The short answer
Guess every game like a coin flip and your chance of a perfect bracket is 1 in 2^63. That is one chance in 9,223,372,036,854,775,808, or about 1 in 9.2 quintillion, for one bracket covering the 63 games from the Round of 64 through the championship in a single tournament.
Knowing the sport helps a lot. It just doesn’t help enough. NCAA.com puts the average Bracket Challenge player (which we call the typical fan) at 1 in 120.2 billion. Picking the higher seed in every game gives about 1 in 73.1 billion by our calculation. The best published models, per Sokol and inpredictable, sit between about 1 in 10 billion and 1 in 50 billion, and the most optimistic estimate Georgia Tech’s Joel Sokol has seen is about 1 in 2 billion.
So a well-informed bracket faces odds in the billions to hundreds of billions. No one has ever filled out a verified perfect bracket. The longest verified perfect start is 49 games, in 2019.
Why the published odds disagree by billions
Every perfect-bracket figure is a model, not a measurement. Each one multiplies 63 game-by-game chances together, so a small change in how likely you think each game is becomes a giant change at the end. Here is the ladder, from longest odds to shortest, with the method behind each rung.
- Coin flip, 1 in 9.2 quintillion. Every game is 50/50. Nobody believes that about a No. 1 seed against a No. 16 seed, but it is the exact answer for pure guessing.
- Typical fan, 1 in 120.2 billion. This is NCAA.com’s figure. NCAA.com took the average Bracket Challenge player’s pick accuracy on first-round games over five years, weighted by how often each seed gap occurs, and got 66.7%. It then applied that rate to all 63 games: 0.667 raised to the 63rd power. That is crude: it uses first-round accuracy for every round, and it gives the same 66.7% to a 1-vs-16 mismatch and to a toss-up title game.
- Always pick the higher seed, about 1 in 73.1 billion. This is our calculation, and picking every favorite like this is called chalk. It uses each seed matchup’s historical win rate: the NCAA’s 1985 to 2025 first-round records, Wizard of Odds’ seed-pair rates for the next three rounds, and 50% for each Final Four game and the title game. Wizard of Odds’ own versions come out at 1 in 56.7 billion on raw historical rates and 1 in 13.6 billion with smoothed seed strengths.
- Market-based optimal bracket, about 1 in 50 billion. The analytics site inpredictable built team rankings from betting markets and found the single most likely bracket for each tournament from 2010 to 2013. Those odds ranged from 1 in 39 billion to 1 in 72 billion, about 1 in 50 billion. In the same years, picking strictly by seed was about 10 times worse, at 1 in 190 billion to 1 in 649 billion.
- Best statistical models, 1 in 10 billion to 1 in 40 billion. Sokol’s benchmark is a model that predicts regular-season games correctly about 75% of the time, and he puts the best models’ perfect-bracket odds in this range. That does not mean 0.75 to the 63rd power, which would be about 1 in 74.3 million. Tournament games are harder to call than that regular-season 75% suggests, so his range is far longer.
- Most optimistic, about 1 in 2 billion. This is the best figure Sokol has seen, and its method was not given. Even at those odds, he notes, you would need roughly 1 billion tournaments for a 50-50 chance of ever seeing a perfect bracket.
Why do two seed-based figures differ so much (about 1 in 73.1 billion here, 1 in 190 billion or worse at inpredictable)? They measure different things. Ours uses long-run averages for a generic tournament. inpredictable priced the seed bracket in four specific years using that year’s betting markets, which could rate a team well above or below its seed.
Older estimates fill in the history. In 2014, Carl Bialik of FiveThirtyEight showed that a flat 70% accuracy on every game gives 1 in 5.7 billion, while noting that real accuracy varies from game to game. His 2006 Wall Street Journal column, using Ken Pomeroy’s ratings, put the chance of a perfect bracket at about 1 in 722 billion.
One more figure circulates and does not belong on the ladder. Betting moneylines priced the brackets that actually happened from 2010 to 2013 at 1 in 571 trillion to 1 in 20,034 trillion. Those are the odds of the real results, upsets included, which no sensible person would have picked in advance. A perfect-bracket figure asks about the best bracket you could pick, not the one that happened.
