Psychology of Addictive Behaviors
© 2021 American Psychological Association
ISSN: 0893-164X

2021, Vol. 35, No. 8, 948–960
https://doi.org/10.1037/adb0000715

A Large-Scale Prospective Study of Big Wins and Their Relationship
With Future Financial and Time Involvement in Actual Daily
Fantasy Sports Contests
Timothy C. Edson1, 2, Matthew A. Tom1, 2, Kahlil S. Philander3, 4, Eric R. Louderback1, and
Debi A. LaPlante1, 2

Division on Addiction, Cambridge Health Alliance
Department of Psychiatry, Harvard Medical School
School of Hospitality Business Management, Washington State University
Department of Psychology, University of Sydney

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

This article was published Online First March 18, 2021.
https://orcid.org/0000-0003-2702-0098
Timothy C. Edson
https://orcid.org/0000-0001-6574-3110
Matthew A. Tom
https://orcid.org/0000-0002-0747-0772
Kahlil S. Philander
https://orcid.org/0000-0002-9754-9790
Eric R. Louderback
Debi A. LaPlante
https://orcid.org/0000-0001-5418-5504
This study was pre-registered on the Open Science Framework (see https://
osf.io/34t9s).
This research was supported by a research contract between the Division
on Addiction and DraftKings, Inc. (DraftKings). DraftKings is a major
provider of daily fantasy sports offerings. DraftKings had no involvement
with the development of our research questions or protocol. They did not see
any associated materials (i.e., retrieved studies, charted data, manuscripts in
preparation) while the study was in progress and will not have any editorial
rights to any resulting manuscripts. DraftKings communication about this
work requires approval from the Division on Addiction.
As of the publication date of this article, the Division on Addiction was
receiving funding from EPIC Risk Management; the Foundation for Advancing Alcohol Responsibility; Entain PLC (formerly GVC Holdings PLC);
Massachusetts Department of Public Health, Bureau of Substance Addiction
Services via St. Francis House; Massachusetts Department of Public Health,
Office of Problem Gambling Services via Health Resources in Action; MGM
Resorts International via the University of Nevada, Las Vegas; National
Institutes of Health (National Institute of General Medical Sciences and
National Institute on Drug Abuse) via The Healing Lodge of the Seven
Nations; Substance Abuse and Mental Health ServicesAdministration via the
Addiction Treatment Center of New England; and Substance Abuse and
Mental Health Services Administration via the Gavin Foundation. During the
past 5 years, the Division on Addiction has received funding from DraftKings, Inc., The Foundation for Advancing Alcohol Responsibility (FAAR),
The Healing Lodge of the Seven Nations via the Indian Health Service with
funds approved by the National Institute of General Medical Sciences,
National Institutes of Health; The Integrated Centre on Addiction Prevention
and Treatment of the Tung Wah Group of Hospitals; the Gavin Foundation
via the Substance Abuse and Mental Health Services Administration
(SAMHSA); University of Nevada, Las Vegas via MGM Resorts International; David H. Bor Library Fund, Cambridge Health Alliance; Massachusetts Department of Substance Abuse Services; Massachusetts Gaming
Commission; Caesars Enterprise; Massachusetts Council on Compulsive
Gambling; Gold Foundation; ABMRF/Foundation for Alcohol Research;
Aarhus University Hospital via Danish Council for Independent Research;
bwin Interactive Entertainment AG; Heineken USA Incorporated; National
Center for Responsible Gaming; Fenway Health; and Entain PLC.
Timothy C. Edson has no financial disclosures or conflicts of interest to declare.
During the past 5 years, Matthew A. Tom has received travel funds from
The Hong Kong Jockey Club via the Tung Wah Group of Hospitals. During

the past 5 years, Tung Wah Group of Hospitals’ Integrated Centre on
Addiction Prevention and Treatment (ICAPT) has received funding from
the Board Management of Tung Wah Group of Hospitals; The Hong Kong
Jockey Club; Beat Drugs Fund which is managed by the Narcotics Division
of Security Bureau of the HKSAR Government; and the Partnership Fund for
the Disadvantaged which is managed by the Social Welfare Department of
the HKSAR Government.
During the past 5 years, Kahlil S. Philander has received research funds
from the Washington State Gaming Commission, Manitoba Gambling
Research Program, UNLV International Centre for Gaming Regulation,
U.S.-Japan Business Council, Wynn Resorts, Wymac Gaming, Board of
Regents of the Nevada System of Higher Education. He has received
consulting payments from British Columbia Lottery Corporation, Responsible Gambling Council of Canada, the Commonwealth of The Bahamas,
Eilers & Krejcik Gaming, West Virginia Lottery, Indiana Gaming Commission, and iDevelopment and Economic Association. He has received reimbursement for travel from the National Council for Problem Gambling,
International Association of Gaming Advisors, National Center for Responsible Gambling, North American State and Provincial Lottery Association,
Evergreen Council on Problem Gambling, Global Gaming Expo Asia, and
Alberta Gambling Research Institute.
During the past 5 years, Eric R. Louderback has received research funding
from a grant issued by the National Science Foundation. Dr. Louderback's
research and conference travel has been financially supported by a Dean’s
Research Fellowship from the University of Miami College of Arts & Sciences,
and he has received travel support funds from the Hebrew University of
Jerusalem. During the past 5 years, Debi A. LaPlante has received speaker
honoraria and travel support from the National Center for Responsible Gaming
(NCRG) and the National Collegiate Athletic Association. She has served as a
paid grant reviewer for NCRG and received honoraria funds for preparation of a
book chapter from Universite Laval. She is a non-paid board member of the
New Hampshire Council on Problem Gambling.
Some of the raw data used for this study (i.e., the 2014 cohort data) have
been used in other published manuscripts, including: (a) an epidemiological
description of DFS player behaviors, (b) a study of longitudinal playing
trends of DFS playing behaviors, (c) a study assessing relationships between
social DFS behaviors and risky play, and (d) a study of natural groups of DFS
players. The current study is unique from these other studies because (a) it
examines novel research questions; (b) uses a matched pairs approach; (c)
tests different outcomes; and (d) uses survival analysis, count models, and
panel regression.
Correspondence concerning this article should be addressed to Timothy
C. Edson, Research & Evaluation Scientist, Division on Addiction, Suite
630, Cambridge Health Alliance, Malden, MA 02148, United States.
Email: tedson@cha.harvard.edu

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

LARGE-SCALE PROSPECTIVE STUDY OF BIG WINS

Objective: Early big wins might have a psychological impact upon gamblers that increases their likelihood
of intemperate gambling; however, there has been a paucity of empirical research examining this effect
using actual gambling data. Method: We assessed the effects of daily fantasy sports (DFS) big wins on
subsequent play by analyzing a prospective dataset from a major DFS provider (N = 34,596 DFS
subscribers) representing over 18 million entries into DFS contests. Results: We found that experiencing
a big win in DFS is associated with subsequently increased DFS engagement (i.e., increased contest entry
fees and contest entries) and losses (i.e., higher net loss). However, the effect of a big win on engagement
and losses decays over time. Whereas theorists have highlighted the effects of early big wins, our analyses
indicated that later big wins had a relatively stronger effect on DFS engagement. Sensitivity analyses
confirmed the robustness of most results, with somewhat greater support for big wins’ effects on
engagement metrics than losses. Conclusion: Our results collectively indicate the existence of a big
win effect in DFS. For some players, big wins might instill unrealistic expectations about future probabilities
of winning and lead to increased—and potentially excessive—engagement. Explanations from cognitive
psychology (e.g., the illusion of control) and behavioral psychology (e.g., operant conditioning) might help
to explain the big win effect.

