Dobozy 1

We Have Candy

Ben Dobozy
Kitchener-Waterloo Collegiate & Vocational School
MDM4UI
Mrs. Ford
Dobozy 2

Contents

1 Introduction 3

2 Instructions and Game Design Philosophy 3

2.1 Game Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.2 Gameplay Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Theoretical Predictions 5

4 Experimental Results 9

5 Analysis 10

6 Successes and Improvements 11

7 Connections 13

8 Conclusion 13

9 Appendix: Photos of the Final Game Design 14

Dobozy 3

1 Introduction

The annual KCI games fair is a mass experiment in statistics. Students enrolled in MDM4UI form small

groups and design games of chance. Students calculated in advance the odds of each outcome in their game

and the game’s overall expected value. One event day, the students set up their games in the cafeteria

and interested students were given ”Data Dollars” to spend on the games. The cost to play a game was

determined by its expected value. The number of occurrences of each outcome at the fair were recorded by

the students and compared to the predicted theoretical results. As a result, it would be ideal to attract as

many people as possible to your game.

After taking an informal survey of KCI students, my partner Dan Duran determined that the most desired

quality in a games fair game is free candy. We designed a minimalist marketing scheme focused on making

it as clear as possible that there would be free candy available. A picture of our final setup is available in

the appendix. We titled our game ”We Have Candy feat ben and dan”(sic). We designed a simple dice game

that ensured people didn’t feel cheated and would want to keep playing while at the same time stacking the

odds in our favour. It was clearly a successful marketing scheme, as we were able to run 85 trials.

2 Instructions and Game Design Philosophy

The game went through several iterations before we settled on the final version. Initially, the game

involved someone playing indefinitely until they lost. There would be a prize pool available that doubled

every round and they could either take the money or continue doubling it. However, the ability to either

”take it or leave it” meant that the player had choice, so it was not a game of purely chance. Also, it would

be impossible to calculate the expected value of such a game using the techniques we had been taught up to

that point.

As a result, we adjusted our design. We still wanted to keep it simple as we thought that a simple game

would be much more likely to both attract and retain players. The central idea of playing until you lost and

rolling dice were retained, but everything else was thrown out. We settled on a five round game where players

were rewarded based on the round they reached, up to a grand prize from winning round five. The specifics

of the game are described below. All that was left was to pick (more or less arbitrarily) the prize values that

would place the expected value within the acceptable range. We selected payout values such that our game

cost 2 data dollars to play. We decided that the values should get larger until the last round, where a win

awarded less money than a loss. The intent was to create tension in the last round of the game.
Dobozy 4

2.1 Game Materials

TM
• Beyblade Metal Fusion Bolt Blast Stadium

• 2 6-sided dice

2.2 Gameplay Instructions

1. The player declares two numbers

TM
2. They roll the dice in the Beyblade Metal Fusion Bolt Blast Stadium

3. If either of the numbers declared appear on the dice, the player advances to the next round (if none

appear, they lose.)

4. The player declares 1 number and rolls both dice.

5. If the number appears on either dice, the player advances to the next round.

6. Repeat steps 4 and 5 for subsequent rounds, up to round 5.

If the player loses at any given round (or wins round 5), consult Table 1 for the prize they win. For a diagram

of standard game progression, see Figure 1. A photo of the game setup is provided in the appendix.

Table 1: Prizes Awarded

Outcome Payout ($)

Loss Round 1 0
Loss Round 2 1
Loss Round 3 3
Loss Round 4 10
Loss Round 5 16
Win Round 5 32

Figure 1: A Diagram of Standard Gameplay

Dobozy 5

3 Theoretical Predictions

Figure 2: Tree Diagram Used to Calculate Probabilities of Each Outcome

Calculating the probability of each outcome was fairly straightforward. For the first round, the player

rolls two dice and declares two numbers. They win when either number declared came up on the dice. The

probability of winning the first round is therefore given by:

2 2
P (Winning Round 1) = +
6 6
4
=
6
2
=
3

And the probability of losing is simply the conjugate of this value:

2
P (Losing Round 1) = 1 −
3
1
=
3
Dobozy 6

For all subsequent rounds, only one number is declared, so the probability of winning is given by:

1 1
P (Winning rounds 2-5) = +
6 6
1
=
3

And:

1
P (Losing Rounds 1-5) = 1 −
3
2
=
3

These are the probabilities for each round in isolation. However, in order to win (or lose) a given round,

one must first have reached that round, meaning they must have won the previous rounds. In other words,

P (Winning Round 2 6= P (Winning on Round 2)). For example, the probability of losing on round 3 is given

by:

P (Losing on Round 3) = P (Winning Round 1) and P (Winning Round 2) and P (Losing Round 3)
2 1 2
= ( )( )( )
3 3 3
4
=
27
≈ 0.148

2
Note that since the probability of winning the first round is 3, the probability of winning any subsequent

round is 13 , the probability of losing any subsequent round is 23 , and the probability of losing on a given round

is the product of the probabilities of winning each previous round times the probability of losing the current

round, it is simple to derive an expression for the probability of losing on any round after the initial round.

