Bitcoin Edge 15-Min
The countdown, how fast bitcoin is moving, and what the price on the board has meant in past windows — for every Kalshi 15-minute bitcoin window. No picks; the paper record fills up in public first.
The arithmetic
P(above) = Φ( ( ln(spot / strike) − σ²τ / 2 ) / (σ√τ) ), τ = seconds left / seconds in a year, σ = realized bitcoin vol (annualized)
Fee per contract = ⌈0.07 × p × (1 − p)⌉ cents, taken on each fill. Breakeven YES price is the highest p with p + fee(p) ≤ P(above). Distance to strike is ln(spot/strike) / (σ√τ), in sigmas.
Quick Answer
Bitcoin Edge 15-Min is a cockpit for Kalshi's KXBTC15M markets — one above/below contract per 15-minute window, 24/7, settling on the 60-second CF Benchmarks BRTI average. It shows the countdown with the settlement-averaging minute, how fast bitcoin is moving (a read of the move, not a call), and what the price on the board has meant in past windows at the same minute. It publishes no picks: across 4674 graded windows the model beats the market's own Brier in only 2 of 10 buckets, and two candidate rules are being paper-tested in public. Window open.
What this cockpit is for
A 15-minute bitcoin window is not a market you research; it is one you sit in front of. The useful things are the ones that change every second — how far spot sits from the locked strike in volatility units, how many seconds remain, whether the last-minute average has started, and what a YES contract has to cost for the arithmetic to work after Kalshi’s fee. That is what is on screen, every window, with no pick attached.
This hour’s call · top-of-hour momentum
Loading the bot’s decision…
The bot’s own decision, not a recomputation: at :55, if Bitcoin moved at least 0.036% over the last three minutes, it takes the side it moved toward at the ask. PLAY means it runs live at small size on our own Kalshi account; the rule itself is still in watch mode on the paper record below. Not advice — trade responsibly.
BTC +0.00% in 3 min — flat (under 0.015%).
1 min −0.01% · 3 min +0.00% · public spot, one-second
What this price has meant
Past KXBTC15M windows · history, not a forecastInside the final 60 seconds — the settlement average is being struck. History for this minute is not comparable.
Sizing arithmetic on the model’s own fair value — not a recommendation. The model has not cleared its gate against the market (scorecard below), so treat its fair value as a second opinion, not a call.
Paper record
Pre-registered 2026-09-23 · paper only · losers publishedWatch mode — not a pass yet.
Top-of-hour momentum has the best paper record we’ve run on Bitcoin 15-minute markets: 186 fills, +3.59¢ per contract after Kalshi’s fee. The sample is still small — 485 of the 500 top-of-hour windows we committed to before it counts, and the 90% lower bound is −0.77¢. Until then it stays labeled watch mode, and if it fails, we publish that too.
The rule: every hour, five minutes before the :00 window closes, if Bitcoin moved at least 0.036% over the last three minutes, take the side it moved toward at the asking price.
Two rules the history flagged as candidates, not findings. Maker fills are graded twice — touch (the book reached the limit) and trade-through (a trade printed through it, the conservative number the pass/fail uses). PASS needs +1.0¢ per contract net after fees, a positive 90% lower bound and both halves positive at 500. The paper ledger is the record; since October 10, 2026 the top-of-hour momentum rule also runs live at small size on our own Kalshi account ($3 max per position).
Model scorecard
not cleared · beats the market in 2 of 10 bucketsEvery window is scored against the Kalshi price the model was shown beside, one reading per minute left. A pick needs the model to beat the market’s own Brier in every populated bucket on a holdout, and it has not. The gate, bucket by bucket →
Go Pro · 26% off your first year
The full strike grid, priced in real time — $150/yr
Pro opens the full silver, gold, oil and bitcoin strike grids that price Kalshi contracts in real time, Thee Oracle, and edge alerts in Discord the moment a strike goes mispriced.
$111 for the first year, then $150/yr. Code applies automatically. Ends Dec 31. Monthly is $14.99/mo.
Every signal graded in public, losers left on the board. See the record →
Related Tools
Related Reads & Boards
Why the last 60 seconds are a different game
The contract does not settle on the price at the close; it settles on the average of the CF Benchmarks index across the final minute. Once that minute starts, every second of price you see is already part of the answer, and the remaining uncertainty is not “which way does bitcoin go” but “what does a partly-known average end up at.” The cockpit draws that band in gold and tells you when you are inside it. Trading through it is a choice; being surprised by it should not be.
