Monte Carlo Strategies to Win 2016 MFL10s – Part I
Editor’s note: This is one of two Monte Carlo simulation articles aimed at solving the best-ball puzzle, each using different assumptions. We believe doing so gives a good idea of the range of possible outcomes. For the other article by Nick Giffen (@RotoDoc), click here.
You MFL vets out there might remember the Monte Carlo article that your humble authors wrote a few years back. Some of you took our advice and increased your winnings by double digits. Kevin Cole wrote these very kind words about our Early-RB Strategy that we advocated for in our original 2014 article:
Around mid-way through my MFL10 drafting, I switched from a 4 RB allocation with one or two taken in the first five rounds, to much earlier RB drafting with three to four taken in the first 5 rounds. I finally “bought in” on the philosophy presented in the excellent Monte Carlo analysis by AJ Bessette and Greg Meade. The improvement in results was dramatic. Luckily, my switch in strategy coincided with a switch to MFL50s & 100s, which ended up turning my losses into a 25 percent gain.
While we try not to strain our arms patting ourselves on the back, the reality is that the 2016 NFL year will be much different from the 2014 one. The Golden Age of Passing continues to overwhelm both real life and fantasy football, and past seasons are dropping further and further out of sample.
Our original sample of 2011-2013 isn’t recent enough to use for decision making this year, so we dug into our Monte Carlo model to update our analysis. After figuring out what the hell we were doing in 2014 (comment your code people), we made a few improvements to develop a new pick order for 2016.
We used a similar Monte Carlo model to our 2014 version, with the big change being using 2014-2015 data instead of 2011-2013. If you care about the specifics of what a Monte Carlo analysis is, check out our old article where we describe it in depth. Tl;dr, it simulates weekly scores based on the positions you pick at each ADP to figure out what positional allocation is optimal.
We also put in some adjustments for injuries which our original article didn’t account for. We looked at historical injury rates by position as well as some third party injury predictions to come up with a likelihood of missing each game by position. This was added into the simulation so that players took zeroes for bye weeks and injuries the correct proportion of weeks. This was modeled in separately from ADP since injury likelihood and ADP have no reason to be correlated, so we assumed the same rate of games missed by position regardless of where they were selected.
The first run of our sim spit out the following as the optimal best ball draft allocation for MFL leagues:
WR WR WR TE WR TE WR RB QB RB QB WR RB RB QB WR TE DEF DEF DEF
Alright, so we’ve got this sexy new model with new data, and it says to go heavy WR early. This makes sense due to the league passing so much…
Dumber analysts (read: Greg) would have taken the output from the model as the big conclusion, but luckily, smarter analysts (read: AJ) realized that our methodology only works if positional ADP stays similar year to year. And you’ve guessed it, that is not the case here.
A cursory glance at the 2016 ADP data will show you that WRs are going earlier than ever. This makes sense, but we believe the market has actually overcorrected, as we’ll show throughout this piece. First, let’s examine just how much ADP is changing. This chart is the first 4 Rounds of ADP.
WRs are being taken much earlier and in higher numbers than in 2014-15. The opposite is true for the RB position, fewer early RBs and fewer RBs overall. This ADP shift means that our simple conclusion to “take lots of WRs really early” ends up needing more nuance. Luckily, this type of opportunity cost analysis is a perfect problem for our corrected model to solve. So what’s the optimal allocation if you adjust to 2016 ADPs?
If you pick 1st-4th: WR RB RB TE RB WR WR TE QB WR WR QB RB WR QB WR WR DEF DEF DEF
If you pick 5th-12th: TE RB RB TE WR WR WR WR WR RB RB QB QB WR QB WR WR DEF DEF DEF
The Pretty Pictures
So why do we believe the optimal strategy shifted? Let’s go position by position to break down the new data compared to the old. Remember, “Old” is 2011-13, “New” is 2014-15.
This is the best example of the ADP overcorrection, so we’re going to show you our original data and the 2016 corrected version. All graphs after this will only show the corrected 2016 ADP version.
In the uncorrected chart, recent WRs are better values at every single draft position than their past counterparts. But in the corrected data, we realize that WRs go so early that they don’t provide the value bump we’d otherwise expect. Instead, the value improvement only comes with greater depth in the later rounds.
