Exploiting Roster Construction and ADP in MFL10s
Is there a way to beat MFL10s without relying on player evaluation and nailing late-round sleepers? I think there is. Over the past few months, I have analyzed data from every MFL10 roster drafted in the past three years (over 52,000 teams across 4,350 leagues) in an effort to answer that question. Ultimately, there is no silver bullet, and each year is unique in its own way. This year, trends in ADP have opened the door for a draft strategy which, through deliberate roster construction, should generate a strong return on investment for your best-ball portfolio. This article outlines that strategy.
Running Back ADP Dislocation
Thanks to changes in ADP, today, for the first time, I can confidently go into an MFL10 draft with a plan to secure a top-5, a top-10 and a top-15 RB with my first three picks – but the simple fact that I can doesn’t necessarily mean that I should.
Lower price does not always equal greater value, and there is logic behind the recent drop in running back ADP which cannot be ignored. Fantasy scoring has been on the decline for top tier running backs for several years, while wide receiver scoring has crept higher.
However, current pricing seems to imply that 2015’s results were the continuation of a trend rather than an outlier event. The chart below illustrates both that the trend of declining scoring from early-round running backs is real, and that the 2015 decline was an exceptional event, beyond anything we have seen in years.
A rebound for elite running back production in 2016 feels like one of the safest bets out there. And, even better, in a best-ball format we aren’t even fighting a negative trend with the top RBs (that nasty 2015 aside).
The chart below shows the total points an MFL10 roster’s two starting RB spots would have accumulated from three running backs with a positional ADP in the top 15 in each of the past five years. To construct the curve for a given year, I generated the total points contributed to two MFL10 starting RB spots for each possible combination of one top-5 RB, one RB6-RB10, and one RB11-RB15 by ADP (125 combos per year) – using each player’s actual weekly performance. The curve represents the probability of getting a given number of points if selecting one of those 125 combinations at random. So, the more area under the curve to the right on the x-axis (Points), the better.
Each year between 2011-2014, the median trio produced 525-550 points between two starting RB slots. In 2015 that number was over 100 points lower – a dramatic outlier.
Our Objective: Score at Least 2,500 points
While the source of points by position has varied, the overall output of winning MFL10 rosters has not changed much over the past few years. Based on the table below, which removes kicker points from 2013 and 2014, it looks like we are trying to hit at least 2,500 total points to win an MFL10. And make no mistake, winning is the objective – second place gets a free MFL10 entry next year, but these are effectively winner-take-all contests.
There is no benefit to generating a relatively high average score across your teams if none of them are scoring enough to take first place. We do not gain any benefit from a high floor strategy here.
Over the past few years, the Zero-RB, antifragile, draft strategy has deservedly grown in popularity, accelerated by a perfect storm of running back flops in 2015. I agree with the tenets of the anti-fragile strategy in most formats, but in a winner-take-all best-ball environment, I am running toward discounted fragility, not away from it. I want to take advantage of the current dislocation in pricing for our “fragile” elite running backs. But that is not as easy as it might sound. Just loading up on RBs without regard for the rest of the roster would be a mistake.
It does not feel good to be “right” on a bet, but still lose because you are exposed to other factors that went against you. In this case, while the bet is on running backs, those factors are the other players on your team. Fortunately, the best-ball format allows us to diversify away a lot of that risk and put our fate almost entirely in the hands of the RBs.
The following outlines a framework which not only should yield a strong ROI for your best-ball portfolio, but is also extremely easy to implement.
Hyper-Fragile Roster Construction
As might be obvious by now, drafting a high upside, fragile team starts with selection of a running back in each of the first three rounds (all of whom are likely within the first 15 RBs off the board). Since we want the success of those players to drive the overall success of the team, the goal with the remainder of the roster is to build a high-average, low-variance range of outcomes to which our RB points will be added. Here’s how we do it, working backwards from the end of the draft.
Note: Going forward, I reference data from both 2014 and 2015, so we can see how the strategy fares in both a more normal RB environment and in an RB-apocalyptic environment.
Three Defenses, Rounds 18-20
By the time the 18th round comes along, most of the quality skill position players have been snatched up by your opponents (and hopefully by you as well), but there remains a source of real points that will contribute to your team’s bottom line. By going with three defenses instead of two, we can use the earlier picks that are required for a “top-5” defense on more valuable skill-position players.
The chart below illustrates what we get by opting for three late defenses. The red and orange dotted lines show the distribution of points that actual MFL10 teams got from defenses in 2014 and 2015 (there is not much difference between the two). The blue and green solid lines show the distribution of points generated by all possible combinations of three defenses, excluding the top-5 in ADP, for their respective years (1540 combinations). In other words, the blue and green distributions show what you’d expect to get when selecting a random trio of later defenses. We can see the median scores for the three-defense combinations are significantly higher than the median team generated in their respective years.
A three defense approach is also supported by the results of existing Monte Carlo simulations.
So far, so good. We have generated a higher average and lower variance profile than the general population for our expected defense production.
