Chasing The FanDuel Sunday Million
*Skip to Week 2 Results if you read the introduction last week.
One of my goals for this NFL season was to find more time to do data analysis and turn that into creative content. I feel like I’ve been able to do that in a few ways this season:
Still, I wanted to do a regular piece that was a bit more fun and hopefully is able to portray how I go about using a lot of the information on the site.
I figured, hey why don’t I try to win a $1million in the process? After all, I’m surrounded by million dollar winners and starting to feel a bit left out.
SportsGrid CEO Jeremy Stein, who is making lineups for the DR Sweat promotion, has done it TWICE.
Each week I’ll go through the process of how I made my lineups, and the macro (roster construction) and micro (specific player pool decisions) strategies employed. I’ll then evaluate the results and how I likely got screwed by variance with no other explanation for my failings.
I’ll document the results in here, although the goal isn’t to grind out weekly profit in this format. The goal is to give myself the best chance to win life changing money, something I feel got lost in my macro strategy this week (more on that later).
In general, I’ll keep my entry fees in the $1,000 – $2,000 range. So, Week 1 that allowed me to max enter 150 times. In Week 2, the entry fee was $15, and I entered 100 lineups.
Here was the Week 1 recap: https://dailyroto.com/chasing-the-fanduel-sunday-million/
Now, onto Week 2.
Week 2 Results
Entry Fees: $15 x 100 entries = $1,500
Entries Cashed: 32 (32%)
Gross Winnings: $1,288
Best Entry: 514 / 246,642 ($150)
Net Profit: -$212 (-14.1%)
Lessons Applied From Week 1
Following a Week 1 where I felt I let opportunities to take heavy stances on James Conner and Rob Gronkowski slip through my fingers and simply had too much of a spread out approach in general, I vowed to be more aggressive moving forward. At first glance, it might look like I ignored this advice.
I owned just two players over 40%, James Conner and Alvin Kamara, at 43 and 42% respectively, and those were basically in line with the field. Boring.
However, I did take heavy stances on some value plays, owning Kenny Golladay at 37% (nearly 4x the field) and George Kittle 39% (15.5 points higher than the field). I’ll talk about them more in depth in a bit.
The issue with locking a player this week in my opinion was the combination of juicy game stack opportunities, more efficient Week 1 pricing, and injuries opening up many different strong values. There wasn’t a potential workhorse back priced at $5,000 like Conner last week. Kittle was an intriguing option to take a very heavy stance on at the TE position, but he didn’t have the overwhelming odds to win the position outright that Rob Gronkowski had last season. Plus, forcing the lock button on Kittle would have left me empty handed in terms of Travis Kelce exposure, a player with Top 3 odds to win the position outright and an important piece of PIT-KC game stacks.
So, instead of taking ultra aggressive stands on one or two individuals, I whittled down my player pool so that I was in line or overweight to all the options I used at the RB position, with the exception of a few stray pieces as a part of game stacks:
|Player||Exposure||Leverage (Exposure% – Field Ownership%)|
Of my core RB group, the only player I really ended up underweight on was Tevin Coleman, and I’ll discuss why shortly. I ended up about even with the field on Conner, Kamara, and Cook. I ended up overweight by an average of 11.4 percentage points on five of my nine core RBs: Gurley, McCaffrey, Gordon, Yeldon, and Williams.
From a more macro perspective, I felt a did good job of having a more consistent theme across my lineups through a combination of taking heavy overweight stances on a couple of individuals (Golladay, Kittle), tightening up my RB pool, and having a more active hand in the stacks I chose.
Last week, I set up some simple stacking rules but didn’t force any particular stacks. As a result, I ended up casting a pretty wide net among the QBs I stacked, using some that popped since they were cheap values but didn’t necessarily carry meaningful upside that was amplified by fun stacking choices, both directly (on the same team) and indirectly (an opponent coming back).
