A draft strategy against FBG common sense? (1 Viewer)

[bigz99]

So, if I could sum up FBG and VBD, I would say "stock pile RBs, WRs are a dime-a-dozen, there are always surprise DEF, don't pick one too early, do not take the top QBs, plenty available later, etc".

Well, I was putting in my point system to the VBD and it surprisingly had almost as many WRs as RBs going in round 1. I just couldn't believe my eyes. It is PPR, but come on! Anyway, I look at the estimated FPTS, and I notice very little separation of these WRs, with more separation with the RBs, so why would I take one of those WRs?

This got me thinking...I compiled all stats of the top QB, WR, RB, TE, DEF over the last 3 years and the projected FBG 2007. I looked at standard deviation between the top player at his position and the 10th best player. I was trying to get an idea of the distribution and which positions were closely distributed and which had large spreads. Here is my data: (note the FBG 2007 projected FPTS is not in line with my other data, so AVG* represents the 3-year average stand deviation of years 2004-2006 without 2007.)

        QB               RB              WR             TE             DEF
2007	27.07241318	35.94101495	10.15718684	20.32562695	22.38914419
2006	42.43824402	72.97582629	12.90056416	19.08578296	40.95593838
2005	27.25937352	58.63654625	20.77488869	28.65816928	18.33151506
2004	64.93092141	37.04167173	16.644305	38.83594555	30.02609976
AVG	40.42523803	51.1487648	15.11923617	26.72638119	27.92567435
AVG*	44.87617965	56.21801475	16.77325261	28.85996593	29.7711844

This tells me that there is more separation (value?) between RBs #1, QBs #2, DEF #3, TE #4, and WR#5

So...

Why don't we all draft for the best QB and DEF higher? Arguably that would construct a better team, but this theory goes against what FBG often preaches.

 

[stevegamer]

Predicatability and positional scarcity.

There are enough defenses to go around in every league, saem with QB's. You need multiple RB & WR.

Also predicting top D's can be somewhat of a crapshoot.

 

[bigz99]

I agree with the DEF predictability theory. But the fact that WRs all finish very close to each other AND are unpredictable would lead me to try out a top QB instead, don't you think? I have largely never gone for the elite QB, but now I am wavering. The beauty of never having to worry and just plugging in a Manning seems nice.

 

[amphibianbri]

I've wondered the same thing at times. Once I read your info above, I thought about the two championship teams I've had the last 3 years. Manning QB once (even though that was in a midseason trade), and Palmer last year. Other years I get into the playoffs then lose mainly to teams with a better QB. With going to an auction this year, maybe going after a top 10 rb, top 5 qb, a couple rb's in 11-20 and bargain basement at WR is the way to go. I'm usually scared not having a top 10 wr starting the season, but if you have several wr in the 15-25 range, you would suppose at least one would end up the year near the top 10....hmm.....

 

[u999spf]

Sounds like a plan. I think grabbing a top 3 QB is a key to having a championship team. Provided he stays healthy. Can you win without a top QB, yes. But you need to be very strong everywhere else to do so.

 

[inca911]

VBD deals with the math behind "value". I disagree with your summary of what VBD is saying, as VBD doesn't "preach" any specific strategy and just deals with numbers. This year, I have Manning at #5 in terms of value, Gates at #16, CHI DST at #42. Getting a top QB or DST can definitely be a winning strategy and is 100% supported by VBD.

Note that I'm talking about true VBD calculations, not the math performed by the current VBD application (which doesn't work quite right).

 

[bigz99]

what is true VBD then? How does the VBD excel sheet or DD not perform correctly?

 

[GregR_2]

what is true VBD then? How does the VBD excel sheet or DD not perform correctly?