Three Quarterbacks, Rounds 12-17
The same logic that we used for defenses also holds for quarterbacks. Rather than spend an early pick on one of the top 10 QBs, and therefore need to be “right” on the pick in order to get what we paid for, we can pull a trio of quarterbacks together later in the draft and get the same, or likely better results. The chart below is in the same format as the one we looked at for defenses. This time we use quarterbacks with ADP between QB11 and QB26 (those roughly going in Rounds 12-17 today).
The results are not quite as compelling here as they were for the three defense model, as 2014’s curves look very similar to each other, but the 2015 three-quarterback distribution does dominate the field. In both cases, we have again achieved the goal of a relatively high average, low variance expected return from the position by dictating how many we pick, and in which part of the draft we take them.
Two Tight Ends, Rounds 1 – 7* and Rounds 15 – 17
The production offered by top tight ends is so compelling that RotoViz’s RotoDoc has recommended taking two TE1s in MFL10s. Here, we select only one top-6 tight end and conserve the other early pick, opting instead for a late round selection in the TE21-TE26 range (rounds 15-17 by current ADP). The chart below shows the points generated for the TE starting position from all such combinations versus what actual drafters were able to get out of their tight ends.
This is where we give some points back. Taking two for this onesie position, instead of three, reduces our expected return and adds some variance to the strategy, but freeing up a spot for an additional WR proves worthwhile, as you’ll see below.
It is also important to note that in this case the comparison is not apples-to-apples, as the “Actual” numbers include tight end points from the Flex position, while the simulated numbers do not – this handicaps the simulated scores, though not to a great degree.
*With a late pick, one can typically take Rob Gronkowski in the first round, or reach for Jordan Reed at the end of the third, and still get a top-15 RB in the early fourth round.
Nine Wide Receivers, Rounds 4-17
Having accounted for 11 roster spots across the other positions so far (3QB, 3RB, 3DEF and 2TE), the remaining nine slots are filled in with WRs. Since we are taking running backs in each of the first three rounds, I assume here that we cannot get a top-24 WR. Without access to those elite options, we compensate through volume. To assess a realistic application of this strategy, I looked at combinations of WRs selected from nine groups of five, ordered by ADP between WR25 and WR65 – that way I excluded combinations such as one including all nine of WR25-WR33, which would be impossible to acquire in a real draft. There are 1,953,125 such combinations. For this analysis, I included the score of the fourth best WR each week as the combination’s “Flex” contribution.1
Again, we have generated a range of scores with a higher average and lower variance than what actual teams got from the position in each of the past two years. So what do we get when we put it all together, excluding RBs?
Through a diversified approach, we have created a high baseline for our roster. Now we just need to bring in the running backs to finish the job.
The Running Backs
You may have surmised from the beginning of this article that 2015 was not a good year for this strategy. That’s evident in the running back version of our distribution chart.
So, 2015 = not pretty. No surprise there. I am more focused on what happened in 2014, which is more in line with what I expect from 2016. For 2014, our three top-15 running back strategy generates expected production in line with what actual MFL10 teams were able to get from their running backs that year. That’s despite the simulations not getting any credit for Flex points (as was the case with the TE charts).
Pulling it All Together
Now that you have gone through the arduous process of ingesting my charts, here’s our payoff. By combining the RB chart with our QB+WR+TE+DEF baseline, we can get a rough idea of how the strategy would have performed, and estimate a theoretical ROI. It’s important to remember that these are conservative calculations, given that our simulations assume zero contribution to the Flex position from running backs or tight ends (an extremely unlikely outcome, particularly for the best lineups). Let’s get the ugly business of 2015 out of the way first.
Even with the exceptionally poor running back year in 2015, our strategy ends up with a higher projected average than that of real teams last year, but unfortunately we don’t care about average here – we only care about wins. The right “tail” of the blue distribution does not stretch very far to the right. In fact, it breaks our 2,500 point target only two percent of the time. Using 2,500 as a proxy for a “win,” the ROI on this strategy would have been a horrific -80 percent. Fortunately, it’s not 2015 – we just have 2015 to thank for the opportunity in front of us. Now let’s take a look at 2014.
That’s more like it! In a 2014 environment, a whopping 20 percent of lineups would have broken the 2,500 point threshold. That’s a +100 percent ROI! Anything approaching 2014 RB results (which, as a reminder, very closely resemble each of the preceding three years), will give us a healthy return on investment.
It’s important to recognize the downside to this approach – 2015 could happen again, and in that case a portfolio of MFL10 rosters constructed this way would get crushed. A return to 2011 – 2014 results is also not a guarantee that you will double your money, but I believe that current ADP has presented a tremendous opportunity to profit from overreaction to one outlier year.
This 3QB/3RB/9WR/2TE/3DEF approach to roster construction stacks the deck in your favor. If this article were not already over 2,000 words, I would show how even the mid-point between 2015 and 2014 performance would generate a strong expected ROI. Maybe next time. Until then, here is an example of an actual 2016 MFL10 team I drafted using this approach, to give you an idea of how it shakes out. Good luck and happy drafting!
Special thanks to Jim Kloet for invaluable technical assistance on this project.
- In other words, we’re assuming a WR will be the Flex every week. (back)