Does this mean wait on WRs? It’s more nuanced than that. The top WRs are still worth a first round pick. If you’re drafting 1-4, the simulation likes taking the big three of Antonio Brown, Odell Beckham, Jr., and Julio Jones as the top picks. It also likes picking a WR fourth based on the value the WR4 by ADP usually generates (we personally like DeAndre Hopkins, but drafting fourth is tough). If you’re drafting fifth or later, the simulation likes Rob Gronkowski to provide better value, essentially saying he’s more valuable than the WR5 by ADP.
Recommendation: If you’re drafting 1-3, take one of the big three. If you’re drafting 5-12, take Gronk. If you can’t get Gronk, take the best value you can. Our sim prefers RBs in the middle to end of the first round at current ADPs. The sim recommends grabbing 7-8 total WRs, with most coming in the middle to late rounds.
As Biggie once told us: mo’ passing, mo’ tight end production. Not only are TEs more productive than ever, the top TEs are big time values. The data strongly suggests that if you can’t get a top four WR, Gronk is your best pick in the first round. No matter when you’re drafting, Jordan Reed in the fourth is a universal value in the sim. The second tier of TEs behind Gronk compare favorably to WRs straight up when assuming new ADPs, and the sim likes taking a second one even if you already have Gronk to be your flex. After the second tier of TEs (Reed, Greg Olsen, Travis Kelce, and Coby Fleener this year) WRs become more valuable again due to the depth at the position. Obviously, if you think this is the year of the TE and more might break out (we tend to lean this way), picking mid-round TEs as a flex is a great value.
Recommendation: Don’t take Gronk first overall, but if you’re drafting later in the first round he’s the best value you can get. Either way, take Reed in Round 4 if he’s there or one of the other tier two TEs in that range. The simulation prefers getting two good TEs and never picking another since you can’t use all three in the same week and late round WRs provide better value.
Quarterbacks are deeper than ever before, which only improves the Late Round Quarterback strategy recommended in our original article. This is helpful as we can grab the higher upside position players early before they drop off in quality.
Recommendation: Rounds 9-15 are the sweet spot rounds for QB. Take three total.
If you believe the above, then the demise of the RB has been overstated. The upside of top backs (Le’Veon, David Johnson, etc.) who get both goal line work and passing game duties remains similar to prior years when those backs are healthy. The position is also deeper with the amount of committees and pass catching backs which have standalone value in the PPR format. Players like Danny Woodhead used to be nigh undraftable, but now regularly go as early as round 4 due to the explosion of passing games. To some degree, this added depth makes grabbing early RBs less important than it used to be. RBs still don’t score as much as WRs, so you have to believe the gap between the RB points and the equivalent WR you could select is the smallest it will be throughout the draft to pull the trigger. Our sim says that RBs come closest to matching their WR counterparts in the early rounds, so that’s where you should take them.
Regardless of where you draft them, we still recommend you only grab four. Otherwise you’d be allocating too much draft capital to a low scoring position.
Recommendation: Take multiple RBs in the early rounds to take advantage of ADP overcorrection. Take four total.
Defenses score less than before due to the explosion of offense. The strategy hasn’t changed since the variance is the same and team defense doesn’t compete for use in your flex spot. Grab 3 to take advantage of the variance and find Defensive TDs.
Recommendation: Take advantage of the relatively flat points curve and wait as late as you can to grab three reasonable defensive units
How does this all come together?
NFL offenses score more points and gain more yards than ever before, with a higher portion of the offense being passing plays. This has created more depth at nearly every position, particularly in PPR, so positional scarcity isn’t nearly the issue it used to be. Drafters took notice of this and have been drafting WRs early and often since they outscore RBs regularly in the new NFL. This ADP over-correction seems to open the door for early RB to return to dominance (with some better injury luck than last year required for success).
That said, the margin of error is small in this analysis. For the most part, your competitors are drafting very rationally. If you believe the league will continue to become more pass-happy than the ’14-’15 seasons, then early WR may still be optimal. We believe early RB and TE to be the smart contrarian choice in drafts, but the market seems to be more efficient than two years ago. The biggest edge to be had is probably from drafting three starting QBs and three defensive units late, as that inefficiency still doesn’t seem to have been priced out yet.