EQUAL RELABELINGS FOR P Q-SIDED DICE A Thesis Presented to the Faculty of California State Polytechnic University, Pomona In Partial Fulfillment Of the Requirements for the Degree Master of Science In Mathematics By Alec Lewald 2017 SIGNATURE PAGE THESIS: EQUAL RELABELINGS AUTHOR: Alec Lewald DATE SUBMITTED: Winter 2017 Department of Mathematics and Statistics Dr. Amber Rosin Thesis Committee Chair Mathematics & Statistics Dr. Berit Givens Mathematics & Statistics Dr. John Rock Mathematics & Statistics ii ACKNOWLEDGMENTS This thesis is dedicated to my family, friends, and professors who have supported me throughout my education. A special thank you to my parents and Samier who always pushed me even when it was abundantly clear that I had no intention of working. iii ABSTRACT For m n-sided dice, we will call the sums m, m + 1, m + 2, . . . , mn the standard sums, and we will say that m n-sided dice have an equal relabeling if the dice can be labeled with positive integers in such a way that the standard sums are equally likely to occur. We will consider the case when n = pq for distinct primes p < q. The m for which n = pq sided dice have an equal relabeling have previously been characterized and it has been shown that all such equal relabelings require stupid dice (dice with 1’s on every face). We find the minimum and maximum number of stupid dice that are needed for an equal relabeling of pq-sided dice and show that such a relabeling is possible for any number of stupid dice between the maximum and minimum. iv Contents Signature Page ii Acknowledgements iii Abstract iv 1 Background 1 2 A Partial Proof of a Previously Unjustified Statement 5 3 The Greatest and Least Number of Stupid Dice 13 3.1 Stupid Dice Facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.2 An Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.3 An Equal Labeling . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 4 Further Questions 23 Bibiliography 24 v Chapter 1 Background Consider rolling a pair of 6-sided dice. The possible sums are 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, and 12 where the likelihood of rolling each of these sums is and 1 36 1 2 3 4 5 6 5 4 3 2 , , , , , , , , , , 36 36 36 36 36 36 36 36 36 36 repectively. Can we relabel the dice so that we still get the sums 2, 3, . . . , 12 but the sums have an equal chance of being rolled? The answer is no due to the fact that there are 6 · 6 = 36 possible die rolls and 11 different sums. Since 11 does not divide 36 it is impossible for each sum to have an equally likely chance of being rolled. Now consider rolling seven 6-sided dice. The possible sums are 7, 8, 9, . . . , 41, 42 where again the sums are not all equally likely. For example, to roll a 7 we would need to roll a 1 on all seven dice, which has a 1 67 chance of occuring. Whereas rolling an 8 would require that six of the dice result in a 1 but the last die results in a 2, which has a 1 66 chance of occuring. Can we relabel the dice so the sums 7, 8, . . . , 42 are equally likely? Consider the following relabeling: 1 Die 1: 1 1 2 2 3 3 Die 2: 1 1 4 4 7 7 Die 3: 1 1 1 10 10 10 Die 4: 1 1 1 19 19 19 Die 5: 1 1 1 1 1 1 Die 6: 1 1 1 1 1 1 Die 7: 1 1 1 1 1 1 If we roll the first two dice, the sums will be 2, 3, 4, 5, 6, 7, 8, 9, 10 with each combination having 4 ways of occuring because there are two choices on each die of which face to use, so there are 2 · 2 = 4 ways of getting each sum. If we then combine these with die 3 we get the sums 3, 4, . . . , 11 if we roll a 1, and we get the sums 12, 13, . . . , 20 if we roll a 10. Each of these sums has 3 · 2 · 2 = 12 ways of occuring. Further combining this with die 4 we get the sums 4, 5, . . . , 21 and 22, 23, . . . , 39 if we roll a 1 or 19 respectively. Each of these sums has 3 · 3 · 2 · 2 = 36 ways of occuring. Finally, when combined with dice 5, 6, and 7, each sum increases by exactly 3. Therefore, this set of dice yield the sums 7, 8, 9, . . . , 42 where each sum has 6 · 6 · 6 · 3 · 3 · 2 · 2 = 7776 ways of occuring. Since there are 67 possible outcomes and each has 65 = 7776 ways of being rolled, there is a probability of 65 67 = 1 62 = 1 36 for each sum to be rolled. This now generates the question of when such a relabeing is possible. More generally we can look at m dice that have n sides each. The question of can m n-sided dice be relabelled so that we still get the sums m, m + 1, . . . , mn with each sum being equally likely was answered in [1] when n = pq where p and q are distinct prime numbers with p < q. We have seen that for n = 6, m can equal 7 but not 2. We will build on the results shown in [1]. 