VBD is a tool that helps you by expressing where value is in the SUPPLY of players. It is not a tool that tells you who to draft. Anyone who uses it that way does not understand what it is doing.Any individual players value to your team is a combination of many things. It includes:1) The supply of players.2) The demand for players in the form of the way your league depletes the positions in the draft.3) The risk associated with the player, which is more in the form of loss of opportunity such as RBBC or that he just won't perform where you think, than it is injury.4) The interaction between #1 and #2. Value is relative. For instance, if your 4th best WR is going in the 10th round, then there is less value in taking your 3rd best WR in the 3rd round where you believe he should probably go.5) Other factors like the existance of a handcuff for your RB that allow you to mitigate risk of injury with a later pick than you would otherwise have had to going after a RBBC backup.6) Whether taking the player now will leave you in a position to be able to poach players who slide and represent good value, vs lock you into having to pick a certain way to fill a need.VBD only shows you anything to do with #1.VBD is not a draft strategy. It's a tool. It helps you determine how valuable Peyton Manning is compared to Steve Smith. It doesn't tell you if your draft that will unfold if you take Steve Smith is better than the draft that will unfold if you take Peyton Manning. It doesn't know or try to give you any of the components of value from 2-6, or any other components of value I didn't mention.To get your optimal team based on the picks you hold, you need to understand the individual value of players, which comes from VBD. You also need to understand what the impact is in later rounds of choices you make now. If you pass on a RB and they go faster than you thought, does it mean you have to spend an extra mid-round pick on a backup RB? And if so, how much does that hurt WR which gets delayed a round by the change? Does taking that extra RB remove your flexibility to pick up a value player who slid unexpectedly, because the RBs will run out if you don't take him? Those are elements of value that no system, no software that exists today, is going to be able to tell you. VBD, and Draft Dominator, are great tools to use to try to figure out the answer to those questions because they do help you compare the value of players for your team. But no coded system I've ever seen understands all the of the components of value you want to take into account. The best draft strategy article I've seen yet on Footballguys is the annual Perfect Draft article by Dodds. He does exactly what I'm talking about.Identify areas that value exists in the draft. Try to create a strategy that navigates through those areas. Look at the team it gives you, and then try something else and see if it turns out better. Put yourself to the task of making a decision in round 3 and then seeing what impact it had on your decision in round 6, or if you find yourself wishing you'd gone another way.

 

[bigz99]

Maybe I am not articulating my point correctly, let me try another angle...

If you miss out on the top RBs (1-3), maybe going QB1-3 is a better play then WR. I think that means I am agreeing with you Greg, because DD or VBD would say to not take the QB (outside of manning).

I am questioning value as represented in those tools. e.g. +/- 26 points gives you 10 WR, 2 RB, 1QB, 1 TE, 2 DEFs based on FBG projects from the #1 spot and my pt system. So why would DD tell me there are 4 WRs in the first 10 most "valuable" players? If 10 are within 26 total points of each other, there should not be 4 in the top 10 overall, right?

 

[Dirty Weasel]

what is true VBD then? How does the VBD excel sheet or DD not perform correctly?

Those are elements of value that no system, no software that exists today, is going to be able to tell you. VBD, and Draft Dominator, are great tools to use to try to figure out the answer to those questions because they do help you compare the value of players for your team. But no coded system I've ever seen understands all the of the components of value you want to take into account. The best draft strategy article I've seen yet on Footballguys is the annual Perfect Draft article by Dodds. He does exactly what I'm talking about.

Doesn't DVBD try to do this? It's VBD with a twist. DVBD tries to recalculate VDB after each pick, taking into position position scarcity.

 

[Loose Moose]

It's also sometimes more difficult to discern who the top WRs will be beyond a precious few studs. That adds some value to those studs even if they are close together.

Also, your analogy would work perfectly in a league with starting requirements of 1 QB, 1 RB, 1 WR, 1 TE, 1 DST, and 1 PK. But when you have to start 2 or even 3 WR, suddenly you're not comparing to the 10th WR, but the 20th or 30th. Then you'll see a drop.

 

[bigz99]

It's also sometimes more difficult to discern who the top WRs will be beyond a precious few studs. That adds some value to those studs even if they are close together.Also, your analogy would work perfectly in a league with starting requirements of 1 QB, 1 RB, 1 WR, 1 TE, 1 DST, and 1 PK. But when you have to start 2 or even 3 WR, suddenly you're not comparing to the 10th WR, but the 20th or 30th. Then you'll see a drop.

Ok, that is a valid point. Here are the numbers based on best starter to worst starter 1qb, 2rb, 3wr, 1te, 1dst

Starters QB	          RB	          WR	          TE	          DEF2007	27.07241318	33.4801135	24.18116394	20.32562695	22.389144192006	42.43824402	71.45293335	26.79651719	19.08578296	40.955938382005	27.25937352	65.0440895	35.13777219	28.65816928	18.331515062004	64.93092141	44.94597342	30.85779098	38.83594555	30.02609976AVG	40.42523803	53.73077744	29.24331108	26.72638119	27.92567435AVG*	44.87617965	60.48099876	30.93069345	28.85996593	29.7711844

To me this still says, get a QB in the top of the range, there is more separation than going for a top WR.

