Common Errors in Counting Problems by Scott Annin and Kevin Lai Mathematics teachers are often asked: What is the most difficult topic for you to teach? Our answer is easy: it's teaching students to count. Although the computations in combinatorics are not necessarily difficult, the concepts can be challenging. Many times, no rigid procedures or formulas can be directly employed to solve the problems, and students simply do not know where and how to begin approaching these problems. Furthermore, subtle differences in questions posed can necessitate entirely different solution techniques. Combinatorics is an important topic in the mathematics curriculum. For example, Kapur (1970) observed that it can help students develop their reasoning skills (making conjectures, generalizing, systematic thinking) and conceptual understanding of mapping, functions, equivalence relations, etc. Combinatorics also has applications in many fields such as physics and biology, and Kapur recommended that it should be taught at all grades. The same sentiment is expressed by Principles and Standards for School Mathematics (NCTM 2000), which observes that although the topic of Discrete Mathematics was removed from the 2000 standards, the topic is actually evident across all the other standards and spans from grades K-12. It recognizes the wide usage of this area of mathematics in business and industry, and encourages students to develop their understanding of combinatorics throughout their school years. A common strategy among Discrete Mathematics textbooks is to attempt to classify counting problems neatly according to various types in the hopes of giving students some clear guidelines and procedures (Brualdi, 2004; Johnsonbaugh, 2009; Rosen, 2007). This way of classification is also reflected in the research literature (Fischbein & Gazit 1988; Batanero et al. 1992). In counting arrangements of objects, for example, one can distinguish problems according to whether repetition of objects is allowed and whether the order that those objects are placed in makes a difference to the arrangement. This results in four general types of counting problems. Provided that students are studying problems that rightly belong in one and only one of these four categories, there does not appear to be anything extraordinarily unusual about what the student must do to successfully learn the material. Figure 1 shows the number of arrangements of n objects in each of the four traditional cases. Repetition Allowed r-permutations of n objects r-combinations of n objects nr Repetition Not Allowed n! n r ! (n r 1)! n! r ! n 1! r ! n r ! Fig. 1: Common table seen in Discrete Mathematics textbooks What we often find challenging about teaching students to count, however, is that most problems do not fall cleanly into one and only one of the standard categories of counting problems. Rather, each problem typically involves different aspects that rely on different techniques. Moreover, subtle differences from one problem to the next lead to a propensity for hesitation and uncertainty in the student's thinking. In this article, we will present a few examples of such problems, illustrate their subtleties, and provide some guidance and practice with this subject. Let us begin with a few problems: Problem 1: How many flags of eight horizontal stripes contain at least six blue stripes, if each stripe is colored red, green, or blue? Problem 2: How many flags of eight horizontal stripes contain at least two blue stripes, if each stripe is colored red, green, or blue? Fig. 2: Examples for Problems 1 and 2 Problem 3: How many 4-card hands consisting of two cards each of two different kinds can be made from a standard deck of 52 cards? Problem 4: How many full house* card hands can be made from a standard deck of 52 cards? Fig. 3: Examples for Problems 3 and 4 *A full house consists of precisely three cards of one kind and two cards of a second kind from a standard deck. All four of these problems involve aspects that must be handled carefully. Many students do not see these delicate points, nor do they see any substantive differences in the way these four problems are best handled. This paper is intended to draw attention to some of the errors in thinking that students fall victim to in solving counting problems such as these, examine the differences between the correct solutions to these problems and the solutions often given by misdirected students, and provide some problems to encourage practice in this challenging area. Over-Counting due to Order In my many years of teaching combinatorics and asking exam questions like Problem 1, I (S. Annin) have often found that students arrive at the following incorrect solution. Incorrect Solution to Problem 1: Consider the flag shown below. Each stripe must be labeled with a color: red, green, or blue. Since at least six blue stripes are required, one can begin by choosing six stripes to color blue. There are C(8,6) = 28 ways to do this. Once the six blue