Category: Free Content

Some Thoughts on Keeper League Strategies

One of the mistakes I most often see fantasy owners make in keeper leagues is keeping players who were studs the year before, but also cost high draft picks to keep.  This is a problematic strategy for a few reasons, not the least of which is the annual attrition that takes place amongst the “studs” of fantasy football.

The easiest way for me to describe my priority order when deciding on my keepers is through a simple matrix where the production of a player is shown across the top (low, moderate, high) and the draft pick cost is down the side (early round, mid-round, late round).  In the boxes of the matrix I have put a number that corresponds with my priority for that action.  So keeping a high production late rounder (Michael Vick) is the #1 priority.

Production
Low Moderate High
Draft Pick Early 9 7 4
Mid 8 5 2
Late 6 3 1

Think of the production levels as being something like a high production guy is your typical fantasy stud, while a moderate production guy would be a guy who would start most weeks in your league.  Maybe LeGarrette Blount is a good example.

As you can see, I have keeping a moderate production guy like a LeGarrette Blount as a late rounder as being preferable to keeping a high production guy with an early pick.  So basically, I would rather keep Blount and have a first round pick, than keep MJD and give up my first round pick.

This strategy generally holds if you have 2-4 keepers (it’s a little different if you just keep one player).  But for now let’s think about why this is sound strategy when you keep let’s say 3 players.

The first reason it works is that your leaguemates will just keep their best players from the prior season, often even when they cost 1st through 3rd round picks.  If you keep guys who cost later round picks, you’re going to be picking a lot in rounds 1-3.  Each of your leaguemates might end up with 3 of the top 36 players.  But you’re going to end up with maybe as many as 6 of the top 50 players.

So you’re going to keep 3 guys who are at a minimum starters in your league, and then you’re going to have a disproportionate number of the picks in rounds 1-3 where you’re going to take the guys who were just not quite good enough to be kept.  So like I said, you’re starting the draft with about 6 of the top 50 players, whereas some of your leaguemates will have as few as 3 of those top 50 players (really just their keepers), and others might have 4 of those top 50 players.

If all you did was execute this simple strategy, you would probably have a playoff team about 3 of every 4 years even if you didn’t draft really well.  You’re at such a numbers advantage to your leaguemates when you do this, and the best part is they might not ever realize it. 

I’ve done this and had a mix of keepers who range from awesome (Jamaal Charles in the 13th round last year) to below average (Pierre Garcon in the 16th round last year) and it comes out the same every time.  I load up on the Arian Fosters, Ahmad Bradshaws, and Hakeem Nicks of the world in those early rounds when the other guys in my league can’t pick because they kept DeAngelo Williams.  Yeah, you’re not super excited about having Garcon, but just from a numbers standpoint you win out due to having a disproportionate number of the top 50 guys.

Introducing Game Level Similarity Projections

One of the things I’ll be rolling out for subscribers when the season starts is Game Level Similarity Projections (GLSP, or since all of these projection systems have to have stupid names of obscure former players, we’ll call ours GILLESPIE after former Vikings and Buccs wide receiver Willie Gillespie).

The easiest thing we can do to get a sense as to what an individual player will do against a defense is to look at what similar players have done in the past against similar defenses.  Before I get into the theory of why we should look at similar players versus similar defenses (instead of running a regression), let’s look at an example.

Let’s take Ben Roethlisberger’s 2011 Week 2 matchup against the Seattle defense.  For purposes of picking out the similar games, we’ll use Ben’s 2010 stats and the Seattle defense’s 2010 stats (when the season starts we will slowly incorporate the 2011 data so as not to be using old data).

Here are their stats that we’ll be using to pick out similar games (the stats should be fairly easy to understand with the exception of Efficiency, which is basically an opponent adjustment):

Att/G Y/G TD/G INT/G Y/A Rush Yards Efficiency
Ben Roethlisberger 32.4 266.7 1.4 0.4 8.2 14.7 1.2
SEA Defense 36.4 264.1 1.9 0.8 7.2 6.0 1.1

We then take those averages and compare them to other players and other defenses, then find the most similar matchups that featured players and defenses with similar stats.  We get a table of similar games as follows:

