Catching Counters

Catching Card Counters
BY: Sarah French
ADVISOR • Jim Bishop
EDITORIAL REVIEWER • Phyllis Schumacher
_________________________________________________________________________________________
Submitted in partial fulfillment of the requirements for graduation with honors in the Bryant University Honors
Program
APRIL 2016
Table of Contents
Abstract ..................................................................................................................................... 3
Introduction ............................................................................................................................... 4
Literature review ....................................................................................................................... 5
Introduction ........................................................................................................................... 5
Background ........................................................................................................................... 6
Review................................................................................................................................. 10
Conclusion .......................................................................................................................... 16
Methodology ........................................................................................................................... 17
Introduction ......................................................................................................................... 17
Simulation ........................................................................................................................... 18
Observations ........................................................................................................................ 19
Betting Strategies ................................................................................................................ 19
Analysis ............................................................................................................................... 21
Conclusion .......................................................................................................................... 22
Findings ................................................................................................................................... 22
Appendix A: Blackjack Basic Strategy ................................................................................... 26
Appendix B: Observation Maximum Bets .............................................................................. 27
Appendix C: 5 Unit Spread Maximum Bets ........................................................................... 27
Appendix D: 10 Unit Spread Maximum Bets ......................................................................... 28
Appendix E: Function of Minimum Bet Maximum Bets ....................................................... 28
Appendix F: Function of Maximum Count Maximum Bets ................................................... 29
Appendix G: Frequency of High Counts ................................................................................ 29
References ............................................................................................................................... 30
Senior Capstone Project for Sarah French
ABSTRACT
The casino industry has been researched through a variety of disciplines including
psychological gambling habits, technological advances, business strategies, and
mathematical simulations. In the vast number of studies that have been conducted, there
are few scholarly articles that focus on the specific aspect of card counting. The majority
of games in the casino are designed to favor the “house”. This study focuses on the game
of blackjack, in which players using a card counting strategy can tip the odds in their
favor. A computer simulation was used to model the betting strategy of a card counter
who would bet methodically. Conversely, the unpredictable betting strategy of a
“normal” gambler was gathered through observations of over one thousands hands of
blackjack. The comparison of the two led to deviations in behavior and betting habits.
An understanding of these differences will provide a casino with additional information
to catch card counters at the table.
3
INTRODUCTION
This study has been created with the intent of providing the casino industry with an
outline for identifying and catching card counters. In the game of blackjack, the goal is
to get a score as close as possible to 21 without going over. Each player is dealt two
cards at the beginning of the hand and individually plays against the dealer or the
“house”. The value of each card two through ten is face value, the jack, queen and king
have a value of ten and the ace can be played as a one or an eleven at the player’s
discretion. The player has the choice to ask for additional cards at his or her own
discretion, however the dealer is required to deal himself another card if the total is below
17.
A card counter would choose whether or not to ask for another card using a method
called basic strategy. The table in Appendix A explains when a person should “hit”
meaning to ask for another card, or “stand” meaning not to ask for additional cards.
Other options include to “split” which is when the player’s cards are the same they
choose to play them as two separate hands and to “double” which is to bet twice as much.
This table shows which choice a player using basic strategy should make based on both
what cards the player has been dealt and what the dealer’s one face up card is.
Blackjack is played with a given number of decks, and therefore a defined number of
cards. Mathematically speaking this is considered sampling without replacement which
means that each card has a certain probability of appearing. As a card is played, this
probability changes in relation to the remaining cards in the deck. To exploit this
4