2 When rolling m n-sided dice we call the sums m, m + 1, m + 2, m + 3, . . . , mn the standard sums. This is due to the fact that if all m die were labeled with the usual labels of 1, 2, . . . , (n − 1), n the possible sums would be m, m + 1, . . . , mn. Note that the number of standard sums is mn − m + 1 = m(n − 1) + 1. We say that m, n-sided dice have an equal relabeling if they can be relabeled so that the standard sums are equally likely. For many equal relabelings, such as the one in the example above, we will need dice that have a label of 1 on every face; we will refer to these as stupid dice. Before continuing the work done in [1], we summarize the main results in [1]. The principal result from the paper is the following: Lemma 1. A set of m pq-sided dice have an equal relabeling if and only if m= nx pγy − 1 n−1 (*) m= nx q γy − 1 n−1 (**) or where n = pq for distinct primes p and q with p < q, x and y are nonnegative integers such that (x, y) = (0, 1), and γ is the multiplicative order of p modulo n − 1. From Lemma 1, we see that the number of standard sums is m(n − 1) + 1 = nx pγy = px+γy q x or nx q γy = px q x+γy . We will use this result later, so we will record this as a lemma. Lemma 2. For m pq-sided dice the number of standard sums is px+γy q x when m= nx pγy −1 n−1 and px q x+γy when m = nx q γy −1 . n−1 Lastly, it is shown that with the exception of a few possibilities for x and y, all equal relabelings require stupid dice. In particular we have the following: 3 Lemma 3. So long as (x, y) = (0, 0), (0, 1) or (1, 0), stupid dice are necessary for any equal relabeling of m n = pq sided dice. This provides a great deal of information, but it does not address how many (or how few) stupid dice are required. We will consider this question. There is also an unjustified statement in [1], namely that an equal relabeling requires that the number 1 appears on a given die 1, p, q, or pq times, and we will look at this statement. 4 Chapter 2 A Partial Proof of a Previously Unjustified Statement As mentioned before, the proof of Lemma 3 given in [1] relies on the statement that: if m pq-sided dice have an equal relabeling, then the number of 1’s on any given die must divide pq. However, no justification for this statement is given. We attempt to justify it. Since we need to have our m dice sum to m each die must have at least one face labled as 1. We are going to look at two specific dice: one being a regular die that we can choose from our given m dice and the other having faces that equal the sums of the remaining dice. Die 1 will have pq sides. We need to prove that each number that appears on die 1 appears 1, p, q, or pq times. For the following proofs we will use the notation that ai and bi are the number of faces with value i on dice one and two respectively. Furthermore, let s and t be the largest values to appear on dice one and two respectively. Finally, we will denote k being on die 1 5 or die 2 as k ∈ D1 and k ∈ D2 respectively. Lemma 4. If the smallest number to appear on die 1 is 1, each other number that appears on die 1 appears less than or equal to the number of times that 1 appears on die 1, and s appears the same number of times as 1 does on die 1. Proof. Given our m dice we are required to find combinations that add up to the standard sums of m, m+1, m+2, . . ., mn. Assume to the contrary that there exists some die, among the original m dice, without a face of value 1. Then consider the sum of the smallest faces on each of the die. Since m − 1 of the dice have at least 1 as their smallest value and one of the dice has at least 2 as its smallest value, we have that the smallest sum we could get is 1 ∗ (m − 1) + 2 ∗ 1 = m − 1 + 2 = m + 1. This means that the smallest sum that we can get