Player Game Date Opp Comp Att Cmp% Yards TD Int Y/A Rushes Yds TD
K. Orton 19-Sep-10 vs SEA 25 35 71% 307 2 0 8.8 3 -5 0
J. Cutler 17-Oct-10 vs SEA 17 39 44% 290 0 0 7.4 2 19 0
P. Rivers 26-Sep-10 vs SEA 29 53 55% 455 2 2 8.6 1 2 0
A. Smith 12-Dec-10 vs SEA 17 27 63% 255 3 0 9.4 4 12 0
A. Smith 12-Sep-10 vs SEA 26 45 58% 225 0 2 5.0 0 0 0
B. Roethlisberger 14-Nov-10 vs NE 30 49 61% 387 3 1 7.9 1 12 0
G. Frerotte 23-Dec-00 vs SF 18 29 62% 205 1 0 7.1 3 23 1
J. Freeman 26-Dec-10 vs SEA 21 26 81% 237 5 0 9.1 4 23 0
E. Manning 07-Nov-10 vs SEA 21 32 66% 290 3 0 9.1 1 -1 0
M. Brunell 04-Nov-01 vs TEN 21 32 66% 261 1 1 8.2 2 21 0
M. Brunell 23-Sep-01 vs TEN 17 27 63% 235 0 0 8.7 6 19 0
M. Schaub 18-Nov-07 vs NO 21 33 64% 293 2 0 8.9 2 7 0
R. Gannon 09-Dec-99 vs TEN 20 28 71% 273 1 1 9.8 1 3 0
M. Ryan 19-Dec-10 vs SEA 20 35 57% 174 3 1 5.0 6 6 0
J. Kitna 19-Dec-10 vs WAS 25 37 68% 305 2 0 8.2 4 11 0
A. Rodgers 10-Oct-10 vs WAS 27 46 59% 293 1 1 6.4 4 30 0
J. Cutler 13-Dec-07 vs HOU 27 39 69% 254 1 0 6.5 2 11 0
M. Schaub 19-Sep-10 vs WAS 38 52 73% 497 3 1 9.6 1 2 0
M. Schaub 25-Nov-07 vs CLE 22 36 61% 256 2 2 7.1 1 7 0
C. Batch 21-Oct-01 vs TEN 25 42 60% 338 3 1 8.1 2 3 0
R. Gannon 22-Dec-01 vs TEN 29 50 58% 249 1 1 5.0 2 11 0
M. Hasselbeck 24-Nov-02 vs KC 25 36 69% 362 3 0 10.1 3 23 0
Comp Att Cmp% Yards TD Int Y/A Rushes Yds TD
Averages 23.7 37.6 63% 293 1.9 0.6 7.8 2.5 10.9 0.0

Let’s look over the table of similar games to see what it gives us.  First, the five most similar games all feature the SEA defense from 2010.  The next most similar game is actually our subject player Ben Roethlisberger against a defense, the Patriots, that was similar to SEA last year.  All of the games should feature QBs who are at least a rough approximation of Roethlisberger, and defenses who are at least rough approximations of the SEA defense.  Not every game will be an exact match, but they’ll all be close.

We can then use the averages of those 20 similar games to create a projection for Big Ben when he faces the SEA defense in Week 2.  In fact, we can create projections for the first 6 weeks of the season using this methodology.  To be clear, the error rates will be a little higher when we’re using last year’s data.  Schemes change and players change.  But over the course of the season we’ll be adding in more data from the current season and the error rates should go down.

Maybe a good discussion to have at this point is why we would use GILLESPIE, instead of using perhaps a simple regression.  First, GILLESPIE is going to overcome issues that you might have with a linear regression when not all of the variables have linear relationships.  In a sense, GILLESPIE is a lazy method of forecasting.  We don’t care whether all of the variables have linear relationships.  It doesn’t matter.  We’re just look for rough approximations of the matchup at hand.

I’m excited about GILLESPIE.  I think it’s just one more tool in the arsenal to help you win your league this year.

Free Preview: Comparing Peyton Hillis and Steven Jackson

Peyton Hillis seems to give drafters the cold sweats when he’s still sitting out there in the third and fourth rounds of fantasy drafts.  All fantasy drafters have essentially assumed that Hillis can’t do what he did last year.  The only question then is how much falloff you will see year over year.

I have Hillis as a value pick because you’re getting a discount in order to take him at his current ADP.  So even if Hillis sees a production dropoff, he could still perform equal to his ADP.  But one thing I always like to do is compare my targets to other guys he was similar to the previous year, but who are going ahead of him in ADP.  For Hillis that guy is Steven Jackson.  Jackson is currently going 2 RB spots in front of Hillis.

Below are Hillis and Jackson’s stats for 2010.  Hillis finished with more total yards, more receptions, and 7 more touchdowns than Jackson.  Hillis did all of this with 40 less touches than Jackson.  Hillis is also 3 years younger than Jackson.  It seems like Hillis should be the one to be drafted higher right?