 

[The Jerk]

It's also sometimes more difficult to discern who the top WRs will be beyond a precious few studs. That adds some value to those studs even if they are close together.Also, your analogy would work perfectly in a league with starting requirements of 1 QB, 1 RB, 1 WR, 1 TE, 1 DST, and 1 PK. But when you have to start 2 or even 3 WR, suddenly you're not comparing to the 10th WR, but the 20th or 30th. Then you'll see a drop.

Ok, that is a valid point. Here are the numbers based on best starter to worst starter 1qb, 2rb, 3wr, 1te, 1dst

Starters QB	          RB	          WR	          TE	          DEF2007	27.07241318	33.4801135	24.18116394	20.32562695	22.389144192006	42.43824402	71.45293335	26.79651719	19.08578296	40.955938382005	27.25937352	65.0440895	35.13777219	28.65816928	18.331515062004	64.93092141	44.94597342	30.85779098	38.83594555	30.02609976AVG	40.42523803	53.73077744	29.24331108	26.72638119	27.92567435AVG*	44.87617965	60.48099876	30.93069345	28.85996593	29.7711844

To me this still says, get a QB in the top of the range, there is more separation than going for a top WR.

We're getting somewhere with these revised numbers. Is it possible to break down the WR numbers individually? Say, group the top 10-12 WRs as column 1, the next group of 10-12 WR as column 2, and the third group of 10-12 WRs as column 3. That will give the best comparison to QBs, TEs and DEFs as each of these positions typically is a single-player position. You might find that there is greater variation in WR1 than in WR3...

 

[bigz99]

Here are the numbers:

        QB	        RB1	         RB2	       WR1	      WR2	         WR3	              TE	           DEF
2007	27.07241318	35.94101495	9.917442547	10.15718684	10.60576363	7.080332698	20.32562695	22.38914419
2006	42.43824402	72.97582629	7.919182057	12.90056416	6.715190078	8.771677909	19.08578296	40.95593838
2005	27.25937352	58.63654625	12.20794368	20.77488869	8.896216424	7.111149305	28.65816928	18.33151506
2004	64.93092141	37.04167173	8.364667756	16.644305	6.568409244	8.287366557	38.83594555	30.02609976
AVG	40.42523803	51.1487648	9.60230901	15.11923617	8.196394845	7.812631617	26.72638119	27.92567435
AVG*	44.87617965	56.21801475	9.497264497	16.77325261	7.393271915	8.056731257	28.85996593	29.7711844

Thoughts?

 

[The Jerk]

Thanks for doing the analysis. These are my favorite type of posts, especially a month before the season begins.

It appears that the numbers support selecting a top QB over a top WR if you base your decision purely on the separation factor. However, there are still contributing factors that could make this data less meaningful than it might appear on the surface.

1. These statistics compare the season-ending top performances. They don't reference the preseason rankings or EXPECTED top performers. To me, the utility of this analysis is linked strongly to whether or not we can rely on the expected top performers to be the actual top performers. The obvious anomaly last season was Drew Brees as a Top 3 QB. Yet he was far from the only shocking development at the QB position last season.

Case Study: I started a poll last year on Matt Hasselbeck's 2006 expectations (view link here) and out of 132 respondents, only two predicted that he would finish below QB 12 in 2006. According to FBG stats, Hasselbeck finished 15th in ppg and 22nd in overall QB points.

2. Using standard deviation is always dicey when evaluating any consistency-based statistics. The lower the overall production of a given player, the more consistent he tends to appear to be. (The ultimate consistent player may be someone who scores 0 points every game.) When you turn that around, you'll see that players at higher scoring positions tend to have higher standard deviations just because they are more volatile. QBs certainly fit that description, and depending on scoring criteria, so might defenses. Tight ends are most resistent to this skewing effect.

