Every David on this page was losing first. We counted 370,713 matches, 286 insurgencies and 24,511 court judgments to find out how often people climb back from the bottom, and what the ones who made it did.
Caravaggio, David with the Head of Goliath, around 1600. Kunsthistorisches Museum, Vienna. Public domain
I cheer for the small side, most of the time, whether it is a team I have never heard of or one person taking on something a thousand times bigger, and when they win it feels a bit like a personal win.
So I wanted to find out how often it actually happens, and whether the numbers could get more people to believe that a Goliath, whatever yours happens to be, can be beaten. There is a line I keep saying to myself these days, and it has really helped me move.
I can do anything, and it will work out.
These are some of the people nobody gave much of a chance, in their own words, from around the time it looked worst, and under each one is a number from the records we counted for people in the same spot.
My flights were booked at the end of qualifying.
Emma Raducanu, 18, ranked 150th
Press conference, 8 September 2021, reported by APShe had to win three qualifying matches just to get into the draw, and she expected to be flying home soon after.
She won ten matches in a row without dropping a set and won the US Open.
I wasn't the biggest lad. It never made me stop.
Jamie Vardy, on his years in non-league football
Premier League interview, 8 October 2019Vardy played non-league football for years while working in a factory, and he was 25 when he joined Leicester.
In 2015-16 he scored 24 league goals for a Leicester side that won the Premier League.
My only concern was then to reach 100.
Kapil Dev, India's captain, walking out at 9 for 4
Gulf News interview, 2008India were 17 for 5 soon after he came in against Zimbabwe, at a World Cup where the bookmakers had them at 66 to 1.
He made 175 not out, India won by 31 runs, and a week later they beat the West Indies at Lord's in the final.
I was supposed to fit a new kitchen but then we were called away to play in Sweden.
Richard Møller Nielsen, Denmark's manager, after Euro 92
UEFA match report, Denmark v Germany, 26 June 1992Denmark had not qualified for Euro 92 and were invited about ten days before it started, when Yugoslavia was banned.
They beat France, the Netherlands and then Germany 2‑0 in the final. He added that a professional decorator finished the kitchen.
I was staggered that a company as huge as McDonald's would take legal action to try to stop people handing out leaflets.
Dave Morris, a former postman, sued with Helen Steel, a gardener
BBC interview, 11 April 2002They defended themselves without a lawyer for 313 days in court, and in 1997 they lost most of the case.
In 2005 the European Court of Human Rights ruled their trial unfair, and the UK paid them €35,000.
I had to write a paper about something, so I decided, let's write a paper about Facebook.
Max Schrems, an Austrian law student on an exchange in California
PBS Frontline interview, 28 March 2018The paper turned into complaints against Facebook, and then into cases at the EU's highest court.
The court struck down the EU-US data agreement in 2015 because of his case, and its replacement in 2020.
I knew that going into the fight no one was giving me a chance, but I believed, my people believed.
Buster Douglas, on fighting Mike Tyson
ESPN, 11 December 2018Tyson was unbeaten, and only one Las Vegas casino would even take bets on the fight, at 42 to 1.
Douglas knocked Tyson out in the tenth round.
I couldn't win more than three games in a set, so I was a little bit lucky.
Juan Manuel Cerundolo, ranked 56th, after playing the world No. 1
AP, 28 May 2026He had lost the first two sets 3‑6 2‑6 to Jannik Sinner and was 1‑5 down in the third, with Sinner struggling in the heat.
He won 7‑5 6‑1 6‑1.
Women like me aren't supposed to run for office.
Alexandria Ocasio-Cortez, then working in a bar in Manhattan
The Courage to Change, campaign video, 30 May 2018She was running against Joe Crowley, the fourth-ranking Democrat in the House, who spent $3.4 million to her $209,617.
She won the primary 57% to 43%.
We took every professional match since 2000 that had a bookmaker's price on it, 370,713 of them in eight sports, and sorted the underdogs by how long their odds were. Each block below is a hundred of them, and the green dots are the ones who won.
Source: sheets.works count of 370,713 matches with a pre-match price, football-data.co.uk, tennis-data.co.uk, nflverse and public odds archives, 2000 to 2026. Hover or tap a block for the exact count.
Hover or tap to look for them
The 5,000 to 1 was a price set before the 2015-16 season on Leicester winning the whole league, a different kind of bet from the single matches above, so it gets a block of its own.
The same underdog has a very different chance depending on what they are playing. In baseball the side the bookmakers expected to lose won 42 games in every hundred, and in European football they won only 23, although football hands them another 27 draws, so half the time the smaller side walks off without losing.
Each row is 100 matches. Green: the underdog won. Light: a draw. Hover or tap a row for the counts.
In football the gap between the best and worst team in a league matters more than the name on the shirt, so the underdog wins more often the further down the pyramid you go. In England's fourth tier the side expected to lose won 27 games in a hundred, and in the Premier League 21. Tennis works the other way at the biggest events, because best of five sets gives the better player more time to recover: at the men's Grand Slams the underdog won 23.5% of matches, and 31.8% everywhere else.