Public Health Significance Statement
Our study suggests that for some DFS players, big wins can promote ongoing and potentially risky DFS
engagement behaviors, including future losses. These findings have important implications for DFS
operators and stakeholders, who should consider the promotion of healthy behaviors for players
following large wins (e.g., setting some winnings aside for necessary expenses). The results also
emphasize the need for more research into the relationship between big wins and DFS-related problems.
Keywords: gambling, big win effect, early big win, daily fantasy sports, prospective study
Supplemental materials: https://doi.org/10.1037/adb0000715.supp

Formative experiences in individuals’ lives can have a lasting
impression on future behaviors (Do et al., 2018; Muthukrishnan
& Kardes, 2001; Tversky & Kahneman, 1971). Gambling studies
have examined whether potentially formative experiences such as
winning a large prize during one’s first few gambling sessions
(Weatherly et al., 2004) might predict future risky gambling
behaviors. Early big wins are defined as subjectively large prizes
won by gamblers during an initial and formative period of their
gambling experience (Custer & Milt, 1985). The early big win
effect on gambling behavior seems plausible and could be rooted
in biases associated with both availability and representativeness
heuristics (Tversky & Kahneman, 1971, 1973). The large size of
the win, for example, could impact the availability heuristic, as
the subjectively meaningful sum of money is likely to make the
outcome of winning more salient, relative to other wagers. The
“early” aspect of the phenomenon could impact the representativeness heuristic, as the limited number of trials relative to the
large win event might falsely lead individuals to believe that the
outcome is likely to reoccur often.
Although there is no consensus as to what constitutes “early”
or “big” (Weatherly et al., 2004), the first writings about early
big wins derive from clinical observations by Custer (1984),
who suggested that early big wins are events occurring during an
initial phase of gambling and on a scale that is nearly equal to or
greater than one’s annual salary. Custer identified an “early big
win effect” as a big win during one’s first few gambling
experiences that instills unreasonable optimism about continued
winning, leading to an increase in gambling, gambling losses,
and eventually, the development of gambling-related problems
(1984, pp. 36–37). Custer and Milt (1985) later described the
definition of “big” in a way that it could be taken to mean any

win that is comparable to some quotient of one’s annual earnings
(i.e., a month’s salary or more; also see Weatherly et al., 2004).
Researchers have observed reports of early big wins among as
many as half of gamblers with problems in clinical settings, and
suggested that they play a role in early persistence in gambling,
due in part to distorting the gambler’s perception of his or her
ability to determine outcomes (Lesieur & Custer, 1984).
Although there is very limited prospective empirical evidence
about these clinical claims, many researchers have suggested
that early big wins play a role in the development of gamblingrelated problems, particularly among young people (Derevensky
& Gilbeau, 2015; Griffiths & Auer, 2016; Turner et al., 2006,
2008; Williams et al., 2015).
In the present study, we used pre-registered hypotheses and
methods to examine the role of big wins in future daily fantasy
sports (DFS) engagement and losses, based on DFS player
records from one of the largest DFS providers (DraftKings,
Inc.). DFS is a partially skill-based online gaming activity which
shares many attributes with popular gambling activities that are
also partially skilled-based such as poker and sports betting. We
studied DFS because it overlaps and shares characteristics with
traditional gambling games (see Nelson et al., 2019), is a relatively new gaming activity that many users are likely experiencing for the first time, and contains large tournaments with jackpot
sized payouts each week, thus providing a reasonable sample of
big winners to analyze. For this work, we introduce two competing definitions of the term “big win”—one that is an adaptation of
Custer (1984)'s original definition (i.e., Custer definition; wins
greater than or equal to $1,000), and another that is based on the
ratio of players’ largest prize won to how much they spent on an
entry fee (i.e., Prize Ratio definition).

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

EDSON, TOM, PHILANDER, LOUDERBACK, AND LAPLANTE

Using these two operationalizations of big wins, our analysis of
the DFS player records proceeds in four primary ways, following
our pre-registered research plan. First, we provide a census of
DFS big wins by examining three cohorts of DFS players who
first deposited on the DraftKings website during 2013, 2014, and
2015, to assess how many experienced a big win by both
definitions. Second, we examine the effects of big wins on
DFS play by identifying whether players who experience a big
win show subsequently increased DFS engagement compared to
those who do not experience a big win. Third, we investigate the
durability of observed big win effects by exploring whether they
diminish over time. Fourth, we study the relevance of time (to big
win) in the effects of big wins on DFS play.

Method
Study Design
We used a longitudinal prospective cohort design with continuous
time points comprising weekly aggregates of players’ online DFS
data from DraftKings (one of the largest DFS operators) between
August 1, 2013 and December 26, 2016 (i.e., the analytic study
period). Because American football (i.e., the NFL) is the most
popular DFS offering (Gillies, 2016), we purposefully selected
cohorts of players who made their first DFS deposits near the
beginning of the 2013, 2014, and 2015 NFL seasons, respectively.
As such, our initial pool of participants includes all players who
made their first deposit on DraftKings between August 1, 2013 and
September 30, 2013 (2013 cohort, N = 12,041), as well as commensurately sized samples of players who first deposited money on
DraftKings between August 1, 2014 and September 30, 2014 (2014
cohort, N = 12,041) and between August 1, 2015 and September
30, 2015 (2015 cohort, N = 12,041). This equates to an initial
sample of 36,123 players. Of these players, 1,527 did not enter any
DraftKings contests during the study period. We excluded these
players and therefore our final analytic sample comprised 34,596
players.
DraftKings provided us with each player’s initial deposit date and
entire contest history (including their number of contest entries,
amount spent on entry fees, and amount won in prizes) from that
initial deposit date to December 31, 2016. In addition to contests that
require an entry fee, DraftKings offers free contests that either pay a
tiny fraction of the people who enter or do not involve any payouts
(i.e., play-for-fun). While we initially specified in our pre-registration
that we would exclude free contests, we ultimately decided to include
free contests in all calculations given their potential for actual cash
payouts, which could affect key metrics such as net loss.

Dataset Creation Procedures
For each player, we generated weekly aggregates of their playing
behavior (i.e., total entry fees, net loss, and number of contest
entries for each week they played). Because we believed that
many players would be participating in primarily NFL-based
DFS contests, for the purpose of this paper, we defined a “week”
as a seven-day period beginning on a Tuesday at midnight
(e.g., 2013-08-07 00:00:00). Under this assumption, for each
player, we calculated weekly aggregates for contest entry fees,
winnings, and number of contest entries. For each week, we also
calculated each player’s net loss (i.e., that week’s total entry fees

minus that week’s total winnings). We considered the players’ “first
week” as the week they made their first deposit on DraftKings.
Players who signed up earlier on during the recruitment period had
more available weeks than players who signed up later (see preregistration for additional details of how we constructed weekly
aggregates).

Measures
Identifying Variables. Player ID: Identification numbers used
to distinguish one player from another.
Week number t: Given a player/Player ID, a week number of 1
means the week that the player made his first deposit into his
DraftKings account. A week number of 2 means the week after
the week of first deposit. “Week t” refers to (t−1) weeks after the
week of first deposit.
Dependent Variables. Our main dependent variables consisted of:
Current week’s entry fees: The total amount a player spent on
entry fees in Week t.
Current week’s net loss: The sum of all entry fees minus the
sum of all cash winnings over the contests the player entered
with contest start times in Week t.
Current week’s number of entries: The total number of entries
submitted into contests in Week t.
Persistence: Persistence is, among big winners, the number of
consecutive months they play (i.e., in which they entered at
least one paid DFS contest) during their active duration
following the big win. We tracked persistence for a total of
six 4-week (i.e., six prototypical months) periods.
Frequency: Frequency is, among big winners, 100 times the
number of active days on DraftKings during the 24 weeks
(i.e., six prototypical months) following their big win divided
by their total active duration (i.e., number of days, inclusive,
between first active day and last active day) during this same
24-week period.
Independent Variables. Our main independent variable was
experiencing a big win. As discussed previously, we defined the
term “big win” in two different ways:
Big Win Measure 1 (Adaptation of Custer & Milt’s definition;
hereafter, Custer big win):
The Custer and Milt (1985) extension of writings by Custer (1984)
on big wins among individuals experiencing gambling problems
informed our first measurement approach to empirically defining a
“big win.” This definition of “big win” is based on the average
monthly discretionary income in the U.S., which we calculated as
(Income after taxes – Average annual expenditures divided by 12)
based on the U.S. Bureau of Labor Statistics data table “Tables 1,800.
Region of residence: Average annual expenditures and characteristics,
Consumer Expenditure Survey, 2013–2014.” The resulting average
annual discretionary income after taxes for all consumer units (average
annual income after taxes minus food, housing, transport, clothing,
and healthcare) is approximately $18,532; dividing this figure by
12 months is approximately $1,544. Taking the smallest possible
value that is within the same order of magnitude as this figure
(i.e., $1,000), we classified a big win as a single prize that equals

