Soccer semifinals’ implications
What can we infer about a tournament’s final from its semifinals’ odds?
If the odds for both semifinals of a soccer tournament are known, we can estimate each team’s probability of reaching the final and, from there, the probability of every possible final pairing. The estimate requires two simplifying assumptions — and, as we will see, the important one can be put to the test. The 2026 World Cup semifinals will accompany us from the first step to the last: a real case, worked from its actual odds.
The odds we start from
On July 13, 2026, one day before the first semifinal, the School of Odds’ default IQM model1 stated the following probabilities for France against Spain:
| Regulation result | Probability |
|---|---|
| France wins | 40.64% |
| Draw | 29.46% |
| Spain wins | 29.91% |
and for England against Argentina:
| Regulation result | Probability |
|---|---|
| England wins | 36.36% |
| Draw | 32.03% |
| Argentina wins | 31.61% |
Six numbers, and at first sight they speak only about the two matches themselves: who wins during regulation, or whether the ninety minutes end level. Our question is what these six numbers can tell us about the match after them — the final.
What the numbers do and do not say
The 40.64% is not France’s probability of reaching the final. It is France’s probability of winning during regulation. But a semifinal cannot ultimately end in a draw: if the match stands level after ninety minutes, extra time and, if needed, a penalty shootout will still send one team through. France can therefore reach the final along two routes — by winning the match outright, or through the drawn branch, the 29.46%, by prevailing afterward. The same holds for Spain.
So the ordinary three-way market answers a different question from ours. It prices the match at the end of regulation; we want the probability of qualification. Some bookmakers price qualification directly, in a market usually called “to qualify”, and when such a market exists it is the better source. But suppose all we have are the ordinary win–draw–win odds — the most common situation. The qualification probabilities can then be inferred, and the inference is worth learning: it is simple, it works wherever a knockout tie is decided in a single match2, and it lands close to what the direct markets themselves say.
Removing the draw
The bridge is one clean idea: set the draw aside and compare the two teams’ winning probabilities directly. In general terms, for a semifinal between teams A and B:
w(A): probability that A wins during regulation
w(B): probability that B wins during regulation
q(A): inferred probability that A qualifies for the final
q(A) = w(A) / (w(A) + w(B))
and symmetrically for B, so that the two figures add up to 100%. Applied to France against Spain:
q(France) = 40.64 / (40.64 + 29.91) = 40.64 / 70.55 ≈ 57.60%
q(Spain) = 29.91 / 70.55 ≈ 42.40%
And to England against Argentina:
q(England) = 36.36 / (36.36 + 31.61) = 36.36 / 67.97 ≈ 53.49%
q(Argentina) = 31.61 / 67.97 ≈ 46.51%
What happened to the draw? It has not been ignored — it has been distributed. Dividing by the sum of the two win probabilities implicitly splits the drawn branch between the teams in proportion to their strength during regulation: France, the stronger side over the ninety minutes, silently receives the larger share of that 29.46%, and Spain the smaller.
That is also the method’s first and central assumption, and it deserves to be stated plainly: we are assuming that the balance between the teams after a regulation draw — in extra time, in a shootout — remains broadly similar to their balance during regulation. Reality can deviate; a team may be unusually strong in penalty shootouts, and the regulation market cannot see it. The figures above are therefore inferred qualification probabilities, not observed ones. We will test how well they hold up in a moment. First, let us collect them:
| Team | Inferred probability of reaching the final |
|---|---|
| France | 57.60% |
| Spain | 42.40% |
| England | 53.49% |
| Argentina | 46.51% |
France is the clearest favorite to qualify, though Spain retains a little more than a two-in-five chance. England against Argentina is closer: England leads, but narrowly.
The four possible finals
Two semifinals, two qualifiers, four possible finals: France–England, France–Argentina, Spain–England, Spain–Argentina. The two matches are played between different teams, so their results can be treated as independent — this is the method’s second assumption, milder than the first — and the probability that two independent events both happen is the product of their probabilities:
P(France–England final) = q(France) · q(England) = 57.6045% · 53.4942% ≈ 30.82%3
Repeating the multiplication for the other three pairings:
| Possible final | Inferred probability |
|---|---|
| France–England | 30.82% |
| France–Argentina | 26.79% |
| Spain–England | 22.68% |
| Spain–Argentina | 19.72% |
The four pairings cover every possible outcome, so the probabilities add up to 100%, apart from rounding. And each of them quietly prices a second match: the two losers meet in the third-place match, so a France–England final implies a Spain–Argentina third-place game, with the same 30.82% probability.
France–England is the most probable final — and it is still far from likely. About 31% means that the other three pairings together hold the remaining 69%. The most probable item in a set can be less probable than all the alternatives combined, and whenever a “most likely final” is announced anywhere, that distinction is worth remembering: most probable does not mean probable.