is m + 1 which is a contradiction due to the fact that we are required to get a summation of m. Therefore each of our dice must have at least one face of value 1. Note that since each die has a 1, then the smallest value that appears on die 2 is m − 1 since we have m − 1 dice remaining after selecting a die for die 1. To generate equally likely standard sums with our dice we must have that each sum appears the same number of times. By way of contradiction let k ∈ D1 with ak > a1 . Then we know that k + m − 1 appears at least ak ∗ bm−1 times since m − 1 appears bm−1 times on die 2. Next consider that the only values we can use that sum to m are 1 ∈ D1 and m − 1 ∈ D2 since these are the minimum values on each die. Therefore the number of times that the sum m appears is exactly a1 ∗ bm−1 . Since a1 ∗ bm−1 < ak ∗ bm−1 the sum k + m − 1 is more likely to occur than the sum m, which is a contradiction. Therefore we must have that for all k ∈ D1 , ak ≤ a1 . 6 Note that the same reasoning applied to die 2 implies that bj ≤ bm−1 for all j ∈ D2 . Finally, we will consider the sum s + t. If we use a value from die 1 that is less than s, say s − j, then the value we would need from die 2 is s + t − (s − j) = t + j. Since t is the max value on die 2 it must be that j ≤ 0 making our value for die 1 at least s. Similarly, if we used t − i for our value from die 2 we would need to use s + i from die 1 making i ≤ 0. Since s is our largest value on die 1 we have that i ≤ 0 making our value for die 2 at least t. Since our value for die 1 is at least s and our value for die 2 is at least t we have that they must be exactly s and t. Therefore the only values that sum to s + t are s on die 1 and t on die 2. Since all sums appear an equal number of times, we have that a1 ∗ b1 = as ∗ bt . Using our above facts we have that as ≤ a1 and bt ≤ b1 . If as < a1 or bt < b1 then we would have that as ∗ bt < a1 ∗ b1 which would contradict our previous statement. Therefore, as = a1 and bt = b1 . It is important to note that since die 1 was arbitrary we have that all of these results hold no matter which die is choosen to be die 1. For the rest of the paper we will need terminology to describe the cases ai = a1 or bj = bm−1 . If this happens, then we refer to i and j as being f ull. Likewise, if ai = 0 or bj = 0 then the value i or j is empty. Furthermore, if a sum, say d, has the same number of ways of being obtained as m then we will say d is maxed. These will be important due to the fact that if a value is full then the die which it is on can no longer have any more faces with said value. Similarly, if a combination is maxed then if there are any other two values that make that combination which have not yet been acounted for, one must be empty. Finally we will refer to values of our dice as mirroring if ai = as−(i−1) and bj = bt−(j−(m−1)) . 7 Lemma 5. Die 1 is mirrored from 1 to 1+k1 where 1+k1 is the smallest non-empty number on die 1 other than 1; furthermore, 1 + k1 ∈ D1 is full. We will also show that die 2 is mirrored from m − 1 to m − 1 + k1 with m − 1 through m − 2 + k1 being full while m − 1 + k1 is empty. Proof. Let 1 + k1 be the smallest non-empty number on die 1 other than 1. Then we know that to get the sums of m, m + 1, . . . , m − 1 + k1 we must use the value 1 ∈ D1 since (1 + k1 ) + (m − 1) is larger than any of these values. Then for any value we just listed, say x, we know that x − 1 ∈ D2 and x − 1 must be full since there are no other combinations that can result in x other than 1 + x − 1. Therefore the numbers m − 1, m, m + 1, . . . , m − 2 + k1 on die 2 are full. Since s + m − 1 is a maxed combination, due to s and m − 1 being full, we have that any number smaller than s on die 1 that sums to s + m − 1 which uses the numbers m, m + 1, m + 2, . . . , m − 2 + k1 as summands from die 2 is empty. This list of empty numbers on die 1 is s − 1, s − 2, . . . , s − k1 + 1. Similarly to the above s − k1 is the largest non-empty number on die 1 other than s. Therefore to get the sums of s + t − k1 + 1, s + t − k1 + 2, . . . , s + t − 1 we must use s ∈ D1 since s − k1 + t = s + t − k1 is too small. Then for any value we just listed, say y, we know that y − s ∈ D2 and y − s must be full since there are no other combinations that can result in y other than s + y − s. Therefore t−k1 +1, t−k1 +2, . . . , t−1 on die 2 are all full. Since 1+t is a maxed combination, due to 1 and t being full, we have that any number larger than 1 on die 1 that sums to t + 1 which uses the numbers t − k1 + 2, t − k1 + 3, . . . , t − 1 as the summands from die 2 is empty. This list of empty numbers on die 1 is 2, 3, . . . , k1 − 1, k1 . Since t − k1 + 1, t − k1 , . . . , t − 2, t − 1, t are full, we have the sums of t − k1 + 2, t − k1 + 3, . . . , t − 1, t, t + 1 being maxed due to the fact that we can add these 8 values to 1 ∈ D1 . Since we know that 1 + k1 ∈ D1 is non-empty but t + 1 is maxed, we know that t − k1 ∈ D2 is empty or else the sum t + 1 would occur too many times. Consider the sum of s + t − k1 . The labels t − k1 + 1, t − k1 + 2, . . . , t − 1, t are too large to sum with s to attain this, but any value on die 1 that is less than s − k1 would be too small to combine with t. Therefore we must use some value on die 1 that is greater than or equal to s − k1 but smaller than s. Since all of these values except for s − k1 are empty we must only use s − k1 for this sum. Thus, s − k1 is full or else s + t − k1 would not be maxed. Also notice that if we sum s − k1 ∈ D1 with m − 1 + k1 ∈ D2 we get s + m − 1 which we have already established is maxed, leaving us with the conclusion that m − 1 + k1 ∈ D2 is empty. This brings us to our final point. Consider the sum m + k1 . This can only be achieved through using 1 or 1 + k1 ∈ D1 since any other number would be too big or is empty. These reqiure the summands of m − 1 + k1 or m − 1 ∈ D2 respectively. Since m − 1 + k1 ∈ D2 is empty we must have that 1 + k1 ∈ D1 is full to make m + k1 maxed. In conclusion, die 1 contains at least 1, 1+k1 , s−k1 , and s which are all full while having 2, 3, . . . , k1 and s−k1 +1, s−k1 +2, . . . , s−1 be empty. Similarly we have that die 2 contains at least m−1, m, m+1, . . . , m−2+k1 and t−k1 +1, t−k1 +2, . . . t−1, t full while having m − 1 + k1 and t − k1 be empty. Using our definition, we have that die 1 mirrors from at least 1 to 1 + k1 while die 2 mirrors from at least m − 1 to m − 1 + k1 . Lemma 6. If there exists a non-empty element of a die that is also not full, then there is at least one die that doesn’t mirror. 9 Proof. Let e and g be the smallest elements on D1 and D2 respecitvely such that they are neither empty nor full. By way of contradiction assume that both dice mirror. We first show that e − 1 = g − (m − 1). Assume by way of contradiction that e − 1 < g − (m − 1) making e + (m − 1) < g + 1. Since e is not full we have that e + (m − 1) is not maxed from just e ∈ D1 and m − 1 ∈ D2 . Since m − 1 is the smallest number on D2 we have that there exists some x ∈ D1 such that 1 ≤ x < e and a y ∈ D2 such that m − 1 < y ≤ t where x + y = e + (m − 1). Due to the fact that e + (m − 1) < g + 1 we have that m − 1 < y < g. Since x < e and y < g we have that both are either full or empty. If one of them is empty then their sum does not contribute to maxing e + (m − 1) but if both are full then e + (m − 1) would occur too many times. Therefore e − 1 ≥ g − (m − 1). We can similarly show that e − 1 ≤ g − (m − 1) making e − 1 = g − (m − 1). Now let f and h be such that f = s − (e − 1) and h = t − (g − (m − 1)). Since the dice mirror ae = af and bg = bh . By the above equations we get g − (m − 1) = e − 1 = s − f which makes g + f − (m − 1) = s. Now if we consider the sum of s ∈ D1 and (m − 1) ∈ D2 , we can see that both of the elements are full making s + (m − 1) maxed. Using our above equation we get that s + (m − 1) = g + f − (m − 1) + (m − 1) = g + f, but since that sum is already maxed we must have that either g or f is empty. By assumption g and e are nonempty and af = ae , therefore f must be nonempty, which gives us a contradiction. Therefore, if there does exist a smallest such e ∈ D1 such that e is neither empty nor full, then D1 can’t mirror. Conjecture 1. Dice 1 and 2 are mirrored with each element being either full or empty. (not finished proving) 10 Proof. (partial ) We will induct on the number of gaps on D1 . Let lemma 