stripe minimum is fulfilled, the remaining two stripes (whichever ones they are) can each be colored in any one of three ways (red, green, or blue). Therefore, there are 3*3 = 9 ways to finish coloring the stripes, for a total of 28*9 = 252 ways to color the stripes of the flag. Fig. 4: Flag Template for Problem 1 Analysis: The solution above appears to be a straightforward application of the Multiplication Principle: to count the total number of ways of doing a task, one can multiply the number of ways of performing the subtasks that comprise the whole task (hence the answer 28*9). However, this answer is nearly two times too large. The main reason for this is that the solution presented above imposes an ordering on the blue stripes (students are unaware of this), when the order in which the stripes are painted is immaterial: the question is only interested in counting finished flags. In particular, the solution above first chooses six blue stripes, and then (possibly) chooses additional blue stripes. Therefore, the solution has ordered the blue stripes into two “groups”: the first six blue stripes and any additional blue stripes after that. To see the difficulty more explicitly, consider, for example, the flag colored with all eight stripes blue. There is obviously only one such flag, yet the solution above counts it 28 times in the 252 total obtained! How is that possible? To see how, consider two different ways to complete the coloring of the flag: (1) choose the top six stripes to be blue, then choose to color both of the last two stripes blue also, and (2) choose the bottom six stripes to be blue, then choose to color both of the first two stripes blue also. In either case, we arrive at the same “all blue” flag, but from the point of view of the solution above, these flag-colorings were obtained differently and are therefore counted as two different flags. In fact, there would be 28 different flags if we kept track of which six of eight stripes were chosen first. For instance, suppose we place a gold star next to each stripe that is chosen among the first six for blue color. For an “all blue” flag, there are C(8,6) = 28 different patterns of gold stars that could occur on the stripes. The solution above incorrectly counts each of these 28 outcomes as a different flag. Correct Solution to Problem 1: While it is possible to do subtraction to correct for the overcounting that has occurred above, it is easier to simply take care to avoid over-counting in the first place. One might observe, for example, that the trouble described in the analysis above does not arise on flags whose last two stripes are colored red or green. The number of such flags (flags with exactly six blue stripes) is precisely C(8,6)*2*2 = 28*4 = 112. Next, the number of flags that have exactly seven blue stripes can be counted. There are C(8,7) ways to choose seven stripes to color blue, and the last stripe must be either red or green. Hence, there are C(8,7)*2 = 8*2 = 16 such flags. Finally, there is one flag with all eight stripes blue. Using the Sum Rule, the total number of flags with at least six blue stripes is 112 + 16 + 1 = 129. 8 8 2 2 2 1 6 7 Exactly 6 blues Exactly 7 blues Exactly 8 blues Fig. 5: Correct Solution to Problem 1 Conclusions: Many counting problems involve a “minimum” requirement (e.g. flags with at least six blue stripes, license plates with at least six A’s, bags of fruit containing at least six apples, etc.) Problems such as these are vulnerable to over-counting in the manner observed above, but this can be remedied by replacing the minimum condition with several cases containing an “exact” condition and summing the total count for each of these cases. Instead of counting the number of flags with at least six blue stripes in one step, for example, we can count the number of flags with exactly six stripes, exactly seven stripes, and exactly eight stripes separately. Each of these three subcases are less prone to error, and as we have done above, the results from each subcase can be added to yield the final result. Correct Solution to Problem 2: This problem is closely related to Problem 1 and can be solved the same way. Since the number of blue stripes can now be any value 2,3,4,5,6,7, or 8, we have seven different cases to take care of. We ask: how many flags of 8 stripes containing exactly k blue stripes? The answer is C(8,k)*28-k, obtained by first choosing k stripes to color blue, and then deciding on either red or green color (two choices each) for each of the remaining 8-k stripes. We must sum this value for all k=2,3,4,5,6,7,8. The total is C(8,2)*64 + C(8,3)*32 + C(8,4)*16 + C(8,5)*8 + C(8,6)*4 + C(8,7)*2 + C(8,8)*1 = 1,792 + 1,792 + 1,120 + 448 + 112 + 16 + 1 = 5,281. Although the final answer requires a lot of terms by this solution, students become very comfortable with it, especially after seeing enough examples similar to Problems 1 and 2. Of course, when the “minimum” requirement on the problem is rephrased as the complement