Bitcoin 5- and 15-minute prediction markets: what resolves
How 5- and 15-minute Bitcoin up/down markets resolve using Chainlink TWAP, why your exchange chart can differ, and what fees and spreads do to short windows.
In this guide
The comparison you are actually buying
In a Bitcoin 5- or 15-minute up/down market you are not buying a forecast of where Bitcoin will go. You are buying one claim: whether a time-weighted average price (TWAP, an average of prices sampled across a period rather than a single last trade) measured at the end of a fixed window is greater than or equal to the TWAP measured at the start of that same window. Source documentation says Chainlink TWAP resolves Up/Down markets, using the same asset stream for both readings: final greater than or equal to start means UP, anything lower means DOWN.
That wording is the whole game. The market title alone cannot resolve anything, because the rules must name the source, the deadline and the edge cases. If the rules point at a different stream, a different window or a different equality direction than you assumed, the same Bitcoin move can settle the opposite way from what your chart suggested.
TWAP is not the candle close on your chart
TWAP averages across time, so the final reading is not identical to the last trade you see on an exchange chart. The documentation explicitly warns that different exchange candles can differ from the resolution stream. Over 5 or 15 minutes this gap is proportionally larger than over an hour, because a few seconds of prints move the average more.
This is why a pattern that looks clean on spot candles can fail live: you are settling against the market's own stream and window, not against the Bitcoin price in general. Read the specific market rules and confirm the stream and boundaries for that exact market instead of assuming every duration or source behaves the same way.
A worked illustrative example
Take an illustrative up/down market whose stated rule is final TWAP greater than or equal to start TWAP means UP. The start TWAP price to beat is exactly 100000. If the end TWAP is also exactly 100000, the stated equal-or-greater rule resolves UP. If the end TWAP is 99999, it resolves DOWN. The example exists to show how the equality direction and the 2 readings interact. It is not a claim about any live market's numbers.
Every evaluation of one of these markets follows the same logic: both readings on the market's own stream, the exact inequality written in the rules, and the exact window boundaries. Drop any of the 3 and you are guessing.
| Start TWAP | End TWAP | Stated equality rule | Outcome |
|---|---|---|---|
| 100000 | 100000 | final >= start is UP | UP |
| 100000 | 99999 | final >= start is UP | DOWN |
| 100000 | 100001 | final >= start is UP | UP |
What the displayed price and spread mean for your fill
The displayed price on the order book is usually the midpoint between best bid and ask. When the spread is wider than 0.10, the platform shows the last trade instead. In both cases you are looking at an observation, not an executable guarantee. What you actually pay depends on the resting quotes and depth at the moment your order arrives.
Short windows amplify this. Market orders consume several price levels, so slippage grows as the book thins. If the spread is wide, the gap between the midpoint you saw and the price you got can exceed any edge the resolution rule seemed to offer. For a slower, more structural view of how these venues are organised, the crypto prediction market guides cover the wider landscape.
Costs inside a 5- or 15-minute window
The documented fee is C times feeRate times p times (1 minus p), where C is the number of shares and p is the price. Makers are not charged; takers pay on fee-enabled markets. Documented category rates include crypto 0.07, sports 0.05, finance 0.04, politics 0.04, economics 0.05, culture 0.05, weather 0.05, general or other 0.05, mentions 0.04, tech 0.04 and geopolitics zero. Crypto, 100 shares at 0.40, gives 1.68, which is 4.2 percent of a 40 notional, not a flat 7 percent of cash. Sports, 100 shares at 0.50, gives 1.25. These are calculations from the current formula, and the market's own parameters should be checked before ordering.
That structure punishes frequent 2-sided taking in very short windows. Resting orders avoid the taker fee but add fill uncertainty, and external wallet or bridge intermediaries may charge separately. The documentation mentions USDC in one place and pUSD in newer product docs, so it is unsafe to claim one collateral asset across every product. A fee comparison across venues is only meaningful once you know which side of the book you will sit on; the prediction market fee guide walks through that distinction.
Resolution edge cases and payouts
Ordinary binary markets pay 1 per winning share and 0 per losing share. The resolution source notes that rare unknown or 50-50 resolutions can pay 0.50 to each side instead of cancelling or refunding. That is not a universal refund mechanism. Read the specific market rules to see how those rare paths are handled.
Up/down markets normally settle by comparing the 2 TWAP readings, but the general framework still applies to anything outside the normal path. If the rule text is silent on a scenario, that silence is itself part of your risk. Markets that resolve on a number rather than a direction raise a related set of questions, and the threshold markets explained covers how those boundaries are defined.
A decision sequence for one market
Start by opening the exact market and reading the rule text: source stream, window boundaries, equality direction, deadline. Compare that stream with the chart you normally watch and judge whether the difference matters at 5 or 15 minutes. Look at real book depth and spread rather than the midpoint alone. Compute your fee under the current formula for the size and price you are considering. The losing case is concrete: end TWAP one tick below start under an equal-or-greater rule, taker fee paid, slippage on entry, and a position that pays zero. If the rule text, the stream comparison, the book depth and spread, or the fee calculation changes your view of the edge, do not trade.
If you plan to automate the read-and-check loop, latency and batching choices matter, and the prediction market bot guide is the right next stop. Execution is only half the problem, though: none of these checks tells you where Bitcoin will be.
Limits of this reading
This is a rule-reading guide, not a price model. It makes no claim about continuously forecasting the coin price, about any latency advantage from batching or LLM-style processing, or about fills at displayed prices. The worked example is illustrative and not an empirical study.
The source documents are checked to a specific date, and their formulas, fee rates and resolution mechanics can change. Verify current documentation and the specific market parameters before acting. The habit that survives all of it is short: read the rule that resolves the market, price the cost of entering it, then decide.
Sources & verification
Sources checked
Sources checked
Sources checked
PolyZeno. Automated review with DeepSeek V4.1 Flash.