How accurate are CollegeVine admission predictions?

Answered

Hi! I'm glad you asked this question, and appreciate the discussion on this thread. I’ll try to outline how the chancing system works as best I can, as well as the proper way to “read” its results. Hopefully that helps make chances that might seem a bit odd in a vacuum make more sense.

First, you hit on something in your original question—while CS might be a competitive major at some schools (though it’s not necessarily more competitive than other engineering majors), being a woman in STEM will actually boost your chances quite a bit. It’s not exactly the same as being a URM, but our data demonstrates that it is a pretty significant hook, and that’s probably playing into your chances a bit. (As a side note, the UCs do not factor race into admissions decisions. So being Asian/not-Asian matters for a lot of competitive schools, but not for Berkeley, UCLA, UCSD, etc.)

Second, for the chances themselves, the way to read them is not as predictions of what will happen, but as a rough gauge of what percentage of students like you will get into that school, based on our profile. Example: If you have a 70% chance for a school, then that means that, based on data from past students/people we've sourced data from/publicly released datasets, about 7 in 10 students with those stats were accepted. That said, we obviously can’t take things like the strength of your essays into account—and that’s typically what separates the 7 people who got in from the 3 who didn’t. If your essays aren’t strong, it will look like the algorithm heavily overestimated your chances. If your essays are extremely good, you might even get the opposite situation where you get into several schools CollegeVine marked as reaches.

And with that in mind, acceptance rate is actually a pretty bad barometer by which to gauge the competitiveness of a school. That’s because schools like Berkeley and NYU (as well as most other extremely competitive schools) get a LOT of applications from students who are below their academic thresholds—which means those applications get sorted out by an algorithm in their admissions office and usually never even get read. That means they get to basically auto-reject a bunch of students and artificially lower their acceptance rate, which then gets them a higher ranking the next year. (Schools will actually advertise themselves heavily to students who they know don’t have the stats to get in for this very reason.) What that means is that, if you’re above the academic threshold, your chances might look significantly higher than the acceptance rate—because, though it might not seem so, you’re actually well above the average student that applies to e.g. Berkeley or NYU.

As for where the data comes from, it’s a combination of 1) publicly available datasets, 2) interviews with students at individual colleges, and 3) several thousand students who went through our previous program between 2013-2019.

I hope that helps these numbers make sense. They’re not infallible, as with any probabilistic system there will always be outliers, but there are thousands of data points behind them!