It shouldn’t be too hard to outsmart the Draw, right? After all, the housing office is happy to explain how it works. The computer systems analyst who maintains the University’s custom housing-assignment software, Rich Wales, is also the author of the notorious “red packet,” the 23-page guide to Draw procedure. It contains 55 bullet points, 23 tables, six “key implementation concepts” and one worksheet.
Wales, ’75, looks a little disappointed when it’s suggested that explanations of the explanations are often required. “I’m a computer nerd. To me, the Draw is transparently obvious,” he says. “I desperately want people to understand how it works, but nevertheless, I’ll admit that I probably still am the world’s expert on the Stanford housing draw.”
An excerpt from the red packet:
Assignment Processing
• The Basic Algorithm
The Draw program starts out by tentatively assigning every student to his/her first residence choice. If (as will normally be the case) some residences are overfilled, each excess student is pulled out and reassigned instead to his/her next choice. This process of assigning, pulling out and reassigning continues until no residence has any excess assignments.
• Focus Priorities and Ethnic Status
After students have been assigned to a focus or cross-cultural residence—but before excess students are pulled out—a quota enforcement pass is made to ensure that the residence’s priority quota is not exceeded.
If more students are trying to claim a focus priority or ethnic status than the residence’s quota allows, the excess students have their priorities deactivated.
• Improvement Processing
After the program has finished assigning, pulling out and reassigning students, a thorough check is performed to ensure that each student is in the best possible assignment. If any students can qualify for assignment to a better choice, a certain number of these students are pulled out of their assignments, and the program reprocesses them. Since this improvement process could, in theory, lead to an endless loop, the Draw program runs improvements carefully and within limits designed to give improvement a reasonable chance, but without taking an inordinate amount of time. In the tests done on the new Draw program so far, most runs require no more than two improvement cycles after the initial processing; the largest number of improvement cycles seen so far is 22.