5 be our base case. For our inductive case let the elements 1, 1 + k1 , 1 + k1 + k2 , . . . , 1 + k1 + . . . + kn ∈ D1 be full where each ki represents a gap between each face labling value, i.e. all values between each 1 + k1 + k2 + . . . + ki and 1 + k1 + k2 + . . . + ki + ki+1 are empty. Furthermore let 1 + k1 + . . . + kn + kn+1 ∈ D1 be non-empty and all other values on D1 between 1 + k1 + . . . + kn and 1 + k1 + . . . + kn + kn+1 be empty. Next, let all elements, m − 1 ≤ x ≤ m − 1 + kn on D2 , be full or empty. Finally, we let all elements of D1 between 1 and 1 + k1 + . . . , +kn and all elements of D2 between m − 1 and m − 1 + kn mirror. We then let m − 1 + kn < v ≤ m − 1 + kn+1 with v ∈ D2 such that j is the smallest such element where it is neither full nor empty. Since it is neither full nor empty the sum of 1+v is not maxed from just adding 1 ∈ D1 and v ∈ D2 . Therefore there exists some non-empty x ∈ D2 such that 1 + k1 + . . . + kj + x = 1 + v for some 1 ≤ j ≤ n. The reason that we have this restriction on j is due to the fact that if j = n + 1 then we would be using 1 + k1 + . . . , +kn + kn+1 ∈ D1 but the smallest sum we could attain from this would be (1 + k1 + . . . , +kn + kn+1 ) + (m − 1) = m + k1 + . . . , +kn + kn+1 but the max that 1 + v could be is m + kn+1 . If we use our formula of 1+k1 +. . . , +kj +x = 1+v we get x = v−k1 −. . .−kj < v. Since x < v we have that x is either full or empty from or hypothesis. If x is full then v is empty, otherwise 1 + v would be over our max. However, if x is empty then contradict our requirement that x is non-empty. Since we assumed that v was the smallest non-empty or full element between m − 1 + kn and m − 1 + kn+1 , we have that all elements on D2 between m − 1 and m − 1 + kn+1 are either full or empty. 11 So far we have shown that 1 appears on all m dice and appears the most often, each die mirrors from 1 to at least 1 + k1 where 1 + k1 is the smallest non-empty number on the die, and if there exists a non-empty element of a die that is also not full, then the die does not mirror. The end goal of these Lemmas was to show that each non-empty ai was equal. To complete the proof, we would need to show that the dice mirror completely. We have not yet been able to show this, so for now we state the following conjecture: Conjecture 2. If m pq-sided dice have an equal relabeling, then the number of 1’s on any given die must divide pq. We will assume Conjecture 2 is true in the next chapter. 12 Chapter 3 The Greatest and Least Number of Stupid Dice 3.1 Stupid Dice Facts By Lemma 3 we know that equal relabelings require stupid dice, but how many stupid dice are needed? Suppose m pq-sided dice have an equal relabeling. Then the number of standard sum is equal to either px+γy q x or px q x+γy by Lemma 2. The number of rolls that sum to each standard sums is calculated by dividing the total number of outcomes by the number of standard sums. In either case the total number of sums is nm = pm q m . In the first case the number of times each sum appears is pm q m px+γy q x sum appears is = pm−x−γy q m−x . In the second case, the number of times each pm q m px q x+γy = pm−x q m−x−γy . We will consider the case where we have px+γy q x standard sums, the other case is very similar. The least standard sum is m, and if we only label faces with positive integers, then in order to achieve a sum of m using m dice, every die must have at least one face labeled as 1. Let ci be 13 the number of times 1 appears on die i. Then each ci must equal 1, p, q, or pq from Conjecture 2. Since the sum m can only be achieved by using a 1 from each die, the number of times the sum m appears is c1 c2 . . . cm = pm−x−γy q m−x . For a die to be a stupid die ci = n = pq . To find the maximum and minimum numbers of stupid dice that are possible in an equal relabeling we need to find the least and greatest possible amount of ci ’s that equal pq. In all cases to get c1 , c2 , . . . , cm = pm−x−γy q m−x , m − x − γy of the ci ’s must have a factor of p, and m − x ci ’s must have a factor q. To find the largest amount of stupid dice, we want as much overlap as possible, so of the m − x ci ’s that have a factor of q, assume