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

LARGE-SCALE PROSPECTIVE STUDY OF BIG WINS

or exceeds $1,000. These players were identified with an indicator
variable, BigWin1, coded as “1” for players who experienced a big win
during the prior week and “0” otherwise.
Big Win Measure 2 (Prize Ratio-Based Measure; hereafter, Prize
Ratio big win)
Our second measure of “big win” begins with the idea of a prize
ratio, the size of a prize divided by the largest entry fee the player had
ever paid up to that week. For example, if the largest entry fee a player
ever paid for was $5.00 at the time the player won a $48.00 prize in a
contest, then the prize ratio would be $48.00/$5.00 = 9.6. Even if the
$48.00 prize came from a $1.00 contest, because the player has a $5.00
entry in his past, the prize ratio for that win would still be 9.6 and not
48.0 (see pre-registration for additional details of the prize ratio
calculation). We excluded prize ratios less than or equal to zero, as
well as instances where the largest entry fee ever paid up to that week
was $0.00. For each player cohort (i.e., 2013, 2014, and 2015)
separately, we conducted a percentile ranking of prize ratios and
classified the largest 1%1 of prize ratios within the cohort as “big wins”
(Five number summary of prize ratio for the full sample of qualified
Prize Ratio big wins: Min: 10.93; 25%: 14.81; Median: 20.00; 75%:
37.04; Max: 20000.00).
Matching Covariates. Prior Week’s Engagement (Total Entry
Fees and Number of Entries)
Prior research has noted a close relationship between current
activity and subsequent activity, particulary in the domain of sports
gambling (Ma et al., 2014). Therefore, we matched players based
on prior DFS engagement in terms of total entry fees and total
number of entries submitted into contests as of the end of the week
prior to the week of the big win.
Cumulative Net Loss
Because past financial engagement and success rates can inform
future expectations and engagement patterns, we also matched
players based on their cumulative net loss (i.e., running net loss
since they first deposited) as of the end of the week prior to the week
of the big win.

Analytic Strategy
Significance Criteria
We regarded p-values less than 0.05 as criteria for statistically
significant results. All reported p-values relied on two-tailed tests of
significance.

Hypotheses and Analytic Plan
Hypothesis 1: A small percentage (less than 0.1%) of DFS
players will experience a big win, diversely defined.
To examine how each definition of big wins classifies DFS
players (Hypothesis 1), we noted which players had a big win
based on just the Custer $1,000 threshold, which players had a big
win based on just the Prize Ratio criterion, which players had a
big win based on both criteria, and which players did not have a
big win. Based on previous research with actual DFS player data
(Wiley et al., 2020), we anticipated that a small percentage (less
than 0.1%) of DFS players would experience a big win, diversely
defined.

Hypothesis 2: Players who experience a big win will exhibit
more DFS engagement (i.e., higher weekly total entry fees,
weekly net loss, and weekly number of entries) in the following week when compared with players who do not experience
a big win.
Our primary interest in Hypothesis 2 was to test how big wins
influence engagement outcomes. However, we expected early big
wins to be very infrequent, leaving us with potentially small sample
sizes, issues with statistical power, limited variation in the big win
measures, and model estimation problems when using conventional
regression techniques (Hair et al., 2010; Lu et al., 2018). Moreover,
we sought to accurately compare how engagement outcomes differ
for people who experience a big win with those who do not, but the
use of observational data prevented random assignment (Shah et al.,
2005) of the treatment variable (i.e., experiencing a big win). To
remedy these issues, we followed a matched pair approach. More
specifically, we used Mahalanobis distance matching (Chang et al.,
2018; Jung & Yoon, 2018) to pair big winners with matching
controls.
First, we identified all big winners for each week of the study
period, starting at Week 1. We began at Week 1 because Turner
et al. (2008) found that an early big win in the first few days of one’s
first real gambling experience was associated with future problem
gambling. For Week 1 big winners (nCUST = 25, nPR = 280), we
picked matched controls randomly because all matching covariates
are equal to zero. Next, we used the matchit() function from the
MatchIt package in R (Ho et al., 2011) to match each big winner
from subsequent weeks with another player who played that same
week but did not experience a big win on the aforementioned
covariates (i.e., total entry fees, number of entries, and cumulative
net loss) based on the Mahalanobis distance (King & Nielsen, 2019;
Stuart et al., 2011). We set the parameter ratio = 1 so that each
treatment case is matched to only one control case. If there were
multiple viable matching controls (i.e., multiple matches), we chose
one randomly. The MatchIt package includes several methods for
matching that are applicable for the present dataset: exact, subclassification, nearest neighbor, optimal, and coarsened exact. We
compared each of these methods and utilized the nearest neighbor
method, which yielded the lowest mean difference between the big
winners and corresponding matching controls. We then checked the
summary of balance table to ensure that, for each of the covariates,
the mean difference between the big winners and their matched
controls was less than the corresponding mean difference between
the big winners and the rest of the unmatched sample of players (see
Supplementary materials). After applying exclusion criteria (see
Matching Filters, below) and conducting the matching process, we
arrived at a final Custer matched sample of 1,654 players (827
Custer big winners and 827 matched controls) and a final Prize Ratio
matched sample of 4,262 players (2,131 Prize ratio big winners and
2,131 matched controls).
Our approach to determining the percentage threshold for prize ratio is
grounded in methods employed in research on actual DFS playing behavior
(Nelson et al., 2019; Edson & LaPlante, 2020) for separating more involved
players. Specifically, we took the percentile rankings of all prize ratios in the
analytic data greater than 1.0 and the minimum value of prize ratio within
each percentile. Graphing these minimums along a histogram, the top
percentile evidenced the largest jump in terms of separation from other
percentiles (see the Supplementary materials for histogram figure).

EDSON, TOM, PHILANDER, LOUDERBACK, AND LAPLANTE

After conducting the matching procedure, we performed matched
pairs t-tests (see Moore et al., 2009, p. 428) and assessed mean
differences in the (a) entry fees, (b) net losses, and (c) numbers of
entries of the big winners and their matching controls during their
week after the big wins (e.g., if the big win occurred in Week 1, then
we compare the big winner and his matched control on Week 2
results; if the big win occurred in Week 43, then we compare the two
players on their Week 44 results). We assessed the effect size of
matched pair differences using Cohen’s d (Cohen, 1988).

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

Hypothesis 3: For players who experience a big win, the big
win’s effect on DFS engagement tends to diminish over time.
To test Hypothesis 3, we tested for differences on the three DFS
engagement measures in the weeks beyond the subsequent week
following the early big win. Specifically, in addition to testing mean
differences of DFS engagement between big winners and their
matched controls in the week following the big win, we also
used matched pairs t-tests to assess mean differences between the
DFS engagement measures of big winners and their matched
controls in the second week following the big win, third week
following the big win, etc., up to the 24th week following the big
win (i.e., at least six prototypical months, or rather six 4-week
[prototypical month] periods). We graphed the Cohen’s d’s derived
from these analyses, where the x-axis reflected the week number
(starting with Week 1 after the week of the big win and ending with
Week 24 after the week of the big win) and the y-axis reflected the
Cohen’s d value for that week number. We also indicated the weeks
in which the associated t-value for each Cohen’s d is significantly
greater than 0. We examined this graph to determine whether the big
win effect diminishes or dissipates over time (in terms of decreasing
values of Cohen’s d or decreasing observations of statistical significance over time, respectively).
Hypothesis 4: Among big winners, shorter time from first
deposit to a big win will be associated with more persistence
(i.e., more months played consecutively during their active
duration) and greater frequency (i.e., percentage of active
playing days during their active duration) following the big win.
To test how persistence may vary depending on how early in a
player’s DFS experience the big win occurred, we first constructed
two new datasets. Dataset 1 included the subgroup of big winners
from the Custer matched sample who experienced an initial Custer
big win during their rookie NFL season (n = 371) and Dataset 2
included the subgroup of players from the Prize Ratio matched
sample who experienced an initial Prize Ratio big win during their
rookie NFL season (n = 1,439). We included only players who
experienced an initial big win during their rookie NFL season in
these analyses, given the potentially confounding effects of NFL
seasonality that exists for this DFS sample (Edson & LaPlante,
2020) on persistence trends. To measure when the big win occurred,
we created a new variable in each of the two datasets that is assigned
a numeric value based on the week number in that player’s DFS
experience in which the big win event occurred.
We then created six new variables, Month 1 through Month 6, for
each player that correspond to each prototypical month (i.e., 4-week
period) following their personal big win event. These variables
corresponded to the weeks 1 through 4, 5 through 8, 9 through 12, 13

