“In what contests should I be playing?” is one of the most common questions we see. Jesse set out to answer it, once and for all.
At Elite Fantasy, there are a handful of questions we are constantly asked in chat rooms, on Twitter, and through e-mail. Mans even gets them over the phone almost every day on SiriusXM:
“What is the difference between a cash lineup and a GPP lineup?”
“How many players should I include in a stack?”
“What contests should I be playing in?”
“Every player I used was listed in the articles and I still lost. This has been going on for over A WEEK! What am I doing wrong!?”
I’ve been seeing these questions, and a dozen others like them, for over 4 years now:
The 1st time, I was asking them,
the 2nd through 300th times I was seeing them answered (basically the same way each time) by Mans, DC, Benny, Vlad, et al.,
and the 301st through 1,117th times I’ve been the one receiving them.
However, in an effort to end 2019 with that ticker below 1,500, I’m setting out to answer each of these questions, one at a time, as thoroughly as I can.
This study is meant for players who are currently focusing on GPPs or other tournament-type contests. If you are mostly playing in cash games and you’re happy , I recommend continuing the “100-1,000 player, single-entry, 50/50s & double-ups” approach we usually preach here at EF (at least until I do a study on cash game ROI!).
If you’re entering anything other than those cash games, though, this study is for you.
“In what contests should I be playing?”
The Usual Answer (short)
In GPP, stick to single-entry contests and 3-entry maxes. In cash, play single-entry contests (50-50s and double-ups) with 100 or more players. Mans will tell you to stick only to 50-50s, while some others will say double-ups are just fine.
The Usual Answer (long)
It depends. Goals, playing styles, risk tolerances, and budgets can vary greatly from person to person, so it’s impossible for there to be one approach that is “optimal” for everyone. The short answer (above) is as good a take as you’ll find if you want one in <50 words, but if we want to get specific, we have to consider the question-asker’s situation:
Do they want to give themselves the best opportunity for a large score, or do they want to generate a steady, more consistent (albeit relatively small) profit?
Amount 1: What’s the most money they could lose over the course of one month (or of one season then divided by the number of months in that season) playing DFS without it negatively affecting their lives or future decisions?
Amount 2: What’s the absolute most money they could lose over the course of one month (or season) playing DFS? This may be an amount that would suck to lose and could possibly affect their lives, but not so severely it causes emotional distress. It’s a loss that could be chalked up to entertainment, but to lose even a little bit more would be personally devastating and unacceptable.
Are they okay with losing “Amount 1” for three months in a row, or would that cumulative loss bother them too much?
The answers to these questions give us good data to start a tailored “contest selection plan”. I won’t go into too much detail, as these questions have an infinite number of answer combos, but I’ll give a few rules of thumb that can guide your decision making:
If you prefer a large score, large-field (>500 player) GPPs are the way to go. If you prefer steady profit, cash games are your thing. As you move from the GPP end of the spectrum toward the middle, you’ll be adding in more mid- and small-field events (<500 players and <100 players, respectively), and when you cross the mid-point and begin moving toward the cash game end you’ll be removing large-field GPPs altogether and replacing them with double-ups, 50-50s, and then head-to-heads (in that order).
As you move from mega-multi-entry GPPs that allow players to each enter 26-150 LUs, to regular GPPs (allowing 5-25 LUs/person), to 3-maxes and single-entry contests, and finally to cash games, your variance will reduce, meaning your win rate (the number of contests you cash in divided by the number of contests you enter) will increase. This doesn’t necessarily mean you’ll make more money; it just means the chances you lose (or double) your entire budget over the course of a day, week, or month, will go down.
There are no right or wrong answers for the questions determining Amount 1 and Amount 2, the only thing that matters is you are being honest with yourself when you answer them. It’s easy to say you’re just fine with losing $100/week during the NFL season, but if going 3, or 4, or even 7, weeks in a row losing it will bother you, it’s the wrong answer. Here’s a test: Take ½ of what you decided on for “Amount 1” and put it in front of you, in cash. Now urinate on it and light it on fire. If you aren’t at least kind of laughing at what you just did, that amount was too high, and you were lying to yourself when you set it.