Where perfect brackets die
View the data table
| Outcome | |
|---|---|
| Perfect after Round of 64 (32 games) | 36,500 |
| After Round of 32 (48 games); 2025 and 2026 streaks ended here | 30,000,000 |
| After Sweet 16 (56 games); record 49-game streak ended at game 50 | 760,000,000 |
| Full bracket (63 games) | 73,100,000,000 |
Run a chalk bracket through our calculation and watch it shrink. A perfect first round, all 32 games, is about 1 in 36,500. Getting the second round right as well, given a perfect first round, is about 1 in 821, so a bracket perfect through 48 games is about 1 in 30.0 million. Through the Sweet 16 it is about 1 in 760 million, and the full 63 games come to about 1 in 73.1 billion.
In log terms the first two rounds do most of the damage, because they hold 48 of the 63 games, but the last 15 games still lengthen the odds from about 1 in 30.0 million to about 1 in 73.1 billion.
In 2026, the one year this page counts, real brackets collapsed in the same early rounds. In 2026, NCAA.com tracked about 36 million brackets across the major online games. After the first round, about 224 were still perfect. The last one, an ESPN entry called ‘christienter,’ lasted 43 games and busted in game 44, when No. 6 Tennessee beat No. 3 Virginia in the second round. In 2025 the last perfect bracket fell in game 43. The record run, 49 games, made it one game into the Sweet 16.
The first-round seed record, updated
A No. 1 seed against a No. 16 seed is the closest thing to a sure pick, and it is no longer perfect. From 1985 through 2025, No. 1 seeds went 158-2 (99%) in that game. The two losses: UMBC over Virginia in 2018 and Fairleigh Dickinson over Purdue in 2023. After the 2026 tournament the record stands at 162-2.
The other higher seeds, 1985 to 2025: No. 2 seeds 149-11 (93%), No. 3 seeds 137-23 (86%), No. 4 seeds 127-33 (79%), No. 5 seeds 103-57 (64%), No. 6 seeds 98-62 (61%) and No. 7 seeds 97-62 (61%). The 8-vs-9 game is the odd one out: No. 8 seeds are 77-83 (48%), so the higher seed is actually the underdog.
Multiply those rates across four regions and a chalk first round comes to about 1 in 36,500. Pick the 9 seeds instead of the 8 seeds and it improves to about 1 in 27,000. An earlier version of this page made a chalk first round look much easier because it used seed records from before 2018, when No. 1 seeds had never lost and several other seeds had better records than they do now.
Upsets are not rare accidents. The NCAA counts an average of 8.3 upsets per tournament since 1985, where an upset means the winner was seeded at least five lines worse. On average, 4.6 come in the first round and 3 in the second. A chalk bracket is counting on none of a year’s upsets (8.3 on average) showing up.
Has anyone ever come close?
The closest verified bracket belongs to Gregg Nigl of Columbus, Ohio, in 2019. He got the first 49 games right: both opening rounds plus the first game of the Sweet 16. His bracket busted in game 50, when Purdue beat Tennessee.
Recent years have ended sooner, in the second round. In 2025 the last perfect bracket died in game 43, when No. 3 Kentucky beat No. 6 Illinois. The higher seed won that one, so the last survivor fell on the upset it had picked. In 2026 the end came in game 44, No. 6 Tennessee over No. 3 Virginia.
No verified perfect men’s bracket has ever been recorded.
Tens of millions of brackets, still no winner
Lots of brackets get filled out. NCAA.com tracked about 36 million men’s brackets across the major online games in 2026 (it normally tracks about 20 to 25 million a year). That count is probably a floor, in our reading, because it leaves out paper sheets and office pools. Survey estimates from the American Gaming Association run higher: 56.3 million American adults planning to enter a bracket contest in 2023, and, in 2015, 40 million Americans filling out more than 70 million brackets.
Now multiply. The expected count is the average number of perfect brackets you would see per year over a very long run: the number of brackets times the chance that each one is perfect. With the 36 million brackets NCAA.com tracked in 2026 at 1 in 120.2 billion, it is about 0.0003 a year, or one perfect bracket about every 3,300 years. NCAA.com says it normally tracks about 20 to 25 million online brackets a year, which at the same odds would be about 4,800 to 6,000 years.
Better pickers barely rescue it. At Bialik’s illustrative 1 in 5.7 billion, the expected count on the 2026 count of 36 million brackets is about 0.006 a year, or one about every 160 years. At Sokol’s most optimistic 1 in 2 billion, it is at most about 0.018 a year on that same count. At coin-flip odds, it is about 4 in a trillion.