through 16, 17 through 20, and 21 through 24 after the week of the
big win, respectively. For example, if a player experienced a big win
in his Week 6, then Month 1 would cover the player’s Week 7
through Week 10, and Month 2 would cover the player’s Week 11
through Week 14. We set each Month variable equal to “1” if the
player had at least one entry into a contest during those weeks and
“0” if they did not enter a contest during that 4-week period.
We first created a survival curve based on the non-parametric
Kaplan–Meier estimator (Efron, 1988) with months on the x-axis
and the probability of persistence on the y-axis. This plot shows the
percentage of big winners who stopped playing at each month,
beginning at the month following the big win and the subsequent
5 months.
To empirically assess how the week of the big win affects
persistence, we created a survival curve with a Cox regression
model and hazard rate (Cox, 1972) with the week of the big win as
the predictor and the number of months of continued play as the
outcome variable. This plot shows how the week number in which
the big win occurred was associated with persistence in the six
months following the big win event. Here, we estimated two Cox
regression models: (a) a model with only the independent variable
(week of big win) predicting the hazard function and (b) a model
with added controls for frequency of play and entry fees per day
(i.e., total entry fees divided by number of active days) in all
available weeks leading up to, and including, the week of the big
win event. This second model thus controlled for potentially confounding effects of time and financial engagement across the sample
of big winners.
For these same subsamples of big winners, to test how players’
general engagement in DFS may vary depending on how early in
their DFS experience the big win occurred, we employed ordinary
least squares (OLS) linear regression to model the frequency of days
players are active during the 24-week period following the big win,
with week of big win as the sole independent variable. Next, we
estimated the same regression model and added control variables for
frequency of play and entry fees per day (i.e., total entry fees divided
by number of active days) in all available weeks leading up to, and
including, the week of the big win. This model thus controlled for
potentially confounding effects of time and financial engagement
across the sample of big winners. We conducted the necessary tests
(e.g., residual plots, quantile–quantile plots) to ensure that all major
assumptions of OLS were satisfied (e.g., homoscedasticity, no
autocorrelation, normality in the distribution of residuals).

Matching Filters
For Hypothesis 1, we considered all big wins that met the
aforementioned Custer & Prize Ratio definitions. For subsequent
hypotheses we needed to consider: (a) the presence of a sufficient
follow-up period; and (b) the potential—for individuals who experience multiple big wins—that a previous big win could bias the
effect of a subsequent big win and/or one’s eligibility as a matched
control. As such, for Hypotheses 2 through 4, we considered players
eligible as big winners if: (a) they experienced their first big win by
the respective definition for the given week and (b) they had at least
24 following weeks where they were eligible to play AND did not
experience another big win by the respective definition during those
weeks. We considered players eligible as matched controls if: (a)
they had not experienced a big win by the respective definition

LARGE-SCALE PROSPECTIVE STUDY OF BIG WINS

before and during the given week, (b) had at least 24 following
weeks where they were eligible to play, and (c) did not experience a
big win by the respective definition during those weeks.

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

Sensitivity Analyses
Supplementing our main analyses, and to test for the robustness of
the findings from the main analyses, we completed sensitivity
analyses that: (a) utilize an alternative methodological approach
to test our hypotheses by using fixed effects time series regression
models and derivations of survival analyses, (b) control for highly
profitable players (i.e., professionals), and (c) test Hypothesis 4 with
a count (i.e., Poisson regression) model. We also extend our planned
confirmatory analyses by conducting an unplanned exploratory
examination of the relationship between experiencing a big win
and a proxy measure for potential DFS-related problems (i.e., voluntary self-exclusion [VSE] from the DFS website; Caillon et al.,
2019; LaBrie & Shaffer, 2011).

Human Subjects
The Cambridge Health Alliance Institutional Review Board
(IRB) determined that this study was exempt from IRB review.

Results

(N = 4,262)2 likewise submitted more entries (Mean[SD]BW =
34.7[92.9]; Mean[SD]MC = 14.8[44.7]; t[2130] = 8.9, p < .05,
Cohend = 0.43), paid more entry fees (Mean[SD]BW = 362.5
[1543.7]; Mean[SD]MC = 112.3[651.4]; t[2130] = 6.9, p < .05,
Cohend = 0.27), and had a greater net loss (Mean[SD]BW =
145.0[672.0]; Mean[SD]MC = 28.9[258.7]; t[2130] = 7.4, p < .05,
Cohend = 0.22) compared to their matched controls in the week
following the big win.

Investigating the Durability of Big Win Effects on Subsequent DFS Engagement
Figure 1 shows Cohen’s d standardized effect sizes for the
number of entries, entry fees, and net loss for the subsequent
week (i.e., the week following the big win event; Week 1) and
for the following 23 weeks, for both big win definitions. For all
outcomes, the trends for both big win definitions indicate a mostly
persistent and statistically significant difference between cases and
controls, but also show a decreasing size of Cohen’s d (i.e., diminishing effect) and/or increased likelihood of non-significant corresponding t-values (i.e., dissipation effect) as time progresses.
Notably, the effect of experiencing a big win by either definition
on all outcomes (including net loss) does not begin to dissipate until
after several weeks. The effect of Custer big wins on entries and
entry fees does not dissipate at all during the 24-week follow-up
period.

Planned Confirmatory Analyses
Providing a Census of DFS Big Wins
Our analytic sample includes 34,596 DFS players who collectively submitted 18,331,156 entries into DFS contests between
August 1, 2013 and December 26, 2016 (i.e., the analytic study
period). Out of the 34,596 DFS players in our analytic sample, 3,335
of them (9.6%) experienced at least one type of big win. Of these
3,335 players who experienced a big win, 1,862 (55.8%) experienced only a Prize Ratio big win, 464 (13.9%) experienced only a
Custer big win, and 1,009 (30.3%) experienced both a Prize Ratio
and Custer big win during the study period. In all, 1,473 players
(4.3% of the analytic sample; 44.2% of big winners) experienced at
least one Custer big win and 2,871 players (8.3% of the analytic
sample; 86.1% of big winners) experienced at least one Prize Ratio
big win during the study period. A descriptive analysis of big wins
indicates that they are not necessarily mutually exclusive to either
Custer’s definition or the Prize Ratio definition. Out of all big wins
in our study (N = 10,600), 5,814 (54.9%) qualified under just the
Custer definition, 3,077 (29.0%) qualified under just the Prize Ratio
definition, and 1,709 (16.1%) qualified under both definitions.

Examining the Effects of Big Wins on Subsequent DFS
Engagement
In the week immediately following their big win, compared to
their matched controls, Custer big winners (N = 1,654) submitted
more entries (Mean[SD]BW = 56.5[121.2]; Mean[SD]MC = 25.0
[84.1]; t[826] = 6.2, p < .05, Cohend = .50), paid more entry
fees (Mean[SD]BW = 550.1[795.8]; Mean[SD]MC = 149.9[325.3];
t[826] = 13.9, p < .05, Cohend = 0.64), and experienced a greater
net loss (Mean[SD]BW = 253.1[526.3]; Mean[SD]MC = 48.0[182.4];
t[826] = 10.9, p < .05, Cohend = 0.45). Prize ratio big winners

Assessing the Relevance of Time to Big Win Effects on
Subsequent DFS Engagement
A Kaplan–Meier survival curve modeling persistence among big
winners at each 4-week period following the big win event is
presented in Figure 2. The subsamples of Custer big winners
(n = 371) and Prize Ratio big winners (n = 1,439) exhibited similar
rates of persistence in the 24 weeks following their big win.
Cox Proportional Hazard models indicated that the week of a big
win has an extremely small, albeit significant, negative effect on the
hazard function for both Custer big winners (Hazard Ratio
[HR] = 0.98, CI = [0.97,0.99]) and Prize Ratio big winners
(HR = 0.98, CI = [0.98,0.99]). These effects remain small but still
significant after controlling for players’ frequency and entry fees per
day in the weeks leading up to, and including the week of the big
win, for both Custer big winners (HR = 0.99, CI = [0.98,0.9952])
and Prize Ratio big winners (HR = 0.98, CI = [0.98,0.99]).
Linear regression models likewise indicated a small but significant positive effect of big win week on subsequent frequency of
play, among both Custer big winners (b = .56, β = .23, SE = .13,
p = .000) and Prize Ratio big winners (b = .55, β = .22, SE = .07,
p = .000). These significant effects also remain after controlling for
play frequency and entry fees per day in the weeks leading up to, and
including the week of the big win among Custer big winners
(b = .61, β = .25, SE = .12, p = .000) and Prize Ratio big winners
(b = .73, β = .28, SE = .07, p = .000). Linear regression models
satisfied all basic regression diagnostics (see Supplemenatry
materials for diagnostic test results).
The five number summary of the prize ratio measure for the full sample
of Prize Ratio big winners in the matched sample is: Min: 10.93; 25%: 14.81;
Median: 20.00; 75%: 36.43; Max: 20000.00.