advantage, card counting was invented to estimate when a given hand is more or less
likely to win.
In a typical hand of blackjack, the house has between .5% and 3% advantage over the
player. However by card counting a player can have a 1%, 2%, or even 3% advantage
over the house (Ma, 2015). A card counter using the Hi-Lo strategy assigns a value of -1,
0, or 1 to each card dealt according to the following table:
Card: Ace 2 3 4 5 6 7 8 9 10 Jack Queen King
Assigned 1 1 1 1 0 0 0 0 0 -1 -1 -1 -1
Value
Table 1: Hi-Lo Strategy
The card counter will keep a mental tally of this count after every card is dealt. The
dealer has the highest chance of going over 21 when there are the most 10s, jacks, queens
and kings left in the deck. This means that as the count increases the player is more
likely to win and should therefore bet more. The next step is to determine the betting
strategy that the player will implement which will be discussed in the Methodology
section below.
LITERATURE REVIEW
Introduction
This research thesis is relevant because it relates to a recent article that reported on events
from the 1980's through the early 2000's. This case study tells the story of the most
widely known incident of card counting; when a group of individuals at MIT used card
counting to make millions (Casey, 2008). It explained the rigorous training process and
5
techniques they used to remain undetected. It followed these individuals from formation
to disbandment, while providing insight into the casino industry. This event has social
relevance because there was both a book written about this team and a corresponding
movie made. The media sensation around card counting in the early 2000’s provides
social interest for this analysis.
Furthermore, Jeff Ma who was a member of this team, spoke at a SAS Global Forum in
2015 about card counting in relation to data (Ma, 2015). This brought card counting into
the big data arena, and he utilized examples from his own card counting to explain the
advantage of leveraging every piece of data available. Ma states that card counting is still
utilized and easy enough to teach anyone. This suggests the need for card counting
detection because many people may be counting without the casino’s knowledge. This
project is also relevant in regards to the use of big data, a popular subject now that
analytics has come to the forefront of mathematics. Ma campaigns that every person
should understand the importance of data and use all available information to make
educated decisions. When this is applied to the casino industry, to best catch card
counters casinos should not merely use one method or area of research analysis. This
shows the impact that this research analysis could have on casinos.
Background
Professors and researchers in the fields of mathematics and science have previously
conducted simulations for various casino games. One researcher considered the game of
blackjack in continuous time, as a theoretical model (Andersson, 2012). The article was
printed in England, as a part of the author's doctorate thesis for Stockholm University.
6
The purpose was to provide the optimal betting strategy in card games, such as blackjack,
that are played with multiple rounds of the same deck prior to shuffling. Andersson was
able to create a stochastic differential equation to estimate the best strategy in continuous
time though it cannot be directly applied while a card counter is in a casino. This data
was one of many studies that incorporated mathematical equations in predicting
outcomes of casino games.
Another analyzed the game of blackjack through computer simulations to determine
changes in various player advantages. This study analyzed the use of multiple decks, a
method casinos implemented to deter card counters (Golden, 2011). It was a study
conducted for the Applied Probability Trust School of Mathematics and Statistics,
produced by Professor Golden at the University of Illinois Chicago. It ran Monte Carlo
simulations to model the potential hands in a game of blackjack with the use of one deck
up to eight decks at a time. It used normal approximations to create each situation and
created graphical presentations of the findings. This article was beneficial to the card
counting equations to provide information to calculate the “true count”. The true count is
an aspect of card counting that accounts for the number of decks left to be played. When
multiple decks are in effect a high count early in the decks has more variability and is less
predictive of success than a high count later in the game. This study was used to
determine the effect of multiple decks on the success of a card counter.
A third study simulated games of blackjack with a specified number of chips to determine
the average length of play (Yang, Li, Zhoa, Liu, & Han, 2011). It was produced by
professors at the Beijing Normal University and has been published as an international
7
scientific study. The researchers ran upwards of a million rounds of simulations to
ensure accuracy. In addition to the numerous trials, they also accounted for different rule
changes and controlled some player decisions. This was a very specific study in which
assumptions of minimum bets were necessary, so it cannot be considered entirely
reflective of a person’s true gambling behavior. This study was originally analyzed as a
potential baseline for the behavior of a normal gambler, however in order to use this
study it would require incorporating all of the controls.
Multiple sources analyzed the application of the central limit theorem to use normal
approximation for certain casino games. One study utilized Monte Carlo simulation and
applied the central limit theorem to replicate slot machine probabilities (Singh, Lucas,
Dalpatadu, & Murphy, 2013). The simulation was conducted by four doctorate level
professors at the University of Nevada Las Vegas. It was funded by the Caesars
Foundation, a leader in the casino industry, but appears to be written objectively without
bias. It explained the mathematical approach to calculating probabilities of success and