This week, instead of running all of my lineups at once, using generic stack settings in our optimizer, I ran my lineups by QB – using more specific settings for each stack I was using. Of my 100 lineups, I ended up using the following QBs:
-Tyrod Taylor (15)
-Ben Roethlisberger (15)
-Patrick Mahomes (15)
-Jimmy Garoppolo (15)
-Matt Ryan (10)
-DeShaun Watson (10)
-Case Keenum (10)
-Alex Smith (10)
I’m considering in future weeks potentially tightening that up even further so that I can own each stack around 15-20 but am undecided on that.
The DeSean Jackson Lesson
As I mentioned, not only did I choose exactly how many of each QB I wanted to play this week, by running my lineups in our optimizer by QB, I was able to set up more focused stacks.
Rather than simply saying, “Stack QB with 1 WR and 1 opposing player”, I let the upside and level of concentration of each team/game stack dictate my settings for that specific QB run. For example, when running my Patrick Mahomes stacks, I boosted KC pass catchers and all opposing Steelers players:
At a minimum, I forced Mahomes to be paired with 1 WR/TE with at least 1 PIT Flex player coming back. Since the total was so high in this game (53.5, highest of the week), I allowed for a triple stack with Mahomes and up to 2 PIT Flex players to be used:
Finally, I used our new “shuffle” function, which helps to get more normally distributed player exposures.
Setting things up this way allowed me to, at a minimum, pair Mahomes with a pass catcher and a PIT Flex player. Utilizing the boost on the player projections and the team player limits, I allowed for a triple stack on the KC side and 2 PIT Flex players coming back, but I didn’t force it. This allowed me to get a more organic mix of stack types within the same game stack without arbitrarily deciding ahead of time that I would play X triple stacks and Y double stacks, etc.
The boosting of skill players on both teams along with our shuffle function also helped to solve “The DeSean Jackson” problem. If you recall, last week I lamented having heavy Godwin exposure in my TB-NO game stacks but having zero DeSean Jackson. The projection gap between the two was clear but not massive (about a full point). That’s something you don’t want to sacrifice in a regular lineup, but if playing a GPP where you expect a game to get somewhat nutty in order to realize Top 1% upside, was that 1-point projection gape really worth owning only Chris Godwin as my TB WR? Probably not, especially in light of the work I did on floor/ceiling projections this past offseason, which showed that the 90th-plus percentile outcomes for WRs with a high team total and high YPT were disproportionately higher, relative to their base projection, than the average WR.
As a result of these settings, I had 6% Sammy Watkins exposure, who scored 16.1 FD points at $5,800 and just 1.8% ownership, and the leverage created there would have been much greater had Watkins simply found the end zone once (not a far cry from reality given his yardage combined with 6 Mahomes passing TDs).
It also allowed me to have a smattering of James Washington, who I used when Justin Hunter was ruled inactive. Washington had a pretty low key day, but it felt like the right type of play given the game stack upside and his ownership (0.2%).
Aside from using some big play options with really low ownership percentages in my stacks, the boost and shuffle approach allowed me to get some Tyreek Hill and JuJu Smith-Schuster into lineups. Had I not done that, the PIT component of both my KC and PIT stacks would have been filled almost entirely by Antonio Brown.
This lineup ended up as my top placing lineup on the week, finishing in the Top 0.2% of all lineups, much better than my top finish a week ago (Top 1.1%):
The final adjustment I made after last week’s analysis was to take a more natural contrarian approach. Last week I was burned by heavy LeSean McCoy exposure and almost burned by heavy Saquon Barkley exposure as well, with both backs in difficult spots. I didn’t want to remove those plays from my lineup set because I’m a believer in playing volume at low ownership percentage, regardless of the matchup. However, it was probably unnecessary for me to have them potentially sinking such a high percentage of my lineups.
This week, I once again used our shuffle feature, which makes it less likely a player is re-used in a subsequent lineup once they are used. This allows you to more organically get a bit deeper into your player pool. I also, rather than forcing heavy exposure to specific contrarian plays as a way of differentiating, made use of our “Max Total Public Owned” feature.
What this feature does is simply add another constraint when making your lineups. Just like you can’t exceed the salary cap when you make a lineup, you can’t exceed a certain cumulative projected ownership in any of your lineups. For most of my builds this week, I set Max Total Public Owned to 110-120:
Using the settings in the above screenshot, every lineup that is produced will have a cumulative projected ownership of 110% or lower.
