Data pulled from Kalshi's public API on October 26, 2026. Every number below can be rebuilt from the same endpoint; the method is at the end.
DeskBot uses proprietary trading logic with customer-selected settings. Its dashboard shows supported market views and app activity. See the current product and support pages for the features and requirements of your release.
Kalshi's Bitcoin 15-minute market (series KXBTC15M) asks one question four times an hour: will Bitcoin's price at the end of this window be at least what it was at the start? We downloaded the last 3,000 settled markets, a little over a month of windows from September 25 to October 26, 2026, and counted what actually happened. How often did it settle Yes? Do results come in streaks? Are some hours of the day different? And how close are the calls?
Short version: in aggregate it behaves almost exactly like a coin flip, the streaks look like a coin flip's streaks, and the size of the move varies a lot by hour and by weekday. None of this is a trading signal on its own. It is a picture of what you are trading.
How a KXBTC15M market settles
Each market's rules, as published in the API, say it resolves Yes if the simple average of the sixty seconds of CF Benchmarks' Bitcoin Real Time Index (BRTI) before the window closes is at least the same sixty-second average before the window opened. The opening average becomes the market's "Target Price". For example, the 5:45–6:00 PM ET window on October 26 had a target of $79,989.64; the closing average was $79,968.93, so it settled No.
Two details matter for anyone automating this market:
- A tie counts as Yes. "At least" means equal is Yes. In our 3,000 markets there was no exact tie, so in practice it never came up.
- It is an average, not a last trade. A spike in the final second barely moves a sixty-second average. More on that in how the BRTI works and how Kalshi markets settle.
Result 1: exactly 50% Yes
Of 3,000 settled markets, 1,500 settled Yes and 1,500 settled No. That is a coincidence of this sample, not a law, but it tells you there was no built-in tilt toward up or down over the month. If someone tells you "the 15-minute market usually goes up", this sample does not support it.
Result 2: streaks are what randomness looks like
Many traders feel that after three or four No results in a row, a Yes is "due", or that a streak tends to keep going. We checked both directions.
- The chance that a window repeated the previous window's result was 49.0% (1,468 of 2,995 back-to-back pairs).
- After two identical results in a row, the next one matched 48.3% of the time. After three, 47.5%. After four, 49.3%.
- After five in a row, it matched 56% of the time, but that is only 166 cases, too few to separate from chance.
- The longest run was 16 No results in a row, ending at 1:45 PM ET on October 2 (four hours of Bitcoin closing each window below where it started).
Now compare the run lengths with what a fair coin would produce. With about 1,530 runs, a fair coin gives roughly half of them length 1, a quarter length 2, an eighth length 3, and so on:
| Run length | Seen in data | Fair coin, expected |
|---|---|---|
| 1 | 771 | ~766 |
| 2 | 389 | ~383 |
| 3 | 201 | ~191 |
| 4 | 99 | ~96 |
| 5 | 36 | ~48 |
| 6 or more | 36 | ~48 |
The match is close. A 16-long run looks shocking, but in a month of 15-minute windows a long run somewhere is normal. A rule like "fade every streak of four" has nothing in this data to stand on, and it costs fees every time it fires.
Result 3: most calls are close
The gap between the closing average and the target tells you how much room a position had. Across all 3,000 markets:
| Closing average within … of target | Share of markets |
|---|---|
| $10 | 11.9% |
| $25 | 25.8% |
| $50 | 44.4% |
| $100 | 69.8% |
| $200 | 88.7% |
The earlier developer-ledger example has been retired. A historical sample is not a current product performance record.
Result 4: the hour of day changes the size of the move
We grouped markets by the hour (Eastern time) in which they closed. Each hour has 111 to 128 markets.
| Closing hour (ET) | Markets | Settled Yes | Median move |
|---|---|---|---|
| 12–3 AM | 484 | 53.5% | $49 |
| 4–7 AM | 491 | 53.2% | $60 |
| 8 AM | 128 | 48.4% | $68 |
| 9 AM | 128 | 32.0% | $77 |
| 10 AM | 128 | 50.8% | $101 |
| 11 AM | 128 | 53.1% | $98 |
| 12–3 PM | 512 | 47.7% | $59 |
| 4–7 PM | 505 | 51.1% | $49 |
| 8–11 PM | 496 | 48.8% | $62 |
Two things stand out:
- The 10 AM and 11 AM hours had the biggest moves, a median near $100, against about $50–60 overnight and in the evening. That lines up with the US stock market opening at 9:30 AM ET, when more traders and more news hit at once. Bigger moves mean fewer coin-flip finishes and faster price swings inside the window.