Games Rushing Receiving
Player Year Age Tm G GS Att Yds Y/A TD Y/G Rec Yds Y/R TD Y/G
Peyton Hillis 2010 24 CLE 16 14 270 1177 4.36 11 73.6 61 477 7.82 2 29.8
Steven Jackson 2010 27 STL 16 16 330 1241 3.76 6 77.6 46 383 8.33 0 23.9
Provided by Pro-Football-Reference.com: View Original Table
Generated 8/14/2011.

I look at these two backs and think that whatever risk there might be in Hillis breaking down (the most often cited knock against Hillis due to his finish last year) has to be equal or less than the risk that the 28 year old Jackson breaks down. 

Some have also pointed to Hillis’ touchdown numbers and predicted regression.  But the reason that Hillis had good touchdown numbers is that the guy is ridiculous in the red zone.  He led all running backs last year in TD Rate in from the 10 yard line and closer, and also averaged the highest yards per carry as a percent of the yards to goal.  Steven Jackson was one of the worst running backs in the league by these two measures.

Running Back Effectiveness Inside the 10 Yard Line

RB Carries TD YPC YTG TD Rate YPC as % of YTG
P.Hillis – CLE 19 9 3.11 4.26 0.47 0.73
A.Bradshaw – NYG 24 6 3.26 5.92 0.25 0.55
M.Tolbert – SD 30 11 2 4.07 0.37 0.49
K.Moreno – DEN 16 5 1.44 3.25 0.31 0.44
B.Green-Ellis – NE 25 9 1.96 4.48 0.36 0.44
A.Peterson – MIN 26 10 1.81 4.19 0.38 0.43
M.Bush – OAK 23 7 2.09 4.91 0.30 0.42
R.Rice – BAL 16 5 2.44 5.81 0.31 0.42
W.McGahee – BAL 16 4 1.94 4.63 0.25 0.42
T.Jones – KC 25 5 1.84 4.64 0.20 0.40
A.Foster – HOU 42 13 1.95 4.95 0.31 0.39
R.Mendenhall – PIT 28 11 1.36 3.68 0.39 0.37
B.Jacobs – NYG 17 7 1.65 4.65 0.41 0.35
M.Jones-Drew – JAC 26 4 1.88 5.65 0.15 0.33
M.Turner – ATL 44 11 1.50 4.70 0.25 0.32
L.McCoy – PHI 16 2 2.06 6.50 0.13 0.32
C.Johnson – TEN 24 6 1.17 3.71 0.25 0.31
M.Forte – CHI 22 2 1.59 5.45 0.09 0.29
C.Benson – CIN 33 6 1.31 4.94 0.18 0.27
C.Ivory – NO 16 4 1.25 5.50 0.25 0.23
M.Barber – DAL 20 3 0.70 3.55 0.15 0.20
S.Jackson – STL 24 3 0.92 5.58 0.13 0.16
M.Lynch – SEA 20 5 0.25 3.55 0.25 0.07

Maybe one thing that has the potential to derail the Hillis train this year is the change in offensive system.  But it’s interesting to note that Hillis will be in the same offense that Jackson was in last year.  Another interesting point is that both RBs will be in offenses led by 2nd year quarterbacks.

Perhaps another element going into the Hillis discount is the unknown situation with the running backs in Cleveland.  Montario Hardesty is returning from injury and Brandon Jackson is also in the mix now.  I sort of doubt that Hardesty’s ACL can stay in one piece for an entire season given the way he runs, and Jackson is just an inferior back to Hillis.  Jackson has averaged under 4 YPC for his entire career. 

Brandon Jackson Career Stats

Rushing Receiving
Year Age Tm G GS Att Yds TD Y/A Y/G Rec Yds Y/R TD Y/G
2007 22 GNB 11 3 75 267 1 3.6 24.3 16 130 8.1 0 11.8
2008 23 GNB 13 0 45 248 1 5.5 19.1 30 185 6.2 0 14.2
2009 24 GNB 12 0 37 111 2 3.0 9.3 21 187 8.9 1 15.6
2010 25 GNB 16 13 190 703 3 3.7 43.9 43 342 8.0 1 21.4
Career     52 16 347 1329 7 3.8 25.6 110 844 7.7 2 16.2
Provided by Pro-Football-Reference.com: View Original Table
Generated 8/15/2011.