Last season, I toyed with what I called a normalized standard deviation (apologies to true statisticians here if this term already means something else), in which I divided a player's standard deviation by their season total points scored. (Link to 2006 thread on Rudi Johnson) I feel this does a better job of capturing consistency. If you don't mind doing one more stat study, and if it's not too hard to do, you might try this out and see if it pushes the trends one way or the other.

3. Record-setting seasons (see LT 2006, SA 2005, PM and DC in 2004) will spike the separation values that you are seeing here. We haven't seen strong WR performances in the past several years. I'd love to see the 2003 WR numbers when Randy Moss had 17 TDs (4th highest total in NFL history) and Torry Holt was second with 12. I suspect the numbers would show greater separation for that season. Which leads to the question, are we seeing an aberration at the WR position, or is this lack of high TD WRs a trend?

Although I have no way of actually knowing, I believe I am more open to selecting a QB early in a draft than most FBGs. However, outside of the numbers, there are certain intangibles that usually work against me actually acting on this belief. Basically, GregR did a great job of outlining many of these. I'll phrase it a little differently. There can be a huge opportunity cost of choosing a QB early. Greg mentioned it in terms of missing out on a RB early, which can multiply into forcing you to miss WRs later. While this is absolutely true, I tend to focus on the bargain hunter side of the evaluation. Even if I could know that a given QB drafted in the top 3 rounds would perform as expected, by selecting the QB, I rule out any possibility of benefiting when a solid QB slides down the board in the 3rd-5th rounds. I've pretty much ruled out selecting another QB until round 8 or so, and what makes it worse is that one of my competitors is going to get this QB at a bargain rate instead of me. Drafting 2 WRs early does not preclude grabbing up a 5th round value that slides to me. Even a 3rd RB in round 5 is great value (and I'm not at all a stud RB proponent).

Add this opportunity cost to the fact that owners in my leagues tend to be more QB-friendly than the norm, and I almost always end up drafting a QB after at least 6 have been taken.

Great discussion starter! Thanks.

 

[GregR_2]

what is true VBD then? How does the VBD excel sheet or DD not perform correctly?

Those are elements of value that no system, no software that exists today, is going to be able to tell you. VBD, and Draft Dominator, are great tools to use to try to figure out the answer to those questions because they do help you compare the value of players for your team. But no coded system I've ever seen understands all the of the components of value you want to take into account. The best draft strategy article I've seen yet on Footballguys is the annual Perfect Draft article by Dodds. He does exactly what I'm talking about.

Doesn't DVBD try to do this? It's VBD with a twist. DVBD tries to recalculate VDB after each pick, taking into position position scarcity.

DVBD focuses heavily on #2, the value created by the depletion of positions. The simplest example is, don't take a QB this round if the same exact guy is available next round. Look at how much a position drops between now and your next pick, and take the position that drops the most. But the view of value it gives shows you less about the entire player pool than does VBD. VBD tries to normalize all players at a position. DVBD looks at one snapshot of the position, between your pick now, and your pick next round or 2 rounds out. It only shows you player value of the players available now vs those available at whatever round you choose to look at.But there are obviously shortcomings from only including that aspect. Let's say it's round 2 in some PPR league that favors TE receptions. You can take Gates now. Or in round 3 you think Heap is your best pick, and he drops off 40 points. Or in round 4 you can take Shockey who drops off 50 points from Gates. That's great for deciding between those players, but what if you have some other TE who you think is only 60 points off Gates and can be had in the 10th round? If you're only looking at rounds 3 and 4 drop offs with DVBD, you are going to miss that the best option for your team might be to wait and take the guy in the 10th as you can gain more than the 60 points at some other position that you lose by not picking Gates.

That's why I'm saying you should learn the consequences of picking a given position by doing mocks to test and see what happens to your decisions later in the draft. You're trying to get at what would be a DVBD that simultaneously accounts for every single pick you have in your draft. But since that is a very difficult problem to solve mathematically, you do it by understanding how the pick will affect the rest of your draft to allow you to make a good decision, instead of relying on a system to give you the final answer. You can still do it quantitatively as well as quantitatively, comparing your final rosters you ended up picking, figuring starter points and also deciding how you like your risk and bye weeks and other things that affect a team's success.

 

[bigz99]

Jerk -

Good post. Its good to hash this out.

