US Trends

how often does a 14 seed beat a 3 seed

A 14 seed beats a 3 seed in the men’s NCAA tournament a little under 15% of the time, so it’s rare but definitely not a once‑in‑a‑lifetime shock.

Quick Scoop

  • Since the field expanded in 1985, 14 seeds are 23–137 vs. 3 seeds, which is about a 14–15% win rate.
  • Flip it around and 3 seeds win roughly 86% of these games (137–23 record).
  • In about half of all tournaments since 1985, at least one 14 seed has pulled the upset in a given year.

So if you’re filling out a bracket, you shouldn’t pick every 14 over a 3, but penciling in one carefully chosen 14‑over‑3 upset is very defensible.

How often in plain terms

Think of it this way:

  1. Every year there are four 3‑vs‑14 games.
  2. Across all years since 1985, 14 seeds have won 23 of 160 total matchups (through 2024), landing that ~14–15% rate.
  1. That works out to roughly one 14‑seed win every 1–2 tournaments , though it’s streaky: some years there are multiple upsets, some years none.

Example: in 2024, No. 14 Oakland beat No. 3 Kentucky 80–76 to continue the trend.

Quick numbers table

Here’s what the matchup looks like overall:

[7][9][1] [3][9][1] [9][1] [1][9]
Matchup Record (14 seed) Win rate (14 seed) Record (3 seed) Win rate (3 seed)
3 seed vs 14 seed 23–137 ~14–15% 137–23 ~85–86%

Bracket‑talk angle

From a bracket strategy point of view:

  • Statistically, you expect 3 seeds to advance most of the time, but history says upsets happen often enough that ignoring all 14s is a mistake.
  • Upsets have come in mini‑clusters (e.g., stretches of several straight tournaments with a 14‑over‑3), so fans now almost “plan” for at least one 14 to be live each March.

In forum terms: if someone asks “how often does a 14 seed beat a 3 seed,” the go‑to answer is:
“About 1 in 7 of those games, and in roughly half of all tournaments, we see at least one 14‑over‑3 upset.”

TL;DR: A 14 seed beats a 3 seed about 14–15% of the time overall, with 23 such upsets in men’s March Madness history since 1985.

Information gathered from public forums or data available on the internet and portrayed here.