First, since:

L(n) = P (Winning Every Previous Round) · P (Losing the Current Round)

And:

   n−2
2 1
P (Winning Every Previous Round) =
3 3
Dobozy 7

Therefore:

   n−2  
2 1 2
L(n) =
3 3 3
 2  n−2
2 1
L(n) =
3 3
   n−2
4 1
L(n) = (1)
9 3

Where n is the round in question and L(n) is the probability of losing on the nth round.

Note that since there are two possible outcomes (a win or a loss), the probabilities don’t change from one

round to another, and (1) describes the probability of a loss after a given number of rounds, (1) describes a

geometric distribution. This formula applies only to the middle rounds (loss on 2 to loss on 5). It doesn’t
1
apply for the special cases of a loss on round 1 or a win on round 5. A loss on round 1 is simply 3 and, while

the probability of winning on round 5 is slightly harder to find, it is given by the equation:

W (n) = P (Winning Every Previous Round) · P (Winning the Current Round)

   n−1
2 1
W (n) = (2)
3 3

Using equations (1) and (2), we generated a table in Microsoft Excel of possible outcomes and their
Pn
respective probabilities. Using the formula for expected value, E(n) = i=1 (xi )(P (i)), where P (i) is the

probability of losing on the ith round, and xi is the payout for losing on the ith round, we calculated the

expected value for the game. We summarized the information in a table (rounded to 2 decimal places for

simplicity):

Table 2: Probabilities of Each Outcome and Expected Value Calculation

Outcome Payout ($) Theoretical Probability Expected Value (x × P (x))

Loss Round 1 0 0.33 0.00

Loss Round 2 1 0.44 0.44
Loss Round 3 3 0.15 0.44
Loss Round 4 10 0.05 0.49
Loss Round 5 16 0.02 0.26
Win Round 5 10 0.01 0.26
Dobozy 8

These values can be plotted to show the following probability distribution:

Figure 3: Theoretical Probability Distribution

The expected value was found by simply summing up the items in the fourth column of Table 2. It was

found to be 1.73. Using this result, we also predicted our expected profit. We used the formula:

P rof its = 2n − E(x) × n

Where n is the number of trials and E(x) is the expected value. Using this formula, we predicted that after

85 trials we would have made a profit of 22.95 data dollars. We could expect our profit to be somewhere

around 23 dollars.
 Dobozy 9

4 Experimental Results

Summarized in a table, our experimental results are as follows:

Table 3: Probabilities of Each Outcome and Expected Value Calculation

fi
Outcome Payout ($) Frequency (fi ) Experimental Probability ( total ) Expected Value (P (i) × P ayout)

Loss Round 1 0 32 0.38 0

Loss Round 2 1 36 0.42 0.42
Loss Round 3 3 10 0.12 0.35
Loss Round 4 10 6 0.07 0.71
Loss Round 5 16 1 0.01 0.19
Win Round 5 10 0 0 0

The calculation of each part of this table is fairly straightforward. The total number of trials was found by

simply summing the number of occurrences of each outcome. The experimental probability of each outcome

was then determined by dividing the frequency of that outcome by the total number of trials, and the expected

value was found by multiplying the experimental probability and payout for each outcome and summing them

together. In other words:

fi
P (experimental) =
total

and:

6
X
E(x) = P (experimental) × P ayout
i=1

We found our experimental expected value to be 1.67. We made a profit of 28 data dollars. However, we

had to ask for extra money to cover debts at multiple points during the games fair. Once that is taken into

account, our final profit was 18 data dollars. Though I will cover this in more detail in the analysis section,

our experimental expected value is remarkable close to our theoretically predicted value. This is possibly

because we had a very large sample size thanks to Gabriel Meissner who simply stood there and played our

game for the entire lunch period. It was strange that there was not a single occurrence of someone winning

round 5. The theoretical probability of someone winning round 5 (P ≈ 0.01) suggests that we probably

should have seen it happen at least once in 85 trials.

Dobozy 10

Figure 4: Experimental Probability Distribution

5 Analysis

Our theoretical and experimental expected values differed by only 0.06. To calculate percent difference,

I used the equation:

|V1 − V2 |
Percent Difference = (V1 +V2 )
× 100%
2

I let V1 equal 1.73 and V2 equal 1.67. The percent error in our expected value is 3.53%. This is a remarkable

similarity, which probably exists because of Gabriel Meissner’s help. Since the AP lunch happened to coincide
Dobozy 11

with the games fair, we had access to a surplus of Chinese food. In exchange for the Chinese food, Gabriel

Meissner agreed to play our game over and over again. This is probably why our experimental results match

our predictions so closely.

