Cup Upsets: Giant-Killings Through the Lens of Implied Probability
Cold open: the night the odds bent
On a wet Tuesday night, somewhere far from the bright lights, a 1-in-20 shot bent the arc of a cup tie. The home keeper dove, the bar shook, and a small stand roared like a jet. The favorite looked shocked. The scoreboard did not care. One goal stood. One crowd believed.
We love these nights. But love can blur numbers. Let’s look straight at the thing the market tells us before kick-off: implied probability. We will keep the math small and the game big.
Why cups breed chaos
Cups are one-off shows. A single match can swing on a corner, a card, a slip in rain. Coaches rest stars. Underdogs run hard. Small grounds press the favorite, and noise travels. Extra time and pens add more coin flips. This is not league play, where slow truth shows over 38 rounds. A cup is sharp, short, and strange.
Each country has its own theater of shocks. In England, the FA Cup competition gives non‑league clubs a crack at giants. In Germany, the DFB-Pokal is a sprint with traps in round one. Spain’s Copa del Rey moved to single ties in early rounds, which helps smaller sides. The U.S. Open Cup mixes tiers and travel, and MLS depth gets tested.
Why does this matter to odds? Team rotation, dense schedules, and fatigue shift risk from week to week. Some of this is known at pricing time; some breaks late. Sports science keeps finding stress points in tight runs of games; see recent fixture congestion research. Cups land right in those gaps.
Math, but only for 90 seconds
Implied probability turns odds into a percent. It says, “Given this price, what chance does the market suggest?” It is not the truth. It is the market’s read, plus a fee the book takes (the margin).
How to convert fast (decimal odds): implied probability = 1 / odds. So 5.00 implies 20%. For fractional odds a/b, the percent is b / (a + b). For moneyline, +N means 100 / (N + 100); −N means N / (N + 100). If you want a short primer, see implied probability on Investopedia.
One more bit: the overround. Books build in margin, so if you add up implied probabilities for Home/Draw/Away, the sum is often 103–108%, not 100%. That extra is the fee. Learn more in the overround entry. To make a fairer read, we “strip” that margin by scaling all three outcomes down so they total 100% again.
Debiasing the odds: what gets priced, what does not
Market prices mix true chance, margin, and the crowd’s bias. Some shifts are known (injury news, travel), others are hard to model (a wet pitch at a small ground). In cups, the news flow can be thin, and lineups can flip near kick-off. That adds noise to pre‑match lines.
So implied probability is a snapshot, not a promise. It can still teach us a lot when we adjust for margin and track the same types of spots over many ties.
A catalogue of giant-killing patterns
Not all shocks look the same. Many share tells you can spot in real time. An early underdog goal often slows the game and forces a siege. A red card to the favorite turns field tilt on its head. Bad weather or a narrow pitch cuts speed and space, which helps deep blocks. A team on a third match in eight days may fade late. Crowds in tight grounds add pressure on set pieces.
Set plays and keeper hot streaks matter a lot in these one-offs. One corner can be the whole thing. One keeper can add 1–2 expected goals saved above norm on the night. For data views of these stories, check Opta’s FA Cup upsets data analysis.
Chaos has local flavors. In Germany, lower-league hosts in the DFB-Pokal’s first round create landmines; see the DFB-Pokal official site. Spain’s Copa del Rey one-leg rounds widen the door. In the U.S., travel and midweek slots shape the U.S. Open Cup.
Casefiles, then a table
Let’s ground this in matches. We pulled historic ties across countries and years. We looked at pre‑match odds for the underdog (decimal), turned those into implied probability, and then took off the margin with a simple, transparent step: proportional scaling so that Home/Draw/Away would sum to 100% if we had all three legs. For match facts, see match data at FBref. Odds here are indicative and based on archived market snapshots at or near kick-off. They are not from a single book.
Now, 10 well-known shocks. Note the small base rates. Note the flags: rotation, cards, weather, set plays, and schedule load.
| FA Cup | 1991–92 | Wrexham 2–1 Arsenal | 15.0 | 6.7 | 105 | 6.4 | set‑piece goal, late surge | FBref match page; market aggregate est. |
| FA Cup | 1971–72 | Hereford Utd 2–1 Newcastle Utd | 17.0 | 5.9 | 105 | 5.6 | muddy pitch, long‑range strike | FBref match page; market aggregate est. |
| FA Cup | 2014–15 | Bradford City 4–2 Chelsea | 17.0 | 5.9 | 104 | 5.7 | rotation, sub impact, late goals | FBref match page; market aggregate est. |
| FA Cup | 2016–17 | Lincoln City 1–0 Burnley | 10.0 | 10.0 | 104 | 9.6 | set‑piece goal, deep block | FBref match page; market aggregate est. |
| FA Cup | 2017–18 | Wigan Athletic 1–0 Man City | 13.0 | 7.7 | 105 | 7.3 | red card to favorite, direct play | FBref match page; market aggregate est. |
| DFB‑Pokal | 2020–21 | Holstein Kiel (p) 2–2 Bayern | 12.0 | 8.3 | 105 | 7.9 | snow, rotation, pens | FBref match page; market aggregate est. |
| DFB‑Pokal | 2020–21 | Rot‑Weiss Essen 2–1 Leverkusen (aet) | 9.5 | 10.5 | 104 | 10.1 | extra time, fatigue, set plays | FBref match page; market aggregate est. |
| Copa del Rey | 2009–10 | AD Alcorcón 4–0 Real Madrid | 21.0 | 4.8 | 106 | 4.5 | heavy rotation, small ground | FBref match page; market aggregate est. |
| Copa del Rey | 2011–12 | Mirandés 2–1 Villarreal | 9.0 | 11.1 | 105 | 10.6 | late goal, direct play | FBref match page; market aggregate est. |
| U.S. Open Cup | 2022 | Sacramento Republic 2–1 LA Galaxy | 5.5 | 18.2 | 104 | 17.5 | rotation, set play, travel | FBref match page; market aggregate est. |
Method note: implied probability = 1/decimal odds. Margin‑adjusted probability scales the raw percent by dividing by (overround/100). This assumes the book’s margin is spread in proportion across outcomes. Values are rounded to one decimal place.