provided table and graphical representation of the findings. The slot machine does not
have the same property of sampling without replacement therefore it is not directly
applicable. It was used by Caesars Foundations and therefore validates the use of
simulation techniques in the casino industry today.
The business applications for this topic consist of the casino success and technology
advances used to detect card counters. The profile of the casino and gambling industry
provides a market definition and segmentation, as well as listings of industry leaders
(Casinos & Gambling Industry Profile: North America, 2014). This was produced by the
8
company MarketLine, which is a Datamonitor business and as such collects and compiles
data on various industries. It provided definitions of each market segment, which
clarified the industry layout. It also provided information on one market leader, Caesar’s.
This data enhanced the information on the size and strength of the casino market and as
the industry grows the opportunities for card counters increase.
There are also articles that include the technical specifications of the industry that already
are and soon will be implemented (Wyld, 2007). One article analyzed the capabilities of
Radio Frequency Identification (RFID) technology that has been and will be
implemented in casinos. It was written by a Professor Wyld at Southeastern Louisiana
University in 2007. It demonstrated the data that can be collected and the improvements
that will lead to preventing counterfeit, cheating, and problem gambling. In recent years,
some of this technology has been implemented and there are future opportunities for
continued utilization of RFID technology. One future possibility includes the use of
RFID chips in the cards for blackjack, this would provide casinos with insight into card
counter strategies and possible card counting suspects. This business research provides
insight into the casino market useful for understanding the measures already used to
detect card counters.
Finally, extensive psychological research has been done analyzing the behavior and
choices of gamblers. Some are specific and provide findings based on small populations
or data (Chau, Phillips, & Von Baggo, Departures from sensible play in computer
blackjack, 2000). A few researchers used a small sample of students, 16 undergraduates
from the University of Hong Kong, to conduct an experiment. This research explored the
9
irrational choices that gamblers make when playing blackjack. It showed that players
will often forgo advice from others in order to proceed with one's own strategy. The
results of the experiment may not be reflective of society as a whole due to the lack of
diversity of the participants and the small sample size. The article itself provided insight
into what questions should be discussed from a psychological standpoint.
Other studies were performed on a larger scale and illustrate valuable information about
the choices gamblers make. An extensive study on the psychology of blackjack which
involved simulation, experimentation and interviews was conducted in the 1980’s by two
psychologists Dr. Keren and Dr. Wagenaar for the American Psychological Association.
The goal was to determine the behavior of gamblers and whether rational thinking was
employed. The study included observation of over one hundred players, and personal
interviews with one hundred and fifty players (Keren & Wagenaar, 1985). It provides
valuable insight into a players mind set when approaching a game of blackjack and
explores the concept of “luck”. These insights are vital to the understanding of normal
player behavior, as compared to strategic card counter behavior.
Review
Many ideas related to the project have been debated from opposing sides, especially due
to the various fields combined for this study. One of the issues that has been disputed is
the validity of simulation, the key method used in this analysis. An article by Claus
Beisbart and John D. Norton, published for international publication in 2012, specifically
analyzed the process of utilizing Monte Carlo simulations for investigative purposes. It
critiqued the pseudorandom generation and the mathematical basis for running such
10
simulations (Beisbart & Norton, 2012). This article claims that Monte Carlo simulations
are inferential rather than experimental. It presents an argument from Michael R.
Dietrich, a Professor at Dartmouth College, who believes that these simulations are the
same as controlled experiments. Beisbart and Norton proceed to describe their main
argument, which states that pseudorandom numbers are derived and no element of a true
experiment can be derived.
In contrast, Professor Paul Humphreys describes the attributes of Monte Carlo simulation
which he believes should be classified as a new type of analysis. In his paper, he says
that it is neither empirical experimentation nor a numeric method, instead it is what he
calls numerical experimentation (Humphreys, 1994). He utilizes the law of large
numbers which explains that probabilities will converge to the statistical estimate as the
number of trials increases. This suggests that the larger amount of simulations that can
be run will lead to more accuracy in the results. He further explains the value of
transparent computations, which suggests that the complexity of the equations will play a
factor in the accuracy of results. Humphreys sees Monte Carlo simulation as a valid
method of accurately running an experiment.
The discussion of simulation as a legitimate method of research is pivotal to this project.
This is the primary method of modeling the card counter’s behavior and therefore
challenges the legitimacy of this entire study. Beisbart and Norton raise the argument
that the derivation of the pseudorandom number generator proves that simulation is
inferential not experimental. However, Humphrey counters that the law of large numbers
illustrates the value behind this method. Furthermore, countless professors and
11
researchers have utilized this method of simulation in papers published in the
