Draft Strategy

Monte Carlo Strategies to Win 2016 MFL10s – Part II

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 A.J. Bessette and Greg Meade, click here.

Almost exactly two years ago, A.J. Bessette and Greg Meade wrote a great piece on using Monte Carlo simulation to solve the best-ball puzzle, aimed at helping you win your MFL10 leagues. There’s no doubt their groundbreaking work led to the popular, and successful, RB-heavy approach employed by many top best-ball drafters in 2014. In fact, their recommended first four round start of RB TE RB RB was, on averagethe 13th best start among 1296 possible combinations of first four round starts.

I italicized on average for a reason. To win an MFL10 you have to place first out of 12 teams, so I tend to look at upside, rather than averages. There’s nothing wrong with averages, but I also am a GPP-oriented DFS player, so upside tends to be my personal focus.

So I’ve decided, rather than giving a roster combination which will — under the given assumptions — give you the best team on average, instead I’ll look at the top scoring rosters among each of the individual simulations to give you an idea of which roster combinations have the most upside, and how often they fall within that upside range.

Assumptions

Like A.J. and Greg, I’ll be using 2014-2015 data (I will also have future articles with Monte Carlo simulations using data from multiple time frames to get a better idea of what the upside combinations are depending on what type of output we might see). This allows us to capture the changing landscape of positional output in the NFL. I am also using the current MFL10 format, which is different than the format when they ran their original article in 2014.

Also like A.J. and Greg (and full credit to them for suggesting this), I’ve adjusted positional ADP to the current 2016 overall ADP, because WRs are now going earlier than their RB counterparts, and QBs are going later than ever. In other words, the QB1 in 2015 went with a 19.54 overall ADP. So if I simulated based off overall ADP, it would treat this year’s QB1, Cam Newton, as approximately the QB3, diminishing his statistical output compared to what it would be historically for the QB1. But we know that’s not the case. Making the adjustment allows us to appropriately account for 2016 ADP.

I also incorporated zeroes for bye weeks and gave zeroes for missed games, which I assigned proportionally based on historic missed game rates for each position, whether due to injury, suspension, or benching. I found that missed games does correlate with positional ADP, but a lot of that is likely due to the fact that players with later ADPs may not have posted stats because they were more likely to be replaced on the depth chart, or they were backups to begin with. I will have more on missed games in a future article.

The other, and most major difference from their prior work, is that I used a Bayesian version of Monte Carlo simulation to simulate the uncertainty in the parameters for the optimal fit to the data. I also allowed for the code to identify the appropriate distribution at each position when sampling weekly points by positional ADP. This lets the simulation pick up the appropriate upside by position. For example, here are the points scored above and below positional ADP expectation for QBs (on the left) and RBs (on the right), drafted in rounds 1-4 the last two years.

QBRB

Methodology

For the sake of time and computational power, I chose to start drafts from the third spot, which is typically the end of the WR trio of Antonio Brown, Julio Jones, and Odell Beckham, Jr., and from the 1.09 spot (for no particular reason other than to have a late draft spot).

Finally, I am looking at upside. To do so, I am taking the top 8.33 percent of all my individual simulations, and looking at the roster combinations that made up those simulations. The reason I chose 8.33 percent is because that is equal to 1/12 — which is where you need to finish in MFL10s to win. Anything else doesn’t matter.1 In other words, I’m not looking at what’s happening on average. Instead, I’m looking at what individual data points put me at the upper tail of a distribution. Here are the results.

Draft Results From 1.03

Here are the top 25 starting draft combinations for three, four, and five rounds that show up in the top 8.33 percent of all simulations, along with their frequency of doing so, based off a 1.03 draft position:

