CSV files

Uploading your own applicants or programs, and the files a run downloads.

Uploading a population

You can upload your own applicants or programs (or both) instead of generating them; runs then use the file. A run's downloaded applicants or programs can be uploaded again unchanged. The whole file is checked first: if any row has a problem, nothing changes and the problems are listed with their line and column.

With the default parameters the columns are:

• Applicants: name,group,strength,board_scores,research,honors,weight:reputation,weight:program_size,weight:location (sample file)
• Programs: name,tier,quality,capacity,reputation,location,weight:board_scores,weight:research,weight:honors (sample file)

The rules:

• The first row names the columns. The attribute columns are the attributes in your parameters, so the simulation page lists the exact columns for your settings.
• strength, quality and the attributes are on the z-scale (0 = average). capacity is a whole number of positions (1 or more); program_size is computed from it, so it is not a column.
• group and tier are optional names made of lowercase letters, digits and underscores.
• The weight:… columns are optional; give all of them or none. Each row's weights are rescaled to add up to 1. Without them the model draws the weights when the run starts.
• Names must be unique within a file. Files may be up to 5 MB and 50,000 rows, in UTF-8 (Excel's "CSV UTF-8" works).

Please use made-up or anonymised data only (see the terms).

A run's downloads

match.csv

One row per applicant: group, strength, applications, signals, interviews, list length, the matched program, where the applicant ranked it, where the program ranked the applicant, and (with Compare both sides proposing) the program they would get if programs proposed.

program_results.csv

One row per program: tier, quality, positions, applications and signals received, invitations, interviews, list length and positions filled.

applications.csv

One row per application: the applicant's pre-interview rank of the program, reach/target/safety (portfolio strategy), the signal, the invitation wave, whether the interview took place, both sides' true utility, pre-interview view, realised utility and post-interview view, both list ranks, and whether it ended in the match.

pairs.csv

Every applicant–program pair before interviews: both sides' true utility, pre-interview view and ranks (offered for markets small enough to recompute every pair).

metrics.json and params.json

The run's diagnostics with its version stamps, and its parameters with the seed used.