fantasy football     JOIN THE HUDDLE    


The Madness of Value Based Drafting
Fritz Schlottman

Never forget that only dead fish swim with the stream.
- Malcolm Muggeridge

One of the things about fantasy football that leaves me shaking my head each and every season is the continued obsession with Value Based Drafting.  The concept was originally conceived in rotisserie baseball  as a tool for mathematically estimating the relative value and draft position of different players playing different positions for any given scoring system.  There is little doubt that the concept was a step-forward for its time and has provided its users with a good tool for determining value.  However, its fable powers of prognostication, like many a good deer story, have grown bigger and more powerful with every passing year until few dare to challenge it without being buried by the flaming posts of the true believers.  Call me stupid for trying, but bring on the heat.

As every serious fantasy football owner has been exposed to the concept, I will not bore my readers to tears with a long essay on the mechanics of setting up a VBD draft board when a short summary will do.  Value Based Drafting is: 1) projecting the future statistics of a player using historical data (see also average value drafting) or through the use of forecasts; 2) calculating the number of fantasy points scored by each player using your leagues scoring formula to determine the values of all players; 3) calculating the difference in value between a given player's value and the projected last starter's value at that position for a league; and 4) ranking all players based on that difference.

On the face of it, Value Based Drafting seems pretty reasonable.  It's defenders can point to examples of how the math works out for any given draft so that the owner using VBD concepts comes out with a better team than the other 11 fictional owners. 

The problem is.often Value Based Drafting doesn't work.  First, it is based on the premise that the VBD user is able to correctly predict an individual player's performance before the start of the season.  Most value a given player based on: 1) last year's numbers; 2) the average of several years numbers; 3) a projection or forecast based on the owners gut feeling, 4) the valuations of others, or 5) a linear regression of historical data. 

And here is where the problems begin.  Let's assume for an example that a VBD value for Player X (a runningback) is based on the assumption that Player X will generate 1,400 total yards and 10 touchdowns.  Assuming that the scoring system allows for 1 point per 10 yards of total offense and 6 points per touchdown, then you could calculate that Player X will earn (1,400/10=140 points for yards and10 x 6= 60 points for touchdowns) for a season total of 200 fantasy points.  And if this happens, there is absolutely no doubt that Player X will score that many points.  The problem is this: just how confident are you that Player X will score 10 touchdowns and 1,400 yards of offense?  Why not 1,200 yards and 8 touchdowns?  Why not 1,700 yards and 6 touchdowns?  And if you can predict that Player X will get exactly 1,400 yards and 10 touchdown, why in blue blazes do you need a formula to tell you where to draft players?

To be perfectly honest, no one and I mean absolutely no one knows exactly how many yards and touchdowns a player will get in a given season.  Not his coach, not his owner, not his agent, not even the player himself knows with any certainty what the season will bring.  If someone could somehow do this, there would be no draft busts, no overpaid free agents and no 7th round draft picks or free agents coming out of college making teams.  All first round selections would end up in the Hall-of-Fame, or at least be a regular feature in the Pro Bowl, and all late round selections would be cut before the end of mini-camp.  Does that happen?  No!  Many early round picks will find themselves on the outside looking in after a few unproductive years and many latter round picks and free agents have found their way to Hawaii in February.

If someone tells you they can accurately project numbers in the NFL, ask them exactly how much money they made on the Patriots last year?  Most experts didn't have them winning the Super Bowl when it was down to two teams, much less predicted New England to win the whole shooting match back in July and August.  Then there is the example of Priest Holmes.  Exactly where kind of numbers did that math major project for Holmes last season?  If he used historical data, Holmes would have been a late round selection at best. 

Scientific Numbers

Remember the excessively boring part of your high school chemistry class where the science teacher talked about using scientific numbers?  If you somehow managed to stay awake long enough to get something out of the class, there was a very important point made somewhere during those long and sleep provoking sessions.  The point of scientific numbers is that you can't have a solution that is more accurate than the input's accuracy. 

For our purposes, if you can't accurately predict how many fantasy points a player will score with any degree of accuracy, you can't know his VBD value either.  The madness becomes immediately obvious to anyone who reads any VBD valuation where the author is telling you he knows the 13th best tightend is better than the 14th best.  First, there is so little statistical difference between the 13th and 14th tightend that the difference the author is hanging his hat on is less than the difference between catching or not catching one ball in the end zone sometime during the season.  And that means the author somehow knows exactly how many touchdowns both players will score, and nobody knows that.  Another example of good math, and bad science.

Draft Run Amok

The second most obvious problem with Value Based Drafting theory is that it fails when you need it most-during the draft.  The basis of VBD theory is that all owners can perceive value equally and will make rational choices. After all, VBD value is based on the difference between your starting quarterback and (in a 12-team league starting 1 quarterback each) the 12th best quarterback on the board.  There are three obvious problems that come to mind: what happens when it's round three and your VBD chart tells you to pick a back-up quarterback; what happens when you have roster limits for given positions; and how do you account for bad drafting, poor play and injuries?