The gate, in public
Every window is graded against the Kalshi price the model was shown beside. A pick would need the model to beat the market’s own Brier score, per minute-remaining bucket, after fees, on data it was not fitted to. Here is where that stands:
| Minutes left | Distance | Windows | Model Brier | Market Brier | Model − market |
|---|---|---|---|---|---|
| 10-15 | <0.5σ | 17249 | 0.2429 | 0.2388 | -0.0041 |
| 5-10 | <0.5σ | 8705 | 0.2368 | 0.2325 | -0.0043 |
| 1.5-5 | >1σ | 8539 | 0.0634 | 0.0573 | -0.0061 |
| 5-10 | >1σ | 7690 | 0.0801 | 0.0774 | -0.0027 |
| 5-10 | 0.5-1σ | 6883 | 0.1730 | 0.1688 | -0.0042 |
| 10-15 | 0.5-1σ | 4419 | 0.1904 | 0.1876 | -0.0027 |
| 1.5-5 | <0.5σ | 3684 | 0.2399 | 0.2295 | -0.0104 |
| 1.5-5 | 0.5-1σ | 3552 | 0.1711 | 0.1591 | -0.0119 |
| 10-15 | >1σ | 1614 | 0.1096 | 0.1101 | +0.0005 |
| 0-1.5 | >1σ | 40 | 0.0003 | 0.0061 | +0.0058 |
| 0-1.5 | 0.5-1σ | 4 | 0.0239 | 0.0047 | -0.0191 |
| 0-1.5 | <0.5σ | 1 | 0.4291 | 0.4290 | -0.0001 |
Lower Brier is better; a positive last column means the model beat the market in that bucket. Gate: not cleared — 2 of 10 buckets with 30+ windows beat the market · 4674 graded windows over 49.322916666666664 days · overall model − market -0.0077. Picks turn on only when every populated bucket is positive on a holdout, and then only as the Pro/Discord layer. Since 2026-09-23 each window counts once per minute left (the first reading in that minute, outside the final 60 seconds), over the last 90 days of readings.
The hourly tool is a different product on a different clock — options-implied probability against Kalshi’s hourly ladder. The two are kept apart on purpose: Bitcoin Edge (hourly) →
See it in your AI app
Frequently Asked Questions
How does a Kalshi 15-minute bitcoin contract settle?
Each KXBTC15M contract asks whether bitcoin is above a strike at the end of a 15-minute window. The strike is the reference price locked when the window opens. It settles on the CF Benchmarks Bitcoin Real-Time Index averaged over the final 60 seconds of the window — not on a single print, and not on a spot exchange price. That last minute is why the cockpit marks it as a separate band: while it runs, the price you see is already being averaged into the settle, and a position opened there is a coin flip on the average, not on the direction.
What does the fair value on this page mean?
It is the probability, from the distance between the live BRTI-constituent spot and the locked strike, the time left in the window, and realized bitcoin volatility, that the window settles above the strike. It is arithmetic once the strike is locked — there is no directional call in it. The Kalshi mid beside it is what the market thinks. When they differ by more than the fee band, the page shows the gap, and nothing more.
Why does this page publish no picks?
Because the model has not earned them. Every 15-minute window is graded in public against the Kalshi price it was shown beside, one reading per minute left, and the promotion gate is strict: the model must beat the market's own Brier score in every populated minute-remaining bucket, on a holdout. Today it beats the market in 2 of 10 buckets with 30 or more readings. The scoreboard is on this page; a pick would be a lie on top of it.
What is the fee-adjusted breakeven?
Kalshi charges a taker fee of 0.07 × price × (1 − price) per contract, rounded up per fill, which peaks near 1.75¢ at 50¢. The breakeven shown is the highest YES price at which buying YES still has non-negative expected value at the model's fair value once that fee is paid. Near a coin flip it takes about 3.5¢ of edge to get in and out — and most windows never offer that.
Is this the same as the hourly Bitcoin Edge?
No, and they are never combined into one number. The hourly tool prices Kalshi's KXBTCD strike ladder against the IBIT options chain — a cross-market comparison against a professional book. Nothing lists a fifteen-minute option, so the 15-minute rung needs a short-horizon model and pays a tax on the spread instead. Two engines, two clocks, two ledgers.
What does a 25¢ Bitcoin 15-minute price with 5 minutes left mean?
Across 178 past Kalshi KXBTC15M windows, a side priced 25–30¢ with five minutes left won 26.4% of the time, and buyers paid 27.2¢ on average — within a cent of what it returned. The market is calibrated here: a 25¢ contract with five minutes to go wins about as often as 25¢ implies. It is history, not a forecast, and the live readout on this page shows the same number for whatever price and minute are on the board (5-second readings).
Is momentum priced into Kalshi 15-minute Bitcoin markets?
Yes, almost to the cent. Buying the side bitcoin's 3-minute move points to wins more often as the move grows — 53.6% for the smallest quarter of moves, 71.5% for the largest — but the ask rises with it (53.7¢ to 70.8¢). After Kalshi's taker fee, following the move returned −1.55¢ per contract across 55,038 window-minutes; the largest moves netted −0.89¢. The velocity gauge on this page is a read of the move, not a call, for exactly that reason.
Should I use limit orders on Kalshi 15-minute markets?
A resting order saves the taker fee (Kalshi charges makers nothing on this series), but when it fills matters more than the fee. A side that first falls to a limit between 5¢ and 45¢ with seven or more minutes left went on to win 1.8 percentage points less often than its price implied, averaged across 28,670 first touches — you get filled because the price is still falling. Fresh drops to 15¢ or below in the last two minutes ran the other way: they won 1.8 points more often than their price across 3,336 first touches. We are paper-testing the one maker rule that survived, in public: 10¢ RUNNING (485 of 500 trade-through fills, −0.31¢ per contract so far); 15¢ retired early (248 trade-through fills, −2.50¢ per contract so far). Nothing here is a recommendation; the rule trades no real money unless it passes its pre-registered line.
15-minute late-flip alerts: when a Kalshi bitcoin window's price moves late, in your inbox — free
No spam. Unsubscribe anytime.