Your “Amount 1” could be $10, or it could be $10,000: No amount is more correct than another. Between DraftKings and FanDuel, every type of contest is available at every buy-in level, and you can create a contest mix that is optimal for your situation regardless of your bankroll size.
If losing Amount 1 three months in a row would be a big deal, you should nudge yourself toward the cash game end of the spectrum. If it wouldn’t be a big deal, feel free to nudge yourself toward the GPP end.
“The Usual Answer (long)” is a paraphrased version of the answer I’ve given many of you over the years in chat and through Twitter DMs. “The Usual Answer (short)” is an almost-verbatim regurgitation of the answer I, Mans, DC, Benny, Vlad, and others will give you 10 times out of 10 if we simply want to make sure we get the main points across.
Now, while I wholeheartedly believe in the validity of both these answers, I’m not sure they’re the best or most impactful ones we could give. The theory behind them is correct, but I believe they may be misleading people as to what one’s expectations should be in regard to the win rate (contests in which you cashed divided by contests entered), ROI (return on investment, calculated by net profit divided by gross entry fees), and variance (here being used colloquially to describe the likelihood of a prolonged winning or losing “streak”) present in the various kinds of contests mentioned.
With that in mind, I conducted an experiment. From Monday, June 10th through Monday, July 29th, I put my usual MLB DFS play on hold and instead focused on collecting data that might illustrate exactly why contest selection is so important.
My belief is, for recreational low- to mid-stakes players, contest selection is the most misunderstood and mis-practiced part of DFS, and it costs them more money than any player selection or lineup construction mistake ever could.
With a better understanding of contest selection, the average Elite Fantasy subscriber (who I am defining as a slightly losing to slightly winning player who may have one or two large scores in their history) could see a boost in their ROI that turns them from a treading-water regular looking to make a big score to someone who generates a consistent profit season after season.
My hypothesis is that, on any given slate, a given lineup will generate a better ROI (and, therefore, more profit over time) as it is entered into contests with fewer and fewer players. This will support the idea that, if profitability is desired, players should be entering a far different mix of contests than they currently are (the assumption about their current mix is based on the relative unpopularity of 3-100 player contests compared to those with 1,000 players or more).
For seven weeks I entered only $1-$5 contests on FanDuel. Every lineup created was entered into a mix of 3-, 5-, 10-, 20-, and 100-man contests, as well as one or two contests of 1,000 or more. The large-field contests were the 3-entry-max Sac Fly, which usually had around 3,500 players, and the Bean Ball, which ranged from 10- to 25-entries max and usually had 4,000-10,000 players. The 3-100 player contests were always FanDuel-created (not user-created) in order to ensure the payout structures remained consistent.
This design allowed me to look at any individual lineup and see how it performed in a 3-man contest compared to a 5-man, 10-man, 20-man, etc. The small buy-ins allowed me to both mimic the contests and competitors many Elite Fantasy subs are familiar with and to enter each lineup into multiple contests at each field size and entry fee so as to reduce the variance inherent in the result of any one contest. I entered around 10 contests per lineup so my goal of 1,000 contests would require roughly 100 unique lineups, and I limited myself to 3 lineups in any given day so the data would take several weeks to accumulate, hopefully further smoothing out short-term variance and mimicking the practices of most of our subs, who are not able to create 25+ lineups, or play in hundreds of contests, every day.
After 50 calendar days, 32 playing days, 82 lineups, and 961 contests, my results were as follows:
“Edge” is simply the difference between my Win Rate and the average percent of the field who got paid.
For example, say we have a 10-man contest that pays 3 places: If everyone in the contest is of equal skill, then each person would have a 3/10 chance (30%) of finishing in 1st-3rd place. Their equal skills would cancel each other out, and over thousands and thousands of contests they would all have roughly breakeven records (actually, they would all be losing players, since the rake taken out of the prize pool by the site would slowly chip away at their overall pool of money, leaving everyone with, after thousands of contests, slightly less than their starting bankroll).
In general, the only way for someone to be profitable in these contests would be to finish in that top-30% more than their 30% share of the time, which would require them to have a skill advantage (an “edge”) over the other players in the field. In my results you see that, although only 30% of the field gets paid in the 10-man contests, I finished in the money 38.46% of the time. That 8.46% difference is what I call my “edge”.