Underdog wins as a share of all league games, 2000 to 2026, 196,150 games in 22 divisions. Source: sheets.works count of football-data.co.uk
The odds are set before anything happens, so we also looked at teams that were already behind. Of 124,906 football teams losing at half-time since 1998, 9,767 went on to win and another 22,325 got a draw, which means one in four losing teams walked off without a defeat.
It happened 15 times in 216,952 games, and one of those teams was four goals down.
Matches are the cleanest place to count, because every one has a winner and a price, but the same question can be asked of wars, courts and elections, although those counts are rougher.
In wars since 1800 where one side was at least five times stronger, the small side won or held out in 12% of those that began before 1850 and in 51% of those that began after 1950. Those figures count a stalemate as a win for the small side, and if you split them out, insurgents fighting a state between 1950 and 2005 won outright 34% of the time, drew 35% and lost 31%.
Arreguín-Toft (2005), figure 2; Lyall and Wilson (2009), our countAt the European Court of Human Rights, 84% of the 24,511 judgments since 1959 found that a government had broken someone's rights, but 94% of applications were thrown out before a judgment was ever written. At the US Supreme Court since 1946, a person facing a company won 57% of the time, and a person facing the federal government 39%.
ECHR Overview 1959 to 2021; Supreme Court Database 2024, our count of 8,088 argued casesIn US House races from 2014 to 2024, leaving out 2022, where both candidates spent over a million dollars, the one who spent less won 35% of the time, but a challenger beat the sitting member only 4% of the time, and a candidate outspent ten to one won once in 791 races. In India's Lok Sabha only 55% of MPs who stood again between 1967 and 2019 were re-elected.
OpenSecrets and FEC, our count; Carnegie Endowment and Trivedi Centre (India)On Lichess in January 2014, a player rated 400 to 500 points lower won 11.6% of 22,483 games, more often than the rating system itself predicted. In the men's college basketball tournament a number 16 seed has beaten a number 1 twice in 164 tries, UMBC over Virginia in 2018 and Fairleigh Dickinson over Purdue in 2023.
sheets.works count of the Lichess open database; NCAA tournament recordsLooking at the comebacks one at a time makes them feel like luck, so we went back to the big datasets to see whether the Davids who won had anything in common. Four things kept turning up, and each of them is something a person could choose to do.
Ivan Arreguín-Toft listed 200 wars since 1800 where one side was at least five times stronger. When the small side lined up and fought the big side's kind of war it mostly lost, and when it refused, usually by avoiding the big battle and making the big side come to it, it won nearly two times in three, though that second group is only 22 wars.
Arreguín-Toft, How the Weak Win Wars (2005)Erica Chenoweth and Maria Stephan collected 323 big campaigns to remove a government, end an occupation or break away, and the ones that used marches, strikes and boycotts instead of weapons succeeded more than twice as often, because far more ordinary people were willing to join them.
sheets.works count of NAVCO 1.1, Chenoweth and Stephan (2011)When companies posted problems their own scientists were stuck on to InnoCentive, a challenge website anyone could join, strangers solved almost a third of them, and some of the winning answers came from people in another field altogether, like a protein crystallographer who cracked a toxicology problem.
Jeppesen and Lakhani, Organization Science (2010)Of the 286 insurgencies Jason Lyall and Isaiah Wilson studied, the ones with no outside help won outright 12% of the time, and the ones with a safe place across a border or support from another country won 46%, which is close to four times as often.
Lyall and Wilson (2009), our countKapil Dev's line was only about getting to 100. Mine is the one near the top of this page, so write yours over it, or keep mine, and we'll turn it into a card you can keep or send to somebody who needs it.
Nothing you type leaves your phone or computer
Most Davids lose, and the people in these datasets are the ones whose fights got big enough to be counted, so somebody crushed in their first week never shows up. Money is the hardest thing to beat: in US House elections from 2014 to 2024, a candidate outspent ten to one won once in 791 races. Long shots also win less often than their odds suggest: sides priced at 10 to 1 or longer won 3.9% of the time when the price said about 6.6%. What this page has is a count of how many people were where you are and got out anyway.
How we did it
Underdogs and comebacks. 370,713 matches with a pre-match bookmaker price across eight sports, 2000 to 2026 (football-data.co.uk, tennis-data.co.uk, nflverse, SportsBookReview archives and others); the underdog is the side with the longer price, with the bookmaker's margin removed. Half-time figures come from 216,952 European league games (1998 to 2026), and two-sets-down figures from 12,379 men's Grand Slam matches (2001 to 2026).
Everything else. War figures from Arreguín-Toft (2005). Campaign figures from NAVCO 1.1. Court figures from the European Court of Human Rights' Overview 1959 to 2021. Election spending from the FEC and OpenSecrets. The tables we computed and the scripts that made them are in one zip file.