Value-Based Drafting and the Value of D/ST (1 Viewer)

[TheMathNinja]

I haven't yet found an article on this, but I'm wondering why D/ST are so undervalued in drafts compared to what their Value-Based Drafting numbers would indicate. Since most leagues start a single D/ST, as well as a single TE and QB, I'll use TE and QB as baselines for comparison.

For a 12-team league, using the 12th starter as the baseline and standard fantasy scoring (1 pt per 25 yds, 4 pts per TD), the value added for QB's and TE's looked like this:

Drew Brees: 337..........98 pts added // Jimmy Graham: 144.....55 pts added

Tom Brady: 329...........90 pts added // Rob Gronkowski: 139....50 pts added

Aaron Rodgers: 329....90 pts added // Tony Gonzalez: 135......46 pts added

Cam Newton: 309........70 pts added // Heath Miller: 125..........36 pts added

Peyton Manning: 304...65 pts added // Jason Witten: 115.........26 pts added

Robert Griffin III: 304....65 pts added // Greg Olsen: 107..........18 pts added

Matt Ryan: 291............52 pts added // Dennis Pitta: 103.........14 pts added

Tony Romo: 271.........32 pts added // Owen Daniels: 100......11 pts added

Andrew Luck: 264........25 pts added // Kyle Rudolph: 99.........10 pts added

Matt Stafford: 263........24 pts added // Brandon Myers: 97........8 pts added

Russell Wilson: 259.....20 pts added // Jermaine Gresham: 94..5 pts added

Andy Dalton: 239........0 pts added // Antonio Gates: 89.........0 pts added

Common drafting practices roughly follow these VBD results, i.e. a QB or TE who is expected to add about 50 points is taken in the 5th or 6th round in a 12-team snake, or valued around $25 in an auction. What still confuses me is how this pattern doesn't seem to come close to holding true for the D/ST position. Here are last year's results:

Bears D/ST: 212...........90 pts added

Broncos D/ST: 186.......64 pts added

Seahawks D/ST: 172....50 pts added

Bengals D/ST: 160.......38 pts added

Chargers D/ST: 159.....37 pts added

Patriots D/ST: 149........27 pts added

Texans D/ST: 147.........25 pts added

49ers D/ST: 146...........24 pts added

Rams D/ST: 134...........12 pts added

Cardinals D/ST: 131......9 pts added

Vikings D/ST: 123..........1 pt added

Steelers D/ST: 122........0 pts added

Yes, in last year's draft, you would have been wise to take the Bears D/ST right between Tom Brady and Aaron Rodgers. This year, I see most people projecting the Seahawks D/ST to be the best; ESPN is projecting 171 points, which would be 53 pts added over their prediction of 118 for the 12th best team (San Diego). My question is this: why, then, is Seattle currently valued at $5 in ESPN auctions and not at $25, where we would expect them to be for such value? Why are they being selected on average in the 9th round in preseason fantasy drafts so far?

If Defenses didn't repeat production with regularity year-over-year (much like kickers...though I think there's value to be had there as well), I could maybe see waiting on them. But though I haven't crunched the hard data for it, my impression is that Defenses are only slightly more volatile in production than QB's and TE's, if at all. And yet, I don't see people gobbling up the Seattle D at the same time as Matt Ryan.

A move like taking the top defense in the 5th round or spending $25 on them would get you some pretty funny looks in most leagues, but this is exactly what VBD would have you do. Why hasn't this caught on yet? D/ST remains the position to get the most value in every draft I've ever participated in; I often take my 2nd defense before most players take their first.

I'd love your insights on this one, sharks. Because I've been feeling a bit baffled on this one for some time.

 

[cdubz]

If you could guarantee that the projections would come true then they would be worth more, but fantasy defense performance is not exactly predictable.

This should give a good summary:

http://www.pro-football-reference.com/articles/defcc.htm

 

[Nucker101]

Great post. IMO, it's always simply based on the thought and strategy of being able to stream D/ST's based on matchups since D/ST's(as long as they're somewhat decent) seem to rely on matchups much more than actual offensive players.

You'd also have to go back and take a look at the actual value that the top ranked pre-season D/ST's provided since usually there's always so much turnover year to year in the top 10. That being said, I agree with you that D/ST's are generally undervalued in fantasy football, and I always like to ask for a D/ST upgrade included in a deal if my trading partner has a better one than me. I also tried pairing up the SEA and NE D/ST's last year since I thought that both could be top 5 D/ST's and weren't going too high. Obviously SEA ended up being gold, but NE was up and down.

And we all know that the top QB/RB/WR/TE's are generally more easy to predict than D/ST's and Kickers. Not to mention that good ones do get dropped in 12 team leagues during bye-weeks.

 

[Slapdash]

I almost always try to get one of the top defenses, even if a little early.

 

[TheMathNinja]

If you could guarantee that the projections would come true then they would be worth more, but fantasy defense performance is not exactly predictable.

This should give a good summary:

http://www.pro-football-reference.com/articles/defcc.htm

These numbers certainly hold true if your only D/ST fantasy stats are turnovers and TD's. But now that most systems rely heavily on points and yards allowed, the correlation for fantasy points year-to-year increases dramatically, as these are considerably more predictable traits for team defenses.

 

[cdubz]

If you could guarantee that the projections would come true then they would be worth more, but fantasy defense performance is not exactly predictable.

This should give a good summary:

http://www.pro-football-reference.com/articles/defcc.htm

These numbers certainly hold true if your only D/ST fantasy stats are turnovers and TD's. But now that most systems rely heavily on points and yards allowed, the correlation for fantasy points year-to-year increases dramatically, as these are considerably more predictable traits for team defenses.

Do you have any numbers on this or just a hunch?

Let's look at ESPN numbers and how well 2011 yards allowed predicted 2012 fantasy points.

Team 2011 Yards Allowed Rank 2012 Fantasy PointsPittsburgh 1 12Houston 2 7Baltimore 3 15San Francisco 4 8NY Jets 5 13Jacksonville 6 30Cincinnati 7 4Philadelphia 8 29Seattle 9 3Cleveland 10 21Kansas City 11 31Atlanta 12 19Washington 13 22Dallas 14 26Miami 15 18San Diego 16 5Chicago 17 1Tennessee 18 16Arizona 19 10Denver 20 2Minnesota 21 11St. Louis 22 9Detroit 23 27New Orleans 24 28Indianapolis 25 25Buffalo 26 24NY Giants 27 17Carolina 28 20Oakland 29 32Tampa Bay 30 23New England 31 6Green Bay 32 14Top 2 in FP were bottom half in yards allowed and only 4 of top 10 in yards allowed were able to finish top 10 in FP. But if you want to bid up defenses because you think they are predictable, be my guest.

ETA: The other part of the argument driving down fantasy defense value is WW replaceability - you can generally find serviceable options on the WW unlike RB for example.

 

[bucsbaby]

Not to mention that injuries to offensive players make having quality depth important, whereas if you have an injury on your DST it doesn't usually make as much of a difference.

What I think would make for a more interesting analysis is not so much VBD of top DSTs relative to baseline of 12th best, but rather to a baseline of streaming ww DSTs, which don't cost you draft picks or auction money (though it may cost you ww bid money, depending on your league rules). Anecdotally, I've found that I can usually come close enough to a top DST that cost like an 8th round draft pick with streaming ww DSTs that I can't see myself spending that pick on a DST.

 

[Anarchy99]

Beacuse it is extremely hard to predict how well D/ST will perform. Projections for them are usually way off. And teams that were great DEF/ST are rarely dominant 2 years in a row. So while there will be a couple of trult valuable DEF/STs at year's end, picking them before the season starts is a total roll of the dice.

IMO, waiting on defenses is the best way to go. Snag one of the leftovers that you might have a hunch on . . . on pick one up off the waiver wire. No point in investing an early pick on a defense IMO.

 

[Drop]

It's a very interesting question. First I would look at some high stakes leagues to see if that shines some light on anything, perhaps there are people who have been exploiting this. Follow the money.

