MyChances.net

Today, I finally put the finishing touches on a form I’ve been working on for a couple of weeks now. It’s the page where you create your college profile. This is an important page, because the computer largely bases its predictions on the information you give it here. Starting today, you’ll be asked to indicate which high school you’re attending (out of a list of about 35,000). If it’s not there, you can easily add it without leaving the page. And for privacy’s sake, you’re invited to hide your high school information before you’re even asked for it.

The benefit in disclosing where you go to high school is that it will allow for better predictions.  Unless you think that it’s an accident that Stuyvesant and Andover flood the Ivies with their graduates every year, you’ll probably agree with me that high school “strength” is a fair indicator of future college acceptance. Well, even if you don’t agree with me, we’ll be testing this hypothesis together over the next few weeks.

Are there any patterns that you think will jump out? How do you think your your high school will compare?

I’m rolling out a big survey update today that should simplify and improve the data that gets collected; as always, this will help with predictions. Expect this around 9:00pm Pacific on Tuesday; I’ll make another note here when the update gets “pushed” out.

Later this week, I’ll be clarifying the privacy policy and beefing up privacy safeguards.

By the way, it looks like MIT has been sending out rejection letters; let me know if you’ve seen any MIT acceptances.

Other bloggers have been discussing which schools place the most emphasis on GPA.

At Mychances.net, at least, it’s our experience that the schools at the “top” are the ones most likely to take into account more than GPA and scores. These are the ones most difficult to “chance” based on numbers alone. The ones in the middle of the pack tend to be pretty well predictable based on GPA, standardized tests, and little more than that.

New feature: community predictions. These allow you to rate fellow members’ chances at any school where they have not yet received a decision (acceptance/waitlist/rejection), and others can rate you.

In short time, I’ll be developing an “application” to go with this. As things stand, the plan is to display a profile based on how often you rate others profiles. The probability that someone will be taken to your profile is determined based on how many profiles you have rated. The system should, therefore, reward you for rating others’ profiles. There may be different incentive structures added on later.

I’ve implemented some survey questions to try to begin making the predictions more accurate at some of the tougher schools. If you’re creating a new account, you’ll automatically be asked 7 easy questions (no writing involved). If you’ve already got an account, why not update your account so the algorithm can improve our predictions for you?

(Note - the prediction algorithm is not using this info yet because the site just started collecting it, but this will change in the weeks to come.)

Today I updated the prediction models for the first time in a few months, so you might see a shuffle in your predictions as the model updates. We’re still working with a relatively small sample size in the low thousands. As the site grows, the predictions will become less labile.

As things currently stand, the site is, on the whole 90% accurate (see the bottom piece of data on that page).

Also, the Admission Game has been updated to properly take into account the predictions made by each school’s model.

Welcome, CCers. After getting some feedback requesting more information, I’ve made a couple changes. First, you will now see the predicted probability of admission in your profile, instead of just a “yes” or “no”. Second, the accuracy of the model that is giving you the prediction will be shown just below, so that you can get a better sense of how much you want to trust the prediction.

Cheers,

James

For the past few weeks, predictions have been displayed on the member list for each school. Starting today, you can just visit your own college tracker to see the predictions at each of the schools on your list. That should save a few clicks. Also, if you hover your mouse over the prediction, extra details about that prediction appear in a box above the mouse.

Use this thread to give feedback, positive and negative.

Quick update - we’ve updated the system to display the predicted admissions outcome for each applicant on each school’s page. For example, if you visit the page of University of Illinois you’ll see that the model predicts admission to that school about 75% of the time.

At the University of Virginia, it predicts admission correctly about 62% of the time, while at the University of Pennsylvania, it’s right over 80% of the time.

What do you think of the accuracy, or of the predictions in general?

So over the past two days I’ve rolled out two new features: private messaging and the Selectivity Index.

Private messaging does what you’d expect: it allows you to communicate with other users directly via the site. This way, when you’re browsing through profiles and you want to ask someone a question or give some words of encouragement, you can do so easily.

The Selectivity Index is a way to gauge, at a glance, how your scores compare to the scores of accepted students at a given school. If you have a higher selectivity index, you are above the 50th percentile for students applying to a given school. By no means will this tell you if you will get in, but it is another tool to help you see how you stack up numerically. Always keep in mind that numbers are not everything - especially at this point in history.