NYC-SIFT's Offer Prediction Model 2025-26

NYC-SIFT News
1

This year’s offer prediction modeler is nearly ready for release! It will be live the morning of Oct. 6.

Its accuracy isn’t where I’d like, but I believe it’s a bit better than last year’s. Below is an explanation of how cutoffs are calculated.

In an ideal world, the DOE would provide Screened Group and Ed. Opt. applicant breakdowns for the previous admissions cycle. Given that we know the number of offers sent to those applicants, we can then make a pretty good estimate of where program cutoffs were for that year. We have this data for the 2023-24 application cycle but the DOE told me that they won’t be able to supply 2024-25 data until late November. Well, that’s not really going to work considering the application deadline is December 3. So, the modeler is making a guess for this year’s applicants based on a combination of 2023-24 applicant data and citywide Screened Group and Ed. Opt. populations.

Let’s use Manhattan Center for Science and Mathematics’s “Science & Mathematics Research” program (a Screened program) as an example, only focusing on GE applicants for now. During the last application cycle (2024-25), there were 1590 applicants and about 560 received offers. We have the Screened Group breakdown of this program’s applicants for the 2023-24 application cycle:

  • G1: 537 (31.70%)
  • G2: 462 (27.27%)
  • G3: 361 (21.31%)
  • G4: 150 (8.85%)
  • G5: 184 (10.86%)

We also know the citywide breakdown of students by Screened Group:

  • G1: 21.98%
  • G2: 19.06%
  • G3: 21.08%
  • G4: 15.08%
  • G5: 22.80%

Just to be safe we average the two percentages, ending up with a new estimated applicant pool breakdown for this year:

  • G1: 26.84%
  • G2: 23.16%
  • G3: 21.195%
  • G4: 11.96%
  • G5: 16.83%

Now we multiply each group by the most recent number of applicants (1590) and get the full applicant pool numbers that will be used for calculating this year’s cutoffs:

  • G1: 427 (26.84%)
  • G2: 368 (23.16%)
  • G3: 337 (21.195%)
  • G4: 190 (11.96%)
  • G5: 267 (16.83%)

Knowing that 560 offers were sent, we can estimate that the cutoff will be about 36% ( 560-427 / 368 ) into Group 2.

If the DOE sends me updated applicant breakdowns for 2024-25, models will be recalculated and cutoffs may change. If this happens, a notification message will be posted here.

2

It’s live! Please let me know if you see any issues or inconsistencies!

3

A few updates:

  1. I’ve been working with the DOE to examine some suspect offer and applicant data, which is still ongoing. 10 schools have been flagged and so far one has been confirmed with bad data and has just been fixed - Francis Lewis High School (26Q430).
  2. Models for last year’s applicants has been changed and is weighted more towards 2023-24 applicant numbers. Instead of citywide and 2023-24 program specific applicant breakdowns having equal weight, program specific applicant breakdowns now carry twice the weight. For example, using the MCSM calculation from the original post, Group 1 is now: ( (31.70% x 2) + 21.98%) / 3 = 28.46%
  3. Related to the MCSM example and item #1, the number of offers is actually 650, not 560. The DOE corrected a change I made on my own because MCSM’s previous offer numbers from 2022-24 (560, 520, 560) were much lower and I thought the DOE made a typo - I guess not!