Start_3 Freq Start_4 Freq Start_5 Freq
WR WR RB 8.5% WR WR RB TE 2.6% WR WR WR TE RB 0.8%
WR RB WR 8.2% WR RB WR RB 2.4% WR WR WR TE TE 0.8%
WR WR WR 7.6% WR WR RB WR 2.4% WR WR RB WR TE 0.7%
WR RB RB 6.5% WR WR WR TE 2.4% WR RB RB TE TE 0.7%
WR WR TE 6.3% WR RB WR WR 2.2% WR WR RB TE RB 0.7%
RB WR WR 6.0% WR RB WR TE 2.1% WR RB WR RB WR 0.7%
WR RB TE 5.9% WR WR WR RB 2.1% WR WR RB TE TE 0.7%
WR TE WR 5.8% WR RB RB TE 2.0% WR WR RB WR RB 0.7%
RB WR TE 5.6% RB WR WR TE 2.0% WR WR WR WR TE 0.7%
WR TE RB 5.5% WR RB RB WR 2.0% WR RB WR WR TE 0.7%
RB RB WR 5.1% WR WR RB RB 1.9% WR WR RB TE WR 0.6%
RB WR RB 5.0% WR TE WR TE 1.9% RB WR WR TE WR 0.6%
WR TE TE 4.2% RB WR TE RB 1.9% WR RB WR RB QB 0.6%
RB TE WR 4.2% WR WR TE TE 1.9% WR WR RB WR WR 0.6%
RB RB TE 3.8% WR WR WR WR 1.8% WR RB WR WR WR 0.6%
RB TE RB 3.3% WR WR TE RB 1.8% WR WR RB RB TE 0.6%
RB RB RB 3.1% WR RB TE WR 1.7% WR WR RB TE QB 0.6%
RB TE TE 2.5% WR TE RB RB 1.6% WR RB RB WR TE 0.6%
TE WR WR 0.5% WR WR RB QB 1.6% WR RB WR RB TE 0.6%
TE WR RB 0.5% WR RB TE TE 1.6% WR RB WR WR RB 0.6%
TE RB WR 0.4% WR TE WR RB 1.6% WR WR WR RB WR 0.6%
TE RB RB 0.3% RB RB WR WR 1.5% RB WR WR TE TE 0.6%
TE RB TE 0.3% RB WR TE WR 1.5% WR RB WR RB RB 0.6%
TE TE WR 0.3% WR RB TE RB 1.5% WR RB WR TE QB 0.6%
TE WR TE 0.3% RB RB WR TE 1.5% RB RB WR WR WR 0.5%

It certainly appears early WR is the way to go. We can boil this down a bit more simply by looking at the starting total positional allocations and their relative frequency of appearance in the top 8.33 percent of all simulations. To do so, we have to account for the number of combinations that can happen for each allocation. For example, if we start 2 WR, 1 RB, there are three combinations:

  • WR WR RB
  • WR RB WR
  • RB WR WR

So I summed up the frequency from each of those and divided by three to get a relative frequency for each positional allocation. Here is the result for the first three rounds:

Start_3 Freq N_Comb Rel_Freq
3 WR, 0 RB, 0 TE 7.6% 1 7.6%
2 WR, 1 RB, 0 TE 22.7% 3 7.6%
1 WR, 2 RB, 0 TE 16.6% 3 5.5%
2 WR, 0 RB, 1 TE 12.5% 3 4.2%
1 WR, 1 RB, 1 TE 22.0% 6 3.7%
0 WR, 3 RB, 0 TE 3.1% 1 3.1%
0 WR, 2 RB, 1 TE 7.4% 3 2.5%
1 WR, 0 RB, 2 TE 4.9% 3 1.6%
0 WR, 1 RB, 2 TE 3.1% 3 1.0%
0 WR, 0 RB, 3 TE 0.1% 1 0.1%

We see similar with the first four rounds:

Start_4 Freq N_Comb Rel_Freq
3 WR, 1 RB, 0 TE, 0 QB 8.0% 4 2.0%
4 WR, 0 RB, 0 TE, 0 QB 1.8% 1 1.8%
2 WR, 2 RB, 0 TE, 0 QB 10.8% 6 1.8%
2 WR, 1 RB, 0 TE, 1 QB 4.3% 3 1.4%
2 WR, 1 RB, 1 TE, 0 QB 16.2% 12 1.4%
3 WR, 0 RB, 1 TE, 0 QB 5.3% 4 1.3%
3 WR, 0 RB, 0 TE, 1 QB 1.3% 1 1.3%
1 WR, 3 RB, 0 TE, 0 QB 4.9% 4 1.2%
1 WR, 2 RB, 1 TE, 0 QB 13.7% 12 1.1%
1 WR, 2 RB, 0 TE, 1 QB 2.8% 3 0.9%
2 WR, 0 RB, 2 TE, 0 QB 5.2% 6 0.9%
2 WR, 0 RB, 1 TE, 1 QB 2.3% 3 0.8%
0 WR, 4 RB, 0 TE, 0 QB 0.7% 1 0.7%
1 WR, 1 RB, 2 TE, 0 QB 8.0% 12 0.7%
1 WR, 1 RB, 1 TE, 1 QB 4.0% 6 0.7%
0 WR, 3 RB, 1 TE, 0 QB 2.4% 4 0.6%
0 WR, 3 RB, 0 TE, 1 QB 0.6% 1 0.6%
0 WR, 2 RB, 2 TE, 0 QB 2.8% 6 0.5%
0 WR, 2 RB, 1 TE, 1 QB 1.3% 3 0.4%
1 WR, 0 RB, 3 TE, 0 QB 1.5% 4 0.4%
1 WR, 0 RB, 2 TE, 1 QB 0.7% 3 0.2%
0 WR, 1 RB, 3 TE, 0 QB 0.8% 4 0.2%
0 WR, 1 RB, 2 TE, 1 QB 0.5% 3 0.2%
0 WR, 0 RB, 3 TE, 1 QB 0.0% 1 0.0%