My hypothesis, that a lineup will generate a better ROI as the field size becomes smaller, was incorrect, but not completely. In fact, I believe my results confirm that smaller fields are more profitable, there is simply the caveat that there exists a tipping point: A contest size that, despite us having an edge on the field, does not pay us enough on winning days to cover the entry fees of our losing days, causing us to effectively “leave money on the table”. On our good days, our lineups are so much better than our competitors’ that to take the money of only 2 or 4 of them (as we would in 3- and 5-man contests, respectively) is simply not enough. Those lineups, because they are so much better than our opponents’, could easily be scooping up wins in the 10- and 20-man contests.
If you look at Table 2, you’ll see my ROI is higher in the 20-man contests than in the 3-, 5-, and 10-mans, which is what makes my hypothesis technically incorrect. However, looking at Table 3, you’ll see my “Contest Types Entered” are not perfect, considering the goal of the experiment. This is because, when I set out to do this, I only had a broad conceptualization of what I was trying to prove.
I knew I was hoping to show people would be more profitable if they put a great majority of their volume into smaller fields, but exactly how I would go about it, and what, specifically, would be the hypothesis, was still uncertain. At the beginning I simply stopped entering my usual contests, jumped down to the low stakes (at that time, mostly the $2 and $5 levels), and started entering 5-100 man contests, intending to simply compare my results at the 5-, 10-, 20-, and 100-man levels. It was only after a few days I realized the experiment had a lot more potential, and if I moved down to just the $1 and $2 levels I could enter far more contests per lineup and introduce the 3-mans and large-field GPPs, which would increase the number of unique opponents with which each lineup was competing (thus, reducing variance). Adding in the new contests would also widen the scope of the experiment to include the large-field GPP data I needed to compare to my small-field results, in order to show which was more profitable.
Unfortunately, this realization came after four of what would be my eight most profitable days of the experiment, which I believe had an effect on the results. Had I entered each of those early lineups into 3-man contests and large GPPs, I believe the -6.94% ROI we see in the 3-mans would become positive, and the -21.42% ROI in the large GPPs would move closer to breakeven. This is supported by the fact I still saw an edge in the 3-man contests and GPPs despite having a negative ROI. This ROI:Edge relationship leads me to believe those ROIs are affected by missing out on those four big days toward the start of the experiment.
At this point, there are a couple of conclusions I believe we can draw from this data:
1. It is not only possible, but actually likely, that if you are only playing in contests of 100+ people you are leaving A LOT of money on the table. In fact, you could currently have the skill to be a profitable DFS player, but have lifetime results well into the red because of poor contest selection. This experiment showed that a profitable player, who had a decent couple of months, was STILL suffering from a -21.42% ROI in large GPPs. If those were all I’d been entered in, I’d likely be looking at a $300-$400 net loss playing nothing more than $2.22 contests. Instead, I had a $316.80 profit, which would have been even greater had that $141.90 in GPP entry fees been put toward 10-man and 20-man contests.
2. A stretch of 5, 10, or even 15 slates where you’re a net loser is not only possible, but extremely common. This experiment, which ended with an ROI of 18.91%, still had a 25 (read: TWENTY-FIVE)-slate stretch where I was a net loser (from 6/19 through the end of the experiment), and that could absolutely have continued for another several slates. If you’re upset and tilting because you’ve been losing for a week or two (or three), it’s likely you are playing above your bankroll or, maybe, just need to adjust your view of what DFS is and what it is capable of doing for your life.
Post-Analysis Thoughts / What We Didn’t Consider
While I was reviewing the results, I uncovered a few things I think are very relevant to the study but that, going in, I didn’t anticipate would be factors.
The most surprising thing about the results was my ROI was higher in the 10- and 20-man contests than it was in the 3- and 5-mans. Now, this could be due to variance, which will be higher when facing fewer opponents, but I suspect, in this case, it isn’t. After all, my edge was also higher in the 10- and 20-man contests. Why were my lineups performing consistently better against a 20-man field (paying 15% of the players) than a 3-man one (paying 33% of the players)? I believe the answer lies in something we haven’t even discussed: lineup construction.