There are some things that are sort of unique to the D/ST position as a fantasy starter.

-D/ST can't get injured like a single player can.

-There are matchups on a schedule that are wise for even stud defenses to sit against, unlike a stud player. IE - D/ST vs elite QB at home.

-Favorable matchups can be gotten off the waiver wire regularly. For instance a solid team that has had 3 tough games then they play OAK or some bad team, perhaps with a backup QB getting the start. This means for 3 weeks that D/ST would be almost unstartable, so are they worth taking up a roster spot when said roster spot could be used on a speculative RB/WR? Or perhaps you have 2 D/ST and 1 of them is clearly the better start for 5 weeks in a row, it sort of hurts the value of the 2nd D/ST for you. I don't know how common these situations are, but they seem to be just thinking off the top. Whereas rostering a positional player for 5 weeks without starting him can be much more reasonable. IE - Ben Tate, Shane Vereen, Peyton Hillis. So you can see how and why positional players are hoarded and because of the injury factor guys who only start half the season or for 1 month stretches can significantly outscore their positional peers and it will not always be represented properly in the end of year VBD ratings unless you account for the startability of a player.

I don't know if the turn over at other positions drives the D/ST value down. So if a D/ST is a 5th round value in terms of VBD is the reason that D/ST goes in the 9th round because the positional players selected 5th-9th are regularly littered with players who finish as 2nd/3rd round type value? If CHI D/ST finished on par with Rodgers in terms of VBD this season in hindsight it would seem mathematically logical to see them equally, but how often does a D/ST finish like that? Is it more common to see even the best D/ST finish as a 4th round value, meaning the upside is historically limited.

The depth of your RB/WR positions is addressed in the middle and later rounds and especially with the RB picks a lot of guys that are selected are real life RBBC players who are a sprain, tweak, or concussion away from being feature backs, even a few handcuff backs have a lot of potential value. I'm not sure how to weight the value of RB/WR depth. There seems to be a lot more answers on the waiver wire if your D/ST needs to be filled in for than if your RB2 does. Maybe you do want to pickup the JAC D/ST and start them against ARI at home instead of having to pickup and spot start Shaun Droughn and pray he gets some touches. The opportunity of waiver wire D/ST seems to be much greater than the opportunity of waiver wire RB/WR/QB/TE fill ins.

This is all anecdotal speculation on my part, that's why I really would first look at people who have the money to spend the time and effort on evaluating these notions mathematically.

 

[Anarchy99]

Here are the DEF/ST scoring totals for each team in each year from the past 10 years to better illustrate what I had mentioned a few posts ago . . .

				2012	2011	2010	2009	2008	2007	2006	2005	2004	200349ers, San Francisco SFO Def	162	186	131	182	111	98	119	130	97	154Bears, Chicago CHI Def		227	180	158	127	158	197	235	203	160	118Bengals, Cincinnati CIN Def	164	128	107	146	105	150	133	145	164	95Bills, Buffalo BUF Def		129	160	108	139	116	146	147	131	212	144Broncos, Denver DEN Def		184	124	77	155	69	137	149	164	110	127Browns, Cleveland CLE Def	133	94	124	133	129	127	122	123	125	119Buccaneers, Tampa Bay TBB Def	123	101	127	134	157	159	106	153	161	185Cardinals, Arizona ARI Def	156	130	164	141	142	139	144	129	154	88Chargers, San Diego SDC Def	167	108	144	139	122	227	157	109	128	93Chiefs, Kansas City KCC Def	72	122	129	129	88	107	132	140	116	182Colts, Indianapolis IND Def	108	87	110	126	142	164	98	179	168	129Cowboys, Dallas DAL Def		95	124	150	147	145	172	151	123	104	163Dolphins, Miami MIA Def		125	114	111	118	151	110	144	143	127	187Eagles, Philadelphia PHI Def	69	158	138	181	201	108	155	116	152	141Falcons, Atlanta ATL Def	128	129	149	139	116	101	135	141	174	137Giants, New York NYG Def	145	131	162	126	160	168	126	184	138	132Jaguars, Jacksonville JAC Def	78	132	77	69	90	145	149	152	143	102Jets, New York NYJ Def		129	163	168	183	161	121	133	127	155	112Lions, Detroit DET Def		84	177	152	106	88	169	113	143	154	143Packers, Green Bay GBP Def	133	164	181	189	149	150	163	111	114	153Panthers, Carolina CAR Def	122	99	107	153	134	112	132	191	148	161Patriots, New England NEP Def	184	142	199	147	115	193	170	107	198	226Raiders, Oakland OAK Def	83	119	164	86	135	107	116	96	85	118Rams, St. Louis STL Def		149	104	129	83	91	114	126	136	106	198Ravens, Baltimore BAL Def	133	176	131	166	202	118	237	148	194	213Redskins, Washington WAS Def	128	101	128	93	88	133	74	146	149	104Saints, New Orleans NOS Def	128	105	131	179	109	128	115	83	141	144Seahawks, Seattle SEA Def	180	166	147	115	112	184	145	159	146	152Steelers, Pittsburgh PIT Def	110	132	186	139	178	151	161	173	181	133Texans, Houston HOU Def		175	152	79	131	103	140	115	96	145	89Titans, Tennessee TEN Def	163	122	138	123	170	151	143	133	124	166Vikings, Minnesota MIN Def	140	117	123	147	163	179	163	148	114	162

 

[Drop]

Beacuse it is extremely hard to predict how well D/ST will perform. Projections for them are usually way off. And teams that were great DEF/ST are rarely dominant 2 years in a row. So while there will be a couple of trult valuable DEF/STs at year's end, picking them before the season starts is a total roll of the dice.

IMO, waiting on defenses is the best way to go. Snag one of the leftovers that you might have a hunch on . . . on pick one up off the waiver wire. No point in investing an early pick on a defense IMO.

I think the point being made is that if you can predict D/ST well there is a huge advantage. Just because it's hard doesn't mean it's impossible and your opinion of the best way to go is irrelevant unless you can show your work. I don't mean to be confrontational, I'm just saying there is math behind this and the prediction code is not unbreakable. Given the size of the edge you can afford to be wrong some % of the time and still be profitable. The numbers are saying conventional wisdom is wrong in the VBD context and just because prediction is difficult doesn't mean proper risk/reward evaluation should be ignored.

 

[cdubz]

Here are the DEF/ST scoring totals for each team in each year from the past 10 years to better illustrate what I had mentioned a few posts ago . . .

2012 2011 2010 2009 2008 2007 2006 2005 2004 200349ers, San Francisco SFO Def 162 186 131 182 111 98 119 130 97 154Bears, Chicago CHI Def 227 180 158 127 158 197 235 203 160 118Bengals, Cincinnati CIN Def 164 128 107 146 105 150 133 145 164 95Bills, Buffalo BUF Def 129 160 108 139 116 146 147 131 212 144Broncos, Denver DEN Def 184 124 77 155 69 137 149 164 110 127Browns, Cleveland CLE Def 133 94 124 133 129 127 122 123 125 119Buccaneers, Tampa Bay TBB Def 123 101 127 134 157 159 106 153 161 185Cardinals, Arizona ARI Def 156 130 164 141 142 139 144 129 154 88Chargers, San Diego SDC Def 167 108 144 139 122 227 157 109 128 93Chiefs, Kansas City KCC Def 72 122 129 129 88 107 132 140 116 182Colts, Indianapolis IND Def 108 87 110 126 142 164 98 179 168 129Cowboys, Dallas DAL Def 95 124 150 147 145 172 151 123 104 163Dolphins, Miami MIA Def 125 114 111 118 151 110 144 143 127 187Eagles, Philadelphia PHI Def 69 158 138 181 201 108 155 116 152 141Falcons, Atlanta ATL Def 128 129 149 139 116 101 135 141 174 137Giants, New York NYG Def 145 131 162 126 160 168 126 184 138 132Jaguars, Jacksonville JAC Def 78 132 77 69 90 145 149 152 143 102Jets, New York NYJ Def 129 163 168 183 161 121 133 127 155 112Lions, Detroit DET Def 84 177 152 106 88 169 113 143 154 143Packers, Green Bay GBP Def 133 164 181 189 149 150 163 111 114 153Panthers, Carolina CAR Def 122 99 107 153 134 112 132 191 148 161Patriots, New England NEP Def 184 142 199 147 115 193 170 107 198 226Raiders, Oakland OAK Def 83 119 164 86 135 107 116 96 85 118Rams, St. Louis STL Def 149 104 129 83 91 114 126 136 106 198Ravens, Baltimore BAL Def 133 176 131 166 202 118 237 148 194 213Redskins, Washington WAS Def 128 101 128 93 88 133 74 146 149 104Saints, New Orleans NOS Def 128 105 131 179 109 128 115 83 141 144Seahawks, Seattle SEA Def 180 166 147 115 112 184 145 159 146 152Steelers, Pittsburgh PIT Def 110 132 186 139 178 151 161 173 181 133Texans, Houston HOU Def 175 152 79 131 103 140 115 96 145 89Titans, Tennessee TEN Def 163 122 138 123 170 151 143 133 124 166Vikings, Minnesota MIN Def 140 117 123 147 163 179 163 148 114 162