This says that 3 WR, 1 RB is the preferred upside play, with WR 4x and 2 WR, 2 RB taking the next two spots.

I also looked at the best upside “Next 3” positional allocations for Rounds 4-6 when the highest upside 2 WR, 1 RB allocation was used to start:

QB Freq N_Comb Rel_Freq
1 WR, 0 RB, 2 TE, 0 QB 6.1% 3 2.03%
0 WR, 1 RB, 2 TE, 0 QB 5.9% 3 1.98%
2 WR, 0 RB, 1 TE, 0 QB 5.9% 3 1.98%
0 WR, 2 RB, 1 TE, 0 QB 5.8% 3 1.95%
1 WR, 1 RB, 1 TE, 0 QB 11.2% 6 1.86%
0 WR, 0 RB, 3 TE, 0 QB 1.8% 1 1.84%
0 WR, 0 RB, 2 TE, 1 QB 5.5% 3 1.84%
0 WR, 1 RB, 1 TE, 1 QB 10.9% 6 1.82%
1 WR, 2 RB, 0 TE, 0 QB 5.0% 3 1.65%
2 WR, 1 RB, 0 TE, 0 QB 4.8% 3 1.61%
1 WR, 0 RB, 1 TE, 1 QB 9.4% 6 1.57%
3 WR, 0 RB, 0 TE, 0 QB 1.5% 1 1.54%
0 WR, 3 RB, 0 TE, 0 QB 1.3% 1 1.27%
2 WR, 0 RB, 0 TE, 1 QB 3.8% 3 1.27%
0 WR, 0 RB, 1 TE, 2 QB 3.8% 3 1.26%
1 WR, 1 RB, 0 TE, 1 QB 6.9% 6 1.16%
1 WR, 0 RB, 0 TE, 2 QB 3.2% 3 1.08%
0 WR, 2 RB, 0 TE, 1 QB 3.2% 3 1.07%
0 WR, 1 RB, 0 TE, 2 QB 3.2% 3 1.05%
0 WR, 0 RB, 0 TE, 3 QB 0.6% 1 0.57%

This says we should probably be avoiding QBs in Rounds 4-6 and grabbing either one or two TEs with preference to WR over RB as the other non-TE picks.

Draft Results From 1.09

Things are a bit different when running my simulation from 1.09. The top 25 three, four, and five round positional combinations are as follows:

