Bitcoin Up or Down statistics: what 8,000 rounds say about strategy
· Lucyd
Every Bitcoin Up or Down strategy rests on a few questions. Does Up win more often than Down? How far does the price usually move in a round? After three Ups in a row, is a fourth more or less likely? We took every settled round on Lucyd from October 7 to October 25, 2026, 8,362 rounds across six lengths, and counted.
Every round settles on the Pyth BTC/USD price: Up wins if the price at the end is strictly above the strike, the price at the start; otherwise Down wins. The rounds, prices and results are public, so you can check every number below with the API and the free data.
Up vs Down: a coin flip
| Round | Rounds | Up won | Down won | Up share |
|---|---|---|---|---|
| 5 minutes | 5,198 | 2,614 | 2,583 | 50.3% |
| 15 minutes | 1,736 | 859 | 877 | 49.5% |
| 30 minutes | 869 | 437 | 432 | 50.3% |
| 1 hour | 434 | 214 | 220 | 49.3% |
| 4 hours | 108 | 46 | 62 | 42.6% |
| 1 day | 17 | 8 | 9 | 47.1% |
Over short rounds Up and Down split almost exactly in half. The 4-hour and daily results lean toward Down, but with 108 and 17 rounds that is the direction Bitcoin took in these weeks, not a rule. One 5-minute round ended exactly at the strike and went to Down; no round in the period was voided.
What it means: buying one side blindly is a 50/50 bet. Any edge has to come from information the price does not yet reflect, or from buying below fair value.
How far Bitcoin moves in a round
The move is the change from the strike to the settlement price, ignoring direction. With Bitcoin near $81,000:
| Round | Median move | Average move | 9 rounds in 10 move less than | Moved under 0.01% |
|---|---|---|---|---|
| 5 minutes | 0.044% (about $36) | 0.065% | 0.145% | 16.4% of rounds |
| 15 minutes | 0.073% (about $59) | 0.109% | 0.250% | 7.5% |
| 30 minutes | 0.103% (about $83) | 0.156% | 0.370% | 5.9% |
| 1 hour | 0.140% (about $113) | 0.222% | 0.504% | 3.5% |
| 4 hours | 0.290% (about $235) | 0.462% | 1.089% | 0% |
| 1 day | 0.928% (about $750) | 1.132% | 2.308% | 0% |
Half of all 5-minute rounds end within about $36 of the strike, and one in six ends within $8. At that distance the result is decided by the last few prints before the end time.
The move grows roughly with the square root of time, as random price paths do: 15 minutes is three times 5 minutes, and the median move is about 1.7 times larger (√3 ≈ 1.73). That is the basis of the simplest fair-value model for these markets, below.
Streaks and the last round
| Round | Same winner as the round before | After 3 in a row, the 4th the same | Longest streak |
|---|---|---|---|
| 5 minutes | 48.7% | 46.6% | 12 |
| 15 minutes | 48.9% | 43.4% | 9 |
| 30 minutes | 47.0% | 50.7% | 12 |
| 1 hour | 46.9% | 43.2% | 7 |
The previous round tells you almost nothing about the next. If anything, results flip slightly more often than they repeat, but the difference is within what chance gives on samples this size. Streaks of 9 to 12 look remarkable on a chart and are expected over thousands of rounds: with a fair coin, a streak of 12 somewhere in 5,000 flips is likely.
What it means: "it has been Up five times, Down is due" and "follow the streak" are both 50/50 bets.
Where an edge can come from
If direction is a coin flip on average, the money in short rounds is made in the gap between the token price and the real chance at that moment.
Time and distance from the strike. Halfway through a 5-minute round with Bitcoin $40 above the strike, Up is no longer a 50/50: there are 150 seconds left and the typical move in that time is smaller than $40. A simple model of that chance:
P(Up) ≈ N( ln(price / strike) / (σ × √(seconds left / 300)) )where N is the normal distribution and σ is the typical 5-minute move. From the data above, σ is about 0.065% (the median move divided by 0.6745). When the Up token trades well below this estimate, it is cheap; well above, it is expensive. The bot guide turns this into code.
Speed. The settlement price comes from Pyth, which aggregates the major exchanges. In the last seconds of a round, the trader who sees the price first and has orders in the book wins the race.
Making, not taking. A maker who quotes both sides around fair value earns the spread from traders who cross it. On Lucyd there are no trading fees for either side, so a small edge is not eaten by costs; see how fees change the maths elsewhere.
Get the data
- Rounds, strikes, settlement prices and winners:
GET https://api.testnet.lucyd.trade/markets?state=settled&period=5m, API docs. - Every trade, with CSV export:
GET /trades?format=csv. - The full L2 order book of every round, as Parquet: order book data.
The statistics cover Lucyd testnet rounds settled on the Pyth BTC/USD price; the price moves are real Bitcoin moves, the same as on any venue that settles on the same feed. Trade the 5-minute rounds.