The Fundamental DFS Misunderstanding
Over the last couple weeks, I’ve examined the current DFS debacle and some reader perspectives. And I keep getting hammered with a fundamental misunderstanding of the “sport.” Sure, any given day of DFS is wildly unpredictable. The same is true of baseball. Does that mean there’s an incongruity between DFS and the need for large sample sizes we espouse here on FanGraphs? No.
At its heart, DFS is a process-driven lottery. As with any lottery – or for that matter, a baseball game – winning on a single day involves a lot of statistical noise. Where DFS differs from your typical state-run money pit is that the right process can create a positive expected value (EV). Theoretically. The wrong process still has a negative EV, but that doesn’t mean you can’t win.
Here’s where the misunderstanding comes into play. Some people say it’s impossible to offer meaningful advice about a single slate of games. The teeny tiny sample size of one game ensures chaos. That’s true-ish, but it’s never been about one game. You’re thinking too small. Succeeding at DFS is about a full season of games. The outcome of one contest is close to random. The outcome of 250 contests reflects talent level. If you have a positive EV process, you’ll probably make money over a full season.
Daily fantasy really isn’t much different than your typical daily roto league. You pay for your team, you draft it, then you set a lineup for 180 straight days. Every one of those days you accrue stats. You have to be better than your fellows with some frequency in order to win the league.
With DFS, you’re simply paid with cash rather than cumulative roto points. You also have the option to skip contests, draft new lineups for every time slot, and explore a new range of strategies and tactics. But it’s still deceptively similar to a roto league. You’re trying to grind out a winning season, not a winning day.
Of course, nothing is simple. Players rarely use the same process over the entire season. For example, I used an aggressive, multi-entry technique last April. I lost a bunch of money. When I reverted to my previous one or two lineup approach built around bargains, I made over $100 in the final five months. I’m not shopping for a yacht, but it’s a sign that the process is probably reliable.
The quality of a process depends on both expected production and the picks of all other players. I had a lot of near misses this year. They came because my lineups had too much overlap with other owners. At least I think that was the problem. It’s hard to tell. When I rostered Joey Votto’s three-home run night, so did 25 percent of the field. I think 50 percent of the field rostered Bryce Harper’s three-home run day (it was a small contest).
In this case, my good process was nullified by its growing popularity. When I participated in 2014, not that many people were using the simple principles that I follow. Sometime over the winter, other venues got a LOT better at offering DFS advice. Bad for me. Good and bad for our readers.
In August and September, I started reaching for more obscure picks with mixed results. It’s something I’ll try more frequently next year. As the average user gets smarter, the large contests (GPPs) have become more about anticipating who will and won’t be popular relative to their actual value. So there are two important angles to analyze: expected versus actual value and popularity.
Ideally, the game would be about picking the lineup with the highest expected value. In actuality, there are more variables to balance. This is partly why algorithms have an advantage. Their process allows them to perform tertiary analysis on expected ownership rates. Perhaps that’s the next frontier for The Daily Grind.
You can follow me on twitter @BaseballATeam
No matter how you spin it, on any single day, DFS is a gamble.
Yes, but then again so is roto and all forms of fantasy.
Congress wrote in an exemption for fantasy leagues, designed to preserve things like NCAA Tournament contests (with prizes). When then law is unclear, it’s not surprising that people would try to figure out what the boundary line is. There already were “second chance” fantasy leagues and the like.
In my opinion, DFS crosses that line and is quite a bit different than season long leagues where ownership of a player is only granted to one team. There is much more uncertainty in the DFS game and far less opportunity for skill to translate into success.
So it’s a lot like like poker, except the structure of the house take is different. But more like poker than like counting cards at blackjack, where you’re directly against the house.
Are there any DFS games which are drafted/auctioned like regular fantasy – in that only one DFS contestant in a contest “owns” a player?
So DFS bills itself as ‘NO COMMITMENT! INSTANT GRATIFICATION! ONE DAY LEAGUES!’ but to actually see any meaningful results, you have to play for 250+ days in the year.
There’s a sucker born every minute, I guess.
And with a month’s investment, you can earn “over $100.” I get that this is recreation, and most recreation doesn’t actually make you any money (quite the contrary, as eg any boat owner will tell you) but if I’m going to invest real cash and a chunk of my analytical time, I’m going to play the stock market. And if I’m going to blow money on recreation, at least with a boat I get to go outside.