It’s also worth mentioning that Saint Louis has brought in two free agents (Jerious Norwood and Cadillac Williams) to help out running the ball as well.  So Jackson’s carries might be under the same pressure that Hillis’ carries will be under.  I think it’s probably fair to say that Hillis and Jackson have similar risk due to competition.

Jackson is only going 2 RB spots in front of Hillis, so it’s not a big difference.  But I do think going through this analysis is helpful in trying to figure out when you’re picking Hillis, if you’re getting a big enough discount to take him.  Readers of this site will know that I believe all players have risk, it’s only whether or not we recognize that risk that is the difference among players.  So if I can draft guys whose risk has already been priced in to their draft spot, that’s the best I can hope for.

Running Backs and Risk of Lost Games

Earlier today I wrote that our perception of risk and the reality are sometimes far apart.  Trying to understand where risks lie is one of the reasons that I use historical comparisons, or similarity based projections, as one of my evaluation criteria.  If you’ve never read any of the historical comparisons, you can check out Chris Johnson’s write up for free right now. 

Basically I compare a guy to other RBs who put up similar stats at similar ages.  This helps us understand how guys who catch passes might differ from year to year when compared to guys who get most of their points from running the ball, or how guys who were touchdown heavy might differ from year to year.  Primarily I’m interested in what happens to a group of similar players in the year after they were similar to the subject player.  I call this Y2.

One of the things we can do with this analysis is to take a rough approximation of the risk of lost games in the following year.  Running backs can lose games, or not play, for a variety of reasons.  They might lose games to injury, or they might lose games due to lack of productivity.  So we’re going to look at last year’s RBs, the players they were similar to, and then how many games the similar players played in the year following. 

The table below shows average games played for the similar players both in Year1 and Year 2.  So to read the table, Steven Jackson similar players played in 15.6 games on average in the year they were similar to Jackson 2010.  Then in the following year they played in 11.7 games on average.  The Delta between Y1 and Y2 for the Jackson similar players was –3.9 games.

Similarity Based Risk of Lost Games for 2010 Running Backs

Player GP Y1 Av GP Y2 Dif Player GP Y1 Av GP Y2 Dif
Steven Jackson 15.6 11.7 -3.9 Ronnie Brown 14.4 7.2 -7.2
Arian Foster 15.6 13.6 -2 Darren McFadden 14.1 14 -0.1
Chris Johnson 15.5 11.4 -4.1 Fred Jackson 14 10 -4
Peyton Hillis 15.5 13.1 -2.4 Felix Jones 14 11.6 -2.5
Ray Rice 15.5 13.8 -1.7 Shonn Greene 14 9.3 -4.7
Cedric Benson 15.4 11 -4.5 Maurice Jones-Drew 13.8 12.8 -1
Rashard Mendenhall 15.4 11.8 -3.6 Tim Hightower 13.5 11.8 -1.8
Jamaal Charles 15.2 12.9 -2.3 Jonathan Stewart 13.4 10.4 -3
Matt Forte 15.2 14 -1.2 Thomas Jones 13.3 7.3 -6
Michael Turner 15.2 12.2 -3 Knowshon Moreno 13.1 11.6 -1.5
Ahmad Bradshaw 15.1 12.5 -2.6 Ricky Williams 13 5.8 -7.2
BenJarvus Green-Ellis 15 11.4 -3.7 Ryan Mathews 12.1 10.7 -1.4
Adrian Peterson 15 13.5 -1.5 Marshawn Lynch 12 7.3 -4.7
LeSean McCoy 14.6 13.7 -0.8 Frank Gore 12 9.7 -2.3
Brandon Jackson 14.5 7 -7.5 Jahvid Best 11.7 7.7 -4
LaDainian Tomlinson 14.5 8.7 -5.8 Chris Ivory 11.3 10.6 -0.7

Some notes based on going through the table:

  • You’ll notice that almost every player has a negative Delta when it comes to the difference between Y1 and Y2 games played.  The reason is simple.  No RB ever plays more than 16 games, so that’s a ceiling, but RBs do sometimes play zero games in the year following.
  • We can draw three primary conclusions from this graph (if we look at the additional underlying data).  The falloff in games played between Y1 and Y2 can be attributed to age, yards per carry, and receptions per game.
  • So if you’re creating the posterboy for an RB who is likely to see falloff in games played, he would be older, have a poor YPC average, and wouldn’t be a part of the passing offense.  The table below illustrates the differences in average for these stats if we break up the group of RBs into the 10 best in terms of year over year delta in games played, and the 10 worst year over year delta in games played.
  Age YPC Rec/G Avg Delta in Games Played
10 Best YOY Games Played Delta 23.7 5.0 2.4 -0.96
10 Worst YOY Games Played Delta 27.6 3.8 1.5 -6.09

Hopefully this analysis helps you understand which guys are risks to falloff and kill your teams, and which guys are less of a risk.  The analysis was extremely eye opening for me and there are a few guys who I will be adjusting my draft guide rating on now.

