Momentum Strategies in Prediction Markets: What Actually Moves a Price

Momentum in a prediction market means buying a contract whose price has already risen, or selling one that has fallen, on the assumption that the move is not finished. It can work, but the mechanism is narrower than in equities: a probability tends to trend only when information arrives in sequence and the remaining traders are slow to process it. In most markets, on most days, that condition is not met.

The second thing to understand is that the payoff is capped. A stock can double; a contract trading at 90 cents can gain at most 10 cents before it becomes a dollar. Fees are charged on the share count and the price, and the spread is paid on entry and exit. Neither scales with the payoff, so the arithmetic of a momentum trade in a binary market is harsher than it looks.

What momentum means when the price is a probability

Every share pays $1 if the outcome occurs and $0 otherwise, which is why the price reads directly as an implied probability. A move from 0.40 to 0.46 is a six-point change in the market's forecast and a 15% gain on a $0.40 stake. The two descriptions are the same event, but they lead to different intuitions.

The asymmetry matters. At low prices, small absolute moves are large percentage moves: 0.02 to 0.03 is a 50% gain. At high prices, the percentage gain is small but it consumes a large share of what is left: 0.95 to 0.97 is a 2.1% price gain, yet two-fifths of the remaining upside. A momentum signal calibrated on percentage returns will therefore prefer longshots; one calibrated on probability points will prefer favorites. Neither is automatically the right frame.

Why a probability can trend

There are three sources of persistence, and only two of them are durable.

1. Sequential information. Vote counts reported over days, a trial with several witnesses, a hurricane forecast updated every six hours. Each update is a new shock in the same direction, so the price can drift toward the eventual outcome.
2. Slow diffusion. A niche market with few watchers can take hours or days to absorb public information. The price moves when someone finally trades on it.
3. Attention and forced flow. A market becomes visible, a crowd buys one side, and the price moves without new information. This is a liquidity event, and it tends to reverse.

Only the first two are a basis for a repeated strategy. The third is the one that produces the charts that look most like momentum.

Research on prediction markets generally finds that they aggregate public information well, but efficiency is a matter of degree and depends on who is watching. Wolfers and Zitzewitz describe prediction markets as unusually accurate forecast aggregators while noting the limits of that accuracy. That is the honest starting point: the average market is hard to beat, and the exceptions cluster in markets where attention is thin.

Catalysts are jumps, not trends

A scheduled event produces a step change, not a drift. If you buy before a court ruling, you own a binary option on the ruling. If you buy after it, most of the move is gone. Momentum needs a sequence of updates, not one.

So the first test before any momentum entry is simple: name the next information update and explain why it is more likely to favor your direction than the current price implies. If you cannot name it, you are trading a chart pattern in a market where the chart is mostly other people's limit orders.

Liquidity, fees and false breakouts

Polymarket runs a central limit order book. Prices come from users trading with each other, the displayed price is the midpoint of the best bid and ask, and large orders can move the price significantly even though the book has no fixed size limit. That last detail matters: if you buy 5,000 shares in a thin book, the "breakout" on your screen may be your own fill.

Fees compound the problem. Polymarket charges takers a fee of share count multiplied by a category rate multiplied by price multiplied by one minus price. For politics and finance the rate is 0.04, for crypto 0.07, for sports 0.05, and geopolitics is fee-free. Makers pay nothing. At 60 cents, 100 shares cost 100 × 0.04 × 0.60 × 0.40, or $0.96 per side. That is 1.6% of the $60 notional, and a round trip is roughly 3.2% before spread.

A two-cent move is not a trade at that cost. A six-to-eight-cent move might be.

A worked trade

Suppose a contract has drifted from 0.55 to 0.60 and you believe the next update pushes it to 0.72. You enter at 0.60, set invalidation at 0.55, and plan to exit at 0.72.

• Entry fee: 0.04 × 0.60 × 0.40 = 0.96 cents per share.
• Exit at 0.72: 0.04 × 0.72 × 0.28 = 0.81 cents.
• Exit at 0.55: 0.04 × 0.55 × 0.45 = 0.99 cents.
• Net reward: 12 cents minus 1.77 cents, or 10.23 cents.
• Net risk: 5 cents plus 1.95 cents, or 6.95 cents.

Reward-to-risk is about 1.47 to 1, so the break-even hit rate is 6.95 divided by 17.18, or 40.5%. If your signal is right 45% of the time, expected value is 0.45 × 10.23 minus 0.55 × 6.95, or about 0.78 cents per share before slippage. Half a cent of slippage wipes most of that out. This is the central problem with momentum in binary markets: the edge, if it exists, is thin enough that execution decides the outcome.

False breakouts

Three artifacts produce fake momentum.

The first is the display rule. The quoted price is the midpoint, but if the spread is wider than $0.10 the last traded price is shown instead. A book that widens can make the screen jump without any trade at the new level. The second is a single print. One market buy hitting a thin ask is not a trend; require the bid to move up and stay up. The third is news that does not touch the resolution rule, which can reprice a market without changing its true probability.

A useful confirmation rule: a breakout is real only if both sides of the book move and the new level holds for more than a few minutes. One-sided prints are noise.

Risk controls and when the edge disappears

• Define invalidation before entry. "The thesis is that the next update favors Yes; if the price falls to 0.55, the market disagrees by more than my estimate can explain."
• Use a time stop. If the catalyst passes and the price does not follow, exit. Do not convert a momentum trade into an unplanned hold-to-resolution bet.
• Size for a total loss. Binary outcomes gap. Risk 0.5% to 1% of bankroll per trade, not 5%.
• Never average down. A falling price after your catalyst is information, not a discount.
• Check exit depth. You need a bid for your size, not just an ask you can hit.

When the edge disappears

Momentum stops paying in deep, heavily watched markets, because the information is in the price within seconds and what remains is noise. It stops paying around single binary events, because there is no sequence to ride. It stops paying when fees and spread exceed the expected drift. And it stops paying when the move you are chasing was created by a display artifact or by your own order.

One caution about evidence: the published record on momentum specific to prediction markets is thin. Equity momentum is documented, and so are momentum crashes, but importing those results into a binary market is a hypothesis, not a finding. Test it on your own fills, net of fees, before you size it.

The realistic version of the strategy is small, catalyst-specific, and cost-aware. It is closer to trading the speed of information diffusion than to riding a trend.