline VBD of 15 points.<br /><br />And a top RB, projected to score 17.4 ppg would have a VBD baseline of 110, or (17.4-7.4) * 11.<br /><br />Is that right or am I missing something? If you did that for each position would you end up with a much stronger VBD baseline?

 

[GregR]

wdcrob, check out MT's article on his method for determining auction values. That's the direction he headed, of figuring out how many games you could expect players to start for you based on where they were in your rankings, and then deriving values from there. That system is built into draft dominator as one of the options for baselines.

 

[Dinsy Ejotuz]

Will do Greg, thanks.

No idea what happened to my formatting there. I think I'm just going to leave it from now on when the software eats my homework.

 

[Lord Fantasy]

A huge thanks to cdubz and Greg Russell for helping me understand this.

You're also making me wonder if many of our message board arguments stem from people having very different scoring systems and league psychologies.

Saying "there's no way Player X is worth a third-round pick" seems a little silly given how much everyone's mileage may vary.

 

[freshly_shorn]

Hello all- OP here..

I was surprised to see this thread resurrected a bit last month! Just wanted you to know that a new version of my Draft Planalyzer spreadsheet is coming out today, the first one built in Libre Office. I will have an Excel version out soon. Note this one has a Baseline generator in it, new for this year.

Note this spreadsheet is for use in planning your draft in an established redraft league- not actually managing one, like Draft Dominator (that's what I use- I just built this spreadsheet to help me plan for my draft).

Check it out, let me know what you think. IT IS FREE and works very well as a complementary tool for other things going on here at FBG.

http://www.draftologist.com

@draftologist on Twitter