Be Careful if You Think You Understand All of the Risks

(Note: this post will not contain any “Adrian Peterson is a stud.  The guy is a beast!! I see 2000 yards and 20 touchdowns this year!!” – type fantasy analysis.  So if that is the sort you enjoy, please leave now.)

Nassim Taleb has essentially become the white angel on our shoulder when it comes to all things risk related.  If you haven’t heard of him, Taleb is the author of “The Black Swan: The Impact of the Highly Improbable”.  The book essentially popularizes Taleb’s theories on risk (don’t worry, I promise I’ll bring this back to fantasy football).  Taleb uses a string of analogies in the book to make the concept of risk more easily understandable.  One of those analogies is the Thanksgiving Turkey.  Taleb says that if you are a Thanksgiving Turkey, your view of your entire life is that things are pretty good.  Someone is feeding you every day.  You have no sense that there is any risk or danger in the world.  In fact, your confidence that no risk exists lasts all of the way up until your head is chopped off and you’re served for dinner.

Taleb’s point is that if you’re a turkey, getting your head chopped off and served for dinner is a low probability (though high impact) event.  You live your entire life with no measurable occurrences of risk.  Then you have one single high impact event that is impossible to predict or measure the probability of it happening.  Your perception of what risks exist for you and the reality of what risks exist for you are entirely different – if you’re a turkey.

I think when we try to predict the future in fantasy football we have a similar problem.  We can easily be lulled into thinking that no risks exist.  Take the case of Randy Moss.  Moss went from being the 2nd WR selected in fantasy drafts, a spot that many would characterize as being due to his extreme safety as a pick, to being droppable in leagues.  That happened in a mere 8 weeks of time.

I see the same issue with overconfidence every year when it comes to the top picks in fantasy football.  Every year the top guys are hyped due to their safety as picks, and every year we see some busts on what were supposedly safe picks.  Every year we see guys who were supposed to be extremely reliable, due to years of track records, end up not being reliable and killing our fantasy teams.  Our perception of where the risks lie, and the reality of where the risks lie, are often separated by great distances.

Moss’ falloff is a great example because receivers typically hone their craft over years, getting better at things like running routes, using their bodies to shield the ball from defenders, and reading defenses.  So they get better each year until late into their careers.  They get better until their bodies stop cooperating and then they seem to fall off a cliff.  It happened with Marvin Harrison.  It happened with Torry Holt.  Now it’s happened with Randy Moss.

I have some additional thoughts on this topic that I’ll be expanding to other players, but first I wanted to get out there the notion that our perception of risk is not necessarily consistent with what the risks really are.  Don’t be so quick to say a guy is “safe” because safety is a more elusive concept that you realize.

Draft Guide Preview: Peyton Manning

Peyton Manning is a good example of where the Draft Guide can offer us a little bit of a reality check. While Manning is being taken as the 3rd QB in ADP right now, I have him listed as a Stay Away.  Here are the broad strokes:

  • Manning’s Yards/Attempt have been in fairly steady decline for the better part of 5 years (see the career trajectory graphs).  So he has to throw more passes to compile the same stats.  While that has happened his INT/G have gone up, reducing the efficiency of the IND offense.
  • Manning’s neck situation is not currently priced in to his ADP.  With any risk we can always ask ourselves if it’s been priced in to what we have to pay.  In Manning’s case, it hasn’t.
  • If you adjust Manning’s Yards/Attempt for the defenses he faced last year he looks even worse.
  • My Similarity Based Projections have 6 other QBs in front of Manning.  That list includes Drew Brees, who had a statistical season that was very similar to Manning last year, except that Brees is 3 years younger and probably is going to be in the more efficient offense this year.

If you want to take your chances by paying this much for a 35 year old QB who just had neck surgery, that’s up to you.  I won’t be.

Manning Draft Guide

Say What You Will About the Tenets of National Socialism, Dude, at Least it’s an Ethos – A Look at the SOS Nihilists

The Strength of Schedule Nihilists are killing me.  If you don’t remember from the Big Lebowski, nihilists care about nothing.