EDSON, TOM, PHILANDER, LOUDERBACK, AND LAPLANTE

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

Figure 1
Cohen’s d Standardized Effect Sizes (Entries, Entry Fees, and Net Loss) for the 24 Weeks Following
a Big Win for Both Big Win Definitions

Note. Cohen’s d’s With Corresponding t-Values That Are Non-Significant (i.e., p ≥ 0.05) Are Censored From
These Trendlines and Thus Are Not Displayed.

Planned Sensitivity Analyses
We ran a series of planned sensitivity analyses to test for the
robustness of the results from the main analyses with different model
specifications. The full results of these sensitivity analyses are
available in the Supplementary materials.
For the first sensitivity analyses, we conducted a series of fixed
effects regression analyses with robust standard errors that test the
overall robustness of the matched pairs analyses (i.e., Hypotheses
2 & 3), as well as derivations of our survival analyses (Hypothesis 4)
that assess the effects of grouped time intervals (i.e., experiencing a
big either within or after a 4-week period), with a 16-week follow up
period. Both the panel analyses and survival analyses control for

players’ cumulative net loss, cohort effects (i.e., week of first deposit),
and contest week. The fixed effect model design allows for the
removal of potential individual-level endogenous error that is constant
over time. Fixed-effect models are commonly used in panel data sets
when there is an immeasurable unobserved effect in each individual
(Wooldridge, 2006). Unlike matching, then, the fixed effects model is
still valid even if we did not measure all person-level variables that
predict having a big win. The drawback relative to matching is that it
makes more restrictive assumptions about the nature of confounding.
Finally, unlike the matched pairs analyses, which exclude big winners
who experience subsequent big wins (i.e., big wins that occur
during the 24 weeks following their initial big win), the fixed effect
sensitivity analyses encompass all big wins/big winners.

LARGE-SCALE PROSPECTIVE STUDY OF BIG WINS

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

Figure 2
Kaplan–Meier Survival Curve of Persistence Among Custer Big
Winners (n = 371) and Prize Ratio Big Winners (n = 1,439)

threshold of $32,480. We consider that a player “appears to be
professional” if their profit over the study period is equal to or
exceeds $32,480. Excluding these players (N = 58; nBW = 44;
nMC = 14) and their corresponding big winners/matched controls, the results for all hypotheses did not change. This finding
suggests that by our operational definition, “professional
players” did not have a substantial effect on the results of our
main analyses.
Finally, we tested Hypothesis 4 with a Poisson count model to
assess for the robustness of the results from the Cox PH analysis.
The results of the Poisson count model were very similar to the
results of the Cox PH models reported in the main analyses. A
Poisson model for Custer big wins did not find a significant effect of
week of big win on mean levels of persistence, while a Poisson
model for Prize Ratio big wins found a significant but weak positive
effect of week of big win on mean levels of persistence. These
results partially confirm the robustness of the survival analyses in the
main analytical approach.

Unplanned Exploratory Analyses

The panel analyses confirmed most of the results of our main
analyses. Specifically, the fixed effect panel analyses found that both
Custer and Prize Ratio big winners evidenced significantly elevated
entries and entry fees both during the week after the big win event,
with decreasing effect size during the subsequent weeks (Figure 3).
However, these analyses did not find significantly elevated levels of
net loss during the follow-up period among big winners, by either
definition.
Notably, unlike the matched control analyses, these panel analyses did not initially filter out big winners who experienced subsequent big wins. As an unplanned exploratory sensitivity analysis, we
applied this same filter (i.e., excluding big winners who experience
subsequent big wins during the 24 weeks following the initial big
win event) to the panel analyses. In these revised sensitivity analyses, both big win definitions had a strong and significant effect on
increased net loss, signifying that the significant effect of big wins
on losses is contingent upon application of that filter.
The sensitivity survival analyses fully confirmed the results of
our main survival analyses. Kaplan–Meier curves indicated that
big wins extend duration, but that early big wins (i.e., within the
first 4 weeks of depositing) have smaller impacts than later big
wins (see Figure 4 for Prize Ratio results). Cox Proportional
Hazard models likewise indicated that big wins are related to
duration of active wagering and more active days, but the net
effect size is larger for big wins coming after the first 4 weeks of
activity.
In another set of planned sensitivity analyses, we re-ran the
analyses used to test Hypotheses 2, 3, and 4 after excluding players
who appear to be professional based on their net profits at the end
of the study period. We started with the federal minimum wage of
$7.25/hr (United States Department of Labor [US DOL], 2019),
scaled it up to its monthly equivalent of $1,160/month, and then
multiplied that value by 28 months (for our study period) to get a

For both big win definitions, all outcome variables evidenced
substantial skewness. To test the appropriateness of parametric
testing in our matched pairs design, we ran two exploratory
sensitivity analyses. First, we conducted Wilcoxon signed-rank
tests between big winners and matched controls for all outcomes. Second, we re-ran the matched pairs t-tests on the natural
log transformations for entries and entry fees. We could not
reliably take the logarithm of net loss because the profiting
players had negative—in some cases, large and negative—
values of net loss (i.e., indicating a net gain). The results of
these analyses confirmed the robustness of the initial matched
pairs t-tests. All signed-rank tests indicated true location shifts
that were significantly different from zero and all matched pairs
t-tests with log transformations were significant with large
corresponding Cohen’s d’s. The details of both of these tests
are available in the Supplementary materials.
Finally, given the importance of understanding whether big wins
might be associated with future DFS-related problems, we tested
whether big winners by either definition were more likely to
voluntarily self-exclude at any future period following their big
win event. We used exact conditional logistic regression (Mehta &
Patel, 1995) to test this assertion. In these analyses, we controlled
for the number of available weeks left in the study period following
the week of the big win to account for the fact that earlier matching
weeks allot players more time to potentially self-exclude.3 These
unplanned exploratory analyses showed that both Custer big
winners (b = .12, CI = [−.85,1.40], p = 1.0) and Prize Ratio
big winners (b = .47, CI = [−1.63,2.56], p = .601) were no
more likely to self-exclude than their respective matching control
cases.

In our Transparent Change document detailing this exploratory analysis
(see Transparent Change 5 document on our OSF project page), we also
stated that we would control for a self-exclusion event before the matching
week. However, none of the players in our samples self-excluded before the
matching week.