Start_3 Freq Start_4 Freq Start_5 Freq
WR WR RB 6.81% WR WR RB RB 2.06% WR WR RB TE RB 0.70%
WR RB WR 6.15% WR WR RB TE 1.92% WR WR RB RB WR 0.67%
WR RB TE 6.12% WR RB RB WR 1.86% WR RB WR TE TE 0.61%
WR WR TE 5.89% WR RB WR TE 1.86% WR WR RB WR TE 0.61%
RB WR WR 5.85% WR RB TE RB 1.76% WR WR RB RB TE 0.57%
WR WR WR 5.70% WR RB TE WR 1.74% WR WR TE WR RB 0.57%
WR RB RB 5.56% WR RB WR RB 1.74% RB WR TE RB TE 0.55%
RB WR RB 5.33% WR WR TE WR 1.70% WR RB RB WR WR 0.55%
RB WR TE 5.15% WR WR RB WR 1.66% WR RB TE RB TE 0.55%
RB RB WR 4.83% WR WR TE TE 1.66% WR RB TE WR TE 0.55%
WR TE RB 4.02% WR WR WR RB 1.63% WR RB TE WR RB 0.53%
WR TE WR 3.85% RB WR WR RB 1.60% WR WR TE TE TE 0.53%
RB RB TE 3.38% WR RB RB TE 1.59% WR RB RB TE TE 0.52%
RB RB RB 3.07% RB WR WR TE 1.57% WR RB TE RB RB 0.52%
RB TE WR 3.07% RB WR WR WR 1.57% WR WR RB TE TE 0.52%
TE WR WR 3.04% WR RB TE TE 1.57% WR WR WR RB TE 0.52%
WR TE TE 2.95% WR WR WR TE 1.54% WR RB WR WR TE 0.50%
TE WR RB 2.89% RB WR TE RB 1.53% RB WR WR RB RB 0.49%
TE RB WR 2.78% WR WR TE RB 1.53% WR RB TE RB WR 0.49%
TE WR TE 2.43% RB RB WR WR 1.51% WR RB TE TE QB 0.49%
RB TE RB 2.37% RB WR RB WR 1.48% WR RB WR RB TE 0.49%
RB TE TE 2.34% RB WR TE WR 1.48% WR WR TE TE RB 0.49%
TE RB TE 1.73% RB WR RB TE 1.47% RB WR RB WR TE 0.47%
TE RB RB 1.63% WR WR WR WR 1.44% RB WR TE TE WR 0.47%
TE TE WR 1.15% RB WR RB RB 1.42% RB WR TE WR TE 0.47%

Notice the top two combinations are the same as the 1.03 draft. However, the WR-WR-WR start drops to fifth here. WR in Round 1 is still the upside play, but there definitely is more of a mix of TE and RB when compared to drafting at 1.03.

Like above, here is the relative frequency table for positional allocation for the first three rounds:

Start_3 Freq N_Comb Rel_Freq
2 WR, 1 RB, 0 TE 18.80% 3 6.30%
3 WR, 0 RB, 0 TE 5.70% 1 5.70%
1 WR, 2 RB, 0 TE 15.70% 3 5.20%
2 WR, 0 RB, 1 TE 12.80% 3 4.30%
1 WR, 1 RB, 1 TE 24.00% 6 4.00%
0 WR, 3 RB, 0 TE 3.10% 1 3.10%
0 WR, 2 RB, 1 TE 7.40% 3 2.50%
1 WR, 0 RB, 2 TE 6.50% 3 2.20%
0 WR, 1 RB, 2 TE 4.10% 3 1.40%

This makes it a bit more clear that either two or three WRs early is the best play for upside, since the all the 2+ WR combinations made up three of the top four spots. We see similar with the first four round positional allocations:

Start_4 Freq N_Comb Rel_Freq
2 WR, 2 RB, 0 TE, 0 QB 10.3% 6 1.7%
3 WR, 1 RB, 0 TE, 0 QB 6.3% 4 1.6%
4 WR, 0 RB, 0 TE, 0 QB 1.4% 1 1.4%
2 WR, 1 RB, 1 TE, 0 QB 16.4% 12 1.4%
3 WR, 0 RB, 1 TE, 0 QB 5.2% 4 1.3%
1 WR, 3 RB, 0 TE, 0 QB 4.9% 4 1.2%
1 WR, 2 RB, 1 TE, 0 QB 13.7% 12 1.1%
2 WR, 1 RB, 0 TE, 1 QB 3.4% 3 1.1%
3 WR, 0 RB, 0 TE, 1 QB 1.1% 1 1.1%
2 WR, 0 RB, 2 TE, 0 QB 5.5% 6 0.9%
1 WR, 1 RB, 2 TE, 0 QB 10.0% 12 0.8%
1 WR, 2 RB, 0 TE, 1 QB 2.5% 3 0.8%
2 WR, 0 RB, 1 TE, 1 QB 2.1% 3 0.7%
1 WR, 1 RB, 1 TE, 1 QB 4.1% 6 0.7%
0 WR, 3 RB, 1 TE, 0 QB 2.6% 4 0.7%
0 WR, 4 RB, 0 TE, 0 QB 0.6% 1 0.6%
1 WR, 0 RB, 3 TE, 0 QB 2.1% 4 0.5%
0 WR, 2 RB, 2 TE, 0 QB 3.1% 6 0.5%
0 WR, 2 RB, 1 TE, 1 QB 1.3% 3 0.4%
0 WR, 1 RB, 3 TE, 0 QB 1.7% 4 0.4%
0 WR, 3 RB, 0 TE, 1 QB 0.4% 1 0.4%
1 WR, 0 RB, 2 TE, 1 QB 1.1% 3 0.4%
0 WR, 1 RB, 2 TE, 1 QB 0.8% 3 0.3%
0 WR, 0 RB, 3 TE, 1 QB 0.1% 1 0.1%

The top five positional allocations all have at least two WRs in the first four rounds, but it should be noted the top allocation was a more balanced 2 WR, 2 RB. This is one of the options in the RB-RB start that Shawn Siegele said would have high upside (followed by two WRs), and this simulation has shown that certainly to be of merit.