The Strength of Schedule Nihilists, like Chris Harris at ESPN, don’t believe there’s any meaning in SOS information.  Here’s the basic argument of the SOS Nihilists: There is variance in defensive stats each year and therefore we will not consider SOS in our fantasy rankings.

The Nihilists argument is odd to me on a number of levels.  (Note, this post will be heavy on theory and light on application, so if you aren’t into the theory, you can stop reading.  I’ll get into application of our model in future posts)

First, as Bill James has said, I am reluctant to dismiss as randomness something that I might simply not understand.  The Nihilists are not willing to admit that they simply might not understand how to create projections for defenses.  Instead they say “Hey, look at Defense X which finished the season as a Top 5 defense one year and then dropped out of the Top 15 the next year, we can’t count SOS in our rankings.”  This isn’t any different than saying that when I play blackjack, I often bust when I take a hit on 16, therefore I will no longer take hits on 16.

Second, the Nihilists either purposely, or because it has not occurred to them, leave out the possibility that you could adjust the prior year defensive numbers for schedule, and that doing so could aid in reducing forecast error.

Third, the Nihilists create a straw man by arguing that those in favor of using SOS want to use the prior year’s defensive numbers to forecast the entire season.  I use SOS in my model, but I only take it out to Week 6.  I do this for a few reasons.  First, because I want to avoid early season busts as their trade value will be dragged down the entire year.  Second, because I believe that if I avoid early season busts and target players who have easier schedules, I give myself the best chance of having trade chips midway through the season.  Third, I am going to have better data each consecutive week to put into my model, so why take it out past Week 6 when I’m going to throw that old data out anyway.

The test of models that include SOS is not whether individual defenses change year to year.  The test is what are the chances that a schedule that looks difficult to start the year turns out to be easy, and vice versa.  Maybe a more intelligent way to look at that same problem is this: If we introduce SOS data into our model, does it improve forecast errors?  Note that in Harris’ piece he never compares projections arrived at using SOS against projections not using SOS in order to gauge the effectiveness of each model.

There is an additional reason why it is a bad faith argument to suggest that SOS projections fail because they do not accurately project defensive finishes across an entire season.  Defenses change during the course of a year.  A defense might start out on pace to finish where it did the previous season and then a key injury (like Troy Polamalu) might hurt the defense.  So forecasting the impact of SOS early in the season, using the best data available to us at the time, is not as dangerous as projecting out the entire season before any games have been played.

There are a number of improvements you can make to the stats which are being dismissed by the Nihilists.  Let’s think about those improvements.  We can use Similarity Based Projections in order to compare the defenses to other defenses over a 20 year period that performed similarly (I will do exactly that in future posts).  For instance, if a defense gave up a lot of rushing yards and not a lot of rushing touchdowns, it’s likely that if we compare that defense against historically similar defenses, we’ll see those touchdowns come up in the following year.  In addition, we can as I mention above, adjust our defenses for quality of opponents faced.  We can also adjust our offensive players for quality of opponents faced.

The primary improvement that I mention above, adjusting both defense and position players for prior year quality of opponent, can be backtested against prior year’s data.  I’ve done that and found that introducing those improvements to the model improves mean absolute forecast errors when compared against the generic method of forecasting, which is to simply use the position player’s prior year average.

Here’s the ironic thing.  A lot of the SOS Nihilists have no problem using prior year finishes for position players when creating their rankings, despite the turnover in rankings at those positions.  Actually, many of the Nihilists don’t just use the prior year finishes, they almost use an average of position players’ prior two or three years finishes.  Somehow, when applied to position players, the variance in year to year performance can magically be ignored.

Don’t Overpay for Gronkowski’s Touchdowns

Rob Gronkowski is currently coming off the draft board about 11 TEs in front of Aaron Hernandez.  This is a great example of an opportunity to take advantage of fantasy players focused on the wrong things.  Gronk had 10 TDs last year to Hernandez’s 6 TDs.  However, they were otherwise statistically very similar.  In fact, in a PPR league Hernandez would have actually averaged more points/game than Gronk despite the TD difference.

Games Receiving
Rk Player Year Age Draft Tm Lg G GS Rec Yds Y/R TD Y/G
1 Aaron Hernandez 2010 21 4-113 NWE NFL 14 7 45 563 12.51 6 40.2
2 Rob Gronkowski 2010 21 2-42 NWE NFL 16 11 42 546 13.00 10 34.1
Provided by Pro-Football-Reference.com: View Original Table
Generated 7/21/2011.