mathematics field, rendering it a substantial method of analysis.
Professor Patrik Andersson of Stockholm University applied simulation for his analysis
of card counting in a continuous manner, mentioned above. Similarly, Professor Leslie
Golden, of the University of Illinois employed this in the analysis of the effects of
multiple decks. In addition, the professors of Beijing Normal University published their
article on simulation in an international journal. Lastly, the professors and statisticians
that wrote the article for the University of Las Vegas Gaming Research & Review
Journal, simulated slot machines patterns, therefore this method has also been proven in
the casino industry. Overall, the dispute over Monte Carlo simulation raised concerns
that have been disproven by the research and continued use of this method for research
analysis.
The business articles that contributed to this research suggest that the best practice in the
casino industry is to acquire data. The casino and gambling industry have a market value
of over $119 billion dollars, 38.5% of that corresponds directly to casinos (Casinos &
Gambling Industry Profile: North America, 2014). Moreover, by the year 2018 this
industry is projected to see a 15.5% increase in market value from its 2013 numbers.
This suggests that with a growing industry the demand for data and technology will
continue to trend upward. Therefore, the articles relating to the business and technology
aspect of casinos are constantly becoming outdated.
12
In 2002, a professor at the University of Nevada Las Vegas presented a study which
provided analysis of all computer systems on the market that provide player tracking and
slot accounting for casinos (Wang & Aquino, 2002). Its purpose was solely to present
the list of attributes for each product on the market to help educate casino management of
their options. The increase in technology in the recent years has decreased the value of
this study because this software is no longer a complete list of the options. However, one
beneficial element of the study is that it illustrates the desired qualities in casino
technology. This article found that casinos will accept high initial cost if it would lead to
a business practice that will increase savings in the long run.
Professor Wyld of Southeastern Louisiana University conducted a study in 2007 that
would once again profile the technology offered. It explains the risks that casinos face in
terms of both counterfeit and theft of casino chips. In response, new technology is being
developed that will put microchips in each poker chip. This will allow casinos to track
the players’ movement through the casino as well as their betting behavior. It directly
mentions the application for gathering analytics with these chips to detect card counting
based on betting patterns. These chips are not currently in use but expected to be
developed in the near future. If this new technology is implemented, this will enhance
the data in the casino industry.
The movement for data has led to the infusion of mathematics into most important
business decisions. Jeff Ma emphasizes the importance of gathering as much data as
possible to make informed decisions. The profile of the casino industry estimates the
future growth and has projected it to rise steadily through 2018. The article for the
13
Gaming Research and Review journal published in 2002 showed that technology is a
large upfront cost that casinos are often willing to invest if it will lead to long term
savings. In 2007, only the concept for smart chips utilizing RFID technology had been
conceived but this will potentially be implemented in the future. Overall, casinos will
continue to collect and analyze data to support their business decisions.
Another point of contention related to many analyses of gambling is based on the concept
of “luck”. This is an idea that has been psychologically reviewed with respect to effects
of luck in group dynamics versus individual play. The concept of luck has also been
referred to as irrational behavior. Each researcher and psychologist have a different
method of analysis to interpret the patterns of winning and losing that some people term
luck. The small sample size in many of these experiments makes it difficult to apply
their proven findings. Each psychological study provides insight into the thoughts and
behaviors of gamblers even though they may not all directly represent the population.
An experiment was conducted by psychology professors Gunnarsson, Whiting, and
Dixon at the Southern Illinois University Carbondale in 2014, however the results may
not be representative of the whole population. The experiment was conducted with
sixteen people and it was based on refined rules of the game. This study was designed to
compare the person's self-perception when played alone or among a group. Each player
was asked to estimate his or her likelihood of winning a hand of blackjack when played
in each environment. The players were also given modified rules to help control the
variables in the experiment. The findings indicate that players' perception is based on
their relative position to others. The results showed that when playing alone the players
14
in the experiment lost 45% of the time, however when playing in a group they lost 63.5%
of the time (Gunnarsson, Whiting, & Dixon, 2014). They drew the conclusion based on
these participants that betting behavior was influenced more by the other players’
proximity to a score of 21 rather than the dealer’s proximity to 21. They also concluded
that this meant players were concerned with comparing their potential for winning to
their peers. Therefore, the typical casino setting which has multiple players may reflect
the group dynamics exhibited in this experiment. However, based on the small sample
size and modified rules these results are less reflective of society.
Another study analyzed player behavior in a game of blackjack to determine whether
gambling behavior is related to a player’s perception of his or her own level of control
