04.09.2013 Views

Algorithm Design

Algorithm Design

Algorithm Design

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

400<br />

Chapter 7 Network Flow<br />

7.12 Baseball Elimination<br />

Over on the radio side the producer’s saying, "See that thing in the<br />

paper last week about Einstein?... Some reporter asked him to figure<br />

out the mathematics of the pennant race. You know, one team wins so<br />

many of their remaining games, the other teams win this number or<br />

that number. What are the myriad possibilities? Who’s got the edge?"<br />

"The hell does he know?"<br />

"Apparently not much. He picked the Dodgers to eliminate the<br />

Giants last Friday.’"<br />

--Don DeLillo, Underworld<br />

/..~ The Problem<br />

Suppose you’re a reporter for the <strong>Algorithm</strong>ic Sporting News, and the following<br />

situation arises late one September. There are four basebal! teams trying to<br />

finish in first place in the American League Eastern Division; let’s call them<br />

New York, Baltimore, Toronto, and Boston. Currently, each team has the<br />

following number of wins:<br />

New York: 92, Baltimore: 91, Toronto: 91, Boston: 90.<br />

There are five games left in the season: These consist of all possible pairings<br />

of the four teams above, except for New York and Boston.<br />

The question is: Can Boston finish with at least as many wins as every<br />

other team in the division (that is, finish in first place, possibly in a fie)?<br />

If you think about it, you realize that the answer is no. One argument is<br />

the following. Clearly, Boston must win both its remaining games and New<br />

York must lose both its remaining games. But this means that Baltimore and<br />

Toronto will both beat New York; so then the winner of the Baltimore-Toronto<br />

game will end up with the most wins.<br />

Here’s an argument that avoids this kind of cases analysis. Boston can<br />

finish with at most 92 wins. Cumulatively, the other three teams have 274<br />

wins currently, and their three games against each other will produce exactly<br />

three more wins, for a final total of 277. But 277 wins over three teams means<br />

that one of them must have ended up with more than 92 wins.<br />

So now you might start wondering: (i) Is there an efficient algorithm<br />

to determine whether a team has been eliminated from first place? And (ii)<br />

whenever a team has been eliminated from first place, is there an "averaging"<br />

argument like this that proves it?<br />

In more concrete notation, suppose we have a set S of teams, and for each<br />

x ~ S, its current number of wins is wx. Also, for two teams x, y ~ S, they still<br />

7.12 Baseball Elimination<br />

have to play gx~, games against one another. Finally, we are given a specific<br />

team z.<br />

We will use maximum-flow techniques to achieve the following two things.<br />

First, we give an efficient algorithm to decide whether z has been eliminated<br />

from first place--or, to put it in positive terms, whether it is possible to choose<br />

outcomes for all the remaining games in such a way that the team z ends with<br />

at least as many wins as every other team in S. Second, we prove the following<br />

clean characterization theorem for baseball elimination--essentially, that there<br />

is always a short "proof" when a team has been eliminated.<br />

(7.59) Suppose that team z has indeed been eliminated. Then there exists a<br />

"proof; of this fact of the following form:<br />

z can finish with at most m wins.<br />

There is a set of teams T c_ S so that<br />

~ wx + ~ gx~, > mlT[.<br />

x,y~T<br />

(And hence one of the teams in T must end with strictly more than m<br />

wins.)<br />

As a second, more complex illustration of how the averaging argument in<br />

(7.59) works, consider the following example. Suppose we have the same four<br />

teams as before, but now the current number of wins is<br />

New York: 90, Baltimore: 88, Toronto: 87, Boston: 79.<br />

The remaining games are as follows. Boston still has four games against each<br />

of the other three teams. Baltimore has one more game against each of New<br />

York and Toronto. And finally, New York and Toronto still have six games left<br />

to play against each other. Clearly, things don’t !ook good for Boston, but is it<br />

actually eliminated?<br />

The answer is yes; Boston has been eliminated. To see this, first note<br />

that Boston can end with at most 91 wins; and now consider the set of teams<br />

T = {New York, Toronto}. Together New York and Toronto already have 177<br />

wins; their six remaining games will result in a total of 183; and !~ > 91.<br />

This means that one of them must end up with more than 91 wins, and so<br />

Boston can’t finish in first. Interestingly, in this instance the set of all three<br />

teams ahead of Boston cannot constitute a similar proof: All three teams taken<br />

togeher have a total of 265 wins with 8 games left among them; this is a total<br />

of 273, and 273 -T- = 91 -- not enough by itself to prove that Boston couldn’t end<br />

up in a multi-way tie for first. So it’s crucial for the averaging argument that we<br />

choose the set T consisting just of New York and Toronto, and omit Baltimore.<br />

401

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!