Michigan's Upset and the Limits of Election Market Wisdom
Abdul El-Sayed's projected Michigan primary win jolted Democrats. What a favorite-beating primary result says about how event contracts price low-turnout races.
Abdul El-Sayed, the progressive former Detroit health director, won Michigan’s Democratic primary, according to US media projections reported by Reuters and carried widely on Sunday. Reuters described the result as a jolt to Democrats — the language outlets reserve for outcomes their own pre-election reporting did not lead readers to expect.
That is the whole of what the source material establishes: a projected winner, a party surprised. It is also enough to make the point that matters for anyone holding event contracts on primaries. Primaries are the part of the election calendar where prediction markets are least reliable, and the reasons are structural rather than accidental.
Why primary markets misprice more often than general-election markets
A general election in a swing state gets polled dozens of times by firms with published methodologies and long track records. Markets on those races have something to anchor to. When the anchor is good, contract prices converge on it fast, and the market’s job is mostly to adjust at the margins for late news.
Primaries have almost none of that. Polling is sparse, often commissioned by campaigns, and built on turnout models that guess at who shows up in an election where 15 or 20 percent participation can be normal. A model that is wrong about the composition of the electorate is not slightly wrong. It is wrong in a way that can flip the result.
Add the other primary-specific problems:
- Name recognition masks softness. An establishment candidate polls ahead because more voters have heard of them, not because more voters prefer them. That lead evaporates as the challenger’s ads land.
- Late deciders break as a bloc. In a primary there is no partisan floor to catch a collapsing candidate. Everyone on the ballot is in the same party.
- Thin markets amplify the anchor. A contract with modest volume tends to sit where the last confident buyer left it. Nobody is paid to fight a price that looks reasonable.
The result is a familiar pattern: the favorite trades at 80 cents for weeks on evidence that would not justify 60, and then settles at zero.
Reading a “jolt” correctly
There is a temptation after a result like this to conclude that markets are broken, or that the crowd is dumb. That is the wrong lesson and an expensive one.
A contract at 80 cents is not a prediction that the favorite wins. It is a claim that the favorite wins about four times in five. Upsets are supposed to happen at that price, and if they never did, the price would be wrong in the other direction. Nothing about a single El-Sayed win tells you the market was mispriced — you would need to know where the contract actually traded, and the source material here does not say.
What the result does justify is a narrower question: in the class of races that look like this one — low-turnout primary, insurgent versus establishment, sparse public polling — do favorites at high prices win as often as those prices imply? That is a question about a sample, not an anecdote, and it is the only version worth acting on.
If you are new to this, the mechanics are simple. An event contract pays $1 if the stated outcome happens and $0 if it does not. The price, between zero and one, is the market’s implied probability. Buy at 0.20 and you are risking 20 cents to make 80 — a good trade if the true chance is better than one in five, a bad one if it isn’t. Our field guide to reading Polymarket odds as probabilities walks through the conversions.
What the underdog side actually pays
The arithmetic explains why contested primaries attract contrarian money even when the contrarian is usually wrong. The table below is generic — no prices from this race are in the source material — but it shows the shape of the bet.
| Underdog price | Implied chance | Payout per $1 risked | Break-even hit rate |
|---|---|---|---|
| $0.10 | 10% | $9.00 | 1 in 10 |
| $0.20 | 20% | $4.00 | 1 in 5 |
| $0.35 | 35% | $1.86 | 1 in 2.9 |
| $0.50 | 50% | $1.00 | 1 in 2 |
At a dime, you need to be right once in ten tries. If the market systematically shades insurgent candidates down — and the “jolt” framing in the coverage suggests conventional wisdom did, whatever the contracts said — that gap is where the edge lives. It is also where position sizing stops being optional; the Kelly calculator will tell you how much of a bankroll a low-probability, high-payout edge actually justifies, and the answer is usually smaller than instinct suggests.
The downstream markets
A projected primary winner does not resolve only the primary contract. It reprices everything conditional on it: general-election markets for the seat, party-control markets that aggregate seats, and any contract on the ideological composition of a caucus. Those secondary markets often move slower than the primary itself, because the traders who follow one are not always the traders who follow the other.
That lag is the practical takeaway. The primary contract settles within hours of a call. The general-election contract that just changed candidates may take a day to find its new level, and during that day the price reflects a race that no longer exists.
What to check before the next one
Three things, in order:
- Whether the market’s price is anchored to public polling or to something better. If it moves only when a poll drops, it knows what the polls know and nothing more.
- Whether volume is deep enough that the price represents disagreement rather than inertia.
- Whether the same contract is priced differently across venues — a spread that widens on primary night is often just one book updating faster than the other. The arbitrage checker surfaces those gaps.
None of this makes primaries predictable. It makes them a place where the market’s confidence and the market’s information can drift apart, which is a different and more tradeable condition. FluxrBot watches for exactly that drift across venues so you do not have to refresh two order books at once.