Thanks for pulling this. Taking this data and looking at year N+1 as a function of year N, fantasy points in prior year only explains 4.185% of variation in the next year.

 

[cdubz]

Beacuse it is extremely hard to predict how well D/ST will perform. Projections for them are usually way off. And teams that were great DEF/ST are rarely dominant 2 years in a row. So while there will be a couple of trult valuable DEF/STs at year's end, picking them before the season starts is a total roll of the dice. IMO, waiting on defenses is the best way to go. Snag one of the leftovers that you might have a hunch on . . . on pick one up off the waiver wire. No point in investing an early pick on a defense IMO.

I think the point being made is that if you can predict D/ST well there is a huge advantage. Just because it's hard doesn't mean it's impossible and your opinion of the best way to go is irrelevant unless you can show your work. I don't mean to be confrontational, I'm just saying there is math behind this and the prediction code is not unbreakable. Given the size of the edge you can afford to be wrong some % of the time and still be profitable. The numbers are saying conventional wisdom is wrong in the VBD context and just because prediction is difficult doesn't mean proper risk/reward evaluation should be ignored.

Football is not a physics function that has one solution. Just because we assign numerical values to certain events, doesn't really mean that there is "math behind it" in the sense we can solve it. Looking at the 2012 dataset, 60% of the top 5 defenses came from teams that were in the bottom half of yards allowed - how do you account for that? And once you fit a model that explains that, you will have over fit the model and it won't be able to predict out of sample. All the kudos in the world if you are able to come up with something that accurately predicts fantasy defense performance, but nothing in this thread leads me to believe anyone is any closer than the thousands that have tried before. ETA: SD is actually 16th, so it is 40% above not 60%

 

[Anarchy99]

Beacuse it is extremely hard to predict how well D/ST will perform. Projections for them are usually way off. And teams that were great DEF/ST are rarely dominant 2 years in a row. So while there will be a couple of trult valuable DEF/STs at year's end, picking them before the season starts is a total roll of the dice.

IMO, waiting on defenses is the best way to go. Snag one of the leftovers that you might have a hunch on . . . on pick one up off the waiver wire. No point in investing an early pick on a defense IMO.

I think the point being made is that if you can predict D/ST well there is a huge advantage. Just because it's hard doesn't mean it's impossible and your opinion of the best way to go is irrelevant unless you can show your work. I don't mean to be confrontational, I'm just saying there is math behind this and the prediction code is not unbreakable. Given the size of the edge you can afford to be wrong some % of the time and still be profitable. The numbers are saying conventional wisdom is wrong in the VBD context and just because prediction is difficult doesn't mean proper risk/reward evaluation should be ignored.

I did studies and wrote articles on defensive scoring when I worked at FBG. I am not just spewing a random opinion.

 

[TheMathNinja]

Beacuse it is extremely hard to predict how well D/ST will perform. Projections for them are usually way off. And teams that were great DEF/ST are rarely dominant 2 years in a row. So while there will be a couple of trult valuable DEF/STs at year's end, picking them before the season starts is a total roll of the dice.

IMO, waiting on defenses is the best way to go. Snag one of the leftovers that you might have a hunch on . . . on pick one up off the waiver wire. No point in investing an early pick on a defense IMO.

I think the point being made is that if you can predict D/ST well there is a huge advantage. Just because it's hard doesn't mean it's impossible and your opinion of the best way to go is irrelevant unless you can show your work. I don't mean to be confrontational, I'm just saying there is math behind this and the prediction code is not unbreakable. Given the size of the edge you can afford to be wrong some % of the time and still be profitable. The numbers are saying conventional wisdom is wrong in the VBD context and just because prediction is difficult doesn't mean proper risk/reward evaluation should be ignored.

I did studies and wrote articles on defensive scoring when I worked at FBG. I am not just spewing a random opinion.

Can you send me your data in something slightly more Excel-friendly? I'd like to analyze it a bit more.

 

[cstu]

The Bears have been remarkably consistent (for a defense) have finished outside of the top 8 one time (2009) since 2003.

 

[cdubz]

Beacuse it is extremely hard to predict how well D/ST will perform. Projections for them are usually way off. And teams that were great DEF/ST are rarely dominant 2 years in a row. So while there will be a couple of trult valuable DEF/STs at year's end, picking them before the season starts is a total roll of the dice.

IMO, waiting on defenses is the best way to go. Snag one of the leftovers that you might have a hunch on . . . on pick one up off the waiver wire. No point in investing an early pick on a defense IMO.

I think the point being made is that if you can predict D/ST well there is a huge advantage. Just because it's hard doesn't mean it's impossible and your opinion of the best way to go is irrelevant unless you can show your work. I don't mean to be confrontational, I'm just saying there is math behind this and the prediction code is not unbreakable. Given the size of the edge you can afford to be wrong some % of the time and still be profitable. The numbers are saying conventional wisdom is wrong in the VBD context and just because prediction is difficult doesn't mean proper risk/reward evaluation should be ignored.

I did studies and wrote articles on defensive scoring when I worked at FBG. I am not just spewing a random opinion.

Can you send me your data in something slightly more Excel-friendly? I'd like to analyze it a bit more.

http://support.microsoft.com/kb/214261 Text to columns should be able to separate out the data items for you.

 

[Drop]

I think the math behind relative positional advantage in the VBD context can be explained better by someone else, but it is there and it's what we want to exploit for an edge.

I mean that the prediction code is not unbreakable in that we don't need a completely predictive model for all teams if we are able to trim out a lot of teams that won't be in the running often. Getting to a partial solution of the teams likely to be the best can allow you to exploit the VBD logic difference commonly seen. Why we see that difference is the question this thread is trying to answer. I aim to predict the top of the range of D/ST and the rest I care much less about as long as the entire distribution shapes out similarly in terms of relative dropoffs as it has in the past. (The distribution being out of whack can throw a wrench in things if the difference between 1 and 12 changes significantly.) I'm saying that the relative advantage of D/ST is seemingly not being properly reflected in drafts very often so if we can get a partial predictive soluation it doesn't even have to be right a large percentage of the time relative to other positions because the VBD gap that can be exploited is so large unlike that gap at the other positions that the draft is efficiently handling.

Payout structure comes into play somewhat, but generally speaking there seems to be a lot of equity in prize money that can be gained by getting this right while others are too risk averse to even attempt it because they percieve it to be extremely difficult to predict the top D/ST. The idea that VBD is telling us that D/ST are being drafted 4+ rounds late can make taking 2 predicted top defenses in those rounds a good idea because it makes us more likely to select the top D/ST and realize the biggest relative advantage as well as not having to play against the other top D/ST amongst other factors. Ideally this can go a long way in helping to scoop the regular season and playoff prize pool which is what we want.