To illustrate what I mean, look again at Table 2 and go all the way to the last column, “Avg Finish %ile”. This column, and stat, is important to our discussion of lineup construction because it shows how vulnerable a set of lineups is to “swinginess”. “Swinginess” is the word I’m using to describe the variability inherent in different lineups and lineup-building strategies. A case study:
It’s no secret that Mans and Schuster are Elite Fantasy’s (and, possibly, the industry’s) cash game experts. They focus almost entirely on cash game contests (those which pay out over 40% of the field) in every sport, and have shown consistently-profitable results, season after season, for several years. Another non-secret is Mans and Schuster have a few rules about lineup construction they almost never break. In any given lineup, they do not use more than 2 players from the same team (no stacking!), they generally don’t use more than 3 from the same game (spread out your risk!), and they focus on selecting the safest (highest floor) and most valuable (expected pts/$) guys on the slate.
Unsurprisingly, this low-risk, high-floor construction yields them a positive ROI in their bread-and-butter contests: 100-man 50-50s and single-entry double-ups with 100-1000 players. All they have to do is beat 50-55% of the field and they’ll cash, and anyone who has tailed them for a few years knows they, very often, find themselves just sneaking across that pay line.
This isn’t because they’re only kind of good, it’s because they’re very good.
When they find themselves around that pay line, they finish just above it far more than just below it. They know they have an edge on the field, so their average lineup is more than enough to cash, and they follow a lineup-building strategy that yields very little “swinginess”, so their scores rarely deviate far enough below that average to cost them a payday.
When their lineup has a bad day it can still sneak across that threshold and make them money because their players are mostly chosen from separate teams and separate games, and the performance of one guy in their lineup is rarely correlated to or dependent on another. Maybe they started Andrew Luck and he had a bad day. It’s unlikely they also started Eric Ebron AND T.Y. Hilton, who would have struggled right along with him. In fact, if their QB decision came down to Luck or Mayfield and they chose Luck, it’s likely they did use Beckham, or Landry, or Njoku. That way, when Luck struggled and Mayfield went bananas, they weren’t left kicking themselves all day: They kept themselves alive with their exposure to Mayfield’s receivers, and maybe still snuck across the pay line. This thoughtfulness is what gives their lineups such little “swinginess”.
However, if their contests only paid 12-33% of the field (like the ones in this experiment), Mans and Schuster wouldn’t have nearly as many close calls. Many, probably most, of their lineups would find themselves just outside of the money, in that 30th-35th percentile. They would rarely finish 80th/100, or 90th/100, because even their bad lineups would have respectable scores, but they would also finish 1st/100, and 2nd/100, less often. Don’t get me wrong, it’s likely they would still be profitable players, but I think their ROI would shrink considerably if they only competed in 100-man GPPs and didn’t change anything about their lineup construction.
If I were to repeat this study and build my lineups in their low-risk, high-floor style, I believe my hypothesis would have been proven correct: Our ROI would have been highest in the 3-man contests (where 33% of the field gets paid) and declined slightly as we increased the size of the field (and therefore decreased the average % of the field who gets paid). This is because our high floors would have translated into higher “Avg Finish %ile” numbers, and the contests paying the largest percentage of the field (the 3-mans) would be the ones we’d have the most success in.
As it stands, though, I built my lineups a little bit more aggressively than the Mans/Schuster model. My intention at the beginning of the experiment was to mimic what I believe are the most common practices among our subs. I tried to mimic their buy-in levels, time restrictions (I usually registered for contests in the mid/late afternoon and built no more than three lineups a day), contest selection habits (even the large GPPs I entered were mostly 3-entry-maxes, with some 10-, 20-, and 25-maxes mixed in), and, finally, lineup constructions.
In the years I’ve been a part of EF, I’ve seen the same kinds of GPP lineups discussed in the chats and posted on Twitter over and over and over. If it’s FanDuel baseball there is a pitcher (mid- to bottom-tier more times than not) and then eight position players broken down into stacks in one of four ways: 4/4, 4/3/1, 4/2/2, or 3/3/2. Very rarely you’ll see 4/2/1/1, 3/2/2/1, or 2/2/2/2, but anything else is almost non-existent. When it comes to the primary stack (the first number listed), it’s almost always 3 or 4 of the first 5 guys in the order, with one straggler from the bottom occasionally making an appearance because of salary restrictions. The mini-stacks of 2 or 3 usually come from the first 6 guys in the order, and any one-offs could be from anywhere (but are usually either guys hitting in the top 5 who are severely underpriced or studs who will be under-owned). Every stack of 3 or 4 comes from a team facing either a weak starter or an average starter and a weak bullpen.