Those figures are generous, because they treat every bracket as an independent shot. Brackets are not independent. NCAA.com says typically about 94% of the entries in its Bracket Challenge Game are unique, but unique is not the same as different. Two brackets that differ only in one 8-vs-9 pick both die on the same early upset. Millions of entries crowd around the same favorites, which pushes the chance that even one survives below the simple multiplication.
Warren Buffett’s billion-dollar bracket and other prize challenges
View the data table
| Outcome | |
|---|---|
| Quicken/Buffett 2014: grand prize, face value | 1,000,000,000 |
| Quicken/Buffett 2014: lump-sum option | 500,000,000 |
| Quicken 2014: expected cost, 15 million entries at 1 in 120.2 billion | 125,000 |
| Quicken 2014: inpredictable fair value (10M entries, 1 in 50 billion) | 200,000 |
| Kalshi 2026: grand prize (ten $100 million installments) | 1,000,000,000 |
| Kalshi 2026: expected cost, 10M entries at 1 in 120.2 billion | 83,000 |
| Kalshi 2026: expected cost at inpredictable's 1 in 50 billion | 200,000 |
| Kalshi 2026: expected cost at Wizard's smoothed 1 in 13.6 billion | 735,000 |
| Kalshi 2026: best-bracket prize if nobody was perfect | 1,000,000 |
| Berkshire employees: yearly prize for life, perfect through Sweet 16 | 1,000,000 |
| Berkshire employees 2025: prize for 30 of 32 first-round picks | 1,000,000 |
If you came here to ask whether a billion-dollar bracket prize is real: yes, and it is also cheap to offer. The tool is expected cost, the sponsor’s long-run average payout if the same contest ran over and over. It equals the number of entries times the chance each one is perfect times the prize.
Quicken Loans, 2014. Quicken Loans and Yahoo Sports ran the Billion Dollar Bracket Challenge, insured by Warren Buffett’s Berkshire Hathaway. A perfect 63-game bracket would have paid $1 billion in 40 annual installments of $25 million, or a $500 million lump sum. Entry was capped at the first 15 million qualified entrants, one per person, U.S. residents 18 and older. The 20 best imperfect brackets would get $100,000 each.
It ended fast. Every entry was eliminated during the Round of 64; the last three perfect brackets fell when No. 8 Memphis beat No. 9 George Washington. Quicken never said how many people entered.
Even with all 15 million slots filled at the typical fan’s 1 in 120.2 billion, the chance of any winner was about 0.012%. Slate, using DePaul professor Jeff Bergen’s 1 in 128 billion for informed pickers and an assumed 15 million entries, got about 0.012% (1 in 8,500). That makes the expected payout about $125,000 at the $1 billion face value, or about $62,000 at the lump sum. inpredictable put the fair value of Berkshire’s insurance at about $200,000 (the original 10 million entry cap at 1 in 50 billion), called that a floor, and guessed the real premium was at least seven figures.
Buffett reportedly priced the risk by the head. In an earlier version of the deal with Yahoo, which fell through, he reportedly wanted an extra premium for every entrant beyond 10 million. We found no repeat of the public contest.
Kalshi, 2026. The prediction market Kalshi offered a free $1 Billion Perfect Bracket Challenge, one entry per person, capped at 10 million entries (Kalshi could raise the cap). Kalshi could choose to pay the prize as ten annual installments of $100 million, with $1 million for the best bracket if no one was perfect. You needed an identity-verified Kalshi account, and residents of Florida and New York could not enter.
Every perfect Kalshi entry was gone in the Round of 32, when No. 9 Iowa, about a 16% underdog in the betting market, beat No. 1 Florida. The $1 million best-bracket prize went to a user named Jackie, who had UCLA as champion. With a full 10 million entries, the expected cost was about $83,000 at the typical fan’s odds, about $200,000 at inpredictable’s 1 in 50 billion, and about $735,000 at Wizard of Odds’ smoothed chalk figure of 1 in 13.6 billion. Each is an upper bound: duplicate brackets, shared prizes and the installment option would all lower it.
