We found no sign of it in the public record. We pulled Kalshi's own settlement data for all 25,370 KXBTC15M windows from January 10 to October 4, 2026 and ran the checks a skeptic would. Every result matched Kalshi's published rule against its published numbers (one exception is a typo in the published strike, and that market settled correctly against the real one). Each window's strike equaled the previous window's settlement value to the cent in 24,906 of 24,910 cases. Up and Down split 49.87% to 50.13%. And the order-book price was an honest forecast: contracts won about as often as they cost.
Two things are real and worth knowing. Results are not perfectly independent: after an Up window the next one went Up 48.03% of the time, and after a Down window 51.69%, a gap well beyond chance. Kalshi's opening price already leaned the same way, by about half the gap, and the rest is smaller than the fee. And your chart is not the settlement source: Kalshi settles on a 60-second average of the CF Benchmarks BRTI, so a close window can settle against what Coinbase shows.
Kalshi is an exchange, so the question has a specific shape.
Kalshi is a CFTC-designated contract market. Its help center says: "On Kalshi, you are always trading against another member of the platform, not the exchange itself," and that there are "a number of market makers on Kalshi, primarily individuals and small institutions." So for these markets to be rigged in a way that costs you, one of a few things would have to be true: results that do not follow the published rule, strikes that are not what the rule says, a settlement source that drifts in one direction, or prices that systematically overstate your chance of winning. Each of those leaves a trace in public data, and Kalshi publishes every market's strike, settlement value and result through an unauthenticated API. We checked them.
Settlement value at least the strike means Yes. Otherwise No.
For the 24,943 windows where Kalshi publishes both a strike and a settlement value, the published result agreed with "settlement value is at least the strike" in 24,942. The single exception, the window closing at 11:15 AM ET on April 13, 2026, shows a strike of 7.243729 in the API, which is the real strike, $72,437.29, divided by 10,000. That window settled at $71,987.73, below the real strike, and resolved No, which is correct. The next window carries the same scale error and also settled correctly against its real strike. We treat both as display errors in the API record, not settlement errors.
The number to beat is not picked; it is the last window's settlement value.
Kalshi's rule sets each window's strike to the 60-second BRTI average before the window opens, which is the same 60 seconds that settle the previous window. In 24,906 of 24,910 consecutive pairs the strike equaled the previous window's settlement value to the cent. Of the four exceptions, two are the April 13 scale errors above, one differs by 7 cents (April 24, 2026) and one by $3.15 (May 19, 2026). We cannot tell from public data why those two differ.
One cosmetic glitch: on March 10 and 11, 2026, 96 markets displayed a template placeholder ("|| Count ||") in the rule text where the strike should be. The strike field itself was populated, and those markets settled normally.
No month leans further than you would expect by chance.
| Month (2026) | Windows | Settled Up | 95% range |
|---|---|---|---|
| January (from Jan 10) | 2,087 | 49.0% | 46.8 to 51.1 |
| February | 2,676 | 50.1% | 48.3 to 52.0 |
| March | 2,928 | 50.5% | 48.7 to 52.3 |
| April | 2,820 | 50.9% | 49.0 to 52.7 |
| May | 2,928 | 49.5% | 47.6 to 51.3 |
| June | 2,835 | 47.9% | 46.0 to 49.7 |
| July | 2,931 | 50.4% | 48.6 to 52.2 |
| August | 2,941 | 50.1% | 48.2 to 51.9 |
| September | 2,848 | 50.2% | 48.4 to 52.0 |
| October (to Oct 4) | 376 | 51.1% | 46.0 to 56.1 |
| All | 25,370 | 49.9% | 49.3 to 50.5 |
Over the whole period 49.87% of windows settled Up (95% range 49.26 to 50.49). The most lopsided month was June 2026 at 47.9% Up, a month when Bitcoin fell. Settlement was fast: the median window was marked settled 10.9 seconds after its close, and 74.9% within a minute.
This is the one pattern that shows up, and it runs the opposite way from the usual hunch.
| Before this window | Windows | This window matched the run | 95% range |
|---|---|---|---|
| Previous window Up: this one Up | 12,637 | 48.03% | 47.2 to 48.9 |
| Previous window Down: this one Down | 12,699 | 48.31% | 47.4 to 49.2 |
| 2 identical results in a row | 12,185 | 46.7% | 45.8 to 47.6 |
| 3 identical results in a row | 5,684 | 46.2% | 44.9 to 47.5 |
| 4 identical results in a row | 2,625 | 44.6% | 42.8 to 46.6 |
| 5 identical results in a row | 1,170 | 45.0% | 42.1 to 47.8 |
| 6 identical results in a row | 526 | 44.1% | 39.9 to 48.4 |
A fair coin would continue a streak half the time. These windows continued a streak less often, and the longer the streak, the more clearly: after four results in a row the fifth matched only 44.6% of the time (95% range 42.8 to 46.6). Across all consecutive pairs the difference between "Up after Up" (48.03%) and "Up after Down" (51.69%) is 5.83 standard errors, far beyond chance, and a runs test on the full sequence agrees (z = +6.03, more switches than a coin produces). It pointed the same way in both halves of the sample but was weaker in the second: a 5.3-point gap from January 10 to May 23, 2026 (z = -5.92) and a 2.1-point gap from May 24 to October 4 (z = -2.34).