That's how I view this situation. The interesting thing to figure out is why the FF draft market hasn't efficiently taken this into account, or has it and what is the problem with VBD theory? I really need to put much more thought into this, but I sense that there really is something to this D/ST VBD phenomenon that can be used to profit.

 

[cdubz]

I think the math behind relative positional advantage in the VBD context can be explained better by someone else, but it is there and it's what we want to exploit for an edge.

I mean that the prediction code is not unbreakable in that we don't need a completely predictive model for all teams if we are able to trim out a lot of teams that won't be in the running often. Getting to a partial solution of the teams likely to be the best can allow you to exploit the VBD logic difference commonly seen. Why we see that difference is the question this thread is trying to answer. I aim to predict the top of the range of D/ST and the rest I care much less about as long as the entire distribution shapes out similarly in terms of relative dropoffs as it has in the past. (The distribution being out of whack can throw a wrench in things if the difference between 1 and 12 changes significantly.) I'm saying that the relative advantage of D/ST is seemingly not being properly reflected in drafts very often so if we can get a partial predictive soluation it doesn't even have to be right a large percentage of the time relative to other positions because the VBD gap that can be exploited is so large unlike that gap at the other positions that the draft is efficiently handling.

Payout structure comes into play somewhat, but generally speaking there seems to be a lot of equity in prize money that can be gained by getting this right while others are too risk averse to even attempt it because they percieve it to be extremely difficult to predict the top D/ST. The idea that VBD is telling us that D/ST are being drafted 4+ rounds late can make taking 2 predicted top defenses in those rounds a good idea because it makes us more likely to select the top D/ST and realize the biggest relative advantage as well as not having to play against the other top D/ST amongst other factors. Ideally this can go a long way in helping to scoop the regular season and playoff prize pool which is what we want.

That's how I view this situation. The interesting thing to figure out is why the FF draft market hasn't efficiently taken this into account, or has it and what is the problem with VBD theory? I really need to put much more thought into this, but I sense that there really is something to this D/ST VBD phenomenon that can be used to profit.

http://sports.espn.go.com/fantasy/football/ffl/story?page=nfldk2k12_vbdwork I think this might be helpful. VBD is a good tool for retrospective analysis but not necessarily for prospective analysis because it relies on projections that have varying degrees of accuracy. Though it can help explain broad brush strokes such as how you compare RB7 to WR3. For fantasy defense projections in particular, the forecasting is so difficult that and volatile week to week that it diminishes most of the value of any kind of VBD analysis.

Looking at FBG projections from last year http://footballguys.com/cs_p1a.htm GB and PHI were the top 2 projected. If you took GB in the 8th round, you really lost out on some value. Sure if you took Chicago as the 3rd defense you were ecstatic - but it is like shooting fish in a barrel. Denver was foretasted 18th. This isn't because FBG didn't do a good enough job projecting, David Dodds is one of the best in the business just go check out fantasypros.

Vegas and Wall Street make a living off of people thinking they can solve the code of slot machines and stocks; maybe in fantasy football we just need to let people think they can crack the code of fantasy defenses.

 

[FUBAR]

The Bears have been remarkably consistent (for a defense) have finished outside of the top 8 one time (2009) since 2003.

And that might add value in taking the bears top 6 or so, but how many people predicted them as the top D, let alone dominant?

 

[DawnBTVS]

I know points allowed & yards allowed are big factors in Team D but how do sacks/turnovers impact end of year standings? Granted the correlation will be much lower but added up, they usually can score 5-8 points a week themselves depending on the weight of the scoring (assuming say 2 per Sack, 3 per INT/FR).

Last year, Chicago had at least 2 turnovers in 13 games. In 5, they had at least 4 turnovers. Note that these don't include sacks.

 

[Anarchy99]

Using the league info I cited earlier, here were how the defenses fared over the past 10 years. The first set is the Top 5 DEF/ST based on ADP from MFL. The second set is how the Top 5 DEF/ST actually ranked at the end of the season.

2012 ADP VBD 2012 ADP VBDSFO 84 17 CHI 124 82BAL 106 -12 NEP 154 39HOU 110 30 DEN 160 39PIT 120 -35 SEA 147 35GB 121 -12 HOU 110 302011 ADP VBD 2011 ADP VBDPIT 79 0 SFO 154 54GBP 87 32 CHI 123 48NYJ 98 31 DET 136 45PHI 102 26 BAL 107 44BAL 107 44 SEA 208 342010 ADP VBD 2010 ADP VBDNYJ 79 21 NEP 146 52MIN 97 -24 PIT 109 39GBP 97 34 GBP 97 34PHI 103 -9 NYJ 79 21BAL 104 -16 ARI 182 172009 ADP VBD 2009 ADP VBDPIT 74 -7 GBP 134 43BAL 84 20 NYJ 124 37MIN 92 1 SFO 189 36NYG 93 -20 PHI 104 35PHI 104 35 NOS 139 332008 ADP VBD 2008 ADP VBDSDC 80 -29 BAL 146 57MIN 81 18 PHI 155 56NEP 94 -30 PIT 113 33CHI 96 13 TEN 161 25DAL 96 0 MIN 81 182007 ADP VBD 2007 ADP VBDCHI 58 46 SDC 80 76BAL 63 -33 CHI 58 46NEP 77 42 NEP 77 42SDC 80 76 SEA 151 33PIT 103 0 MIN 143 282006 ADP VBD 2006 ADP VBDCHI 68 88 BAL 106 90CAR 84 -15 CHI 68 88PIT 90 14 NEP 140 23IND 104 -49 MIN 173 16BAL 106 90 GBP 187 162005 ADP VBD 2005 ADP VBDBAL 67 2 CHI 135 57NEP 81 -39 CAR 105 45PIT 86 27 NYG 158 38BUF 87 -15 IND 135 33PHI 100 -30 PIT 86 272004 ADP VBD 2004 ADP VBDBAL 70 40 BUF 155 58NEP 72 44 NEP 72 44CAR 93 -6 BAL 70 40PHI 102 -2 PIT 136 27TBB 110 7 ATL 157 182003 ADP VBD 2003 ADP VBDTBB 58 32 NEP 121 73PHI 82 -12 BAL 97 60MIA 82 34 STL 148 45CAR 90 8 MIA 82 34BAL 97 60 TBB 58 32Obviously there were 50 Top 5 picks made (based on the ADP data). Those added up to a VBD score of 537 or basically an average of 10-11 VBD points. However, given that those picks were usually feast or famine, I'm not sure it's remotely close to investing the draft capital on a chance you might hit a home run.

36 of the 50 ADP picks were in the Top 100 picks. Of those, 14 had 0 or negative VBD scores. Out of the entire group of 50, 23 had scores of 0 or negative VBD. Another 4 had VBD scores of under 10 points. Essentially, 27 of 50 picks yielded a return of +10 VBD points or less.

On the flip side, 37 of the 50 teams that ranked in the Top at year's end were picked 101 or lower. 14 were taken 150th or lower.

Maybe Dodds and FBG projected defenses better than the ADP data indicated. If I were really motivated I would look into that, but IMO the net result would be pretty similar. All in all, I don't see any real trends or secret tip offs as ways to better interpret the data to get a competitive advantage at the DEF/ST position. If someone can identify that with any degree of certainty, I'm all ears . . .

 

[Anarchy99]

And clearly individual scoring systems and number of fantasy teams in the league will greatly vary the VBD results. My league is close to a typical scoring league and does not have huge swings for points allowed or yardage allowed and does not penalize defenses by having negative scoring.

 

[cstu]

The Bears have been remarkably consistent (for a defense) have finished outside of the top 8 one time (2009) since 2003.

And that might add value in taking the bears top 6 or so, but how many people predicted them as the top D, let alone dominant?