This is how I ended up underweight on Tevin Coleman. Despite allowing for a 35-40% exposure to Coleman in my settings, his high projected ownership prevented him from cracking more of our optimal lineups.
Now, I did make a mistake here. I ran my lineup set on Saturday night, when we had Coleman projected at around 35% ownership and James Conner in the high 20s. However, our final projected ownership (courtesy of Unsourced Fantasy Collective) had Conner at 38.7% and Coleman at 26.7%. Had I rerun my lineup set with those updates on Sunday morning, I likely would have been even with the field or a little overweight on Coleman and a little bit underweight on Conner.
That is admittedly one of the flaws of running lineups the way I did, by QB with changing stack settings depending on the QB/game. The marriage of exposure flexibility and control over each QB set is great, but it’s not something where you can easily re-run eight different stack scenarios on a hectic Sunday morning as new information comes in.
This approach to allowing more natural contrarian lineups also allowed me to have single digit ownership to David Johnson, a similar type of play to McCoy/Barkley a week ago. I was happy to have the low owned upside in my lineups and even happier when it didn’t crash and burn my entire week when it failed miserably.
All in all, though, I was much happier with my lineups this week, placing three in the top 1,000 of lineups. Despite losing essentially what equates to the rake, I felt I put out lineups that gave me as realistic of a shot as you can hope to have at winning such an elusive grand price.
For the second straight week, we were higher on Thomas than the field. In the FanDuel Sunday Million, I had 30% exposure to Thomas, nearly double the field (17.2%). Admittedly, we were concerned that our catch rate expectation of Thomas was too high (78%), but we felt it was warranted given his career catch rate and Brees’ projected completion percentage at home in a plus matchup. Thomas went on to catch 12 of 13 targets, and the regression in Pass TD% for the Saints, which we expected, continued. Having him projected in line with Brown was a win.
Golladay was our top value at sub-$6,000 on FD. Early in the week it seemed like he would be chalky after a good Monday night performance that wasn’t reflected in his Week 2 salary. While most of the DFS community rightfully has an affinity for Golladay’s talent, size, and athleticism, we were a bit more optimistic on his volume. We were actually wrong on the market share projection (22.8%, actual of 17.3%), but felt there were multiple outs. Two of those other outs hit: an efficient game from Golladay (catching a score and averaging 9.9 YPT, even with Stafford missing him on an easy long TD in the second half) and a pass heavy approach for the Lions that still led to 9 Golladay targets. Golladay was my highest leverage player of the week as my 37% ownership was 27.2 points higher than the field.
Whiffed On It
San Francisco 49ers Values
While a heavy 49ers game stack in the second half leading to just 26 total pass attempts was somewhat bad luck, we overestimated how condensed the Niners passing attack would be with Marquise Goodwin out:
Here were the actual market share of targets:
I had heavy exposure to Garcon, Kittle, Pettis, and Taylor viewing them as very high ceiling plays at their price tags, and figuring if one or two saw less volume, it would see-saw to the other options making them even more valuable.
However, we over projected the MS of targets on all four of those players, while Garrett Celek and Kendrick Bourne caught 2 of 2 touchdowns. We had projected those players for a combined MS of REC TDs of just 9.6%.
Was there some bad luck here? Yes. Is it clear we overestimated how condensed the passing attack would be? Yes. Having 39% exposure to George Kittle, 26% exposure to Pierre Garcon, 22% exposure to Dante Pettis, and 8% exposure to Trent Taylor killed a lot of lineups.
Excluding Minnesota Stack
I don’t have much intellectual insight here other than to share my pain of excluding Kirk Cousins from my QB list this week.
Each week we are also sweating alongside our subscribers as CEO Jeremy Stein puts together lineups. This week our lineups posted a -12% ROI as we max entered the Fanduel Sunday Million. Our best lineup finished in 403rd place as 16% exposure to Travis Kelce helped provide separation at the TE position, but ultimely lack of exposure to Patrick Mahomes proved costly.