Applying to college is kinda hard. You have to figure out which schools you like,
as well as which schools you actually have a good chance of
getting into. That’s the **informational problem.**

Once you have this information, it’s tough to decide how to allocate
your limited applications across desirable, competitive “reach schools”
and less attractive “safety schools” where admissions is a safer bet.
That’s the **strategic problem.**

There are boatloads of “chance me!” websites that can help you with
the informational problem. But when it comes to the strategic problem,
research shows that people often resort to folk wisdom,
risk-averse heuristics, and gut feelings.
**Mulberry treats the strategic problem as a math problem.** It
complements those introspective techniques with dispassionate, numbers-driven advice.

**Mulberry is an algorithm.** You input estimates of your admissions
chances at each school and the amount of utility you associate with going there,
and Mulberry sorts the schools in the order you should apply
to them to maximize your expected utility. (“Utility”?)

We’ve randomly generated a few colleges to get you started.

… will appear here.

Your optimal application order:

If you can only apply to $n$ colleges, you should apply to the first $n$ colleges listed above. Your expected utility with this application strategy, assuming you enroll in the best college you get into, appears in the second column.

Your goal, in applying to college, is (probably) to maximize the utility of
the *best* school you get into. If you apply to just one school, then
your expected utility is its admissions probability times its utility value.
But if you apply to many schools, your expected utility is a conditional
probability expression that depends on whether you get into your first choice,
your second choice, and so on.

It turns out that
**
the optimal application strategy can be characterized
by a certain permutation (ordering) of the schools.
**
If you can only apply to one college, you should apply to the first school
in that permutation; if you can apply to two colleges, you should apply to
the first two, and so on. Mulberry takes your admissions probabilities
and utility parameters and computes this permutation.

The interesting thing is that the optimal permutation is *not*
determined by simply sorting the schools by the products of their admissions probability
and utility values. **Mathematically, we can show that “obvious“ idea is
incorrect**—it can produce
a suboptimal result. To get the true optimum, you need to delve into
the conditional probabilities described above, and that’s exactly what
Mulberry does.

The underlying algorithm is detailed in my
MS thesis.
My thesis also addresses a more general form of the problem in which colleges
have different application *fees* and you have a finite *budget*
to spend on applications. If you want to solve instances of this harder problem,
and are comfortable with the command line,
check out my Julia
package.

A utility
value is a number that summarizes
**
how much you like each school
in light of your own priorities.
**
The college’s student experience, career outcomes,
the cost of tuition, and the graduation rate will
all contribute to the utility value. You can consult
the College Scorecard for
factual data about colleges and universities in the US.

Once you’ve read up on a target school, a good way to estimate its
utility value is to ask yourself the following question:
**
If letters of admission to this college were sold
at Best Buy,
how much money would I be willing to pay for one?
**
Then, if necessary, divide your answer by some constant so you can
fit it in the scale above.

This website is hosted using GitHub pages. The algorithm is
written in JavaScript. You can read all the code
here. This website does not use
cookies, so **all the data you input lives only on your computer,**
and if you refresh the page it will disappear.
The code is provided under a GPL license, which means that you can reuse it in your
own projects, but only if they are released under a similar open-source license.

**Bugs, questions, comments:** Submit a pull request on
GitHub or email me at maxkapur@gmail.com.
My homepage is maxkapur.com.

The logo and theme colors are derived from a Wikimedia Commons image by user Geo Lightspeed7.