IMO defenses are impossible to predict accurately due to the randomness of how they score points and it's therefore a waste to spend a high pick on them. However, there are teams like the Bears whose system creates turnovers and are a little more predictable. The Packers, 49ers, Eagles, Patriots, and Ravens have the type of defenses that produce good fantasy numbers. The Steelers and Bucs used to but have slipped off the last couple years.

If you can consistently predict which defense will be #1 every year then you belong in Vegas, not playing FF.

 

[TheMathNinja]

Using the league info I cited earlier, here were how the defenses fared over the past 10 years. The first set is the Top 5 DEF/ST based on ADP from MFL. The second set is how the Top 5 DEF/ST actually ranked at the end of the season.

2012 ADP VBD 2012 ADP VBDSFO 84 17 CHI 124 82BAL 106 -12 NEP 154 39HOU 110 30 DEN 160 39PIT 120 -35 SEA 147 35GB 121 -12 HOU 110 302011 ADP VBD 2011 ADP VBDPIT 79 0 SFO 154 54GBP 87 32 CHI 123 48NYJ 98 31 DET 136 45PHI 102 26 BAL 107 44BAL 107 44 SEA 208 342010 ADP VBD 2010 ADP VBDNYJ 79 21 NEP 146 52MIN 97 -24 PIT 109 39GBP 97 34 GBP 97 34PHI 103 -9 NYJ 79 21BAL 104 -16 ARI 182 172009 ADP VBD 2009 ADP VBDPIT 74 -7 GBP 134 43BAL 84 20 NYJ 124 37MIN 92 1 SFO 189 36NYG 93 -20 PHI 104 35PHI 104 35 NOS 139 332008 ADP VBD 2008 ADP VBDSDC 80 -29 BAL 146 57MIN 81 18 PHI 155 56NEP 94 -30 PIT 113 33CHI 96 13 TEN 161 25DAL 96 0 MIN 81 182007 ADP VBD 2007 ADP VBDCHI 58 46 SDC 80 76BAL 63 -33 CHI 58 46NEP 77 42 NEP 77 42SDC 80 76 SEA 151 33PIT 103 0 MIN 143 282006 ADP VBD 2006 ADP VBDCHI 68 88 BAL 106 90CAR 84 -15 CHI 68 88PIT 90 14 NEP 140 23IND 104 -49 MIN 173 16BAL 106 90 GBP 187 162005 ADP VBD 2005 ADP VBDBAL 67 2 CHI 135 57NEP 81 -39 CAR 105 45PIT 86 27 NYG 158 38BUF 87 -15 IND 135 33PHI 100 -30 PIT 86 272004 ADP VBD 2004 ADP VBDBAL 70 40 BUF 155 58NEP 72 44 NEP 72 44CAR 93 -6 BAL 70 40PHI 102 -2 PIT 136 27TBB 110 7 ATL 157 182003 ADP VBD 2003 ADP VBDTBB 58 32 NEP 121 73PHI 82 -12 BAL 97 60MIA 82 34 STL 148 45CAR 90 8 MIA 82 34BAL 97 60 TBB 58 32Obviously there were 50 Top 5 picks made (based on the ADP data). Those added up to a VBD score of 537 or basically an average of 10-11 VBD points. However, given that those picks were usually feast or famine, I'm not sure it's remotely close to investing the draft capital on a chance you might hit a home run.

36 of the 50 ADP picks were in the Top 100 picks. Of those, 14 had 0 or negative VBD scores. Out of the entire group of 50, 23 had scores of 0 or negative VBD. Another 4 had VBD scores of under 10 points. Essentially, 27 of 50 picks yielded a return of +10 VBD points or less.

On the flip side, 37 of the 50 teams that ranked in the Top at year's end were picked 101 or lower. 14 were taken 150th or lower.

Maybe Dodds and FBG projected defenses better than the ADP data indicated. If I were really motivated I would look into that, but IMO the net result would be pretty similar. All in all, I don't see any real trends or secret tip offs as ways to better interpret the data to get a competitive advantage at the DEF/ST position. If someone can identify that with any degree of certainty, I'm all ears . . .

Thanks for this awesome analysis on the data, Anarchy; it's much appreciated. But I actually think you're making my point here with your data. Historically, we're seeing an average of 10-11 VBD for the top 5 defenses, and, according to my analysis a mean and median both at +20 VBD for the top D selected in any given year. Is this performance remarkably consistent? No. But in the drafts I've observed, guys who are projected at +20 VBD (guys like Jordy Nelson, Wes Welker, Lamar Miller Colin Kaepernick this year) get selected in Round 4, and guys projected at +10 VBD (guys like Ryan Matthews, Eddie Lacy, Danny Amendola this year) get selected in Round 5.

In fact, the fact that we can expect top Defenses to average this production, but possibly "hit the home run" like the Chicago Bears did this past year, combined with the fact that even if my top-projected Defense bombs, I can probably find a servicable waiver-wire pickup, unlike I can at RB, makes me even more likely to take a risk on a top defense. I think your careful analysis has actually served to provide excellent evidence for why team managers are wise to consider taking a top Defense in Rounds 4 or 5. Granted, this doesn't account for the question of "how many weeks do I actually start my +10 VBD-yielding defense?"...i.e., even if they score 10 higher than the baseline defense, do they actually provide my team with 10 more points throughout the season over and above a waiver-wire strategy, since the "baseline/replacement" is probably not a single 12th-ranked defense, but some kind of shuffling strategy. But even so, I think you gave great numerical justification for taking a top-projected defense in the early rounds.

 

[Late225]

If you took a defense early last year, you took the 49ers. They finished 8th among defenses (based on the first post in this thread).

 

[GregR]

If you could guarantee that the projections would come true then they would be worth more, but fantasy defense performance is not exactly predictable.

This should give a good summary:

http://www.pro-football-reference.com/articles/defcc.htm

These numbers certainly hold true if your only D/ST fantasy stats are turnovers and TD's. But now that most systems rely heavily on points and yards allowed, the correlation for fantasy points year-to-year increases dramatically, as these are considerably more predictable traits for team defenses.

Just did a quick correlation coefficient of yards allowed in Year N and in Year N+1 from 2000 through 2012 and got 0.39

 

[Anarchy99]

Using the league info I cited earlier, here were how the defenses fared over the past 10 years. The first set is the Top 5 DEF/ST based on ADP from MFL. The second set is how the Top 5 DEF/ST actually ranked at the end of the season.

2012 ADP VBD 2012 ADP VBDSFO 84 17 CHI 124 82BAL 106 -12 NEP 154 39HOU 110 30 DEN 160 39PIT 120 -35 SEA 147 35GB 121 -12 HOU 110 302011 ADP VBD 2011 ADP VBDPIT 79 0 SFO 154 54GBP 87 32 CHI 123 48NYJ 98 31 DET 136 45PHI 102 26 BAL 107 44BAL 107 44 SEA 208 342010 ADP VBD 2010 ADP VBDNYJ 79 21 NEP 146 52MIN 97 -24 PIT 109 39GBP 97 34 GBP 97 34PHI 103 -9 NYJ 79 21BAL 104 -16 ARI 182 172009 ADP VBD 2009 ADP VBDPIT 74 -7 GBP 134 43BAL 84 20 NYJ 124 37MIN 92 1 SFO 189 36NYG 93 -20 PHI 104 35PHI 104 35 NOS 139 332008 ADP VBD 2008 ADP VBDSDC 80 -29 BAL 146 57MIN 81 18 PHI 155 56NEP 94 -30 PIT 113 33CHI 96 13 TEN 161 25DAL 96 0 MIN 81 182007 ADP VBD 2007 ADP VBDCHI 58 46 SDC 80 76BAL 63 -33 CHI 58 46NEP 77 42 NEP 77 42SDC 80 76 SEA 151 33PIT 103 0 MIN 143 282006 ADP VBD 2006 ADP VBDCHI 68 88 BAL 106 90CAR 84 -15 CHI 68 88PIT 90 14 NEP 140 23IND 104 -49 MIN 173 16BAL 106 90 GBP 187 162005 ADP VBD 2005 ADP VBDBAL 67 2 CHI 135 57NEP 81 -39 CAR 105 45PIT 86 27 NYG 158 38BUF 87 -15 IND 135 33PHI 100 -30 PIT 86 272004 ADP VBD 2004 ADP VBDBAL 70 40 BUF 155 58NEP 72 44 NEP 72 44CAR 93 -6 BAL 70 40PHI 102 -2 PIT 136 27TBB 110 7 ATL 157 182003 ADP VBD 2003 ADP VBDTBB 58 32 NEP 121 73PHI 82 -12 BAL 97 60MIA 82 34 STL 148 45CAR 90 8 MIA 82 34BAL 97 60 TBB 58 32Obviously there were 50 Top 5 picks made (based on the ADP data). Those added up to a VBD score of 537 or basically an average of 10-11 VBD points. However, given that those picks were usually feast or famine, I'm not sure it's remotely close to investing the draft capital on a chance you might hit a home run.