Using historical comparisons allows us to see how different stats tend to hold up year over year.  Let’s look at guys who were similar to the rookie campaigns for Gronk and Hernandez.  However, for the sake of brevity and in order to not spill too many pixels on TEs, I’m only going to look at Y2, or the year after they were similar to our subject TEs.

Primarily we’re interested in whether Gronkowski similar players caught more TDs in Y2 than did Hernandez similar players.

Gronkowski Similar Players – Year 2

Player Year G Rec/G Yds/G TD/G FP/G S FP/G PPR
Eric Green 1991 11.0 3.7 52.9 0.5 8.6 12.3
Ben Coates 1994 16.0 6.0 73.4 0.4 10.0 16.0
Anthony Fasano 2009 14.0 2.2 24.2 0.1 3.3 5.5
Heath Miller 2006 16.0 2.1 24.6 0.3 4.3 6.5
Rickey Dudley 2000 16.0 1.8 21.9 0.3 3.7 5.5
Heath Miller 2008 14.0 3.4 36.7 0.2 5.0 8.4
Vernon Davis 2007 14.0 3.7 36.4 0.3 5.4 9.1
Cam Cleeland 1999 11.0 2.4 29.5 0.1 3.5 5.9
Johnny Mitchell 1994 16.0 3.6 46.8 0.3 6.2 9.8
Fred Davis 2010 16.0 1.3 19.8 0.2 3.1 4.4
Randy McMichael 2003 16.0 3.1 37.4 0.1 4.5 7.6
Bubba Franks 2002 16.0 3.4 27.6 0.4 5.4 8.8
Kevin Boss 2010 15.0 2.3 35.4 0.3 5.5 7.9
Daniel Graham 2005 11.0 1.5 21.4 0.3 3.8 5.2
John Carlson 2010 15.0 2.1 21.2 0.1 2.5 4.6
Kevin Boss 2009 15.0 2.8 37.8 0.3 5.8 8.6
Kerry Cash 1993 16.0 2.7 25.1 0.2 3.6 6.3
          Average 4.9 7.8

Hernandez Similar Players – Year 2

Player Year G Rec/G Yds/G TD/G FP/G S FP/G PPR
Cam Cleeland 1999 11.0 2.4 29.5 0.1 3.5 5.9
Ben Coates 1994 16.0 6.0 73.4 0.4 10.0 16.0
Jermichael Finley 2010 5.0 4.2 60.2 0.2 7.2 11.4
Chris Cooley 2007 16.0 4.1 49.1 0.5 7.9 12.0
Mark Chmura 1996 13.0 2.2 28.5 - 2.8 5.0
Heath Miller 2008 14.0 3.4 36.7 0.2 5.0 8.4
Todd Heap 2003 16.0 3.6 43.3 0.2 5.5 9.0
John Carlson 2010 15.0 2.1 21.2 0.1 2.5 4.6
Johnny Mitchell 1994 16.0 3.6 46.8 0.3 6.2 9.8
Fred Davis 2010 16.0 1.3 19.8 0.2 3.1 4.4
John Carlson 2009 16.0 3.2 35.9 0.4 6.2 9.4
Greg Olsen 2009 16.0 3.8 38.3 0.5 6.8 10.6
Jason Witten 2006 16.0 4.0 47.1 0.1 5.1 9.1
Keith Jackson 1991 16.0 3.0 35.6 0.3 5.4 8.4
Freddie Jones 1998 16.0 3.6 37.6 0.2 4.9 8.5
Heath Miller 2006 16.0 2.1 24.6 0.3 4.3 6.5
Johnny Mitchell 1995 12.0 3.8 41.4 0.4 6.6 10.4
          Average 5.5 8.8

First, I should probably point out that these TEs are very close to each other.  But if we’re splitting hairs here we see that the Hernandez similar players continued to catch balls and accrue yards as a slightly better rate than the Gronkowski players.  The Gronkowski similar players saw falloff in TD numbers. 

In Y2 the Hernandez similar players were better in both PPR and standard formats.

Adding to the head scratching nature of this situation is the fact that Hernandez was targeted more than Gronk last year (64 targets to 59 for Gronk) even while appearing in fewer games.

So if you take Rob Gronkowski as the TE11, be warned that there is a TE sitting there, who will probably go off about 10 spots later, who has about equal chance for success this year.

Michael Vick and First Round Strategy

I had a brief exchange on Twitter with a guy who asked whether or not I thought taking Mike Vick with the 6th pick in the first round of a redraft league was a good idea.  I don’t think it’s a terrible idea, but I also know that I won’t probably be a Vick owner this year unless he somehow becomes oversold and represents some kind of value (seems unlikely right?).