EDSON, TOM, PHILANDER, LOUDERBACK, AND LAPLANTE

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

Figure 3
Estimates of Lagged Big Win Effects on Entry Fees (models 1 & 2) and Entries (models 3 & 4). Regression Coefficients Are Presented as
Dots and 95% Confidence Intervals as Vertical Lines. Full Results Appear in Supplementary materials

Discussion
Key Findings
In this study, we examined how “big wins,” based on two
operational definitions, were associated with future financial
and time involvement in DFS, based on a sample of 34,596
subscribers to DraftKings. We also tested how the timing of the
big win might have influenced persistence and frequency of play
following the big win event. Several key findings are important to
highlight. First, we found that although a minority of players
experienced a big win by either definition (1,473 for Custer and
2,871 for Prize Ratio), these figures are still considerably higher
than what we had hypothesized (i.e., less than 0.1% of the time),
suggesting that big wins, while still indeed a rare occurrence in
DFS, happen with greater frequency than we initially expected.
Second, matched pairs analyses showed that big winners’ total
entries (i.e., the time involvement dimension of DFS) as well as
entry fees and net loss (i.e., the financial involvement dimensions
of DFS) were elevated in the initial weeks following a big win
when compared to matched controls. Third, an examination of the
effect sizes of these comparisons during the 24 weeks following

the big win event points to a diminishing effect of big wins over
time in DFS. Fourth, while the originators of the big win effect
(Custer, 1984; Custer & Milt, 1985) and initial research on the
subject of big wins (Weatherly et al., 2004) point to early big wins
in particular as having a more meaningful effect, the results of our
analyses of persistence and frequency among DFS big winners
suggest that later big wins have a slightly stronger effect on future
engagement. These findings collectively suggest that big wins in
DFS tend to promote continued engagement, which for some
players might potentially be at levels that are both excessive
and unhealthy. More research is needed that assesses the relationship between big wins and unhealthy engagement behaviors.
However, it would not be premature at this point to suggest that
DFS and gambling operators take steps to reduce potential harms
among big winners. This could include promoting responsible
decisions and behaviors for big winners to follow, such as withdrawing some of their win to use for neccessary expenses (e.g.,
housing, education) or other leisure activities, or to deposit into
their savings account.
Notably, both definitions of big win (Custer and Prize Ratio) had a
strong effect on subsequent DFS engagement and losses. Although

LARGE-SCALE PROSPECTIVE STUDY OF BIG WINS

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

Figure 4
Kaplan–Meier Survivor Functions Describing Account Activity Length. (1) No Controls; (2) Adjusting for Number of Entries in First Week;
(3) Adjusting for Number of Entries in First Week and Separate Plots of Prize Ratio Big Winners and non-Big Winners; (4) Adjusting for
Number of Entries in First Week and Separate Plots of Prize Ratio Big Winners During First 4 Weeks, After 4 Weeks, and non-Big Winners

these definitions were not necessarily mutually exclusive, only a
minority of big wins met the criteria for both, which suggests that
they are likely tapping different dimensions of “big.” Specifically,
whereas the Custer definition might represent an absolute or objective measure of “big,” the Prize Ratio definition could represent a
measure of “big” that is more subjective and relative to an individual’s traditional wager size. This study provides an indication that
both types of big wins are meaningful, however, so future research
might be needed to determine which definition carries more meaning for certain players.
We conducted sensitivity analyses to test the robustness of our
main analyses. These analyses confirmed most of the results from
our main analyses, with the exception of big wins having an effect
on elevated net loss. Further examination of these sensitivity
analyses indicated that one of the exclusions we applied in the
matched pairs analyses (i.e., removing big winners/matching candidates who experience a second/initial big win during the 24 weeks
following the matching week) might have inadvertently removed
more skilled players, or at the very least removed players who were
more likely to experience a net win, from the matched samples. We
applied this filter to account for the potential confounding effects of
experiencing multiple big wins, in order to better understand the
effects of a singular big win on future DFS engagement and losses.

Therefore, we feel that this exclusion was appropriate. We should
also note that DFS is a partially skill-based activity (Meehan, 2015),
and that we were less interested in the effects of big wins on the
engagement behaviors of skilled or professional players, for whom
big wins might be expected as part of their efforts to grow their
bankroll. Reflecting this goal, we had pre-registered and conducted a
separate sensitivity analysis to identify and filter out “professional
players,” which confirmed our initial results. A potential reason for
this could be that we had already effectively removed professional
players in our main analyses by excluding repeat big winners. This
unique finding has substantial implications for how our main
analyses should be interpreted. Specifically, rather than be generalized to DFS players as a whole, our main results possibly generalize
only to less-skilled DFS players, who would be expected to behave
more like a typical gambler.

Theoretical and Conceptual Implications
Given the frequent resurfacing of the concept of “big wins” in the
gambling studies literature and popular media, our study’s empirical
findings have important implications for the substantive and theoretical understanding of this concept. We demonstrated that big wins
do indeed influence future DFS behaviors, and that this effect is

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

EDSON, TOM, PHILANDER, LOUDERBACK, AND LAPLANTE

consistent across time (i.e., number of entries) and financial (i.e.,
entry fees and net loss) dimensions. Why, then, does this significant
effect exist? There are two major explanations for our results. First,
users who receive a “cash windfall” from a big win event also
experience a rapid increase in the amount of money in their DFS
account, and by extension a larger “bankroll” of reserve funds that
increases availability and accessibility to more contests and/or larger
stakes contests. It follows that this influx of cash might lead to
increased spending on DFS, rather than a transfer of this money to
other parts of a person’s budget (e.g., leisure pursuits, life necessities, savings). Indeed, research from behavioral economics (Kuhn
et al., 2011) shows that a greater supply of money after a lottery win
led to an increase in discretionary spending. Although these ideas
might explain a portion of the big win’s effect on future DFS
involvement, we should note that the fixed-effect regression sensitivity analyses found a significant effect of big wins on subsequent
engagement and losses each week, even while accounting for the
prior week’s cumulative loss. These results therefore suggest that
big wins effects might provide motivation for continued engagement, beyond simply the means to do so.
Second, explanations from cognitive psychology might help to
explain the big win effect (Goodie & Fortune, 2013; Michalczuk
et al., 2011). Within this perspective, big wins might impact future
gambling involvement because big winners develop cognitive distortions that make them think that big wins occur frequently when
they actually do not. Said another way, a big winner might have an
inflated perceived probability (i.e., personal probability; see Moore,
2009, example 4.32) of winning. This perception might lead big
winners to continue to apply their perceived superior skills, attempt
to exploit their inflated personal probabilities of winning, and seek
more big wins. As a result, big winners might develop an “illusion of
control” over their winning (Stefan & David, 2013), making them
feel as though they are more in control of their performance when in
fact chance plays a large role in winning, and especially in winning
big. In this way, the big win effect might fit well into the pathways
model of gambling (Blaszczynski & Nower, 2002), which argues
that: (1) increased accessibility, potentially facilitated via a big win,
will lead to (2) operant/classical conditioning effects (i.e., continued
engagement in order to repeat the big win experience, facilitated by
an illusion of control), which can in turn lead to (3) problematic
gambling behaviors (e.g., chasing losses), and eventually the development of gambling problems. Although our study provides strong
evidence for Steps 1 and 2, the results of our analyses of big wins
predicting a proxy for problems (i.e., voluntary self exclusion) do
not suggest this pathway necessarily continues to Step 3. However,
those analyses were exploratory and should be qualified by the
potential limitations of self-exclusion as a proxy for DFS problems.
Specifically, some gambling researchers have expressed skepticism
about self-exclusion as a proxy for gambling-related problems,
given the diverse reasons individuals might choose to self-exclude,
some of which might not be related to experiencing gamblingrelated problems (Griffiths & Auer, 2016). These arguments also
might apply to DFS. Further investigations are needed that employ a
more concrete measure of harm in understanding the potential
ramifications of big wins and health outcomes.
Third, it is possible that big wins represent an extreme form of
operant conditioning (Skinner, 1938; Staddon & Cerutti, 2003).
Whereas much of the research on intermittent reinforcement as it
relates to operant conditioning principles concerns the timing of