36 of the 50 ADP picks were in the Top 100 picks. Of those, 14 had 0 or negative VBD scores. Out of the entire group of 50, 23 had scores of 0 or negative VBD. Another 4 had VBD scores of under 10 points. Essentially, 27 of 50 picks yielded a return of +10 VBD points or less.

On the flip side, 37 of the 50 teams that ranked in the Top at year's end were picked 101 or lower. 14 were taken 150th or lower.

Maybe Dodds and FBG projected defenses better than the ADP data indicated. If I were really motivated I would look into that, but IMO the net result would be pretty similar. All in all, I don't see any real trends or secret tip offs as ways to better interpret the data to get a competitive advantage at the DEF/ST position. If someone can identify that with any degree of certainty, I'm all ears . . .

Thanks for this awesome analysis on the data, Anarchy; it's much appreciated. But I actually think you're making my point here with your data. Historically, we're seeing an average of 10-11 VBD for the top 5 defenses, and, according to my analysis a mean and median both at +20 VBD for the top D selected in any given year. Is this performance remarkably consistent? No. But in the drafts I've observed, guys who are projected at +20 VBD (guys like Jordy Nelson, Wes Welker, Lamar Miller Colin Kaepernick this year) get selected in Round 4, and guys projected at +10 VBD (guys like Ryan Matthews, Eddie Lacy, Danny Amendola this year) get selected in Round 5.

In fact, the fact that we can expect top Defenses to average this production, but possibly "hit the home run" like the Chicago Bears did this past year, combined with the fact that even if my top-projected Defense bombs, I can probably find a servicable waiver-wire pickup, unlike I can at RB, makes me even more likely to take a risk on a top defense. I think your careful analysis has actually served to provide excellent evidence for why team managers are wise to consider taking a top Defense in Rounds 4 or 5. Granted, this doesn't account for the question of "how many weeks do I actually start my +10 VBD-yielding defense?"...i.e., even if they score 10 higher than the baseline defense, do they actually provide my team with 10 more points throughout the season over and above a waiver-wire strategy, since the "baseline/replacement" is probably not a single 12th-ranked defense, but some kind of shuffling strategy. But even so, I think you gave great numerical justification for taking a top-projected defense in the early rounds.

To each his own, I guess, but I would much rather, at worse, draft more depth than take a poor shot in the dark in taking a defense early. YOU KNOW there will be just as productive (if not BETTER) defenses after the first one or two . . . and usually MUCH later. Better stated, rather than blow a pick in the 50-80 pick range on a defense, I would rather get a starting worthy or top back up and roll the dice later in the draft on a defense. IMO, it's MUCH harder to find fantasy start worthy position players in the 50-80 pick range than it is in the 150 pick range.

I think one thing you are missing is that you are still able to hit the home run at DEF/ST 5, 10, 15, or sometimes 20th defense off the board. But in most years, there is a 40-50% likelihood of busting taking a defense early.

 

[cstu]

Thanks for this awesome analysis on the data, Anarchy; it's much appreciated. But I actually think you're making my point here with your data. Historically, we're seeing an average of 10-11 VBD for the top 5 defenses, and, according to my analysis a mean and median both at +20 VBD for the top D selected in any given year. Is this performance remarkably consistent? No. But in the drafts I've observed, guys who are projected at +20 VBD (guys like Jordy Nelson, Wes Welker, Lamar Miller Colin Kaepernick this year) get selected in Round 4, and guys projected at +10 VBD (guys like Ryan Matthews, Eddie Lacy, Danny Amendola this year) get selected in Round 5.

In fact, the fact that we can expect top Defenses to average this production, but possibly "hit the home run" like the Chicago Bears did this past year, combined with the fact that even if my top-projected Defense bombs, I can probably find a servicable waiver-wire pickup, unlike I can at RB, makes me even more likely to take a risk on a top defense. I think your careful analysis has actually served to provide excellent evidence for why team managers are wise to consider taking a top Defense in Rounds 4 or 5. Granted, this doesn't account for the question of "how many weeks do I actually start my +10 VBD-yielding defense?"...i.e., even if they score 10 higher than the baseline defense, do they actually provide my team with 10 more points throughout the season over and above a waiver-wire strategy, since the "baseline/replacement" is probably not a single 12th-ranked defense, but some kind of shuffling strategy. But even so, I think you gave great numerical justification for taking a top-projected defense in the early rounds.

This is a big problem I have with defenses - how often do you end up playing matchups even if you draft a defense early? For example, the Seahawks were #14 going into week 14 against the Cardinals where they scored 40+. That seems like an obvious must start but they only scored 6 points the first time they faced them. What if you had picked up another defense like the Broncos who were facing the Raiders who they had previously nearly shutout?

 

[ghostguy123]

This is a big problem I have with defenses - how often do you end up playing matchups even if you draft a defense early? For example, the Seahawks were #14 going into week 14 against the Cardinals where they scored 40+. That seems like an obvious must start but they only scored 6 points the first time they faced them. What if you had picked up another defense like the Broncos who were facing the Raiders who they had previously nearly shutout?

I think a HUGE factor in where defenses will be drafted or should be drafted will depend on the number of teams in the league and roster sizes.

Some leagues have 15 defenses on waivers every week. Some only have a few and they suck badly, so picking one semi-early there isnt a bad move.

 

[donkshow]

In my 14 team league, every single team had 2 DST for probably 12 of the 16 weeks. There was never really any more than 8 DST on the wire on any given week. Typically the Titans, Raiders, Lions, Chiefs, Saints, Jags. Not teams I'd particularly even want in strong match-ups.

 

[Drop]

I think this article relates to the discussion:

https://www.profootballfocus.com/blog/2013/05/24/fantasy-football-tight-ends-a-fools-game/

 

[cdubz]

I think this article relates to the discussion:

https://www.profootballfocus.com/blog/2013/05/24/fantasy-football-tight-ends-a-fools-game/

How so? Not that I agree with both, but I could see impact on draft slot of defenses going either way depending on your takeaway from the article.

 

[MDSkinner]

Anarchy99 makes the point on this really well. I think if there were a guarantee that by choosing the first defense that you would be adding the value described in the original post then it would be worth taking a defense fairly early.

The problem is that there is very little chance that the defense that comes off the board first will end up adding that kind of value. I would rather be the guy taking a shot on a defense late and having the 10th-12th best defense (with the potential of having top 3) than taking a defense in the 5th round and knowing that most likely the defense will not be the best defense by the end of the year even with the likelihood that week to week it will be a strong defense.

For me, it is just not worth it.

 

[CalBear]

In fact, the fact that we can expect top Defenses to average this production, but possibly "hit the home run" like the Chicago Bears did this past year, combined with the fact that even if my top-projected Defense bombs, I can probably find a servicable waiver-wire pickup, unlike I can at RB, makes me even more likely to take a risk on a top defense.