of a “maximum” requirement, a shorter answer can be found: How many flags with eight horizontal stripes contain at most one blue stripe? Since there are 28 flags with no blue stripes and C(8,1)*27 flags with exactly one stripe, there are a total of 28 + C(8,1)*27 = 256 + 1,024 = 1,280 flags that do not contain at least two blue stripes. Since there are a total of 38 = 6,561 flags with eight horizontal stripes (each independently colored either red, green, or blue), there are precisely 6,561 – 1,280 = 5,281 that do contain at least two blue stripes. Order Issues in Another Context Perhaps no subject matter is more common in combinatorics (and probability) problems than playing cards. Despite the heavy analysis that playing cards have received, they can elicit some tricky aspects that are easily overlooked. The pitfall we are about to describe arises in the context of many other counting problems as well, and as with many difficulties encountered in basic Combinatorics, ordering of objects is at the core of the issue. In particular, there is a basic difference in how Problems 3 and 4 (stated above) are solved, and it has everything to do with order. Solution to Problem 3: The tasks required to construct a 4-card hand consisting of two pairs can be described as follows: (1) choose two kinds of cards (i.e. ace, king, queen, …, 3, 2) to build pairs from, and (2) choose two suits for the two cards from each of the two kinds chosen in the first step. There are C(13,2) ways to complete task (1), and there are C(4,2)*C(4,2) ways to complete task (2). This gives us a total of C(13,2)*C(4,2)*C(4,2) = 78*6*6 = 2,808 such hands. 13 4 4 2808 2 2 2 Ways to choose 2 kinds Ways to choose 2 cards from one kind Ways to choose 2 cards from another kind Fig. 6: Correct Solution to Problem 3 Analysis: By simultaneously choosing the two kinds of cards that the pairs will be built from, we avoid the common mistake of putting the two kinds of cards in a specified order. Faulty reasoning here would suggest that there are 13 ways to choose the first kind, C(4,2) ways to choose two of the four suits of that kind, 12 ways to choose the second kind, and C(4,2) ways to choose two of the four suits of the second kind. Multiplication of the various choices available here would give the answer 5,616, twice as large as it should be. This faulty reasoning, in which the two kinds are placed in order (we have emphasized this point by referring to them as “first kind” and “second kind” above) when they should not be, is extremely common. Students often successfully solve Problem 3 merely because everyone’s experience with playing cards is that the order of the cards does not matter. That is, cards can be played from a hand in any desired order, so the “left-to-right” order that the cards appear in the hand makes no difference. Therefore, most people are trained to use combinations (rather than permutations) to solve card problems. A natural corollary of this, of course, is that in card problems where order is in some way significant, students often falter. Only a minor change to Problem 3, for example, leads to the more troublesome Problem 4. Solution to Problem 4: This problem appears to be almost identical to Problem 3. Again, we must (1) choose two kinds of cards, and (2) choose an appropriate number of cards of each kind. In fact, if one were to try a solution technique similar to that given for Problem 3 above, the expected answer might be C(13,2)*C(4,3)*C(4,2). But this answer is incorrect, because it does not distinguish in any way between the two chosen kinds. In reality, however, one of the two chosen kinds of cards must have three suits selected from it, while the other should have only two. The correct answer in this case would therefore be: 13 4 4 2 3744 2 3 2 Ways to choose 2 kinds Ways to choose the kind from which 3 cards are taken Ways to choose 3 cards from first kind Ways to choose 2 cards from second kind Fig. 7: Correct Solution to Problem 4 Alternatively, we could solve it as: 4 4 13 12 3744 3 2 Ways to choose a kind for 3 cards Ways to choose 3 cards of that kind Ways to choose another kind for 2 cards Fig. 8: Alternative Solution to Problem 4 Ways to choose 2 cards of that kind Conclusions: In Problem 3, the order of the two kinds used to create two pairs is unimportant, while in Problem 4, we must differentiate between the kind that three cards are chosen from and the kind that two cards are chosen from. This distinction appears to be the result of the fact that the same number of cards of each kind are chosen in Problem 3, whereas this is not the case in Problem 4. Here are some related problems that also demonstrate this subtle distinction in the problems. Try to solve the two exercises here before the reading the solutions that follow. Exercise 1: Alice, Bob, Candice, Douglas, Ellen, Freddie, George, Henrietta, and Irvin are instructed to break into one group of size 2, one group of size 3, and one group of size 4. In how many ways can they do