that m − x − γy of them also have p as a factor. Therefore we have have m − x − γy dice where ci = pq, m − x − (m − x − γy) = γy dice where ci = q, and m − (m − x − γy) − γy = x dice with ci = 1, making the making the maximum amount of stupid dice m − x − γy. To find our least number of stupid dice we want as little overlap as possible. Since m − x − γy ci ’s have p as a factor, there are m − (m − x − γy) = x + γy ci ’s that don’t have p as a factor. If these x + γy ci ’s have q’s as factors then there are still (m − x) − (x + γy) = m − 2x − γy factors of q remaining. Since each ci can only have at most 1 factor of q, the remaining factors of q must be factors of ci ’s which also have p as a factor. Therefore, there are at least m − 2x − γy dice whose ci = pq, making the minimum amount of stupid dice m − 2x − γy. Thus, the minimum number of stupid dice is m − 2x − γy and the maximum number of stupid dice is m − x − γy. 14 3.2 An Example Before tryinig to find a general relabeling, we consider a specific example. Let p = 3, q = 5, x = 4, and y = 2. To find γ, consider 31 ≡ 3 (mod 14) 32 ≡ −5 (mod 14) 33 ≡ −1 (mod 14) 34 ≡ −3 (mod 14) 35 ≡ 5 (mod 14) 36 ≡ 1 (mod 14), thus γ = 6. This makes n = 15 and m= 154 36·2 − 1 = 1, 921, 728, 616. 14 The largest amount of stupid dice we could use is m − x − γy = 1, 921, 728, 600 and the least amount is m − 2x − γy = 1, 921, 728, 596. We will show a labeling for 1, 921, 728, 599 stupid dice. Consider the following labeling: The first 14 dice are labeled as 15 Die 1: 1,2,3,4,5 Each value appears 3 times Die 2: 1,6,11 Each value appears 5 times Die 3: 1,16, 31 Each value appears 5 times Die 4: 1, 46, 91 Each value appears 5 times Die 5: 1, 136, 271 Each value appears 5 times Die 6: 1, 406, 811 Each value appears 5 times Die 7: 1, 1216, 2431 Each value appears 5 times Die 8: 1, 3646, 7291 Each value appears 5 times Die 9: 1, 10936, 21871 Each value appears 5 times Die 10: 1, 32806, 65611 Each value appears 5 times Die 11: 1, 98416, 196831 Each value appears 5 times Die 12: 1, 295246, 590491 Each value appears 5 times Die 13: 1, 885736, 1771471 Each value appears 5 times Die 14: 1, 2657206, 5314411 Each value appears 5 times Combining dice 1 and 2 we can see that the labelings from die 1 fill in the gaps of die 2 creating the sums 2, 3, . . . , 16 with each sum appearing 15 times. Consider the fact that the gaps between the faces on die 3 are each of size 15 and dice 1 and 2 combine for 15 consecutive sums. This means that when we combine die 3 with our previous combination the gaps will get filled in and the new sums will be 3, 4, . . . , 47. We can continue this process for all 14 dice and get that the sums 14, 15, 16, . . . , 7971628 occur 3 · 513 = 3662109375 times. The values on the next 3 dice appear exactly once each. The 15th die is labeled as 1, 7971616, 15943231, 23914846, 31886461, 39858076, 47829691, 55801306, 63772921, 71744536, 79716151, 87687766, 95659381, 103630996, 111602611, 16 the 16th die is labeled as 1, 119574226, 239148451, 358722676, 478296901, 597871126, 717445351, 837019576, 956593801, 1076168026, 1195742251, 1315316476, 1434890701, 1554464926, 1674039151, and the 17th die is labeled as 1, 1793613376, 3587226751, 5380840126, 7174453501, 8968066876, 10761680251, 12555293626, 14348907001, 16142520376, 17936133751, 19729747126, 21523360501, 23316973876, 25110587251. Similar to before, the gaps between the faces on die 15 are of size 7971616. When we combine this with our 7071615 consecutive sums from the first 14 dice we get the sums of 15, 16, 17, . . . , 119574239 where each sum still appears 3662109375 times. Repeating this with dice 16 and 17 makes the sums 17, 18, 19, . . . , 26904200641 where each sum appears 3662109375 times. The remaining 1, 921, 728, 599 dice are stupid dice. When combined with our other 17 dice we get the sums 1921728616, 1921728617, 1921728618, . . . , 28825929240 where each sum appears 3 · 514 · 151921728599 times. 