Figure 5: Comparison of Theoretical and Experimental probability distributions

There are slight variations in the data due to our sample size. While it was relatively large, it would

ideally be in the 100s. It terms of profit, we expected that our profit would be somewhere around 23 dollars,

and ended up with a net profit of 18 dollars. This is slightly lower than expected, but can also be attributed

to low sample size and random stastical fluctuations. It is also possible that we simply missed taking money

from Gabe one of the times he played our game, or failed to record his results once or twice, which could

more than account for the discrepancy between our predicted and experimental profits.

6 Successes and Improvements

Overall, I believe that our group was very successful. We had a good (but not ideal) sample size, and even

ignoring the fact that we hired Gabriel Meissner to play our game for us, we succeeded in attracting many

people to play our game. Our minimalist design and very clear, no-nonsense marketing resulted in many

people coming to try our game. We also created a very straightforward set of rules that were still complex

enough to be mathematically interesting. The simplicity of the game was another factor in our success, as it
Dobozy 12

was easy to play again and did an excellent job of creating tension.

One of the best decisions we made was to offer candy as a prize for winning the first round. It didn’t

cut into our profits at all, and we could select a prize for the second round that would ensure we would still

make a profit if someone lost there. People don’t care about data dollars, they care about candy, so they

kept playing to win more and more candy, totally missing the fact that they were losing money every time.

Also, the odds of making it to the second round were high enough that most people who played won candy.

This helped attract people to our game.

However, our record keeping system could probably have been improved, and we could have payed more

attention to the money we distributed. I’m not sure whether or not those were major factors in our profit

discrepancy, but it would help ensure our results are accurate. We could also have adjusted the payout values.

We needed to ask for extra money several times as a consequence of people winning relatively easy prizes

that were valued too high. Specifically, the relatively large number of people who made it to rounds 3 and

4 resulted in us running out of money. It would have run more smoothly if the prize value for round 4 had

been slightly lower, and the values for a loss and win on round 5 had been made higher to compensate.
TM
Another problem with our game was the design of the Beyblade Metal Fusion Bolt Blast Stadium .

The staduim was designed for BeybladeTM battles, not rolling dice, and as a consequence had holes on both

sides. In a BeybladeTM battle, being knocked into these holes counts as a loss. However, the holes acted as

traps for the dice, which would fall in on occasion, sometimes being prevented from landing on a specific face

by the walls of the hole. A potential design improvement would be to use an alternate BeystadiumTM such

at the one pictured below:

Figure 6: Beyblade Burst Turbo Slingshock Rail Rush Battle SetTM

Dobozy 13

7 Connections

The most important thing I learned from this project is the importance of good marketing. We surveyed

the student body, determined the the thing they wanted most out of a games fair game, and centered our

marketing on that. And it worked. Our advertising made it clear that we had candy, the candy was easy to

win, and that’s all you need to know about our game. A high contrast, black-on-white sigh was eye-catching,

easy to read, and got the message accross. The simplicity of our game also worked in our favour. Since our

game was easy to pick up and play and gave the player the illusion of control it was highly addictive. Most

people who came to play it played it at least twice.

These lessons can be easily extended to apply to real life. In the future, if I need to market something,

I’ll make sure to keep my marketing as simple and to the point as possible. When I’m designing a game, or

any other activity, I’ll make sure to keep the rules simple. If people can pick up and play a game easily, then

they’re far more likely to both start, and continue playing.

8 Conclusion

Overall, our predictions were incredibly accurate, with the percent error in our expected value being only

3.53%. Our high accuracy can be attributed to our relatively large sample size (thanks again to Gabriel

Meissner). Theere was also a discrepancy between our predicted profit and our theoretical profit as well as

the experimental and theoretical probabilities of each outcome (Figure 5). However, both of these can be

expected with this kind of stastical experiment. These deviations can be expected to shrink with a larger

sample size. That may not be fully necessary however, as if there’s one thing I’ve learned over the course of

this project, it’s not about how much money you make or how accurate your predictions are. What matters

is the friends we make along the way.

Dobozy 14

9 Appendix: Photos of the Final Game Design

Figure 7: A photo of our final game setup. Note the instructional ”Roll Here” sign. The dice are not pictured.

Figure 8: A photo of the game during the actual games fair. Note the highly visually appealing
”We Have Candy” sign

Though neither photo makes it obvious, the player rolls both dice inside the pictured BeystadiumTM
Dobozy 15

References

Ms Ford, MDM4UI Course Notes - Grade 12 University Level Data Management.