I actually think Vick has really high odds of finishing the season as the top QB.  Because QBs can get a point for 10 yards rushing, along with their throwing points, running QBs are an often overlooked source of value.  Vick has some risk due to injury, but that’s also not why I won’t be a Vick owner this year.

I won’t be a Vick owner because if I take Vick in the first round as the top QB, then I’ve paid a #1QB price for the #1QB.  But I would rather do something like pay a #9QB price for the #2QB.  Or pay #20QB price for the #7QB. 

The key is that I want a discount.

If I take that thought to its conclusion, it surfaces another reason why taking Vick in the first round isn’t what I want to do.  What if there are RBs available who are offered at a discount to their value?  What if I can get the #2RB for #9RB price.  Or even if I can get the #2RB for #6RB price.  If I take Darren McFadden at either of those spots, that’s possible. 

I can’t do that if I take Vick, and it won’t make sense to make my value QB pick later either.  What would I do with that value QB if I have Mike Vick?

So if I abstain from taking Vick, I can pick for value at RB in the first round, and then pick for value at QB in a much later round.  If I take Vick, I don’t get any value picks.  I just get one guy who I paid fair price for.  Again, I’m passing on Vick even though I think he’s going to be good again.  I’m passing because I want to find value guys at every single draft spot and I can’t do that if I start the draft by paying fair value.

Why You Have to Draft for Depth

A quick rundown of the RBs who will be available to be your RB1 shows that none are without risk.  To illustrate this point I’ve annotated a table of the top 12 RBs coming off the board with what I think their risks are.

ADP Name Risk
1 Adrian Peterson Offense
2 Arian Foster Usage
3 Chris Johnson Offense
4 Jamaal Charles Injury/Usage/Schedule
5 Ray Rice Schedule
6 LeSean McCoy  
7 Rashard Mendenhall Talent
8 Maurice Jones-Drew Age/Injury/Wear and Tear
9 Michael Turner Age
10 Darren McFadden Injury/One Hit Wonder
11 Frank Gore Age/Injury/Wear and Tear
12 Steven Jackson Age

Some might look at that and say “Well, that’s a good argument to take Vick as my first round pick” or “Fine, I’ll go back to back WR in the first round”.  But I think that’s a recipe for disaster. 

You need guys on your roster who are capable of performing at an RB1 level.  Actually, that’s not right.  You need MULTIPLE guys who are capable of performing at RB1 level.  The only way you’re going to ensure that you have those guys is by taking RBs early and often.  Think of it like each RB you take has a reduced chance of performing at an RB1 level.  Maybe if you wanted to put those chances into a table it would look like this:

RB Drafted % Chance of Performing at RB1 Level
1 60%
2 20%
3 10%
4 5%
5 2%

Let’s think about where those RBs were picked last year.  In the table below, each RB has their ADP spot as a reflection of which RB they would have been on a standard team in a 12 team league.  So if they have a 1, they were drafted in the top 12.  If they have a 2, they were 13-24, etc.

ADP Name 2010 RB#
1 Adrian Peterson 1
2 Arian Foster 3
3 Chris Johnson 1
4 Jamaal Charles 2
5 Ray Rice 1
6 LeSean McCoy 2
7 Rashard Mendenhall 1
8 Maurice Jones-Drew 1
9 Michael Turner 1
10 Darren McFadden 4
11 Frank Gore 1
12 Steven Jackson 1

So despite our desire that all RB1s that we draft turn out to be RB1s at the end of the year, it’s just not going to happen.  Some of the RB1s will underperform and other guys we never expected will emerge.

This is why we need to stockpile those RBs.  It’s in order to give ourselves the best chances of having legitimate #1 options.  When we evaluate options, we should be thinking about whether we can draft a guy as our RB3 when he has some realistic chance of performing as an RB1.

So if you really believe in Chris Johnson, Adrian Peterson, or any of the other question marks this year, that’s fine.  But cover your downside by taking options lower in the draft who have the potential to perform as your #1 if your first round pick misses.

Too many fantasy owners try to pick the perfect team.  They try to pick the perfect QB, the perfect tight end, the perfect defense.  But in doing so they leave themselves vulnerable to the unexpected.  When you draft your RB2, pretend as if you need him to be your RB1.  When you draft your RB3, pretend that you don’t have an RB2.  If you do that you’ll have a robust team that will withstand bye weeks and injuries and you’ll be right there at the end of the season.