HYP3
01Divergent02Scatter03Timeline04Bracket
March Madness 2026 Tournament

HYPE vs. PERFORMANCE

HYP3 measures the distance between how loudly the internet talked about each team vs. how far they actually went that season. See how we measured 2026.

Most overhyped
NC State
gap -35
Most underhyped
Utah State
gap +47
Overhyped flameouts
11
of 68 teams
Underhyped sleepers
16
of 68 teams
About HYP3

We rank every team twice, then subtract.

Twitter, ESPN, and Google all settle on a handful of favorites weeks before tipoff. The teams that actually win the bracket are usually a partial overlap. HYP3 plots both lists for the same 68 teams and shows where they don't match.

For each team, we pull daily Google Trends search interest two ways: across the 15-day window around Selection Sunday (tournament mode), and across the full season from Nov 1 onward (season mode). Rank the field by hype, then again by wins, and the difference is the gap. Negative means overhyped. Positive means underhyped. Zero means the internet got it right.

API & Data

The whole dataset is one JSON file.

No backend, no auth, no rate limits. The site bundles the year's file at build time. Fetch it yourself and do whatever you want with it, notebook, sketch, dashboard.

GET/data/2026.jsonBundled with this app.
GET/data/2025.jsonFlorida championship year.
Schema per team
teamstring
School name. Matches the NCAA bracket.
seednumber
Tournament seed, 1 to 16.
regionenum
East / West / South / Midwest.
winsnumber
Tournament wins, 0 to 7. First Four wins count.
hype_normalizednumber
Hype index, 0 to 100 across the field.
hype_ranknumber
Rank by hype. 1 is most hyped.
performance_ranknumber
Rank by wins. Tied teams share the lower number.
gapnumber
hype_rank minus performance_rank. Negative is overhyped.
story_tagenum
overhyped / underhyped / as_expected / noise.
hype_dailyarray
Daily hype value across the 15-day window.
hype_accelerationnumber
In-window mean over pre-window mean. Above 1 means surging into the tournament.
season_*various
Same shape, computed over the full season (Nov 1 through Selection Sunday + 9).
Full Pipeline
External Sources

Where the data comes from.

Three upstream sources, all cached locally. The live site doesn't hit any of them at request time.

Google Trendstrends.google.com
Daily search interest per team. Five teams per batch, each batch anchored to a reference team so values stay comparable across batches.
pytrendsgithub.com/GeneralMills/pytrends
Unofficial Google Trends API wrapper. Drives every Trends pull in the pipeline.
NCAA bracket + standingsncaa-api.henrygd.me
Third-party wrapper over the public NCAA endpoints. Source for seed, region, tournament wins, and full-season win-loss.
FAQs

Questions, answered.

We pull daily Google Trends search interest for each team two ways. Tournament mode looks at the 15-day window around Selection Sunday. Trends scores are 0 to 100 within a single query, so values from different queries aren't directly comparable. We fix that by anchoring each batch of teams to a reference team (a year-round national program with reliable signal) and rescaling every other curve against the anchor's standalone curve, so the field is comparable. Season mode looks at the full season (roughly Nov 1 through Selection Sunday + 9 days) and pulls each team standalone, normalized within the team's own history. That makes season values comparable to a team's own past but not to other teams in absolute magnitude. Either way, a team's hype is the mean of its daily series.