Berkshire’s employee contest. Since 2016, Buffett has run a contest for Berkshire Hathaway employees (about 377,000 of them in 2018). Picking every game through the Sweet 16, meaning all 16 teams that reach it, paid $1 million a year for life, with $100,000 for the best bracket. In 2024, employees could skip the eight games involving No. 1 and No. 2 seeds. According to CNBC, anyone who picked the winners of at least 30 of the 32 first-round games was eligible for the $1 million.
That finally produced a winner. A FlightSafety International employee went 31 for 32 and took the $1 million, and 11 others who also went 31 for 32 got $100,000 each. The winner’s only miss in the first 45 games was Illinois beating Xavier: a superb bracket, and still not perfect.
So what is your entry worth? Its slice of an $83,000 or $125,000 expected payout, split across millions of entries, which is next to nothing. A free entry is harmless fun, and the prize you could realistically win is the consolation money for the best imperfect bracket. Sponsors still buy insurance because of tail risk, the small chance of a huge loss: the expected cost is tiny, but one perfect bracket would cost $1 billion. Insurers take that risk for a fee, and If inpredictable’s guess of a seven-figure premium against a $200,000 fair value is right, the margin was comfortable.
Putting the odds in perspective
One Powerball ticket has a 1 in 292,201,338 chance of hitting the jackpot in a draw, about 1 in 292 million. A typical fan’s perfect bracket is 1 in 120.2 billion, and an all-chalk bracket about 1 in 73.1 billion. Even Sokol’s most optimistic 1 in 2 billion is longer odds than the jackpot.
The comparison has limits. Powerball’s figure is exact arithmetic for a fair draw, while the bracket figures are models built on assumptions about basketball. A ticket covers one draw; a bracket covers a whole tournament. For the lottery side of the comparison, see our look at the strategies Powerball players use.
Perfect bracket vs. winning your pool
These are two different goals, and they call for different brackets.
If you are chasing perfection, pick the most likely winner of every game. That means mostly chalk, which our calculation puts at about 1 in 73.1 billion, with one exception the record leans toward: take the 9 seeds in the 8-vs-9 games. The 9 seeds lead 83-77 from 1985 through 2025, and picking them improves a perfect first round from about 1 in 36,500 to about 1 in 27,000. Rankings based on betting markets did better than seed picks in inpredictable’s test years, about 1 in 50 billion.
If you are trying to win your office pool, perfection is not the target. You only need to beat the other brackets in your pool, and, in our reading, most of them lean on the same favorites. With an average of 8.3 upsets per tournament, a bracket with a few well-chosen upsets could pull away from the crowd if those upsets land. That is our reasoning, not a sourced finding. How many to pick depends on your pool’s size and scoring, and we have no data source for that, so we won’t invent a number.
What not to do, either way:
- Don’t pick a No. 16 seed to beat a No. 1. It has happened twice in 164 games.
- Don’t treat a free billion-dollar contest as a retirement plan. With every entry at the typical fan’s odds, Kalshi’s whole expected payout across 10 million entries was about $83,000.
What these numbers assume
- Games are independent. Every model here multiplies per-game chances as if one game tells you nothing about another. In reality a team’s strength shapes every game it plays, so results are linked. No source we found measures how much that changes the odds, so treat it as an open question, not a correction.
- History stands in for this year. Seed rates are long-run averages: first-round records through 2025, and later-round rates from Wizard of Odds’ historical data, which runs to about 2021. A given year’s field can be stronger or weaker at the top.
- The typical fan figure is a flat rate. NCAA.com’s 1 in 120.2 billion applies a flat 66.7% per game, the average Bracket Challenge player’s first-round accuracy, to all 63 games. A real fan is far better than 66.7% on a 1-vs-16 game and worse on a coin-flip matchup.
- The First Four is left out. Most bracket contests start at the Round of 64. Include the four First Four games and a coin-flip bracket becomes 1 in 2^67, about 1 in 147.6 quintillion.
- Expected counts and costs are ceilings. They assume every bracket is an independent, distinct shot and that every contest slot was filled. Quicken never disclosed its entrant count.
- Older figures are retired. An earlier version of this page used seed records from before 2018, when No. 1 seeds were unbeaten against No. 16 seeds. Every chalk figure on this page uses records updated through 2025.
Correction, October 18, 2026: We had said we had not seen how NCAA.com derived its 1 in 120.2 billion figure, and offered our own reconstruction. NCAA.com published its method, and the page now explains it. We also tied the 36 million bracket count to 2026, the year NCAA.com counted them.