this? Exercise 2: Alice, Bob, Candice, Douglas, Ellen, and Freddie, George, Henrietta, and Irvin are instructed to break into three groups of three. In how many ways can they do this? To solve Exercise 1, there are C(9,2) ways to choose two people for the group of size 2. Next, there remain 7 people, and hence C(7,3) ways to choose three people for the group of size 3. Finally, the last four people form a group of 4 by default. The total is C(9,2)*C(7,3) = 36*35 = 1,260. Exercise 2 is rather challenging for students if attempted immediately after solving Exercise 1. In solving Exercise 2, many students will follow the same reasoning exactly to obtain the answer C(9,3)*C(6,3) = 84*20 = 1,680. This answer however, is six times too large! To see why, let us consider a smaller example first: In how many ways can Alice, Bob, Candice, and Douglas break into two groups of two? One might be tempted to give the answer C(4,2) = 6, the number of ways to choose two of the four individuals to form a group (with the other two then being paired up by default thereafter). However, we can directly count only 3, not 6, pairs of two: Alice and Bob in a pair, Alice and Candice in a pair, and Alice and Douglas in a pair. The answer 6 would only be correct if we wanted to place the two groups chosen in order. Perhaps, for example, the first two individuals chosen to form a group are also designated as officers for the whole group. If we wanted to keep track of this additional information, then there would be 6 possible outcomes. Returning to Exercise 2, the answer 1,680 obtained above involves unintentionally putting the three groups in order. This was acceptable for Exercise 1, but not for Exercise 2. For example, the group formation ABC | DEF | GHI is being counted differently from the group formation GHI | ABC | DEF (note that the order of the three letters comprising a group is immaterial here). There are 3! = 6 ways to permute the same three groups of three (see Figure 9), and thus, each way of grouping nine individuals into three groups of three is being counted six times. Therefore, the correct answer here is 1,680/6 = 280. (1) ABC DEF GHI (4) DEF GHI ABC (2) ABC GHI DEF (5) GHI ABC DEF (3) DEF ABC GHI (6) GHI DEF ABC Fig. 9: The same three groups of three occur six times when an order on groups is imposed One of the most important questions to resolve in solving counting problems is: Does the order of the objects occurring in an arrangement matter? As we have seen in this section, however, this question may not be as black and white as it appears at first. Great care must be exercised. Conclusions In this article, we used examples in a variety of contexts to show how minor differences in combinatorics problems can result in major misconceptions in a novice solver. While the standard categorization of these problems into different types can ease this difficulty to some extent, many problems do not belong to one and only one category. A variety of problems need to be presented to students so that they can gain a better conceptual understanding of the problems and become familiar with the common pitfalls that must be avoided. Teachers can readily and specifically address the issues we have raised in this article by giving students any number of problems like the ones we have discussed here. In so doing, they will equip and train their students to be mindful of the potential errors that frequently arise in solving these problems. To that end, we conclude by providing the reader with an even greater variety of examples, similar to those we have discussed, that may be useful in the combinatorics classroom: Homework #1: How many license plates consisting of 9 capital letters contain at least six M’s? Homework #2: How many 6-card poker hands can be formed that contain three cards from each of two different suits? How many 6-card poker hands can be formed that contain four cards from one suit and two cards from another? Homework #3: (Modification of 2002 American Invitational Math Exam-Version 2, Problem 9). Let S = {1,2,3,4,5,6,7,8,9,10}. In how many ways can two disjoint, non-empty subsets of S be formed? References Batanero, C., Godino, J. D., Navarro-Pelayo, V. “Effect of the implicit combinatorial model on combinatorial reasoning in secondary school pupils.” Educational Studies in Mathematics 32 (1997): 181-199. Brualdi, R.A. Introductory Combinatorics, Prentice Hall, Upper Saddle River, NJ, 2004. Fishbein, E. and Gazit, A. “The combinatorial solving capacity in children and adolescents.” Zentralblatt fuer Didaktitk der Mathematik 5 (1988), 193-198. Kapur, J. N. “Combinatorial Analysis and School Mathematics,” Educational Studies in Mathematics 3 (1970): 111-127. Johnsonbaugh, R.J. Discrete Mathematics, Prentice Hall, Upper Saddle River, NJ, 2009. National Council of Teachers of Mathematics (NCTM). Principles and Standards for School Mathematics. Reston, VA: NCTM, 2000. Rosen, K.H. Discrete Mathematics and Its Applications, McGraw-Hill, New York, NY, 2007.