Putting the inference to the test
The qualification figures rest on the central assumption about the drawn branch, and we do not have to take it on faith: on the same date, the bookmaker Svenska Spel4 was pricing the qualification question directly.
| Team | Decimal odds to qualify |
|---|---|
| France | 1.68 |
| Spain | 2.15 |
| England | 1.80 |
| Argentina | 1.98 |
A decimal odd implies a probability of 1 divided by the odd — but a bookmaker’s prices also contain its margin, its built-in profit. For France against Spain: 1 / 1.68 ≈ 59.52% and 1 / 2.15 ≈ 46.51%, which add up to about 106.04% rather than 100%; the excess is the margin. Dividing each figure by that total removes the margin proportionally:
q(France) = 59.5238 / 106.0354 ≈ 56.14%
q(Spain) = 46.5116 / 106.0354 ≈ 43.86%
The same procedure for England against Argentina (1 / 1.80 ≈ 55.5556% and 1 / 1.98 ≈ 50.5051%, a total of about 106.0606%) gives England ≈52.38% and Argentina ≈47.62%. Now the two columns can face each other:
| Team | Inferred from semifinal odds | Svenska Spel direct market | Difference3 |
|---|---|---|---|
| France | 57.60% | 56.14% | 1.47 points |
| Spain | 42.40% | 43.86% | 1.47 points |
| England | 53.49% | 52.38% | 1.11 points |
| Argentina | 46.51% | 47.62% | 1.11 points |
Every inferred probability lands within a point and a half of the market that prices qualification directly. Two different routes, starting from different sources, arrive at nearly the same answer: the inference tracks the direct market closely.5
Recalculated from Svenska Spel’s qualification probabilities, the four possible finals tell the same story:
| Possible final | From semifinal odds | From the direct qualification market |
|---|---|---|
| France–England | 30.82% | 29.40% |
| France–Argentina | 26.79% | 26.73% |
| Spain–England | 22.68% | 22.98% |
| Spain–Argentina | 19.72% | 20.89% |
Same ranking, and no estimate strays by more than a point and a half.
What the semifinal odds cannot tell us
Each semifinal market compares only the two teams playing in it. France–Spain measures France against Spain; England–Argentina measures England against Argentina. Neither says anything about how France compares with England or with Argentina, or Spain with either — and that missing comparison is precisely what deciding the final requires.
From the two semifinal markets alone, then, we can estimate each team’s probability of reaching the final, its probability of playing the third-place match instead, and the probability of every final and third-place pairing. We cannot estimate each team’s probability of winning the tournament. That requires information from elsewhere: direct tournament-winner odds, hypothetical prices for each possible final, or a model of team strength.
Svenska Spel’s tournament-winner prices, from the same day, show the missing layer at work. Their raw implied probabilities add up to about 108.11%; removing the margin proportionally, as before:
| Team | Decimal odds to win the tournament | Margin-free probability |
|---|---|---|
| France | 2.50 | 37.00% |
| Spain | 4.25 | 21.76% |
| England | 4.25 | 21.76% |
| Argentina | 4.75 | 19.47% |
Look at England and Spain. In Svenska Spel’s own qualification market, England is the likelier finalist — ≈52.38% against Spain’s ≈43.86% — yet the same house prices both teams identically at 4.25 to win the tournament, a margin-free ≈21.76% each. If England reaches the final more often but wins it no more often, the market must believe that Spain, once there, is the more likely winner. The division makes it concrete: winning divided by reaching gives England a ≈41.5% chance of winning a final it reaches, and Spain ≈49.6%. Reaching the final and winning it are different questions, and the second needs information that the two semifinal markets do not contain.
The method in short
— Take each semifinal’s margin-free win–draw–win probabilities.
— Remove the draw and renormalize: each team’s qualification probability is its regulation-win probability divided by the sum of the two win probabilities.
— Multiply one qualification probability from each semifinal to price each possible final — and, by complement, each third-place pairing.
— Prefer a direct “to qualify” market when one exists; the inference substitutes for it when it does not, and can be tested against it when both are available.
In the 2026 World Cup, the inference landed within a point and a half of a direct qualification market — and when the check was repeated inside a single bookmaker, its own three-way odds against its own qualification prices, the gaps shrank to about a point at most. What semifinal odds cannot do, however carefully they are read, is name the champion. They tell us who is favored to reach the final, not who wins it.
Notes
- As displayed on odds.school on July 13, 2026, at 21:30 UTC. At that moment the site was tracking prices from 67 sportsbooks; the model analyzes a vetted subset, removes each bookmaker’s margin, and combines the estimates through the interquartile mean (IQM) — an average that discards the highest and lowest quarters of the estimates, so that a few extreme values cannot dominate the result. The snapshots are preserved on the two linked match pages. ↩
- And not only in soccer: the method fits any sport whose ordinary market prices a draw that the knockout format must eventually resolve — ice hockey, for example. Soccer is simply where the situation arises most often. ↩
- The calculations carry the unrounded values throughout, which is why more decimals appear here; repeating them with the rounded figures shown in the tables may shift the last decimal. ↩
- Svenska Spel serves as the benchmark for its standing in the School of Odds’ public calibration ranking: as of July 13, 2026, its odds sampled 24 hours before kickoff — close to the horizon of these snapshots — ranked among the sharpest for soccer, both over the last three months (Brier skill score 0.0815 across 1,205 matches) and during the last month (0.1524 across 69 matches), which is the World Cup’s own period. The prices were recorded on July 13. ↩
- A stricter version of the check keeps everything inside one house. Svenska Spel’s own three-way odds that day — France 2.45 / draw 3.30 / Spain 3.15, and England 2.70 / draw 2.90 / Argentina 3.10 — put through the same draw-removal, give France 56.25% and England 53.45%: within 0.11 and 1.07 points of the same bookmaker’s direct qualification prices. The residual there cannot come from disagreement between sources; it measures the draw-splitting assumption itself. ↩