(Chau & Phillips, Effects of perceived control upon wagering and attributions in
computer blackjack, 1995). The study was limited to computer based blackjack which
was then manipulated to determine players' responses to the winning and losing. It was
conducted with only twelve participants from the University of Hong Kong in 1995. The
manipulated hands make it difficult to determine whether a player would have made
rational decisions in a truly random game of blackjack. Overall, this study showed
insight into a small sample of players' opinions throughout the duration of game. The
professors who conducted this study also ran an experiment on the gambling behavior.
The second study conducted by Professors Chau, Phillips, and Baggo looked into the
rationality of decision making in gamblers. In this experiment the players were given
advice on what moves rational play would suggest. However, in many cases the player
would ignore the advice based on their own personal strategies or emotional feelings.
15
The professors concluded that it was not a lack of intelligence that caused the stray in
player choices from the sensible decision, instead they believe that gambling leads to a
lack of control (Chau, Phillips, & Von Baggo, Departures from sensible play in computer
blackjack, 2000). This corresponds to the findings from the study by Professors
Gunnarsson, Whiting, and Dixon who showed that group dynamics increase the loss
ratios of players. The studies both suggest that group dynamics decrease a player’s
success when playing blackjack.
The study by Professors Keren and Wagenaar examined the patterns in which people
believe they experience both good and bad luck. In fact, the survey responses reflected
that 68.2% of people believed the game was determined by luck. Other players believed
that the results of the games were predetermined rather than probabilistic. This relates to
the study by Professors Chau, Phillips, and Baggo which suggests the irrational choices
gamblers make. The concept of luck may lead gamblers to the notion that their own
personal strategies will return better results than the calculated probabilistic choices. Ma
remarks that a card counter should always bet the way the count tells him (Ma, 2015).
The psychological research illustrates personal choices that deviate from the logical
choices a card counter would make.
Conclusion
There are a number of concepts discussed around card counting which include
mathematical, technological, and psychological ideas. One of the key components of this
capstone is simulation, therefore it is necessary to consider the arguments around this
subject. The psychological argument states that simulation should be considered
16
inferential rather than experimental. However, the professors of mathematics proved that
in their field this is considered an adequate form of research. Another area discussed is
that businesses want the most current technology and data. This concept requires
mathematics in conjunction with analytics to interpret data that can inform casino
business decisions. In respect to the psychological research that has been conducted, a
gambler generally exhibits behavior that would deviate from sensible play. This may be
the result of luck, or strategy but psychologically speaking the average gambler will not
bet systematically as logic dictates.
This capstone is an addition to the field of mathematics and the casino and gambling
industry. This project will include aspects of previous mathematical studies to be as
accurate as possible, such as considering the multiple deck effects and applying the
central limit theorem to find a baseline for card counter betting style. A thorough
understanding of the business field will provide the best recommendation to casinos
based on technological advances and preventative measures. The psychological
examination of the player’s choice will allow for the maximum accuracy of the baseline
normal behavior, especially when used in combination with the observations. Overall,
this project is focused on mathematical principles with the inclusion of studies from all
relevant fields to increase the probability of catching card counters.
METHODOLOGY
Introduction
The method of research for this project consisted of a literature review, blackjack
simulation, and player observations. The idea behind this capstone was to determine
17
what differences exist between a normal person playing blackjack and a person card
counting at blackjack. The hypothesis was that if a casino monitors betting habits then
the casino will be able to determine whether a player is card counting. In order to prove
this hypothesis, data was gathered from current studies that pertained to casinos and card
counting. This review included literature from mathematics, business and psychology
fields to provide a complete analysis of the casino industry. Next, a simulation was
created to model the dealer in a game of blackjack. Then, the player observations were
collected to compare against a card counter. Afterward, an analysis was conducted
comparing a variety of betting strategies for each. Finally, the simulated data was
evaluated in contrast with the player observations to reach these findings.
Simulation
The simulation is an excel file that was originally developed by a student at Bryant
University, Jean-Paul Saggal, who volunteered it for this project. At this stage of the
program, the user would enter the desired number of hands to be played using a single
deck and that many hands of blackjack would be dealt. It was able to calculate the total
number of wins and losses for a player using basic blackjack strategy over the entire
interval. The rules for basic strategy can be found in Appendix A. Next, Professor Jim
Bishop modified the program to incorporate the fact that casinos use eight decks and
reshuffle after every six decks (Golden, 2011). This was information gathered in the
literature review that provided a more accurate simulation. Professor Bishop also
adapted the program to display the card count after each hand and illustrate the respective