When the market misprices
There is a known quirk in odds called the favorite–longshot bias. In some markets, the longshot is a bit too cheap or too dear, and the favorite is the other way. Cups can flip which side gets the “tax,” since lineups and stakes differ from league play. A review of research abstracts on favorite–longshot bias gives the broad map.
We also see cases where model priors miss cup context. Team strength models are great over long runs. In one-off cups, a small change (say, three backups in the back line) can swing the matchup far more than the model weight. For stories and data threads on upsets, see upsets analysis at FiveThirtyEight.
Small samples add risk. Some clubs just have one famous night. Some coaches lean hard into rotation one year and not the next. If you read implied probability as a hard promise, you will get burned.
Toolbox: build your own upset lens
Here is a simple path you can use for any cup tie.
- Collect prices near kick-off from more than one source. Take an average or a median line.
- Convert to implied probability. For decimal odds d, use 1/d. Keep the numbers in percent.
- Strip the margin. If the 1X2 sum is 105%, divide each leg by 1.05 so the sum is 100%.
- Log context: lineup changes, travel, rest days, weather, pitch size, likely plan on set plays, and press strength.
- Compare cups across countries. Host rules, draw rules, and tie formats matter a lot.
You can download historical results and odds from downloadable odds/data and build a small sheet. If you study risk sizing as a concept, the Kelly criterion shows why edge and bankroll must align. This is not advice to bet; it is a lens to read risk.
Editor’s note (for readers in Colombia): before you engage with any real‑money site, check the license and terms. If you need to compare online casinos in Colombia, use that as basic due diligence. We do not endorse any operator. Please act within your local law.
A few honest caveats
This piece explains probabilities and patterns. It is not betting advice, and cup upsets remain rare, high‑variance events. One match is noise‑heavy. Even a 10% event will happen, on average, one in ten times. That is the point, not the flaw.
If you or someone you know struggles with gambling, get help. See responsible gambling resources. Keep sport fun. Know your limits. Never chase.
What this means for fans, media, and models
For fans: when you hear “twenty-to-one,” hear “about five percent.” That small number is not zero. On cup nights, the crowd can make that five feel like fifty for a few minutes. Enjoy the ride, but read the base rate.
For media: say the percent, not just the price. Add plain context: rotation, cards, set plays. Explain that odds include margin. When a shock lands, point out the pattern, not fate.
For modelers: mark cups as a different task. Add flags for rest, rotation, and set‑piece weight. Keep priors light in early rounds. Do not overfit to one famous tie.
Quick FAQ
Are cup upsets more common than league upsets?
Yes. One-off ties, rotation, and pens raise variance. In leagues, strength shows across many games. In cups, a single break can decide it.
Do books underprice or overprice underdogs in cups?
It shifts by season and round. Bias can change with lineup news and crowd money. The key is to adjust for margin and track a large sample.
What’s the fastest way to convert odds to implied probability?
For decimal odds, use 1/odds. For other formats, see a short primer on Investopedia (search “implied probability”).
How should I talk about risk when I cite implied probabilities?
Say the percent and the context. Add that the number is a market estimate with a margin, not a sure forecast.
Sources, methods, update policy
Data sources: archived market lines from public screens; match facts from FBref; competition info from FA, DFB, RFEF, and U.S. Soccer sites linked above. Method: convert odds to implied probability (1/odds), then scale by overround so 1X2 sums to 100%. We used proportional scaling and rounded to one decimal place. For wider methods in sports analytics, the sports analytics methods posts at HSAC are a good read.
Limitations: some historic odds are sparse or vary by book; we used aggregates where possible. Context flags come from match reports and public logs.
Update policy: we revisit this page each season after the round of 16 in major cups, and after notable upsets. Last updated: June 2026.
Disclosure: External links open in a new tab. Some commercial resources may be marked “sponsored.” We do not offer betting advice.
About the author: Sports data analyst with 7+ years in football data, model building, and editorial review. Focus on clear methods and fair context.