continued reinforcement on subsequent behavior (e.g., variable
ratio schedule; see Zeeb et al., 2017), less has been said about
how the size of the reinforcer affects response. The results of our
study (in particular the lasting effects of big wins on continued
engagement while controlling for prior net loss, in sensitivity
analyses) suggest that the size of a single reinforcer might have a
strong effect on continued behavior even after accounting for the
duration and size of additional reinforcers in the domain of DFS
play. Future research should further investigate how the timing,
duration, and size of reinforcements independently affect continued
behavior, both in context of DFS as well as other activities with
similar reinforcement schedules.
The diminishing effect of big wins we observed over time appears
to be consistent with exposure and adaptation models of gambling
and gaming-related behaviors (LaPlante & Shaffer, 2007) previously observed in DFS and online gambling environments (Edson &
LaPlante, 2020; LaPlante et al., 2008). Here, novel activities or
events such as a big win encourage an initial increase in the desire to
gamble (i.e., exposure). However, the novelty of the activity or
event will eventually begin to diminish over time, resulting in a
decrease and plateauing effect (i.e., adaptation). Alternatively,
adaptation could have resulted from big winners’ desire, but failure,
to repeat their big win experience, resulting in their departure from
DFS. This particular mechanism could reflect rational behaviors in
response to new information, consistent for example with typical
behaviors observed on the Iowa Gambling Task (Bull et al., 2015).
Even still, the diminishing effect we observed could simply be
unique to DFS, or even our specific DFS sample, for whom
engagement might have naturally diminished following each cohort’s respective NFL season. Indeed, prior research using these data
has observed strong NFL seasonality, an indication that the cohorts
are very NFL-centric (Edson & LaPlante, 2020).
The results of our study also call into question the central idea of
the big win hypothesis that big wins need to happen early on to have
a significant effect on future engagement, at least in one product
domain (DFS). However, despite later big wins having a stronger
effect, we would stop short of suggesting there is a late big win
effect in DFS, considering the extremely small effect sizes of later
big wins on persistence and frequency of play. Rather, the results of
our study suggest that big wins in DFS have a strong effect on future
engagement regardless of when they happen.
Finally, although this study examined the big win effect in one
activity (DFS), researchers should consider its wider application to
other activities, including areas outside of gambling, especially
given that many of the fundamentals potentially underlying the
big win effect (e.g., illusion of control) have broader applications.
Scholars should consider, for instance, whether conspicuously large
successes, monetary or non-monetary, in any competitive activity
(e.g., publishing in a high-ranking journal in academia; securing a
major lucrative deal in business) are associated with bolstered levels
of immediate subsequent, and potentially long term, engagement in
these activities, as well as whether the timing of the large success
matters at all.

Study Limitations
We highlight six limitations. First, we only had access to activity
records from one major DFS provider (DraftKings). DFS tends to
skew heavily toward males (Nelson et al., 2019) and is at least a

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

LARGE-SCALE PROSPECTIVE STUDY OF BIG WINS

partially skill-based activity (Ehrman, 2015), therefore the findings
of our study might not be representative of gamblers more generally,
for whom the big win effect was originally conceived. Furthermore,
although we considered “early” big wins in terms of players’ DFS
(and specifically their DraftKings DFS) experience, some players
might have experienced a big win earlier on with other fantasy
sports, DFS, internet gambling, and land-based gambling platforms,
that informed their experiences and expectations with DraftKings.
Second, we only examined big win effects over a limited period of
time and did not control for all potential seasonality, so it is possible
that there may be effects beyond the range of our data. For example,
involvement might increase again following a certain period of time,
especially if there are particular sporting events that occur and might
stimulate increases in involvement (e.g., the NFL Playoffs or FIFA
World Cup). Third, we did not have access to a measure of financial
means for gaming or gambling (e.g., household income), so it is
possible that our definitions of big win (especially Custer’s definition) might be perceived differently by individuals with more
income or wealth compared to individuals with less income or
wealth. Fourth, there are other potential variables that we were
unable to measure due to data limitations that might have influenced
engagement and losses following a big win, including skill level at
DFS and experiencing a big win in other modes of gambling (e.g., a
lottery jackpot). Fifth, our findings potentially should be considered
generalizable to less skilled DFS players who experience a big win, not
DFS players as a whole. Sixth, we recognize the inherent arbitrariness
of trying to define a threshold for what types of wins can be considered
“big.” We designed this study to test the theoretical assertion that there
exists some arbitrary threshold by which a gambling win is sufficiently
“big” enough to have a large effect on subsequent behavior, grounding
those thresholds in prior theory and research with an Open Scienceinformed, pre-registered research plan. Although the results of this
study support this assertion, they do not establish that wins need to be
necessarily “big” in order to have a large effect. Put more plainly, this
study’s results do not preclude the possibility that any win might have
an effect on subsequent behavior that is proportional or roughly
proportional to its size. This remains a competing explanation of
the big win effect that deserves consideration.

Conclusions
Our study empirically demonstrated the big win effect in DFS,
with diminishing effects over time, but not uniformly dissipating
effects over time (i.e., the effect was persistent for some outcomes).
Big wins in DFS lead to increased engagement for most players;
however, less skilled players who are unable to quickly repeat the
big win experience tend to experience increased losses, an indication
that, for these players, big wins might instill unrealistic expectations
about future probabilities of winning. These results suggest that big
wins could potentially lead to excessive and even unhealthy levels of
engagement. Notably, our observations of the big win effect in DFS
do not suggest that the timing of the big win has a substantial effect
on subsequent engagement. We were unable to fully ascertain
whether experiencing a big win in DFS leads to the development
or continuation of DFS-related problems, which remains an important area for future research to consider. Overall, we believe that the
findings from this study can help researchers to develop a better
understanding of the big win effect, including its potential for wider
application to other scientific disciplines.

References
Blaszczynski, A., & Nower, L. (2002). A pathways model of problem and
pathological gambling. Addiction, 97, 487–499. https://doi.org/10.1046/j
.1360-0443.2002.00015.x
Bull, P. N., Tippett, L. J., & Addis, D. R. (2015). Decision making in healthy
participants on the Iowa Gambling Task: New insights from an operant
approach. Frontiers in Psychology, 6, Article 391. https://doi.org/10.3389/
fpsyg.2015.00391
Caillon, J., Grall-Bronnec, M., Perrot, B., Leboucher, J., Donnio, Y., Romo,
L., & Challet-Bouju, G. (2019). Effectiveness of at-risk gamblers’ temporary self-exclusion from Internet gambling sites. Journal of Gambling
Studies, 35(2), 601–615. https://doi.org/10.1007/s10899-018-9782-y
Chang, W. W., Hsiao, P. K., Qin, L., Chang, C. L., Chow, J. M., & Wu, S. Y.
(2018). Treatment outcomes for unresectable intrahepatic cholangiocarcinoma: Nationwide, population-based, cohort study based on propensity
score matching with the Mahalanobis metric. Radiotherapy and Oncology:
Journal of the European Society for Therapeutic Radiology and Oncology,
129(2), 284–292. https://doi.org/10.1016/j.radonc.2018.09.010
Cohen, J. E. (1988). Statistical power analysis for the behavioral sciences.
Lawrence Erlbaum.
Cox, D. R. (1972). Regression models and life-tables. Journal of the Royal
Statistical Society. Series B. Methodological, 34(2), 187–202. https://
doi.org/10.1111/j.2517-6161.1972.tb00899.x
Custer, R., & Milt, H. (1985). When luck runs out: Help for compulsive
gamblers and their families. Facts on File Publications.
Custer, R. L. (1984). Profile of the pathological gambler. The Journal of
Clinical Psychiatry, 45(12), 35–38.
Derevensky, J. L., & Gilbeau, L. (2015). Adolescent gambling: Twenty-five
years of research. The Canadian Journal of Addiction, 6(2), 4–12. https://
doi.org/10.1097/02024458-201509000-00002
Do, E. K., Prom-Wormley, E. C., Fuemmeler, B. F., Dick, D. M., Kendler,
K. S., & Maes, H. H. (2018). Associations between initial subjective
experiences with tobacco and self-reported recent use in young adulthood.
Substance Use & Misuse, 53(14), 2291–2298.. https://doi.org/10.1080/
10826084.2018.1473435
Edson, T. C., & LaPlante, D. A. (2020). Longitudinal playing trends among
daily fantasy sports players. Computers in Human Behavior, 104, 106165.
https://doi.org/10.1016/j.chb.2019.106165
Efron, B. (1988). Logistic regression, survival analysis, and the Kaplan–
Meier curve. Journal of the American Statistical Association, 83(402),
414–425. https://doi.org/10.1080/01621459.1988.10478612
Ehrman, N. (2015). Out of bounds? A legal analysis of pay-to-play daily
fantasy sports. A legal analysis of pay-to-play daily fantasy sports (March
17, 2014). The Sports Lawyers Journal, 22, 79.
Gillies, T. (2016). Fantasy sports: The lucrative market that may be legal. On
the Money, February 7. CNBC. https://www.cnbc.com/2016/02/05/
fantasy-sports-the-lucrative-market-that-may-be-legal.html
Goodie, A. S., & Fortune, E. E. (2013). Measuring cognitive distortions in
pathological gambling: Review and meta-analyses. Psychology of Addictive Behaviors, 27(3), 730–743. https://doi.org/10.1037/a0031892
Griffiths, M. D., & Auer, M. (2016). Should voluntary “self exclusion” by
gamblers be used as a proxy measure for problem gambling? MOJ
Addiction Medicine & Therapy, 2(2), 31–33. https://doi.org/10.15406/
mojamt.2016.02.00019
Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. (2010). Multivariate
data analysis (7th ed.). Pearson.
Ho, D. E., Imai, K., King, G., & Stuart, E. A. (2011). MatchIt: Nonparametric
preprocessing for parametric causal inference. Journal of Statistical
Software, 42(8), 1–28. https://doi.org/10.18637/jss.v042.i08
Jung, E., & Yoon, H. (2018). Is flood risk capitalized into real estate market
value? A mahalanobis-metric matching approach to the housing market in
Gyeonggi, South Korea. Sustainability, 10(11), 4008. https://doi.org/10
.3390/su10114008

This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All rights, including for text and data mining, AI training, and similar technologies, are reserved.

