An application of integer programming to playoff elimination in football championships

WHITE PAPER

Submitted by Shilpa Subramanian (Roll No 146) Shruthi Treasa Jose (Roll No 150) Subin Mathew Mammen (Roll No 161)

THE CASE and Our Understanding of it Brazil is the Mecca of Football. Football means big business and big bucks. This paper analyses how one can use integer programming to determine which teams will enter the play offs and which teams will get eliminated from the tournament even before all the matches are completed. Teams which are not qualified to the playoffs lose a lot of money and are even forced to dismantle their structure. We present two integer programming models which are able to detect in advance when a team is already qualified to or eliminated from the playoffs. Results from these models can be used not only to guide teams and fans, but are also very useful to identify and correct wrong statements made by the press and team administrators. This tournament is organized in two stages: the first one is the qualification stage, while the second one is the playoff stage. During the qualification stage, each team plays every other in a single game. A team obtains three points for a win and one point for a tie. At the end of the qualification stage, the teams are ranked by the number of points they obtained along the tournament and the first eight teams in the standing table are qualified to the playoffs. Also, the last four teams are moved to a lower category and do not play in the coming year. In 2002, this championship as played between August and December by the 26 major Brazilian teams. The 325 games played at the qualification stage were organized in 29 rounds of 13 or fewer games. Not all teams played at every round. What we learnt from this paper? How integer programming can form an important predicting tool That complex situations can be easily handled with mathematical modelling

Problem Solving The first problem we handle in this work is the Guaranteed Qualification Problem (GQP). It consists in calculating the minimum number of point any team has to win to be sure it will be qualified, regardless of any other results. This Guaranteed Qualification Score (GQS) depends on the current number of points of every team in the standing table and on the still missing games to be played. The second problem we tackle is the Possible Qualification Problem (PQP). It consists in computing how many points each team has to win to have any chance to be qualified. This Possible Qualification Score (P QS) also depends on the current

number of points of every team in the standing table and on the still missing games to be played. Of course, PQS GQS for any team at any time of the qualification stage. The value of GQS for any team cannot increase along the competition, while that of P QS cannot decrease. Winning a minimum of GQS (resp. P QS) points is a sufficient (respectively necessary) condition for qualification. A team is mathematically qualified to the playoffs if and only if its number of points won is greater than or equal to its GQS. Only at this point its qualification to the playoffs can be announced without any risk of misinformation. Analogously, a team can be said to be mathematically eliminated from the playoffs when the total number of points it still has to play (i.e., the number of remaining games multiplied by three) plus the number of points it already won (M N P ) is less than its P QS. PROBLEM FORMULATION Let n be the number of teams in the championship and m the number of teams which qualify to the playoffs. A team gets 3 points for winning a game and 1 point for a tie. We denote by pi the total number of points accumulated by each team i =1, . . . , n at any particular moment of the championship, with pi = 0 at the beginning. For any pair (i, j) of teams, with i, j = 1, . . . , n and i = j, let gij = gji be the number of remaining games still to be played between teams i and j. Since the rules of the Brazilian national football championship establish that each team plays against every other team exactly once, then ei- ther gij = 0 in case the game between teams i and j was already played, or gij = 1 otherwise. At any time, a valid assignment is a set of triples A(i, j) = (p1 (i, j), p2 (i, j), p3 (i, j)) of non-negative integers for each pair (i, j) of different teams, such that p1 (i, j) + p2 (i, j) + p3 (i, j) = gij , p1 (i, j) = p3 (j, i), and p2 (i, j) = p2 (j, i), where p1 (i, j), p2 (i, j), and p3 (i, j) represent respectively a possible number of victories of team i over team j, a possible number of games between team i and j which end up with a tie, and a possible number of victories of team j over team i along the remaining gij games. Notations used in the paper: Aij Feasible solution of games between teams I and J

p1 - Number of wins of team I over team J

p2- Number of draws p3 Number of wins of team J over team I.

Total points accumulated by a team at the end of a championship Ti = pi+ 3. p1 (i ,j)+ 1.p2 ( i, j)

INTEGER PROGRAMMING MODEL At any point of time during the tournament

PQS <= GQS A team is qualified for the playoffs if its

score> GQS A team is eliminated if

3 x No. of games left to play < PQS

Some Assumptions: GQS^k is the maximum points a team can accumulate and still not be qualified. Therefore, for qualification, the minimum points to be scored: GQS^k+1, in order to be in the top 8 out of m teams.

Defining the Variables: Xij = 1, if team i wins over j 0, otherwise Yj = 1 , if team j is ahead of I

0, otherwise Approach to Objective function: Max GQS^k Min PQS^k

Constraints 1. Determine that only one team can win a game. 2. Establish the total number of points obtained by each team at the end of the championship. 3. determines the maximum difference in points between any pair of teams 4. enforces that team k is not qualified among the first m = 8 teams.

Team k is qualified for the playoffs when GQP(k) turns out to be infeasible. Let M be an upper bound to the maximum difference between the number of points obtained by any pair of teams. Since there are 26 teams in the championship and each of them plays exactly once against every other, M3 25 = 75 is a valid upper bound to |tj tk| for any pair (j, k) of teams, with j, k = 1, . . . , n and j 6= k.

Constraints (3) plays the same role as (3) in the formulation of GQP(k). It enforces that if tj > tk, then zj = 1 (i.e., team j is ahead team k in the standing table Constraint (4) states that there are at most m1 = 7 teams ahead of k in the standing table. The infeasibility of PQP(k) means that this team is mathematically eliminated, i.e., it cannot be qualified anymore to the playoffs. CONCLUSION: This paper helps to determine the winning teams even before the championship come to a close. It is a helpful tool for teams to formulate their strategies, for betting agents and for the fans.