This indicates flawed logic, and possibly a flawed understanding of VBD, especially once the VBD numbers get low. The idea of VBD is to attempt to measure how much value you are getting compared to a replacement player at the same position. If you think you can get a servicable defense on the waiver wire, the value of defenses is lower than other positions, because you can find a cheap replacement.

When heading into the fourth and fifth round, you're comparing projections for the 22nd RB with the 24th RB, and projections for the 1st D with the 12th D. The value of the 24th RB is zero, according to VBD if you're using worst-starter as a baseline. But, it's pretty clear that the value of the 24th RB is not zero. Last year the 24th RB by ADP was Shonn Greene, who finished with 51 VBD points as the #15 RB, a solid RB2. The potential of getting a solid RB is worth a lot more than the potential of getting a solid defense (even if you believe defenses are relatively predictable).

 

[Leonidas]

If you took a defense early last year, you took the 49ers. They finished 8th among defenses (based on the first post in this thread).

Yup. That's just what I did. I took the 49ers in round 4 of my league, which had elevated DEF scoring compared to the standard. It did not work out very well for me.

Perhaps, as mentioned elsewhere in this article, I should have taken ANOTHER top ranked defense a little later. The 5th DEF off the board was 8.03.

At the time you couldn't have convinced me to take ANOTHER pick on a DEF, but I'm seeing the logic behind it.

 

[Drop]

I think this article relates to the discussion:

https://www.profootballfocus.com/blog/2013/05/24/fantasy-football-tight-ends-a-fools-game/

How so? Not that I agree with both, but I could see impact on draft slot of defenses going either way depending on your takeaway from the article.

I posted the article because it delves into the TE position in a similar way. People can take from it what they will and perhaps add new insight into the discussion. I don't totally agree with the article, but I think it may have some merit and adds to the discussion. I don't know if I can get into exactly why I disagree with the article without going off on a TE tangent that will probably hijack the thread.

 

[cdubz]

I think this article relates to the discussion:

https://www.profootballfocus.com/blog/2013/05/24/fantasy-football-tight-ends-a-fools-game/

How so? Not that I agree with both, but I could see impact on draft slot of defenses going either way depending on your takeaway from the article.

I posted the article because it delves into the TE position in a similar way. People can take from it what they will and perhaps add new insight into the discussion. I don't totally agree with the article, but I think it may have some merit and adds to the discussion. I don't know if I can get into exactly why I disagree with the article without going off on a TE tangent that will probably hijack the thread.

I guess I don't see how it explicitly relates to the DEF position, just that it deals with value and where you should draft players. That is the name of the game, knowing where you can find value late so you can load up on value at other players early. And as many above have pointed out, DEF is one area where you can find value late so you can load up elsewhere.

 

[TheMathNinja]

In fact, the fact that we can expect top Defenses to average this production, but possibly "hit the home run" like the Chicago Bears did this past year, combined with the fact that even if my top-projected Defense bombs, I can probably find a servicable waiver-wire pickup, unlike I can at RB, makes me even more likely to take a risk on a top defense.

This indicates flawed logic, and possibly a flawed understanding of VBD, especially once the VBD numbers get low. The idea of VBD is to attempt to measure how much value you are getting compared to a replacement player at the same position. If you think you can get a servicable defense on the waiver wire, the value of defenses is lower than other positions, because you can find a cheap replacement.

When heading into the fourth and fifth round, you're comparing projections for the 22nd RB with the 24th RB, and projections for the 1st D with the 12th D. The value of the 24th RB is zero, according to VBD if you're using worst-starter as a baseline. But, it's pretty clear that the value of the 24th RB is not zero. Last year the 24th RB by ADP was Shonn Greene, who finished with 51 VBD points as the #15 RB, a solid RB2. The potential of getting a solid RB is worth a lot more than the potential of getting a solid defense (even if you believe defenses are relatively predictable).

I'm not sure it's my logic that's flawed here. When we break down the math, it all comes down to expected value and risk aversion.

Let's use your example of the fifth-round scenario. Ultimately, I'm paying for points. Even the 25th RB has 0 VBD, he's still not worthless to me...he's ultimately worth how many points I expect him to give my team. If he's going to fill in for a bye week for a starter, I evaluate him against his next-lowest replacement on my bench, then I consider that maybe he has a breakout season, as well as the possibility that my starter gets injured week 1 and he starts all year, and the possibility that he gets injured week 1 and is worthless. So perhaps my expected value looks like this:

Event A: Everyone stays healthy, he fills in a couple weeks when my starter is on bye and when he plays a weak defense. Expected Points gained: 8...probability: 40%

Event B: He has a breakout season and becomes my starter. Expected points gained: 30. Probability: 20%

Event C: Starter gets injured and he starts all year. Expected points gained (from this player): 30. Probability: 20%

Event D: He gets injured and I never start him. Expected points gained: 0. Probability: 20%.

So adding it all up, I expect Mr. RB25 to be worth (.4)*(8)+(.2)(30)+.2(30)+.2(0) = 15.2 expected points.

Let's say I do the same with the defense, thinking there is a 50% chance they are the top defense and get me +40 over a waiver-wire strategy (which will have varying values depending on how the league hordes Defenses), and there's a 50% chance they totally tank and are useless.

Then my expected value for this defense is (0.5)(40) + (0.5)(0) = 20 expected points.

Now, pure expected value says take the Defense. But if you're particularly risk-averse, you may decide that you're not willing to pay an added 30% of 0 production for 5 more expected points...but these are things for economists to study...risk-aversion is variable person-to-person and is ultimately what makes drafts so interesting...you get to see how people value MJD, and how much they're willing to risk an injury for great production. In this case, I think taking the risk is pretty worthwhile in evaluating a top Defense against a RB22.

 

[Anarchy99]

What you are missing in evaluating RB25 is that if your other RBs get hurt, RB25 becomes a starter on your fantasy team. Guess what? Team defenses don't get zeroes by having an injury (and there also are usually other defenses you can pick up and play and KNOW they will score something). In my leagues, there usually are not great RB options to claim off of waivers, as I still play in leagues where RBs are hoarded.

By taking that defense in the 5th or 6th, you are essentially passing on a player (whatever the position) that is an NFL starter at their position. But by taking a DEF early, you will end up with less depth and have to settle for a part time NFL player later on (ie someone in a job share or someone that needs help to get more playing time).

And you seem more interested at looking at the AVERAGE VBD score of taking the first defense while leaving out the part about swinging and missing. In that scenario, you are potentially WORSE OFF at defensive scoring AND with less depth. If that defense REALLY bombs, then you could end up dumping your 6th round pick and picking up another defense anyway . . . which to me sounds like there's a decent chance of wasting a 5th or 6th round pick in the long run.

Maybe the fact that in the league that I mentioned, there are TONS of defenses on the wire and few owners carry two of them (limited roster spots available) skews my thinking on this. There are usually 12-14 defenses on the wire to pick from.

 

[Borden]

Great discussion fellas. I love the OP finding and using stats for a different look on things and the stats and opinions on the other side are great.

I have a few questions for everyone in general that I think might be getting over looked.

- How much does strength of schedule affect DST?
On a Year to Year Basis: Usually, teams with top defenses are also teams that finish higher in the NFL. Which means going into the next season they will have to play stronger teams the next year. For teams that win their division it means they have to play two extra games against the other division champs in their conference.

On a Drafting Basis: I'm pretty sure every team plays a divisional game week 17. Which means if you draft Denver this year, you get Houston in your Championship game (week 16) and miss out on a must start in Oakland (week 17).

- TDs and fantasy points have to be thrown out?
I think most of us can agree defensive and special teams TDs are pretty random which has a major impact on fantasy points. I am not saying that the TDs shouldn't be a consideration when drafting the Bears but I do believe that it throws things out of whack when using those numbers in this situation.

- APD doesn't matter and neither does last year.
Players move, schemes get solved, new schemes show up and so on. ADP is way to dependent on other people's views and too many people buy into hype and use last years numbers.