wins and losses for each count.
18
Observations
The player observations were conducted on December 12th, 2015 at the Foxwoods Casino
in Mashantucket, Connecticut. There were seven individuals that participated in
observing player behavior during the hours of 4:00pm to 10:30pm. Each person observed
a table of blackjack players for ten consecutive hands of play before moving on to
another table. The observer noted the table minimum as well as the maximum bet for
each player in ten hands. A total of 1,260 hands of blackjack were observed in this
experiment. The gamblers were observed in different areas of the casino, at different
tables, and playing at different table minimums. These observations were used to create a
baseline for normal gambler maximum bet in 10 hands.
Betting Strategies
There are a number of betting strategies that can be employed by card counters, however
all of the methods used involve betting based on the count (Tamburin, n.d.). This study
focused on four main ways to bet based on count which include a five unit spread, a ten
unit spread, betting as a function of the minimum bet and betting as a function of the
maximum count.
Betting by Card Count
 Five Unit Spread: When using the Hi-Lo card counting strategy, as outlined in
Table 1 on page 4, the method of betting should correlate to the true count. The
true count is calculate by taking the observed count and dividing by the number of
decks remaining to be played. This method begins with a minimum bet and
increases based on the count up to five times that minimum (Golden, 2011). It
19
does not exceed five times the minimum bet to help the card counter avoid
detection. The table below illustrates the pattern to follow for a minimum bet of
15 and can be adapted for any minimum bet.
Betting Amount
Count Bet Unit
Below 0 Don’t Play
0-3 15
4-7 30
8-11 45
12-15 60
16+ 75
Table 2: Five Unit Spread
This method does have a long term advantage over the house but may lead to
short term losses in the process.
 Ten Unit Spread: Similar to the five unit spread, this method begins with a
minimum bet and increases based on the count up to ten times that minimum
(Betting Spread, 2016). The table below illustrates the pattern to follow for a
minimum bet of 15 and can be adapted for any minimum bet.
Betting Amount
Count Bet Unit
Below 0 Don’t Play
0-2 15
3-4 30
5-6 45
7-8 60
9-10 75
11-12 90
13-14 105
15-16 120
17-18 135
19+ 150
Table 3: Ten Unit Spread
20
This method also will have long term advantage over the house, with the wider
spread it is less likely to lead to short term losses but still possible.
 Function of the Minimum: Unlike the other methods, this method follows a
formula which changes the bet based on each individual true count.
(True Count-1)*(Minimum Bet)
The formula is to subtract one from the true count and multiply that by the
minimum bet (Bloch, 2016). This method further spreads the bet range and
exploits the high counts.
 Function of the Maximum: This method requires knowledge of the maximum
true count of the deck, and the maximum amount that the card counter is willing
to bet. The equation is as follows (Blackjack Betting Spread Explained, 2015):
𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝐵𝑒𝑡
(𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝑇𝑟𝑢𝑒 𝐶𝑜𝑢𝑛𝑡) * True Count
This method creates a range that allows the card counter to choose the maximum
amount he or she will bet and therefore the largest possible amount he or she can
win.
Analysis
The analysis consisted of a side by side comparison of each betting strategy run over
125,000 times in the simulation. In a separate excel worksheet the card count and
win/loss for each hand were recorded. Next, the formula for each betting strategy was
run to calculate the bet made each hand for each of the four methods. Finally, these
games were broken into sections of ten hands each and compared to the maximum bets
21
found in the observations. This combination of data led us to the determining when a
card counter could be detected.
Conclusion
This project focuses on methods that will increase the probability of catching card
counters within a casino. It analyzed current studies that relate simulation data to the
game of blackjack. These provided insight to the research that has previously been
conducted to ensure that this study enhances the field, rather than repeating previous
works. It utilized simulation to run multiple hands of blackjack and determine the
likelihood of catching a card counter by focusing on the betting strategies of each type of
player. Finally, it compared this data to the player observations to determine key
differences in behavior.
FINDINGS
Simulations and Observations
In order for the observation data to be compared to the four card counting methods, the
simulation data had to be in the same format. The observation data was recorded based
on the maximum bet of ten consecutive hands. Therefore, the over 125,000 simulations
that were run with each of the four methods had to be broken down into groups of ten as
well. Once the simulation data was grouped into ten hands the maximum of those ten
hands was recorded. The maximum bets for the observation data and each of the four
methods were graphed and can be found in Appendices B-F.
Statistical Tests
To determine whether the card counter methods were significantly different from the
observations of normal gamblers, statistical tests were run. In order to compare the data
22
further grouping was required and the maximum bets were broken into 5 groups as
follows: bets from 15-20, bets of 25-35, bets of 40-55, bets of 60-70 and bets of 75 or
more. The first test was a chi-square test to determine the significance of each group
appearing. The table below shows the percentage of bets from each method that belong
in each category compared to the percentage of bets in each category from the observed
data.
Table 4: Chi-Square Results
In order to prove significant the P-value would have to be <.1 and at the bottom of the
table it is clear that none of the methods here are considered significantly different. At