This. The amount of time invested in DFS would be far better spent reading quarterly reports if people were truly interested in generating supplemental income. I would agree that it falls closer to recreation than investing on that spectrum. I think much of the pushback that commenters leave on this site related to DFS is due to the personality types you can expect to find on a site such as Fangraphs. People here tend to be analytical, non-emotional decision makers. The risk-reward balance of full season fantasy sports will likely be more appealing to this type of personality relative to DFS.
As for personal recreation choices, that is obviously up to the individual. But I’ll be hanging with joser and a few beers on his boat after a round of golf.
People who are actually good make quite a bit more.
“In the first half of the 2015 MLB season, 91 percent of DFS player profits were won by just 1.3 percent of players.
The sharks split into two groups (see chart):
■ The top 11 players paid on average $2 million in entry fees and profited $135,000 each. They accounted for 17 percent of all entry fees. The winningest player in our sample profited $400,000 on $3 million in entry fees.
■ The rest of the top 1.3 percent of players paid on average $9,100 in entry fees and profited $2,400 each, for a 27 percent ROI, which is extremely impressive. These contestants accounted for 23 percent of all entry fees and 77 percent of all profits.”
So that’s two groups of people among the winners: people doing it professionally (who made $135k in profit over half a season by playing in basically every game that they could), which is pretty good though admittedly not as much as necessarily an extremely skilled trader, and people who made a 27% ROI and $2400 with somewhat limited investment.
It’s certainly like other forms of gambling, with a somewhat stronger skill component than poker, and there are people who make a decent amount of money, again like poker.
The recreation component is key. I’m a risk averse individual and find DFS to be a much easier game to stomach. I played currency futures in college and even made a couple thousand, but the whole experience was terrifying.
27% ROI is impressive, but you have to be in the top 1.3% to get it, and there may be no guarantee that the winning strategy one year will work next year. How much consistency is there at the top? How accessible is that top 1% to people without either insider knowledge of betting patterns or access to the cuurent best DFS algorithms?
This is okay with stocks, since missing the top 1% just means settling for a 5%-10% return. With DFS, missing the top means winning basically nothing, and there is a high risk of losing lots of money.
I do hate the way they advertise it. Not only is it disingenuous, it’s not even the type of user they want. I can only assume their market research shows that people who have fluke wins early in their playing career are more likely to become long term players. So they’re just trying to coax something like five new users out of every 100 who sign up.
The comments on these posts always run so sour.
I’M SO SMART IF I DONT WIN THEN ITS GAMBLING AND ITS NOT MY FAULT I AM GOOD I SWEAR
I guess one of the places your argument hangs in the balance is whether one accepts that right process can create positive EV. Can you elaborate on what exactly you mean by right process? And is there data suggesting that any proposed “right” process has been shown as providing superior, non-hypothetical results? How can you be sure that your success in 2014 and lesser success were actually caused by the reasons you offer and not by random, coincidental action?
I suppose the same kinds of questions could be asked about season long fantasy. In both instances, anecdotally it seems like some competitors succeed more often than others, but that’s an assumption that is rife with the possibility of confirmation bias. I think it may ultimately become important in the case of a contest like the NFBC to be able to articulate exactly what skills are demonstrated in the course of season long fantasy play.
“The Right Process” here appears to mean entering thousands of contests a day with server banks dedicated to automatically optimizing lineups based on projected ownership percentages and pricing inefficiencies
Asleep yet?
The basic “right process” is taking advantage of the fact that you don’t want to maximize expected points, but expected chance of winning a tournament. So you want high variance lineups (lots of people on the same team), relatively rarely owned players, etc.
Process also isn’t a one-size-fits-all thing. There are days when it’s almost imperative to use Kershaw. There are days when you cannot use Kershaw. Not because of his price or matchup but because of others prices and matchups.
Without insider data, I can’t verify that a sound process still creates a positive EV (short of using a server farm). And as Thacker points out below, the best process might not necessarily produce the highest expected point total.