Lastly, I looked at the best upside “Next 3” positional allocations contingent on using the highest upside 2 WR, 1 RB start. Here’s what the simulation gave:

Next_3 Freq N_Comb Rel_Freq
0 WR, 0 RB, 3 TE, 0 QB 2.2% 1 2.2%
0 WR, 2 RB, 1 TE, 0 QB 6.6% 3 2.2%
0 WR, 1 RB, 2 TE, 0 QB 6.3% 3 2.1%
1 WR, 0 RB, 2 TE, 0 QB 6.2% 3 2.1%
3 WR, 0 RB, 0 TE, 0 QB 2.0% 1 2.0%
1 WR, 1 RB, 1 TE, 0 QB 11.6% 6 1.9%
0 WR, 0 RB, 2 TE, 1 QB 5.7% 3 1.9%
1 WR, 2 RB, 0 TE, 0 QB 5.6% 3 1.9%
2 WR, 1 RB, 0 TE, 0 QB 5.2% 3 1.7%
0 WR, 1 RB, 1 TE, 1 QB 9.6% 6 1.6%
0 WR, 3 RB, 0 TE, 0 QB 1.6% 1 1.6%
1 WR, 0 RB, 1 TE, 1 QB 9.1% 6 1.5%
1 WR, 1 RB, 0 TE, 1 QB 8.2% 6 1.4%
2 WR, 0 RB, 1 TE, 0 QB 3.8% 3 1.3%
0 WR, 0 RB, 1 TE, 2 QB 3.7% 3 1.2%
0 WR, 2 RB, 0 TE, 1 QB 3.5% 3 1.2%
0 WR, 1 RB, 0 TE, 2 QB 3.3% 3 1.1%
2 WR, 0 RB, 0 TE, 1 QB 2.8% 3 0.9%
1 WR, 0 RB, 0 TE, 2 QB 2.2% 3 0.7%
0 WR, 0 RB, 0 TE, 3 QB 0.7% 1 0.7%

Look at all those TEs at the top!!! To me, it looks like you avoid QB altogether and grab at least one TE in Rounds 4-6 and you’ll be set for upside. The other legitimate alternative is three more WRs.

Conclusion

My favorite part of these simulations is that they take all of my personal biases out of it. The numbers tell the story, and they certainly tell an interesting story from a league-winning upside perspective.

These simulations back up some of my claims in my MFL10 series of articles that starting with WR seems to be the way to go as a league winning strategy. Weighing heavier on the consistent WRs early, then leveraging higher and higher variance as you go through the draft, seems to give the most upside. We want our early picks not to miss, then our later picks to give us enough upside to win leagues.

Another thing is that early QB doesn’t seem to have a ton of upside, even adjusting for their historically low ADP. A shot here or there as Joshua Lake recommends isn’t out of the question, especially when they fall far enough. But overall Late Round QB still seems to be the dominant MFL10 strategy, even for upside.

This exercise certainly gives me more inclination to pick a greater number of earlier TEs than I previously have been, especially in Rounds 4-6. I’ve often been starting WR5x as my strategy, but I believe now I’ll certainly mix in TEs in the first six rounds, and yes, even some running backs for this #ZeroRB guy. But my focus will still be grabbing WRs.

Oh, and just for fun, here was the single highest scoring individual simulations from picks 1.03 and 1.09.

1.03: WR RB WR RB RB WR TE TE TE QB QB WR QB RB RB WR WR Def Def Def

1.09: RB WR RB TE RB TE TE QB WR WR RB WR WR WR RB QB WR Def Def Def

Subscribe for a constant stream of league-beating articles available only with a Premium Pass.

  1. Okay, second place matters to a degree, but the big prize is winning.  (back)
By RotoDoc | @RotoDoc | Archive

Comments   Add comment