the point it was determined that if the casino monitored the maximum of all of the bets
and compared each player it would be difficult to determine a difference between the card
counter and the average gambler.
The next test focused only on the bets in the highest group, which is the bets of 75 or
more. This test was a normal approximation to the binomial formula and it was run
comparing the card counter methods to the observation data to produce the table below.
23
Table 5: Normal Approximation Results
The results of this test were significant for each of the four methods. This proves that if
the casino focused on only the maximum bets that are 75 and higher than the casino can
determine whether the person is card counting. In the first method, the five unit spread,
the card counter would bet over 75 dollars significantly less than the average player and
in the other three methods the card counter would bet over 75 dollars significantly more
than the average player.
The final element that was examined, was the frequency in which a high count comes up
during the game of blackjack. In order to determine this the 125,000 simulations were
used as the baseline. The graph in Appendix G shows how often each count appeared in
the simulations. For this test a high count was determined as a true count of 14 or higher
because this count would yield a bet of more than half of the maximum bet in every
method. The results showed that a high count would come up in an 8 deck game about
every 200 hands dealt. Therefore when playing at a table with 5 players and a dealer, the
typical casino table layout, a player would see a high count approximately every 33
hands. This was also compared to the player observation data to determine how many
high bets would be seen in a normal game. In the 1,260 hands observed a card counter
would have been high 38 times, whereas the observed data only yielded 22 high bets.
24
Conclusion
Overall, this study examined the game of blackjack to determine how a casino could
catch a card counter. After reviewing the techniques in the field, the current practice is
that the floor manager or dealer to watch any players exhibiting suspicious behavior or
winning abnormally often. This study illustrates a new method that would require
additional effort on the part of the casino but is statically proven to determine which
players are card counters. In this method the casino would monitor how often high bets
of five times the minimum or more were placed in the overall hands played. If a software
was developed or the dealers were trained to observe this behavior card counters would
be much easier to catch.
25
APPENDIX A: BLACKJACK BASIC STRATEGY
26
APPENDIX B: OBSERVATION MAXIMUM BETS
APPENDIX C: 5 UNIT SPREAD MAXIMUM BETS
27
APPENDIX D: 10 UNIT SPREAD MAXIMUM BETS
APPENDIX E: FUNCTION OF MINIMUM BET MAXIMUM BETS
28
APPENDIX F: FUNCTION OF MAXIMUM COUNT MAXIMUM BETS
APPENDIX G: FREQUENCY OF HIGH COUNTS
29
REFERENCES
Andersson, p. (2012). Card counting in continuous time. Journal of Applied Psychology,
184-198.
Beisbart, C., & Norton, J. (2012). Why Monte Carlo simulations are inferences and not
experiments. International Studies In The Philosophy Of Science, 403-422.
Betting Spread. (2016). Retrieved from Online Blackjack:
http://www.onlineblackjack.com/betting-spread/
Blackjack Betting Spread Explained. (2015). Retrieved from Counting Edge:
http://www.countingedge.com/blackjack-betting-spread/
Bloch, A. (2016). Card Counting and Ranging Bet Sizes. Retrieved from Instructables:
http://www.instructables.com/id/Card-Counting-and-Ranging-Bet-
Sizes/?ALLSTEPS
Casey, R. (2008). The MIT blackjack team and motivation theory. Annual Advances In
Business Cases, 141-150.
Casinos & Gambling Industry Profile: North America. (2014). Casinos & Gambling
Industry Profile: North America, 1-32.
Chau, A. L., & Phillips, J. G. (1995). Effects of perceived control upon wagering and
attributions in computer blackjack. Journal Of General Psychology, 253-269.
Chau, A. L., Phillips, J. G., & Von Baggo, K. L. (2000). Departures from sensible play in
computer blackjack. Journal of General Psychology, 426-438.
30
Clark, T. L. (1986). Cheating terms in cards and dice. American Speech, 3-32.
Golden, L. M. (2011). An analysis of the disadvantages to players of multiple decks in
the game of twenty-one. Mathematical Scientist, 57-69.
Gunnarsson, K. F., Whiting, S. W., & Dixon, M. R. (2014). The near-miss effect in
blackjack: group play and lone play. Analysis Of Gambling Behavior, 87-94.
Humphreys, P. (1994). Numerical Experimentation. Patrick Suppes: Scientific
Philospher, 103-118.
Keren, G. B., & Wagenaar, W. A. (1985). On the psychology of playing blackjack:
normative and descriptive considerations with implications for decision theory.
Journal Of Experimental Psychology, 133-158.
Ma, J. (2015, May 18). SAS Global Forum. Retrieved from YouTube:
https://www.youtube.com/watch?v=wYPOIvRyZdE
Pepper, J. (2016). Blackjack Basic Strategy. Retrieved from Wizard of Odds :
http://wizardofodds.com/games/images/blackjack/wizard-simple-exceptions.gif
Rehagen, T. (2010). Raising the Stakes. Indianapolis Monthly, 34(2), 78-83.
Singh, A., Lucas, A. F., Dalpatadu, R. J., & Murphy, D. J. (2013). Casino games and
central limit theorem. UNLV Gaming Research & Review Journal, 45-61.
Tamburin, H. (n.d.). How to Bet at Blackjack. Retrieved from American Casino Guide:
http://www.americancasinoguide.com/blackjack/how-to-bet-at-blackjack.html
31
Wang, Z., & Aquino, H. (2002). Casino technology: player tracking and slot accounting
systems. Gaming Research & Review Journal, 43-55.
Wyld, D. C. (2007). The ace in the hole: how smarter chips - and cards - can enable
casino management to gain insight into player behavior and better compete in
today's increasingly crowed entertainment marketplace. Competition Forum, 166-
172.
Yang, T., Li, Q. X., Zhoa, E., Liu, G., & Han, Z. (2011). How long can you enjoy
blackjack with 100 chips? International Journal Of Modern Physics & Physical
Computation, 1161-1171.
32