My conclusion:
It makes complete sense to take a defense early. I could argue taking a DST before a QB to be honest. However, because it is so difficult to predict which DSTs will be top 5, the risk/reward too much to justify taking a defense over depth (or starters) at other positions.

 

[cdubz]

Great discussion fellas. I love the OP finding and using stats for a different look on things and the stats and opinions on the other side are great.

I have a few questions for everyone in general that I think might be getting over looked.

- How much does strength of schedule affect DST?

On a Year to Year Basis: Usually, teams with top defenses are also teams that finish higher in the NFL. Which means going into the next season they will have to play stronger teams the next year. For teams that win their division it means they have to play two extra games against the other division champs in their conference.

On a Drafting Basis: I'm pretty sure every team plays a divisional game week 17. Which means if you draft Denver this year, you get Houston in your Championship game (week 16) and miss out on a must start in Oakland (week 17).

- TDs and fantasy points have to be thrown out?

I think most of us can agree defensive and special teams TDs are pretty random which has a major impact on fantasy points. I am not saying that the TDs shouldn't be a consideration when drafting the Bears but I do believe that it throws things out of whack when using those numbers in this situation.

- APD doesn't matter and neither does last year.

Players move, schemes get solved, new schemes show up and so on. ADP is way to dependent on other people's views and too many people buy into hype and use last years numbers.

My conclusion:

It makes complete sense to take a defense early. I could argue taking a DST before a QB to be honest. However, because it is so difficult to predict which DSTs will be top 5, the risk/reward too much to justify taking a defense over depth (or starters) at other positions.

I think this thread may be more beneficial than any of us could have thought,

 

[Drop]

I think this article relates to the discussion:

https://www.profootballfocus.com/blog/2013/05/24/fantasy-football-tight-ends-a-fools-game/

How so? Not that I agree with both, but I could see impact on draft slot of defenses going either way depending on your takeaway from the article.

I posted the article because it delves into the TE position in a similar way. People can take from it what they will and perhaps add new insight into the discussion. I don't totally agree with the article, but I think it may have some merit and adds to the discussion. I don't know if I can get into exactly why I disagree with the article without going off on a TE tangent that will probably hijack the thread.

I guess I don't see how it explicitly relates to the DEF position, just that it deals with value and where you should draft players. That is the name of the game, knowing where you can find value late so you can load up on value at other players early. And as many above have pointed out, DEF is one area where you can find value late so you can load up elsewhere.

It obviously is about TEs, but it provides some insight into the "fool's game" which relates to the concept of taking D/ST early as is being discussed in this thread. People are saying it's too unpredictable both in that article and in here, that's the parallel. The article is an example of how to approach such a problem and lessons can perhaps be taken from the methodoloy. I don't totally agree with the article, but I see that is has merit and may be helpful in this thread.

 

[Drop]

So how do we get better at predicting top D/STs to be rewarded enough to make up for the risk in the long run?

 

[TheMathNinja]

So how do we get better at predicting top D/STs to be rewarded enough to make up for the risk in the long run?

If someone can find a good-looking data set for me, I can run some good numerical analysis and get back to you guys pretty quickly. What I'd be interested in seeing in the data set is: Yards allowed, points allowed, turnovers forced, TD's, Dodds' preseason estimate (or some other top picker), ADP, final fantasy points for the year. What I can then do is create a formula based on Multiple Linear Regression that best estimates a team's fantasy value for the year as a function of last year's points, last year's TD's, last years yards and points allowed, last year's TO's, and preseason expectation (and perhaps ADP). The formula will uncover which metrics are actually useful and which are basically totally random; and I can also tell you the strength of its predictive power. Analysis is easy for me; finding data is not.

 

[Dinsy Ejotuz]

So how do we get better at predicting top D/STs to be rewarded enough to make up for the risk in the long run?

If someone can find a good-looking data set for me, I can run some good numerical analysis and get back to you guys pretty quickly. What I'd be interested in seeing in the data set is: Yards allowed, points allowed, turnovers forced, TD's, Dodds' preseason estimate (or some other top picker), ADP, final fantasy points for the year. What I can then do is create a formula based on Multiple Linear Regression that best estimates a team's fantasy value for the year as a function of last year's points, last year's TD's, last years yards and points allowed, last year's TO's, and preseason expectation (and perhaps ADP). The formula will uncover which metrics are actually useful and which are basically totally random; and I can also tell you the strength of its predictive power. Analysis is easy for me; finding data is not.

I'll save you some time. It's been done. And it doesn't change what Anarchy said upthread.

 

[cdubz]

So how do we get better at predicting top D/STs to be rewarded enough to make up for the risk in the long run?

If someone can find a good-looking data set for me, I can run some good numerical analysis and get back to you guys pretty quickly. What I'd be interested in seeing in the data set is: Yards allowed, points allowed, turnovers forced, TD's, Dodds' preseason estimate (or some other top picker), ADP, final fantasy points for the year. What I can then do is create a formula based on Multiple Linear Regression that best estimates a team's fantasy value for the year as a function of last year's points, last year's TD's, last years yards and points allowed, last year's TO's, and preseason expectation (and perhaps ADP). The formula will uncover which metrics are actually useful and which are basically totally random; and I can also tell you the strength of its predictive power. Analysis is easy for me; finding data is not.

Yeah, you aren't exactly special in that regard. The main challenge in doing statistical analysis on FF is getting the data set into a workable format. Given all the tools that are available (even on this site!) you should maybe do your own work on that front rather than expect others to do your data wrangling.

 

[cstu]

What you are missing in evaluating RB25 is that if your other RBs get hurt, RB25 becomes a starter on your fantasy team. Guess what? Team defenses don't get zeroes by having an injury (and there also are usually other defenses you can pick up and play and KNOW they will score something). In my leagues, there usually are not great RB options to claim off of waivers, as I still play in leagues where RBs are hoarded.

By taking that defense in the 5th or 6th, you are essentially passing on a player (whatever the position) that is an NFL starter at their position. But by taking a DEF early, you will end up with less depth and have to settle for a part time NFL player later on (ie someone in a job share or someone that needs help to get more playing time).

And you seem more interested at looking at the AVERAGE VBD score of taking the first defense while leaving out the part about swinging and missing. In that scenario, you are potentially WORSE OFF at defensive scoring AND with less depth. If that defense REALLY bombs, then you could end up dumping your 6th round pick and picking up another defense anyway . . . which to me sounds like there's a decent chance of wasting a 5th or 6th round pick in the long run.

Maybe the fact that in the league that I mentioned, there are TONS of defenses on the wire and few owners carry two of them (limited roster spots available) skews my thinking on this. There are usually 12-14 defenses on the wire to pick from.

Just to give people an idea how little defenses are valued - in my 12 team dynasty league with 26 roster spots there are currently 27 defense on the waiver wire.

 

[Dinsy Ejotuz]

You can do make good Ds more valuable and even a little more consistent by adding scoring for yards and increasing scoring for points against. But there will still be a handful of solid/good WW defenses available each year. So there's not much scarcity.

 

[Drop]

So how do we get better at predicting top D/STs to be rewarded enough to make up for the risk in the long run?

If someone can find a good-looking data set for me, I can run some good numerical analysis and get back to you guys pretty quickly. What I'd be interested in seeing in the data set is: Yards allowed, points allowed, turnovers forced, TD's, Dodds' preseason estimate (or some other top picker), ADP, final fantasy points for the year. What I can then do is create a formula based on Multiple Linear Regression that best estimates a team's fantasy value for the year as a function of last year's points, last year's TD's, last years yards and points allowed, last year's TO's, and preseason expectation (and perhaps ADP). The formula will uncover which metrics are actually useful and which are basically totally random; and I can also tell you the strength of its predictive power. Analysis is easy for me; finding data is not.

Most of these stats are readily available on NFL.com. I don't know your methods, but it seems that it's been done and isn't particularly groundbreaking according to some of the responses in this thread. There may be other stats that are not being used that can be useful as well as things that aren't so easily quantified.