3.3 An Equal Labeling We now show that for any m − 2x − γy ≤ λ ≤ m − x − γy, an equal albeling with λ stupid dice is possible. Let λ = m − 2x − γy + t where 0 ≤ t ≤ x. Label the first x − t dice as follows: 1, 1 + q i , 1 + 2q i , . . . , 1 + (q − 1)q i for 0 ≤ i ≤ x − t − 1 where each label is repeated p times. To find the sums this group of dice produces, first subtract 1 from each of the labelings. The new labeling would be 17 0, q i , 2q i , . . . , (q − 1)q i for 0 ≤ i ≤ x − t − 1. These are exactly the numbers we need to write the numbers 0, 1, . . . , q x−t − 1 in base q. Since each number can be written uniquely in base q, each of the sums 0, 1, 2, . . . , q x−t − 1 occur the same number of times. Note that this is a set of q x−t consecutive numbers. Adding the 1’s back to the labeling adds x − t to each sum yeilding the equally likely sums x − t, x − t + 1, . . . , x − t + q x−t − 1. Since there are p choices of which face to use on each die, each of these sums occurs px−t times. Label the second x − t dice as follows: 1, 1 + pj q x−t , 1 + 2pj q x−t , . . . , 1 + (p − 1)pj q x−t for 0 ≤ j ≤ x − t − 1 where each label is repeated q times. To find the sums this group of dice produces, first subtract 1 from each of the labels and divide each label by q x−t . This new labeling would be 0, pi , 2pi , . . . , (p − 1)pi for 0 ≤ i ≤ x − t − 1. These are exactly the numbers we need to write 0, 1, . . . , px−t − 1 in base p. Since each number can be written uniquely in base p, then each of their sums, 0, 1, 2, . . . , px−t − 1, occur the same number of times. Note that this is a set of px−t consecutive num bers. After multiplying each label by q x−t , we get the sums 0, q x−t , . . . , q x−t (px−t −1). Then adding 1’s to each label, we get the sums x − t, x − t + q x−t , . . . , x − t + nx−t − q x−t which is px−t numbers each separated by a gap of q x−t . Thus, when we combine the the q x−t consecutive sums from the first x − t dice with the sums from the second x − t dice, the gaps from our second sums will be filled in giving us the nx−t 18 consecutive sums 2(x − t), 2(x − t) + 1, . . . , 2(x − t) + nx−t − 1. Since there are q choices on each die of which face to use, these sums occur px−t q x−t = nx−t times. Label the next γy dice as follows: 1, 1 + pk nx−t , 1 + 2pk nx−t , . . . , 1 + (p − 1)pk nx−t for 0 ≤ k ≤ γy − 1 where each label is repeated q times. To find the sums this group of dice produces, first subtract 1 from each of the labels and divide each label by nx−t . This new labeling would be 0, pi , 2pi , . . . , (p − 1)pi for 0 ≤ i ≤ γy − 1. These are exactly the numbers we need to write the numbers 0, 1, . . . , pγy − 1 in base p. Since each number can be written uniquely in base p, then each of their sums, 0, 1, 2, . . . , pγy − 1, are attained the same number of times. Note that this is a set of pγy consecutive numbers. After multiplying each label by nx−t , we get the sums 0, nx−t , 2nx−t , . . . , nx−t (pγy − 1). Then adding 1’s to each label, we get the sums γy, γy + nx−t , . . . , γy + pγy nx−t − nx−t , which is pγy numbers each separated by a gap of nx−t . Thus, when we combine the nx−t consecutive sums from the first 2(x − t) dice with the sums from the next γy dice, the gaps will be filled in giving us the pγy nx−t consecutive sums 2(x − t) + γy, 2(x − t) + γy + 1, . . . , 2(x − t) + γy + pγy nx−t − 1. Since there are q choices of which face to use on each die, the sums from this group occur q γy times so the above sums occur nx−t q γy times. Label the last t dice as follows: 1, 1 + nδ pγy nx−t , 1 + 2nδ pγy nx−t , . . . , 1 + (n − 1)nδ pγy nx−t for 0 ≤ δ ≤ t − 1 19 where each label appears on exactly one face. To find the sums this group of dice produces, first subtract 1 from each of the labelings and divide each labeling by pγy nx−t . The new labelings would be 0, nδ , 2nδ , . . . , (n − 1)nδ for 0 ≤ δ ≤ t − 1. These are exactly the numbers we need to write the numbers 0, 1, 2, . . . , nt − 1 in base n. Since each number can be written uniquely in base n, then each of the sums 0, 1, 2, . . . , nt − 1 occur the exactly once. Note that this is a set of nt consecutive numbers. After multiplying each label by pγy nx−t we get the sums 0, pγy nx−t , 2pγy nx−t , . . . , pγy nx−t (nt − 1). Then adding the 1’s to each label, we get the sums t, t + pγy nx−t , . . . , t + pγy nx − pγy nx−t , which is nt numbers each separated by gaps of pγy nx−t . Thus, when we