UFC Analytics · Grappling
Gravity Always Wins
Strikers get the highlight reels. Wrestlers get the wins. Thirty-two years of fight data on the sport's oldest weapon — and why nobody has ever really figured out how to make it stop working.
What is a takedown actually worth?
One of them, on its own: nothing you could measure. A fighter who lands exactly one takedown wins 52.3% of the time, and the range that sample supports runs from 50.6% to 54.0% — it cannot rule out a coin flip. Everything the takedown is famous for lives in the second one, and the third.
The takedown is older than the sport. Before there were rounds, before there were gloves, before there were weight classes, there was one man deciding the fight would happen on the floor and another man unable to stop him. Everything the UFC has become — the striking clinics, the spinning elbows, the pay-per-view theater — is built on top of that first, brutal transaction.
Thirty-two years in, everyone knows it's coming. Every camp drills sprawls. Takedown defence is a line item on every scouting report. And yet the reward for landing them has not budged an inch. Out-takedown your opponent and you win roughly two fights in three — 68.7% before 2010, 68.4% through the 2010s, 67.3% in the 2020s. A number that steady across three decades looks less like a trend and more like a law of physics.
This is the story of that law.
Act I — The volume era nobody saw coming
Track takedown attempts per fight from 1994 to today and the shape is not the one the sport tells about itself. The wrestling boom is real, and it is early: attempts peak at 6.83 per fight in 2003. Landed takedowns peak later, at 2.66 in 2010. Neither record belongs to the modern era — the 2020s run 5.37 to 6.05 attempts and 1.96 to 2.19 landed, under both. Scroll through it.
The beginning. 1.32 attempts per fight, landing 0.77 of them. A low base, but a real one: nobody has worked out yet that the takedown is the shortest route to a decision.
The wrestlers arrive. Attempts peak at 6.83 per fight — the highest in the whole record — and accuracy falls to 33.2%. Everyone is shooting; the sprawl is being invented in real time.
The gap in the tape. Recorded attempts collapse to 0.43 per fight in 2006. That's a hole in the data, not a hole in the sport — a caveat this chart wears honestly.
The high-water mark for landing them: 6.21 attempts and 2.66 landed per fight, the most productive wrestling year on record. Nothing since has beaten it.
More shots than ever. 2024, 2025 and 2026 all clear 6 attempts per fight — with around 2.2 landed. Three decades of everyone learning takedown defence, and the wrestlers just shoot more.
Where the record starts
Every archive has a first entry. In the UFC's takedown ledger, it belongs to the man the whole sport was built to showcase.
Act II — What a takedown actually buys you
Here is the exchange rate, computed across every fight in the record with a decisive winner. Land nothing, and you win less than half your fights. Every takedown after that buys you a bigger share of the outcome — and control time compounds with it, from 27 seconds on average to over eight minutes. Tap a number for the receipts.
Win rate when you out-takedown your opponent
2 in 3Split every decisive fight three ways — out-takedown your opponent, break even, or get out-taken-down — and the picture barely moves across three decades. Before 2010: 68.7% win rate for the fighter with more takedowns. Through the 2010s: 68.4%. In the 2020s, with the best takedown defence in history: 67.4%. The whole sport learned to wrestle, and the advantage shrank by less than two points.
Act III — Yes, defence got better. It doesn't matter.
Takedown accuracy has fallen — from 43.9% before 2010 to 36.4% in the 2020s. That is the professionalisation of the sprawl, right there in the record. And wrestlers answered it the obvious way: they shot more, 4.79 attempts per fight before 2010, 5.64 in the 2010s, 5.74 today.
Here is what those two did to each other. Takedowns landed per fight: 2.10 before 2010, 2.07 in the 2010s, 2.09 today. Control time per fight: 260 seconds, then 257, then 252. And each takedown still buys the same two minutes it always did — 123.8 seconds then, 120.3 now.
Thirty-two years of better sprawls and harder shots, and the amount of wrestling that actually happens in a UFC fight has not moved. The extra volume bought back exactly what the defence took away, and not a second more. The ground game is not dying, and it is not compounding. It is in equilibrium.
1993–2009 · 1,265 fights
43.9%
takedown accuracy · 4.79 attempts and 2.10 landed per fight · 123.8s of control per takedown · 260s control per fight.
2010–2019 · 4,196 fights
36.7%
takedown accuracy · 5.64 attempts and 2.07 landed per fight · 124.3s of control per takedown · 257s control per fight.
2020s · 3,335 fights
36.4%
takedown accuracy · 5.74 attempts and 2.09 landed per fight · 120.3s of control per takedown · 252s control per fight.
The great walls are real, for the record: Gleison Tibau turned away 70 of 76 takedowns attempted against him — a 92.1% defence rate, the best of anyone who faced 75+ attempts. José Aldo is a whisker behind on 91.6% from 107 attempts against; Petr Yan sits at 86.9% on a massive 160 attempts against; Max Holloway at 84.0%. Individual fighters can beat gravity. The division-wide average never has.
Accuracy fell seven points.
Volume rose to cover it.
Gravity kept the change.
