Underhook → wrestle-up61.2%
Live now — BJJ · MMA · Wrestling
HellsData turns every roll, every round, every real exchange into a denominator. Tap in what happened, and the system does what a highlight reel never can — tell you what actually wins, with a real interval around the number, not a vibe.
Bottom side control
MMA · GI OFF · n=214 · 90-DAY WINDOW
The actual problem
Every highlight reel shows you the finish. None of them show you the ninety attempts it failed for first, or the guy it doesn't work on at all. "This works" and "this worked once, on camera" have never been the same claim — until now they've been priced the same.
01 — How it works
No cameras, no sensors, no film review. You tap what happened right after it happens — offline-friendly, works over bad gym wifi, syncs when it can.
From-position, what you tried, what happened. One tap per step. Built one-handed for right after the round ends, not for a desk later.
Wilson score interval, never raw percentage. Empirical-Bayes shrinkage toward the pooled rate so a thin MMA sample still means something instead of showing an empty screen.
Every technique from a position, ranked, by ruleset, with an honest confidence band — and an arrow only when a real trend clears two statistical gates, not one good week.
The divergence card
Segment by ruleset and the rankings don't just shift — they invert. The technique that wins in sport BJJ can be the one most likely to get you hit the moment strikes are legal. That's not a footnote. That's the entire reason a single number was never enough.
Real shape of the data, not an illustration — this is what the divergence card looks like when a technique's rank genuinely flips across rulesets. That crossing line is the most shareable thing the product produces.
02 — One engine, every sport
Every position, every transition, every fight-ending question — Wilson intervals, verification weighting, shrinkage, display floors — is discipline-agnostic by design. Content ships sport by sport, real terminology, hand-checked before it goes live.
03 — Two ways in
Same underlying data, two completely different views — an athlete never sees another member's stats, and a gym owner never sees anything below the diversity floor that would out a single student's numbers.
Your own dashboard — your positions, your rates, your gaps against people at your belt and weight. Not the internet's opinion. Yours, with a real denominator behind it.
Which students are quietly stalling out, what the room actually struggles with, engagement without exposing anyone's personal rate. The dashboard is built to tell you who needs a conversation before they quit — not to rank your students against each other.
04 — Why any of this is trustworthy
One viral "this app says a move works 80% of the time and it's garbage" thread costs more than the feature is worth. Every rule below is enforced server-side — the client is never trusted to hide a number it shouldn't have received.
Never raw math
One success out of one attempt reads "100%" as raw math and "somewhere between 21% and 100%" honestly. We show the second one.
Display floor
Below the floor, no percentage renders at all — not a low-confidence guess, silence. n≥200 is the only tier we call confident.
Verification tiers
Video-confirmed weighs 1.0, gym-witnessed 0.7, self-reported 0.4 — inflated self-reports can't quietly dominate a rate.
Two-gate trend
p<0.05 AND a swing bigger than 3 points — both, always — or it's flagged flat. One good week never becomes a headline.
Every tap is one more real data point in a sport that's never had one. Start tonight, on the mat.