I’ll share my experience. I managed to win something almost 50% of my lineups, about double the average. Big winnings are far more elusive since the odds are like 3000 to 1. I only played the 3 dollar contests so the winnings past the top 20 are slim
Knowing what any individual will do in any 1 game is almost impossible. To win you generally have to have almost every player (well, 8 of 10) having good or great games. 3 or more sub-par performances and you are probably out of the money. The best position players will disappoint in 1/2-2/3 of their games. The odds are better playing the hot players with favorable matchups, but DFS pricing goes way up on these guys, and pretty much everyone plays them (well 25% & up).
At the end of the day, if you play the best or hottest players at each position, you find yourself with at least 2-3 spots where you have to take the bottom of the barrel. Since a large percentage of teams have the best and hottest players as well, it frequently comes down to guys like Stephen Drew hitting 3 HR in a game to put you in real money. Thats all luck
Another tactic is stacking. Picking most of the position players from a really good hitting team against a real bad pitcher. Many other contestants use the same tactic, so with the obvious stacks, it comes down to the 2 other position players you must choose from other teams that can put you ahead of the pack. The obvious stacks also come with a price premium. Sometimes the best results is picking a team at random that does not have an obvious stack, and hoping for the best as fewer contestants will be stacking this team
To make room for better hitters, given the high price of pitching, you may also resort to picking one lower cost pitchers and hoping for a good outing, which happens maybe 1/3 of the time. However, K’s really dictate a lot of a pitchers points, so you need high K guys. Just picking the 2 best pitchers, or highest priced pitchers does not always work since these guys disappoint as much as 1/3 of the time, and you have little left for a lineup of solid hitters.
There are a large number of combinations for each of the tactics. Which ones work is a matter of luck. The more lineups you can play, the better your odds of getting into some real money. I have seen some contestants with 200-300 lineups pretty much every day, which makes me wonder if they are employee’s being paid their salary in lineups or DFS own teams to hit the max entry number, which is the basis for the fixed prizes set before the contest(required by law to be considered non-gambling). Only DFS knows who they are. No regulation or supervision, so you just have to trust them that this is not the case
Perhaps some of us small thinkers might be less likely to “fundamentally misunderstand” the above framing of DFS, as a season-long process, if the barrage of DFS ads didn’t breathlessly portray outcomes such as one guy winning a million dollars or more based on a single play in one game.
In the big picture, under the above framing, the DFS sites are misleading, if not outright illegal, in their marketing. Therefore, in the big picture, under the above framing, actually, DFS isn’t getting hammered enough.
That pretty much is the way it’s played. Even Brad wrote that one day of big winnings accounted for much of his seasonal total. Everyone loses more than they win, so you play a lot of days and hope that you get the handful of relatively large payouts to make up for the majority of days where you lost money.
There are basically two ways to play DFS. I play large tournaments exclusively (GPPs). These offer a big reward to few users. So yes, you make your money via a couple big scores a season.
The other way involves a variation of H2H where you bet $1 for a 50/50 chance to make $1.80. There are a hundred variations of this with slightly different odds and rewards. In these, you’re aiming to win about 65% in order to profit.
Personally, I can’t overcome the rake in those latter contests. One winning day will ensure you beat the rake in GPPs.
but the things is, DFS takes the same percentage in each game, so your expected return in each game is the same. It really comes down to whether you are good at making at taking risks and making amazing lineups or if you play it safe and try to finish just top half or top third.
*sigh* Brad, we’re not the ones misunderstanding DFS. You are. Picking lineups of 23 different players 162 times is mathematically nothing whatsoever like picking 23 players and having their accumulated production over a 162 game sample. In the aggregate we expect Mike Trout to be phenomenal. Someone playing daily fantasy might get him on days when he’s phenomenal or days when he’s valueless, and because they’re not picking him every day, they’re in no way assured of getting a return representative of his seasonal value. We have great confidence in Trout’s performance over a full season barring serious injury. You cannot possibly have confidence in that same level of performance picking out fifteen of his games at random. You’re in no way guaranteed to balance the bad days with the good days simply by selecting him more than once if you’re not selecting him every time. You’re peddling a variation of the gambler’s fallacy where one day’s bad results supposedly will be balanced out by some other day’s good results, but that is not the case when you’re presumably picking different lineups each day. It’s fine that you enjoy DFS, but your rationalizations for it are not valid.