Catching Counters

Uploaded by

Copyright:

Available Formats

Catching Counters

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Catching Counters

Uploaded by

Copyright:

Available Formats

Catching Card Counters

BY: Sarah French

ADVISOR • Jim Bishop

EDITORIAL REVIEWER • Phyllis Schumacher

psychological gambling habits, technological advances, business strategies, and

who would bet methodically. Conversely, the unpredictable betting strategy of a

An understanding of these differences will provide a casino with additional information

to catch card counters at the table.

cards. Mathematically speaking this is considered sampling without replacement which

0, or 1 to each card dealt according to the following table:

Card: Ace 2 3 4 5 6 7 8 9 10 Jack Queen King

Table 1: Hi-Lo Strategy

social interest for this analysis.

was one of many studies that incorporated mathematical equations in predicting

outcomes of casino games.

Another analyzed the game of blackjack through computer simulations to determine

determine the effect of multiple decks on the success of a card counter.

scientific study. The researchers ran upwards of a million rounds of simulations to

assumptions of minimum bets were necessary, so it cannot be considered entirely

study it would require incorporating all of the controls.

bias. It explained the mathematical approach to calculating probabilities of success and

simulation techniques in the casino industry today.

provides a market definition and segmentation, as well as listings of industry leaders

data on various industries. It provided definitions of each market segment, which

the industry grows the opportunities for card counters increase.

implemented in casinos. It was written by a Professor Wyld at Southeastern Louisiana

detect card counters.

blackjack, 2000). A few researchers used a small sample of students, 16 undergraduates

into what questions should be discussed from a psychological standpoint.

player behavior, as compared to strategic card counter behavior.

are inferential rather than experimental. It presents an argument from Michael R.

experiment can be derived.

that it is neither empirical experimentation nor a numeric method, instead it is what he

calls numerical experimentation (Humphreys, 1994). He utilizes the law of large

method of accurately running an experiment.

The discussion of simulation as a legitimate method of research is pivotal to this project.

researchers have utilized this method of simulation in papers published in the

mathematics field, rendering it a substantial method of analysis.

of card counting in a continuous manner, mentioned above. Similarly, Professor Leslie

article on simulation in an international journal. Lastly, the professors and statisticians

aspect of casinos are constantly becoming outdated.

a business practice that will increase savings in the long run.

Professor Wyld of Southeastern Louisiana University conducted a study in 2007 that

the data in the casino industry.

business decisions. Jeff Ma emphasizes the importance of gathering as much data as

continue to collect and analyze data to support their business decisions.

referred to as irrational behavior. Each researcher and psychologist have a different

An experiment was conducted by psychology professors Gunnarsson, Whiting, and

Another study analyzed player behavior in a game of blackjack to determine whether

success when playing blackjack.

choices a card counter would make.

capstone is simulation, therefore it is necessary to consider the arguments around this

subject. The psychological argument states that simulation should be considered

bet systematically as logic dictates.

industry. This project will include aspects of previous mathematical studies to be as

based on technological advances and preventative measures. The psychological

relevant fields to increase the probability of catching card counters.

collected to compare against a card counter. Afterward, an analysis was conducted

evaluated in contrast with the player observations to reach these findings.

wins and losses for each count.