When the shot lands
Takedown accuracy decays as the fight goes on — 39.6% in round one, sliding every round to 32.9% in round five. Fresh legs shoot better. But attempts don't fall with it: round three is the busiest of all at 1.38 attempts per fighter-round, because tired opponents are the ones worth shooting on, even at worse odds. Control time per round peaks there too.
Round 1
1.13 att/rd
Round 2
1.19 att/rd
Round 3
1.38 att/rd
Round 4
1.09 att/rd
Round 5
1.30 att/rd
Rounds four and five draw from championship and main-event fights only (768 and 668 fighter-rounds, versus 17,204 in round one).
The lighter you are, the more you fall
Takedowns are a small-fighter's economy. Flyweights land 2.61 per fight — nearly twice the heavyweight rate of 1.38. The gradient is almost perfectly monotonic down the scale: less mass means more shots, more scrambles, more chain wrestling per minute. Heavyweights don't shoot much, but note the exception in the data: heavyweight title fights convert at 43.4%, the best accuracy of any division listed — when a big man commits, it counts.
The emblem · Most takedowns in UFC history
The Machine
If this story has a face, it's Merab Dvalishvili's. 119 takedowns landed across 17 UFC fights — seven per fight, well clear of the 90 of his closest pursuer, Georges St-Pierre. He is everything the data says a takedown artist should be in the defence era: his accuracy is a mediocre 35.2%, almost exactly the modern league average, and it does not matter at all, because he has attempted 338 of them. Nobody in the record comes close to that volume. He's banked 6,445 seconds of control — over 107 minutes of other men carrying his weight — and won 82.4% of his UFC fights doing it.
The signature exhibit is the night he met Petr Yan — the third-stingiest takedown defence in the dataset at 86.9%. Merab shot 49 times, the most attempts anyone has recorded in a single fight, landed 11 at a grim 22.4%, and won anyway. Accuracy is a stylist's stat. Attempts are a believer's.
Only one man has ever landed more in a night: Khabib Nurmagomedov's 21 takedowns against Abel Trujillo at UFC 160 in 2013 — on 27 attempts, at 77.8%. Merab's answer came twelve years later, with the belt on the line.
More takedowns landed than anyone in the UFC record — nearly double second place.
Merab put Cory Sandhagen down 20 times on 37 attempts in their bantamweight title fight — the second-most in a single UFC bout, ever — and kept his belt by unanimous decision.
The oldest weapon
Thirty-two years of adaptation, and here is what the ledger says: the sprawl got better, the accuracy fell, and the takedown won anyway. It won by volume where it lost on percentage. It won two of every three fights where it out-landed the other side — in 1997's chaos, in 2015's boom, in this decade's chess matches, at almost exactly the same rate.
The first recorded entry in the book is Royce Gracie dragging Remco Pardoel to the mat in 1994. The most recent chapters belong to a Georgian bantamweight who shoots fifty times a night and doesn't care what the accuracy column says. Between them, everything about the sport changed. The transaction at the bottom of it never did. You can defend a takedown. You cannot defend all of them.
Gravity always wins.
Men and women
The ladder is the same in both games
Women’s bouts are a tenth of this database, begin in 2013, and sit only in the lighter classes — enough differences that a shared finding is worth checking rather than assuming.
Win rate on two takedowns
1,164 fighter-performances. The step up from one takedown to two is the story’s central claim, and this is where it was measured.
The same, slightly higher
231 fighter-performances. Nominally ahead of the men, but the difference is −2.9 points with a range of −9.7 to 3.9 — comfortably consistent.
The takedown behaves identically in two games that share very little else. Women reach a knockout at less than half the men’s rate and go to the cards nineteen points more often, and none of that changes what a second takedown is worth. A mechanism that survives that much difference in context is a sturdier mechanism than one measured in a single population — so the advice below applies to both, which is not something this site can say about everything.
What this changes
Train the second takedown, not the first
The most useful thing in this data is not that takedowns win fights. It is the shape of the curve: flat at one, and steep after.
The re-shot, not the entry
One takedown is worth 52.3%, which this sample cannot separate from a coin flip. Two is worth 61.0%, and the jump between them is real — 8.7 points, with a range of 6.4 to 12.8. Five or more is worth 82.2%. Drill chains, re-shots and the scramble after the first attempt fails, because that is where the entire effect lives.
Your best shot is the early one
Takedown accuracy peaks in round one at 39.6% and falls every round after — 37.1%, then 35.2%. Attempts move the other way, rising to 1.38 per fighter per round by the third. Late takedowns are chased rather than set up, and the numbers show it from both directions at once.
The takedown as a single event
Game-planning around getting the takedown treats it as a switch. The data treats it as a dose. Land none and you win 41.1%; land one and almost nothing changes. A camp built around one perfect entry is a camp built on the flattest part of the curve.
Every figure here rests on between 696 and 11,241 fighter-performances, and all of them clear the no-effect line except the single-takedown number — which is precisely why it is included. What none of it can tell you is whether takedowns cause the wins, or whether the fighters landing five of them were winning that fight regardless.