Tacking onto your good point, it just feels wrong that DFS All-Stars are Brad Miller or Stephen Drew having a random 3 HR day, whereas traditional fantasy All-Stars are guys like Donaldson, Arrieta, and Goldschmidt.
In a site that is first and foremost dedicated to discovering truth in baseball, I’m pretty sure I’ll always prefer the league that rewards Trout ownership over the one that endorses Zack Cozart because it’s an afternoon game at Coors.
Nope, it’s *mathematically* exactly like it. In terms of probability it’s exactly the same. Just imagine that you play in a league where you can add and drop players at will based on a budget, and multiple players can own the same guy– many full season soccer fantasy leagues are based this way. Brad’s merely arguing in the law of large numbers, that playing a great number of games with positive EV will over the long run almost surely make a profit. Not the gambler’s fallacy at all.
You seem to be the one peddling the gambler’s fallacy, saying that “we have great confidence in Trout’s performance over a full season.” So you think that if Trout has a bad run, he’s “guaranteed” to make it up over the next few months? That’s silly– if you reject the gambler’s fallacy you should definitely argue that each day of a fantasy season should be approached independently, without considering last week. And that means that the fantasy season is exactly like 162 daily contests back to back. To disagree with this is to invoke the gambler’s fallacy.
Your assumption that US fantasy baseball leagues are set up like some european fantasy soccer leagues is your weakness.
That’s wrong, as the size of the pool you’re drawing from absolutely has a dramatic effect on the probability of the results. With “n choose k,” in season long leagues “n” is relatively small whereas “n” in daily leagues constitutes the entire player pool.
I never said anything of the sort, so please do not intentionally misrepresent what I have written in a misguided attempt to suck up to Brad. We have great confidence in Trout’s performance over a full season because of the large sample size, which should outweigh any individual outlier performances. In DFS you’re not dealing with large sample sizes for any players, as you’re changing almost all of your lineup every day. That causes far, far more variance in results.
No, to disagree with that means having sufficient understanding of mathematics and probability to understand why that statement is ludicrously wrong. The size of the pool matters tremendously to probability calculations and there will obviously be far more variance in results when turning over your lineup daily. And that’s before we even get to the effects on probability from being able to choose when to play and having the option to cash out. Anyone who says that DFS and traditional fantasy baseball are the same has absolutely no idea what they’re talking about.
Support this statement if you’re going to say it. So far you’ve offered your opinion and some comments about probability that seem accurate but entirely irrelevant.
Roto and DFS aren’t exactly alike as Thacker says, but they are very similar. They’re both large sample baseball management sims measuring comparable statistics. The differences come in the smaller details. You have to split many hairs to say they aren’t similar games.
It doesn’t matter if I’m taking Trout in just 15 or 50 or 162 contests in a year. Presumably, I’m selecting him in scenarios where he’s expected to outperform his season averages. We have the necessary data and sample sizes to do this with confidence. We’ve been using the same principles to select streaming options in The Daily Grind since before DFS was even a thing.
*facepalm* Yes, it most certainly does. This past season Mike Trout went hitless in 46 games and hit a home run in 36 games. Those who had him in traditional fantasy leagues got both sets whereas daily fantasy players did not. If you were extraordinarily unlucky you could have gotten him just for the hitless games with no home run games. If you were on the opposite end you may have gotten him for all of the home run games and none of the hitless games. Let’s say that you selected Bryce Harper instead of Trout on the days when Trout hit his home runs, only Bryce Harper theoretically went hitless on all of those days.
In daily fantasy it is possible to select All-Star level hitters who accumulate Mendoza-line production if you’re unbelievably unlucky on when you pick them. It is also possible to select All-Star level hitters who dwarf Barry Bonds’ most ‘roided up season if you are unbelievably lucky on when you pick them. That’s because the variance in picking different lineups every day is so, so, so much greater than when most of your lineup stays the same throughout the season.
But that same argument can be applied to any individual player. I can select Trout first next year and get a .200/.300/.300 line in 56 PA if I’m extremely unlucky. In fact, it’s more likely that I get screwed in roto than DFS since I can simply not pick injured players.
Rather than focus on Trout, you should consider all top players as an asset class. Let’s say hitters over $5,000. My 15 Trout days COULD be extremely unlucky, it’s much less likely with the entire asset class especially because I can avoid most of the factors (namely injury and role reduction) that affect bad luck in traditional fantasy.