As far as STDEV and consistency goes, that wasn't exactly where I was going. What I was trying to do is understand drop off from #1 starter to #10/20/30 starter. My initial hypothesis was that the greater the drop off from 1 to 10, the more urgency to be in the upper portion of that range. I was really only considering a starting lineup and not considering risk/injury/backups/etc. In the end I wanted to answer the question, "Do I need to have a top 3-5 QB in order to win?".

Off to work now, but let me look at your request and try to dice up the numbers some more.

Also, I would love to have the analysis of actual vs. expected, but I fear that may take more time than I have! I agree there is a hindsight element to this analysis, but I figured this wasn't about specific players or surprises rather about ranges of 1-10 and if you were better trying to get into the top half or not. Now, that is not to say you can't get lucky, get a sleeper and still end up there, but it is with greater risk than grabbing Manning et al.

 

[bigz99]

Let me also add...

I usually never take a QB early, opting for the favre or vick pick late and assuming they would be ok. I even grab waivers like grossman and think I am ahead of the game. Inevitably what happens is I either a) try to play matchups with a platoon and get burned and lose a bench spot or b) those sleeper picks drop off (see grossman) or c) the sleepers never come through. I know Brees is out there, but I never get him and how often is there really a Brees vs. how often is their a Colston?

Now, who did I get by not spending a 3/4/5 rounder on a QB? I pick up some wildly inconsistent WR who usually isn't winning me any games. Alternatively I could get a backup RB that I never play or never pans out. The QB pays dividends every game of the season (assuming no injury)

The theory in waiting on a QB only pays when the pick you would of used pans out. I may argue the chance to get a good value non-qb pick is higher than to get a value QB. There is something to be said of getting a good qb, plugging him in, forgetting about it, and keeping the extra bench spot for other sleepers like RBs/WRs.

 

[raven1911]

what is true VBD then? How does the VBD excel sheet or DD not perform correctly?

Those are elements of value that no system, no software that exists today, is going to be able to tell you. VBD, and Draft Dominator, are great tools to use to try to figure out the answer to those questions because they do help you compare the value of players for your team. But no coded system I've ever seen understands all the of the components of value you want to take into account. The best draft strategy article I've seen yet on Footballguys is the annual Perfect Draft article by Dodds. He does exactly what I'm talking about.

Doesn't DVBD try to do this? It's VBD with a twist. DVBD tries to recalculate VDB after each pick, taking into position position scarcity.

DVBD focuses heavily on #2, the value created by the depletion of positions. The simplest example is, don't take a QB this round if the same exact guy is available next round. Look at how much a position drops between now and your next pick, and take the position that drops the most. But the view of value it gives shows you less about the entire player pool than does VBD. VBD tries to normalize all players at a position. DVBD looks at one snapshot of the position, between your pick now, and your pick next round or 2 rounds out. It only shows you player value of the players available now vs those available at whatever round you choose to look at.But there are obviously shortcomings from only including that aspect. Let's say it's round 2 in some PPR league that favors TE receptions. You can take Gates now. Or in round 3 you think Heap is your best pick, and he drops off 40 points. Or in round 4 you can take Shockey who drops off 50 points from Gates. That's great for deciding between those players, but what if you have some other TE who you think is only 60 points off Gates and can be had in the 10th round? If you're only looking at rounds 3 and 4 drop offs with DVBD, you are going to miss that the best option for your team might be to wait and take the guy in the 10th as you can gain more than the 60 points at some other position that you lose by not picking Gates.

That's why I'm saying you should learn the consequences of picking a given position by doing mocks to test and see what happens to your decisions later in the draft. You're trying to get at what would be a DVBD that simultaneously accounts for every single pick you have in your draft. But since that is a very difficult problem to solve mathematically, you do it by understanding how the pick will affect the rest of your draft to allow you to make a good decision, instead of relying on a system to give you the final answer. You can still do it quantitatively as well as quantitatively, comparing your final rosters you ended up picking, figuring starter points and also deciding how you like your risk and bye weeks and other things that affect a team's success.

Understanding DVBD is difficult for the beginning DD user. So if I am correct you see one DVBD number per position/player in the Best Value Pick Box and you can base who to take off that number? So if the DVBD number is very large then that means taking that player for better value, right? Is there a way of knowing DVBD for other players not listed in the Best Value Pick Box besides the player that DD suggests presents the best value? I am confused when you say "take the position that drops the most", do you mean the drop value number??