These same lineups get churned out, day after day, and for a pretty good reason: They’re what I would call “tight” constructions. They are safe, with very few off-the-wall plays (guys hitting 8th or 9th, stacks of guys facing solid or elite starters), but include enough 5 hitters, 6 hitters, and low-owned one-off studs to have a little bit of upside. Their “swinginess” is pretty low. It was with these guidelines that I built my lineups.
Most of my lineups followed those four main constructions, and only a few times did I throw in twists like a 4/2/1/1 or 3/2/2/1. One thing I did stray from, however, was the player selection.
As a game theory nerd and firm believer in focusing more on my opponents’ decisions than my own, I had to allow myself to do a few more aggressive things I knew would increase my edge in $1 and $2 contests:
1. I probably played catchers in the C/1B spot four or five times as often as most people (which still isn’t very often on FanDuel).
2. I looked for opportunities to use my Flex as a way to stack 2 guys from a team who were listed at the same position, like Hiura/Moustakas (later Shaw/Moustakas), Sano/Arraez, Lindor/Freeman, etc.
3. I routinely used “wraparound” stacks from AL teams when the 8 and/or 9 hitters were on-base machines or speedsters.
Those three changes moved my lineups up to what I would call “tight-aggressive”. The main stacks were pretty much always from the “obvious” teams on the slate, and the smaller stacks and one-offs would come from a slightly bigger pool but were rarely so contrarian I would call them “wild”.
This “tight-aggressive” strategy has a moderate amount of “swinginess”, and I believe it’s why my results favored the 10- and 20-man contests instead of the 3- and 5-mans of my hypothesis.
In the 3-man contests, where your “Finish %ile” can only be 33.3%, 66.6%, or 100%, the additional swinginess my strategy creates will leave me at the 66.6% and 100% more often than would the Mans/Schuster strategy. Even though both strategies likely have an overall “Avg Finish %ile” in the upper 30s or low 40s (mine was 41.5% in the results [see Table 1]), the fact that my lineups will have more variation (some capable of finishing top 1%, but also capable of finishing dead last in a 100-man field) means I’ll grab that 66.6% spot (25.1 percentage points worse than our average) more often than they will, and they’ll get the 33.3% spot (only 8.2 percentage points above our average) more often than I will.
Since that one-spot difference is also the difference between cashing and not cashing, I believe the Mans/Schuster strategy would be more profitable in the 3- and 5-man fields, but less profitable in the 10-man-and-larger fields, than my “tight-aggressive” strategy. Had I built less aggressive lineups, possibly just matching the “tight” strategy of most of our subs, I believe we would have been able to confirm the hypothesis.
First and foremost, I don’t think this is arguable: Damn near every person reading this would make more money playing DFS if they started entering smaller contests.
Before you apply that lesson to your own play, however, there are two caveats you must consider:
1. Your success in contests of any given field size will depend greatly on your lineup construction style. As your style becomes more aggressive, the “optimal” field size for you will increase. It is extremely unlikely, however, that anyone’s “optimal” field size is greater than about 500, with about 80% of people’s falling between 5 and 100.
2. This is meant to apply to tournaments, not cash games. While some of the analysis here applies to both, the experiment itself, the results, and the analysis are focused on contests that pay <40% of the field and have an escalating prize structure (players receive more money the higher up the field they finish). A study of cash games, which feature flat payouts (everyone who cashes gets the same amount, regardless of their position), could possibly yield an entirely different conclusion in regard to the relationship between field size and ROI.
Well, that’s all I have for today. Hopefully, this will end up being the first of many articles I write concerning game theory, contest selection, lineup construction, and any other abstract concepts you want to read about, and we’re on the verge of opening up all-new discussions, producing all-new theories, and making all-new discoveries, here at Elite Fantasy.
Until next time,