You’ve also referenced two scenarios that never happen. NEVER. No amount of good or bad luck will result in top hitters performing at the mendoza line or above the Bondsian Divide over a 250 contest sample.
Yes, that is precisely my point. Daily fantasy leagues have huge levels of variance for each spot in the lineup, meaning that you’re compounding that variance with each player that is required. That’s why it blows my mind that anyone would ever think that the two games are at all similar, because mathematically they’re enormously different. And again, this is before we even get to the enormous structural differences between the two.
Denying reality because you find it threatening to your argument is not a valid response. You’re on FanGraphs, so you cannot possibly tell me that you’re unaware of sequencing. Clearly it is possible to select “$5,000+” hitters who produce Mendoza line production over the course of a full season, and clearly it is possible for a different sequence to dwarf Barry Bonds’ best season. Those particular outcomes are extraordinarily unlikely, but those are merely extreme possibilities used to illustrate the vastly increased variance when picking between a pool of “$5,000+” players than when picking one for a full season. With extremely high variance and a fairly large number of competitors, someone is going to get lucky and post a huge total but there’s no way of telling who that will be.
If you played 162 DFS games and somehow only used Trout in games that he got 0 hits that was pretty close to an impossible outcome.
How do we know it’s not just luck or random noise? The algorithm players. The point of those algorithms is that they let people use a mathematical advantage over time based on player performance versus the pricing model of the sites. If they didn’t work, people would stop using them and you wouldn’t see players with dozens of entries. The algorithm players would not play anymore if they were not getting paid.
Whether it is “just like season long” or not exactly like season long misses the more salient point Brad is making — don’t look so much at the single day iteration, look at the application of a process to the game over the whole season. If someone only plays once, that’s a lottery ticket. If someone sets aside a bankroll to play the whole season and devotes time every day, it’s less like a lottery ticket.
The question is whether, going up against the Machine players, can an individual use a sophisticated enough process to win over time? Or at least a good enough process that the psychic value of playing is not outweighed by the losses. That is and should continue to be an individual decision.
The “algorithms” you reference aren’t primarily about player performance, they’re mostly about ownership rates and lineup manipulation. It’s gaming the system, not being better at predicting the sport of baseball.
Also, lots of people play the lottery every day. That doesn’t magically make it something other than a lottery. DFS has some skill involved, so it is not purely a lottery, but playing something 90% luck and 10% skill every day doesn’t actually increase the skill quotient.
“Clearly it is possible to select “$5,000+” hitters who produce Mendoza line production over the course of a full season, and clearly it is possible for a different sequence to dwarf Barry Bonds’ best season. ”
You’re referencing outcomes that are like 20 or 50 standard deviations away from average. Let’s change my previous of impossible to “approaching impossible.” Feel better?
jdbolick, I think you’re assuming way too much lineup stability in traditional roto baseball. A lot of leagues have bench spots and daily lineups. So often, only a few mainstays will be in the lineup for all of their PA’s. The others are dropped/added during the season, traded or traded for, or benched on certain days. So, e.g., if I used Michael Taylor for 50 PA’s this year, it’s mostly luck what I got out of him. But the sum of all my individual lineup decisions and add/drops is a much bigger sample. That means that if I make good decisions on individual days and individual lineups, it usually adds up to a good season. It is similar with DFS – if all the individual decisions are wisely made, after a while the results will reflect it.
Traditional roto baseball doesn’t have bench spots. As far as I know I was the first ever columnist on shallow Yahoo! and ESPN style leagues that do have bench spots (way back in ’98), so I’m definitely aware of their existence. Certainly they have substantial roster turnover, and even traditional rotisserie leagues have some as well, but we’re talking about multiple orders of magnitude less roster turnover than daily leagues where almost all if not all players are changed from one day to the next.
As for “good decisions,” you’re overestimating the ability to make those in daily fantasy. No one ever has reason to be particularly confident in any player on any particular day. It is certainly possible in DFS to improve your odds by using certain metrics and targeting certain players, but for the most part that is accomplished with lineup manipulation and stacking rather than doing a great job of picking the players. So “good decisions” don’t really accumulate in daily fantasy leagues the way you’re suggesting.