Let's take the first problem head-on.  I'm in the third round of my draft, my first two selections were quarterback/runningback (let's say Warner/Dillon).  I'm on the clock and look down at my VBD chart for an answer.  Let's assume that according to the numbers, the best player on the board is Brett Favre, the best runningbacks have been picked over, and there is a string of wide-receivers that are all about the same, and Tony Gonzales was taken just before my selection.  What do I do?  I know that Brett Favre will play every game, but I don't need him to if I have Kurt Warner.  In fact, Brett Favre doesn't have  much value to me because, barring great injury, Kurt Warner is going to be my starter.  At this position, maybe another runningback has more value to me regardless of what by VBD sheet tells me. 

And there is the problem, value is relative and it changes all the time.  A player on your bench has almost no value unless the starter has an injury, plays poorly, or has a much worse match-up than the back up.  And if any of these examples comes into play, then the starter does not have as much value to me as he did on draft day.  Value is not a known, not for a game, not for a season.  The only value a player on the bench or in the free agent pool has is that he doesn't play against you.  In the example above, the only reason you would take Brett Favre is that you felt that he had more value if he played against you (trade bait) or you wanted to withhold him from the enemy (negative value).  In that case, you have to trade him or Kurt Warner to get third round value from your pick.  If your third round pick sits on the bench nest to your 14th round selection, how valuable is he?

Now, let's assume I follow my VBD principles to the letter and draft Brett Favre.  What does that do to the value of Kurt Warner?  Well, with Brett taking a powder on the bench it's very doubtful that the 12 best quarterbacks will be playing on any given week.  If that's true, then the VBD value of Warner went up because you are comparing him against the 13th best quarterback not the 12th best (an example of negative value).  What if other owners in my league followed the same logic?

It's quite possible that 7th, 8th or 9th best quarterbacks are also on someone's bench, raising the value of Warner that much more.  Now let's assume there is an injury in preseason to one of the other team's starting quarterbacks.  Now the last starting quarterback is that much worse and Warner's value is that much better.  Did I know any of this when I was making out my draft sheet?  No!  Did I correctly value Warner according to VBD principles when I drafted him?  No, and no one could because it would require knowing in advance all the selections of all the other owners and knowing in advance who will be benched or injured.

Then there is the problem with roster limits.  The biggest public leagues have roster limits for drafting certain positions.  For Sportsline managed leagues, the limits are 2QB, 3RB, 3WR, 2TE, 2K, and 2 defenses on a 14-man roster.  The best way to explain this is to think about bullets in a gun.  With draft roster restrictions, you only have so many bullets to use on a given position.  Let's assume that it's draft day and everyone's using VBD to draft.  We are in the middle rounds and everyone has taken their starters at the offensive positions.  Some of the owners have taken a third running back. 

Assuming that my VBD draft board tells me to select a running back (and I have a sleeper in mind), do I make the pick or do I play chicken?  Because I know that some owners have taken their third running back and can't select another one no matter what their VBD board says (all out of bullets), I may be tempted to take another player at a different position.  Why?  Because, if everyone else is using VBD drafting, you can bet that runningbacks are still going to be flying off the board (because of shortages at the position).  If I ignore my VBD chart, and everyone else selects a runningback, I can hold off to my very last pick to get my sleeper.  At that point, I get a player that I valued in the middle rounds and that everyone else paid a middle round pick to get a player of equal value, for the incredible value of a last round selection.  The truth is that Value Based Drafting is not a substitute for a quick mind and good draft strategy.

The final shot I will take at VBD is its use of, or disuse of, historical data to make projections.  The underlying assumption for using historical data is that a player will perform much like he had done in the past. Plainly stated a player that posts good stats year after year will out perform a player that has not posted good numbers.  The problem is you're comparing apples to oranges.  Historical data does not isolate the variable it wishes to measure (player performance) and instead captures wide array of other influences that affect performance (e.g., strength of schedule, coaching, team support, offensive strategy, team and opponent health, etc.).  Given the turnover in coaching staffs, players, the frequency of injury, and strength and strategy of opponents, comparing results from one year to the next with any more accuracy then the insight each of us are born with seems problematic at best.  I have yet to see how any of the mathematical models used for VBD control for these numerous variables. 

The other problem with using historical data is it does not capture what is happening now and in the future.  If everything can be placed on a trend line, then every website, including those featuring VBD, should never have to answer the question "Whom do I start?"  If a VBD user really believed in what it was putting out, they could tell you before the season began not only who should be starting each and every game but any given player's value that week.  And that's the rub, no one knows how a player is going to perform on any given Sunday.  If a player gets injured, his statistics should go down.  If a player gets hot, his value increases.  The league has training camps and preseason games because they (the teams) don't know how good their team is or how well a player is going to perform.  Some teams will play terribly early in the season, then jell and win their last seven games.  Some will take a dive right into the crapper and never come up for air.  You never know.  The point is right now, in July, no one, not Las Vegas, not the teams, not the players know who is going to win it all.  And if they don't know, we certainly don't know.

If you don't use historical data, you're left with making a projection.  This is essentially, using your gut to tell you what kinds of numbers a player will put up.  And if you're using your gut, you may as well skip the step of coming up with performance numbers you know are ball park figures; just slot the players into where your gut tells you they should be on your draft rankings.

Which brings us to "Modified VBD".  Modified VBD is the acknowledgment that VBD gives you an approximation of how well a player will perform, but is not a substitute for an owner's own intelligence and skill.  In other words, its not any better than your best guess- and that I suppose is the whole point of this article.