Glossary
Consistent definitions for metrics, model assumptions and evidence quality, without universal success promises.
This glossary defines CupidLeaks terminology. Product names and internal platform rules can differ.
| Term | Meaning on this website |
|---|---|
| Algorithm | A sequence of computational steps; recommendation systems can combine several |
| Benchmark | External reference with documented population and measurement; an arbitrary model value is insufficient |
| Bias | Systematic distortion in data, measurement or decisions; not every difference |
| Burnout index | Synthetic frustration in the lab, not a health measurement |
| Collaborative filtering | Recommendations based on interaction patterns across people |
| Dark pattern | Design that can steer decisions against users’ interests; assumed scenarios in the lab |
| Elo | A rating principle, not evidence about current Tinder |
| Export reference date | Time reference for file analysis, not necessarily today |
| Gini | Concentration of a specified distribution, from 0 to near 1 |
| Ghosting | Unexplained cutoff; the two-week metric detects only visible inactivity |
| Like-to-match rate | Attributable matches from own likes divided by all own likes |
| Lorenz curve | Cumulative shares ordered by ascending values |
| MDC | Match-to-date conversion; requires attributable dates and matches from one cohort |
| MRR | Opener response rate in analysis; only a message proxy in simulation |
| Popularity bias | Preference for already sought-after profiles in recommendations |
| Reciprocal | Considering both directions of a possible connection |
| Seed | Random-generator starting value; reproduction also needs identical version, parameters and actions |
| SMR | Historical lab abbreviation for a match/like counter ratio, not a personal success score |
| Visibility | Actual impressions; not derivable from matches alone |
Further context
Tinder explicitly describes its current system as no longer using Elo. 9 Celdir, Cho and Hwang provide a specific, contextual study of popularity bias and business objectives. 14
Chapter references state dates and limitations. Older research can remain useful without becoming a measurement of 2026.
Sources
- [9] Tinder — Powering Tinder: The Method Behind Our Matching 2026-09-01 · Checked: 2026-09-10
Current provider account; not an independent algorithm audit.
- [14] Celdir, Cho & Hwang — Popularity Bias in Online Dating Platforms: Theory and Empirical Evidence 2023 · Checked: 2026-09-10
Model and data from one Asian platform; not a universal claim about every app.