EDSON, TOM, PHILANDER, LOUDERBACK, AND LAPLANTE

King, D. L., & Delfabbro, P. H. (2016). Early exposure to digital simulated
gambling: A review and conceptual model. Computers in Human Behavior, 55, 198–206. https://doi.org/10.1016/j.chb.2015.09.012
King, G., & Nielsen, R. (2019). Why propensity scores should not be used for
matching. Political Analysis, 27(4), 435–454. https://doi.org/10.1017/pan
.2019.11
Kuhn, P., Kooreman, P., Soetevent, A., & Kapteyn, A. (2011). The effects of
lottery prizes on winners and their neighbors: Evidence from the Dutch
postcode lottery. The American Economic Review, 101(5), 2226–2247.
https://doi.org/10.1257/aer.101.5.2226
LaBrie, R., & Shaffer, H. J. (2011). Identifying behavioral markers of
disordered Internet sports gambling. Addiction Research & Theory,
19(1), 56–65. https://doi.org/10.3109/16066359.2010.512106
LaPlante, D. A., Nelson, S. E., & Gray, H. M. (2014). Breadth and depth
involvement: Understanding Internet gambling involvement and its relationship to gambling problems. Psychology of Addictive Behaviors, 28(2),
396–403. https://doi.org/10.1037/a0033810
LaPlante, D. A., Schumann, A., LaBrie, R. A., & Shaffer, H. J. (2008).
Population trends in internet sports gambling. Computers in Human
Behavior, 24(5), 2399–2414. https://doi.org/10.1016/j.chb.2008.02.015
LaPlante, D. A., & Shaffer, H. J. (2007). Understanding the influence of
gambling opportunities: expanding exposure models to include adaptation. Am J Orthopsychiatry, 77(4), 616–23. https://doi.org/10.1037/00029432.77.4.616.
Lesieur, H. R., & Custer, R. L. (1984). Pathological gambling: Roots, phases,
and treatment. The Annals of the American Academy of Political and Social
Science, 474(1), 146–156. https://doi.org/10.1177/0002716284474001013
Lu, B., Cai, D., & Tong, X. (2018). Testing causal effects in observational
survival data using propensity score matching design. Statistics in Medicine, 37(11), 1846–1858. https://doi.org/10.1002/sim.7599
Ma, X., Kim, S. H., & Kim, S. S. (2014). Online gambling behavior: The
impacts of cumulative outcomes, recent outcomes, and prior use. Information Systems Research, 25, 511–527. https://doi.org/10.1287/isre
.2014.0517
Meehan, J. C. (2015). The predominate goliath: Why pay-to-play daily
fantasy sports are games of skill under the dominant factor test. Marquette
Sports Law Review, 26, 5.
Mehta, C. R., & Patel, N. R. (1995). Exact logistic regression: Theory and
examples. Statistics in Medicine, 14(19), 2143–2160. https://doi.org/10
.1002/sim.4780141908
Michalczuk, R., Bowden-Jones, H., Verdejo-Garcia, A., & Clark, L. (2011).
Impulsivity and cognitive distortions in pathological gamblers attending
the UK National Problem Gambling Clinic: A preliminary report. Psychological Medicine, 41(12), 2625–2635. https://doi.org/10.1017/
S003329171100095X
Moore, D. S., McCabe, G. P., & Craig, B. A. (2009). Introduction to the
practice of statistics (6th ed.). W.H. Freeman.
Muthukrishnan, A. V., & Kardes, F. R. (2001). Persistent preferences for
product attributes: The effects of the initial choice context and uniformative experience. Journal of Consumer Research, 28(1), 89–104. https://
doi.org/10.1086/321949
Nelson, S. E., Edson, T. C., Singh, P., Tom, M., Martin, R. J., LaPlante,
D. A., Gray, H. M., & Shaffer, H. J. (2019). Patterns of daily fantasy sport

play: Tackling the issues. Journal of Gambling Studies, 35(1), 181–204.
https://doi.org/10.1007/s10899-018-09817-w
Shah, B. R., Laupacis, A., Hux, J. E., & Austin, P. C. (2005). Propensity
score methods gave similar results to traditional regression modeling in
observational studies: A systematic review. Journal of Clinical Epidemiology, 58(6), 550–559. https://doi.org/10.1016/j.jclinepi.2004.10.016
Skinner, B. F. (1938). The behavior of organisms. Appleton-Century.
Staddon, J. E., & Cerutti, D. T. (2003). Operant conditioning. Annual Review
of Psychology, 54, 115–144. https://doi.org/10.1146/annurev.psych.54
.101601.145124
Stefan, S., & David, D. (2013). Recent developments in the experimental
investigation of the illusion of control. A meta-analytic review. Journal of
Applied Social Psychology, 43, 377–386. https://doi.org/10.1111/j.15591816.2013.01007.x
Stuart, E. A., King, G., Imai, K., & Ho, D. (2011). MatchIt: Nonparametric
preprocessing for parametric causal inference. Journal of Statistical Software,
42(8). Advance online publication. https://doi.org/10.18637/jss.v042.i08
Turner, N. E., Jain, U., Spence, W., & Zangeneh, M. (2008). Pathways to
pathological gambling: Component analysis of variables related to pathological gambling. International Gambling Studies, 8(3), 281–298.
https://doi.org/10.1080/14459790802405905
Turner, N. E., Zangeneh, M., & Littman-Sharp, N. (2006). The experience of
gambling and its role in problem gambling. International Gambling
Studies, 6(2), 237–266. https://doi.org/10.1080/14459790600928793
Tversky, A., & Kahneman, D. (1971). Belief in the law of small numbers.
Psychological Bulletin, 76(2), 105–110. https://doi.org/10.1037/h0031322
Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging
frequency and probability. Cognitive Psychology, 5(2), 207–232. https://
doi.org/10.1016/0010-0285(73)90033-9
United States Department of Labor. (2019). Minimum wage. https://www.dol
.gov/general/topic/wages/minimumwage
Weatherly, J. N., Sauter, J. M., & King, B. M. (2004). The “big win” and
resistance to extinction when gambling. The Journal of Psychology,
138(6), 495–504. https://doi.org/10.3200/JRLP.138.6.495-504
Wiley, R. C., Tom, M. A., Edson, T. C., & LaPlante, D. A. (2020). Behavioral
markers of risky daily fantasy sports play. Journal of Sport & Social Issues,
44(4), 356–371. https://doi.org/10.1177/0193723520919819
Williams, R. J., Hann, R. G., Schopflocher, D., West, B., McLaughlin, P.,
White, N., King, K., & Flexhaug, T. (2015). Quinte longitudinal study of
gambling and problem gambling [report]. Ontario Problem Gambling
Research Centre. https://opus.uleth.ca/bitstream/handle/10133/3641/
QLS-OPGRC-2015.pdf
Wooldridge, J. M. (2006). Introductory econometrics: A modern approach.
Thomson South-Western.
Zeeb FD, Li Z, Fisher DC, Zack MH, & Fletcher PJ. (2017). Uncertainty
exposure causes behavioural sensitization and increases risky decisionmaking in male rats: toward modelling gambling disorder. Journal of
Psychiatry Neuroscience 42(6):404–413. https://doi.org/10.1503/jpn
.170003.

Received October 20, 2020
Revision received December 16, 2020
Accepted December 17, 2020 ▪