We see no sign that this is manipulation: it shows up in the index values themselves, and it looks like a property of Bitcoin's 15-minute moves in this period. We cannot prove its cause. It also is not an opening for the reader: the market already prices most of it. On 18,395 windows from March 24, 2026 with Kalshi price history, Up won 48.33% after an Up window and 51.32% after a Down window, and the Yes midpoint one minute in was already 49.28¢ and 50.91¢. At the ask, Yes cost 51.44¢ after a Down window, more than the 51.32% it won. No cost 51.26¢ after an Up window and won 51.67%, 0.4¢ ahead before Kalshi's 2¢ single-contract fee. At a 50¢ price Kalshi's taker fee is 2¢ on a single contract, so a buyer of the side the pattern favors would need it to win about 52% of the time just to break even, and it won 51.69% after a Down window and 51.97% (100 minus 48.03) after an Up one. Treat "it's due" and "it's on a run" as equally weak reasons.
If prices were rigged against buyers, cheap and expensive contracts would both lose more than they cost.
Our settlement study grouped 15,090 logged asks about three minutes before the close by price. In most groups the side won within a point or two of what it cost, usually a little under, which is the spread a buyer pays. Our 99¢ study found the same at the top of the range: 99¢ contracts lost 0.99% of the time. Prices that are this close to reality are the opposite of a rigged book, and also the reason no price group in our data beat its own price after fees.
Two mechanics produce most "that can't be right" moments.
Public data has limits, and we would rather say so.
A measurement, not a recommendation.
Related: what happens at exactly the strike, the market by time of day, and fees and breakeven.
We found no sign of it in Kalshi's public record. Across 25,370 KXBTC15M windows from January 10 to October 4, 2026, every result matched Kalshi's published rule against its published strike and settlement value (one exception is a strike published 10,000 times too small, and that market settled correctly against the real strike), each strike equaled the previous window's settlement value in 24,906 of 24,910 cases, and 49.87% of windows settled Up.
Kalshi's help center says: "On Kalshi, you are always trading against another member of the platform, not the exchange itself." It says market makers on Kalshi are "primarily individuals and small institutions." Kalshi is a CFTC-designated contract market.
The CF Benchmarks Bitcoin Real Time Index (BRTI). The market resolves Yes if the simple average of 60 seconds of BRTI before the close is at least the same 60-second average before the open. Kalshi publishes both numbers and the result for every market through its public API.
Neither overall: 49.87% of 25,370 windows settled Up. But consecutive windows were not fully independent: after an Up window the next went Up 48.03% of the time, and after a Down window 51.69%. That gap is well beyond chance, but it is smaller than the 2¢ taker fee on a single 50¢ contract.
A little more likely than a coin flip, not much. After four identical results in a row, the next window matched only 44.6% of the time in our sample, and the effect was weaker from late May 2026 onward. It is a small lean, not a prediction you can count on.
Kalshi settles on a 60-second average of the BRTI, which is built from several exchanges. In our data a Coinbase/Kraken spot read at the close pointed the wrong way on 4.40% of windows overall and 39.3% of windows that finished within $5 of the strike.
Anyone can rerun these checks.
Every settled KXBTC15M market in Kalshi's public API (live and historical tiers) that closed from January 10, 2026, when both sides of the rule became 60-second averages, to October 4, 2026: 25,370 windows, read October 5, 2026. 2,451 earlier markets used a point-price strike and are left out.
Check 4's price comparison uses Kalshi's public one-minute candlesticks: the midpoint of the closing Yes bid and Yes ask at the end of each window's first minute, for windows from March 24, 2026 with price history.
Ranges are Wilson 95% intervals. The streak test compares the share of Up results after Up and after Down windows (two-proportion z), plus a Wald-Wolfowitz runs test on the contiguous sequence.
Citations welcome; the CSV above is the primary source and this page is the canonical writeup. Past performance does not guarantee future results.
The free, verifiable part comes first.
Every signal in our BTC 15-minute feed goes into a public results log and a downloadable CSV, graded against Kalshi's official result, losses included, with nothing pulled. Our free Telegram proof channel carries MLB and NFL calls; Bitcoin signals are graded in the log instead.
This is a free trial that becomes an automatically renewing subscription. A payment card is required to start, and you are not charged today. Your trial runs 30 days. At the end of it your card is charged $4.99 and then $4.99 every month until you cancel. Cancel any time before the trial ends and you pay nothing: use your account page, send /manage in the bot, or email [email protected]. Cancelling later stops future renewals at the end of the current paid month and does not refund it. No refunds except as required by law. By starting a trial you authorize these recurring charges and agree to the Terms and Privacy Policy. Signals are informational and for entertainment only, are not financial or betting advice, and any single signal can lose 100% of your stake. Must be 18+ (some platforms require 21+). If gambling is a problem for you, call 1-800-GAMBLER.
These are 15-minute binary prediction-market contracts. Any single window can lose 100% of your stake. They settle on a 60-second index average, not on the exchange chart you may be watching. A 99¢ contract can still lose 99¢. Nothing on this page is financial, investment, or betting advice.
Past performance does not guarantee future results. Trade only with money you can afford to lose. Must be 18+ or of legal age where you live. If gambling is a problem for you, call 1-800-GAMBLER.