Actually I should clarify since what I said was wrong. They do have reserve spots, but in traditional A.L. or N.L. only leagues they are compromised of players you do not want to be using at all barring injury, so the lineup doesn’t change in that fashion very often.
I used “traditional” wrong. I simply meant year-long fantasy baseball.
What do you mean when you say “but for the most part that is accomplished with lineup manipulation and stacking rather than doing a great job of picking the players. So “good decisions” don’t really accumulate in daily fantasy leagues the way you’re suggesting?” If wisely-deployed stacking or lineup manipulation work over the long haul, why shouldn’t they be considered good decisions that accumulate?
Because they aren’t something the average person can replicate, as they involve paying for large numbers of entries to manipulate your odds of winning.
Anonymous made this point earlier by complaining that Brad Miller and Stephen Drew are the stars of daily fantasy. If Player X was the top first baseman on a particular day but was rostered on every entry, then choosing him gained no advantage. If Player X was rostered on half of all entries then an advantage was gained on half of the competition. If Player X was rostered on no other entries then an advantage was gained on all of the competition. Success in DFS isn’t primarily about picking great players, it’s about getting the ones no one else is using on the days when they happen to perform unusually well.
Thank you for the responses, jdbolick. I think I understand your point better now.
The folks that have brought us platforms in which to play season-long fantasy games have been master marketers of these daily leagues. They probably have to offset enormous investments in bandwidth, talking heads, and database management. It helps their revenue that the product is addictive.
The problem is that virtually all DFS sites have enormous rakes that make it almost impossible to achieve an actual +EV, at least without employing strategies requiring huge bankrolls that involve “covering” a tournament with dozens of overlapping entries per night.
That’s not likely to get better as DFS gets “better” regulated, as casinos often use regulatory scrutiny as a way to fix prices to prevent new competitors from undercutting them in a race to the bottom (which is why all Las Vegas casinos, for instance, employ the same odds for slots and table games and keep a standardized -110 rake on ATS sports bets).
I’d be very interested to see if in the legal investigation of DraftKings/FanDuel the NY AG uncovers evidence of collusion between the sites to not compete with one another via lower rakes. Wouldn’t shock me in the least.
The rakes are a problem, but I don’t think they need collusion to leave them so high. I’ve been told that the industry is not profitable because of the cost of customer acquisition. The rakes are essentially the only source of income these sites can capitalize. Yes, they want you to “deposit” money, and they’ll use it too, but it’s not quite theirs.
I’m interested in what’s keeping a hypothetical zero rake site from happening. If there is no regulation of the industry it seems like there is tacit collusion keeping the rake at what it is.
How does a zero rake site profit? It would have to be Yahoo or CBS. Nobody else has the necessary “free” advertising to attract users. And no other site could actually benefit from selling ad space on the DFS platform.
1. I saw something on fanduel that seemed to be addressing the big sharks in the minnow pool problem, but I can’t find it on the site. Seemed to be placing a daily limit on an individual players’ entries in low stakes games. There may be ways around it and I apologize for not having the details, but at least it is addressing the issue.
2. You should be playing these games for fun. If you are trying to play these games to win real money you should have a bankroll and an algorithm. Because that is who ou are competing against.
The ‘recreational value’ argument loses credibility when the singular message of the barrage of advertisements for DFS is not ‘Hey this is a fun thing to do’, but rather ‘BECOME A MILLIONAIRE IN ONE DAY!’
While I would be interested in a low- or no-stakes version of the game for bragging rights, the DFS juggernauts and industry as a whole have chosen to pin the appeal of their game to ‘WIN BIG MONEYS’.
So that is what I will judge it on.
I’m here late but if you notice during the MLB season especially on those tie-in ads for DK on MLB network they would say things like “turn your knowledge into money” and they focused on value picks and things.
It was during the NFL season that this “get rich quick” aspect of it. Ultimately you can play freerolls and $1 contests until the cows come home.
Sure you might hit on a GPP and win some money but you can have lots of fun playing DFS for little or no money and just try to prove how much you know baseball in a different process from your regular league
This may have already been posted, sorry if I missed it, but one HUGE difference between DFS and traditional roto is that in DFS you have to account for the Casino’s take.
It’s kind of like playing on CBS where you have to pay a bunch of money to use their lousy platform.