- The 9 AM hour settled Yes only 32% of the time (41 of 128 markets). That is the largest outlier in the table. Be careful with it: we looked at 24 separate hours, and when you test that many groups one of them is likely to look unusual by chance. One month is also one market regime. Treat it as a question worth re-checking each month, not as a rule. If it does not show up again next month, it was noise.
Result 5: weekends are quieter
The median move on weekend markets (960 of them) was $36, about half the weekday median of $73 (2,040 markets). On a quiet weekend more windows finish within a few dollars of the target, so the result is closer to a coin toss right up to the last minute, and a late entry at a high price has less cushion.
What this means for fees and breakeven
If the market is close to a fair coin, the price you pay and the fee decide most of the outcome. The KXBTC15M series lists a quadratic fee with a multiplier of 1. Under Kalshi's published general fee schedule, a taker order pays about 0.07 × contracts × price × (1 − price), rounded up to the next cent. A worked example:
- Buy 10 Yes contracts at 50¢: cost $5.00. Fee: 0.07 × 10 × 0.50 × 0.50 = $0.175, rounded up to $0.18.
- Total paid $5.18 for a $10.00 payout if Yes. Breakeven win rate: 5.18 ÷ 10 = 51.8%.
- A single contract at 50¢ pays a $0.02 fee after rounding, so its breakeven is 52%.
In a market that settled Yes exactly 50% of the time, buying at 50¢ without any edge loses about 1.8–2 cents per contract to fees. A strategy has to be right more often than its breakeven, over a large sample, to be worth running. Full fee math is in Kalshi fees and breakeven win rate.
What this data does not tell you
- It does not show the prices traders paid during each window, only how each window ended. A 50/50 outcome rate says nothing about whether the price was fair at the moment you would have traded.
- It covers about one month. Volatility changes, and so do the hours that matter.
- It is not a strategy. "Bitcoin moved more at 10 AM" does not tell you which side to take.
Ethereum too: we ran the same analysis on 3,000 ETH 15-minute markets and matched them to these Bitcoin windows. ETH settled the same way as BTC 83.6% of the time. See Kalshi Ethereum 15-minute results.
How we got the numbers (rebuild it yourself)
Kalshi's market data endpoints are public; no account or API key is needed to read settled markets. We requested GET https://api.elections.kalshi.com/trade-api/v2/markets?series_ticker=KXBTC15M&status=settled&limit=1000 and followed the cursor for three pages, waiting between requests to stay well inside the rate limits. For each market we used result, close_time, floor_strike (the target) and expiration_value (the closing average). "Move" is the absolute difference between the last two. Hours are Eastern Daylight Time. The window ran from September 25, 2026, 4:15 AM ET to October 26, 2026, 6:00 PM ET, with four short gaps where a window was missing from the results. If you want to write the code, start with our Kalshi API Python tutorial and historical data guide.
How often does Kalshi's Bitcoin 15-minute market settle Yes?
In the 3,000 KXBTC15M markets that settled between September 25 and October 26, 2026, exactly 1,500 settled Yes (50.0%). Over that month there was no tilt toward up or down.
Do Kalshi 15-minute results come in streaks?
They come in streaks the way coin flips do. A window repeated the previous result 49.0% of the time, and the distribution of run lengths was close to what a fair coin produces. The longest run in the sample was 16 No results in a row.
How close are Kalshi Bitcoin 15-minute results?
Usually close. The median gap between the closing sixty-second BRTI average and the target was $58.81, and 25.8% of markets finished within $25 of the target.
What time of day does Bitcoin move most in Kalshi's 15-minute markets?
In this sample, markets closing in the 10 AM and 11 AM Eastern hours had the largest median moves, about $100, roughly double the overnight and evening hours. Weekend moves were about half the size of weekday moves.
Is there a winning pattern in this data?
Not one this data can show. The outcome rate was 50/50 and the streaks matched chance. After fees, a 50¢ Yes needs to win about 52% of the time just to break even. Trading can lose money.
What does a tie mean in a KXBTC15M market?
The rules say the market resolves Yes if the closing average is at least the opening average, so an exact tie settles Yes. No market in this sample tied.
Is DeskBot affiliated with Kalshi?
No. DeskBot is independent and is not affiliated with, endorsed by or sponsored by Kalshi.
Risk disclaimer: Trading involves risk of loss. Only trade money you can afford to lose. DeskBot does not guarantee profits, returns, win rates or income. Past results, including the data on this page, do not predict future results. Nothing on this page is financial advice. DeskBot is not affiliated with Kalshi.
Current product information: Kalshi · Robinhood · Setup and support. Check the relevant product page for current downloads, pricing, renewal and cancellation terms. Desktop and mobile products have separate availability and offers.
Educational content, not financial advice. Trading can lose money. DeskBot does not guarantee profits, returns, win rates or income and is not affiliated with or endorsed by the named exchanges or brokers.