combine the pγy nx−t consecutive sums from the first 2(x − t) + γy dice with the sums from the next t dice, the gaps will be filled in giving us the pγy nx consecutive sums 2x − t + γy, 2x − t + γy + 1, . . . , 2x − t + γy + pγy nx − 1. Each of these sums appears nx−t q γy times. Finally, label the remaining m − 2x − γy + t dice with 1’s on every face. This yields the sum m − 2x − γy + t. Since there are n choice of which face to use on each of the stupid dice, the sums m, m + 1, . . . , m + pγy nx − 1 all appear nx−t q γy nm−2x−γy+t = nm−x−γy q γy = pm−x−γy q m−x times. We need to show that this labeling results in the standard sums. It is clear that the lowest sum we can get is m and that each successive sum is increased by 1. This means all we need to show is that the largest sum, m + pγy nx − 1, is in fact equal 20 to mn. Since we know that mn − m + 1 = nx pγy we have that mn = m + nx pγy − 1 making the sums created by our dice the standard sums. We will now show that the standard sums are equally likely by showing that there is only one way to achieve each sum. Let δ be one of the standard sums minus m − 2x − γy + t (i.e. without the stupid dices’ contribution). From our above sums we know that the first x − t dice always result in a sum of the form x − t + r where 0 ≤ r ≤ q x−t − 1, the next x − t result in a sum of the form x − t + sq x where 0 ≤ s ≤ px−t − 1, the next γy dice result in a sum of the form γy + wnx−t where 0 ≤ w ≤ pγy − 1, and the final t result in a sum of the form t + vpγy nx−t where 0 ≤ v ≤ nt − 1. So δ can be written as δ = (x − t + r) + (x − t + sq x−t ) + (γy + wnx−t ) + (t + vpγy nx−t ). To show that this sum is unique assume there exists two sums that yield δ, i.e. (x − t + r1 ) + (x − t + s1 q x−t ) + (γy + w1 nx−t ) + (t + v1 pγy nx−t ) = (x − t + r2 ) + (x − t + s2 q x−t ) + (γy + w2 nx−t ) + (t + v2 pγy nx−t ). By subtracting constants that appear on both sides, we obtain r1 + s1 q x−t + w1 nx−t + v1 pγy nx−t = r2 + s2 q x−t + w2 nx−t + v2 pγy nx−t . Then we can consider these sums mod q x−t . Since q x−t |nx−t we get that r1 ≡ r2 (mod q x−t ). Then, since 0 ≤ r1 , r2 ≤ q x−t − 1 we have that r1 = r2 . Next we consider our sums mod nx−t . Since p = 1 we have that nx−t |q x−t making r1 + s1 q x−t ≡ r2 + s2 q x−t (mod nx−t ). Using r1 = r2 we get that s1 q x−t ≡ s2 q x−t (mod nx−t ). Finally, since 0 ≤ s1 , s2 ≤ px−t we have that 0 ≤ s1 q x−t , s2 q x−t < nx−t − 1 making s1 q x−t = s2 q x−t which shows s1 = s2 . 21 We will now consider our sums mod pγy nx−t . Since p = 1 we have that pγy nx−t |nx−t making r1 + s1 q x−t + w1 nx−t ≡ r2 + s2 q x−t + w2 nx−t (mod pγy nx−t ). Using r1 = r2 and s1 = s2 we get that w1 nx−t ≡ w2 nx−t (mod pγy nx−t ). Finally, since 0 ≤ w1 , w2 ≤ pγy − 1 we have that 0 ≤ w1 nx−t , w2 nx−t < pγ y nx−t making 21 nx−t = w2 nx−t resulting in w1 = w2 . Looking back at our original equation r1 + s1 q x−t + w1 nx−t + v1 pγy nx−t = r2 + s2 q x−t + w2 nx−t + v2 pγy nx−t and using the fact that the r’s, s’s, and w’s are equal we get that v1 pγy nx−t = v2 pγy nx−t making v1 = v2 . Therefore δ is a unique sum from these groups of dice. Thus the minimum number of stupid dice is m − 2x − γy, the maximum number of stupid dice is m − x − γy, and it is possible to find an equal relabeling that uses any number of stupid dice between m − 2x − γy and m − x − γy. 22 Chapter 4 Further Questions To continue this work we have left a few open ends that are worth investigating. Can Conjecture 1 be proven? How many relabelings are there for a given m? The labeling in 3.3 can be modified if we let our exponents be pm−x q m−x−γy instead of pm−x−γy q m−x , is it a substantial difference? Are there other substantially different ways to make general relabelings? What about the case when n is not a product of 2 distinct primes? 23 Bibliography [1] F. Bermudez, A. Medina, A. Rosin, E. Scott. Are Stupid Dice Necessary?, The College Mathematics Journal, Vol. 44, No. 4, September 2013. [2] J. Gallian, Contemporary Abstract Algebra, Houghton Mifflin Company, 2008. [3] W. Gasarch, C. Kruskal, When Can One Load a Set of Dice so That the Sum Is Uniformly Distributed?, Mathematics Magazine 72, 1999. 24
© Copyright 2026 Paperzz