Kalshi has started listing gold and silver contracts that live for fifteen minutes. KXGOLD15M and KXSILVER15M each open a single up-or-down market every quarter hour — 96 of them a day — under a new product certification covering 15-minute commodities.
They are not smaller versions of the weekly metals markets. The weeklies are a strike ladder: a dozen or more contracts at different prices on the same expiry, which is what makes an options-implied second opinion possible. These are one binary per window. There is no ladder, and at a fifteen-minute horizon there is no options chain to lean on either — no listed option expires that soon, so there is nothing to extract a fifteen-minute implied volatility from.
That sounds like it should make them unpriceable. It doesn't, and the reason is worth understanding.
The reference price locks, and that changes everything
Each contract resolves against the close of a 1-minute Pyth candle. Specifically: YES if the candle close at the window's end is at or above the candle close at the window's open.
The important consequence is that the second the window opens, the reference price is fixed and public. Kalshi shows it as the target price — $4,241.34 on a gold window, $62.144 on a silver one. From that moment, the contract is not a question about where gold is going. It is a question about whether the price finishes above a number you already know, with a clock you can read.
At the instant of the open, that genuinely is a coin flip. The reference is set to the current price, so up and down are symmetric and 50¢ is right.
Three minutes later it is not. If gold is $1.80 above the locked reference with eleven minutes left, the odds are no longer even — and they are not a matter of opinion. They follow from three things you can measure: how far the price sits from the reference, how long is left, and how much gold has actually been moving lately.
That is the whole calculation:
P(YES) = Φ( (ln(S / K) − σ²τ / 2) / (σ√τ) )
where S is the live price, K the locked reference, τ the time remaining, and σ realized volatility. There is no drift term. We are not forecasting the direction of gold over the next eleven minutes, and neither should anyone else — when we tested a momentum-driven directional model at the hourly horizon, it scored worse than the market's own prices. At fifteen minutes it would be worse still. Setting drift to zero is not a simplification we apologize for; it is what keeps the number a measurement instead of a guess.
We built Gold Edge 15-Min and Silver Edge 15-Min to show that number live for whichever window is currently open.
Pyth is the part that makes this honest
These contracts settle on Pyth's metals feeds — Metal.Index.GOLD/USD and Metal.Index.SILVER/USD, named directly in Kalshi's contract terms.
We already poll those exact feeds every ten seconds. For metals, our price source and Kalshi's settlement source are the same publisher, which means the fair value on those pages is computed against the number the contract will actually resolve on, not a proxy that usually tracks it.
That is not true of the 15-minute bitcoin market, which settles on a 60-second average of the CF Benchmarks Real-Time Index. It is a good index and the bitcoin contracts are fine products, but pricing them means modelling a feed you are not reading. Metals are, for now, the only 15-minute series where that gap closes completely.
One caveat worth stating plainly: settlement reads candle closes, while a live price feed is a stream of ticks. For watching a window unfold that distinction does not matter. For grading what actually happened after the fact, it does, and the candle is what counts.
The fee is the story
Here is the part that decides whether any of this is tradeable.
Kalshi charges a quadratic taker fee on these series: 0.07 × P × (1 − P) per contract. In cents:
| Price | Fee per contract |
|---|---|
| 50¢ | 1.75¢ |
| 60¢ | 1.68¢ |
| 70¢ | 1.47¢ |
| 80¢ | 1.12¢ |
| 90¢ | 0.63¢ |
The fee is largest exactly where these contracts spend most of their lives. Near 50¢, entering and exiting costs about 3.5¢ before you pay anything for the spread — and the books quote a few cents wide. Put together, the market has to be roughly two percentage points wrong, after the spread, before a round trip breaks even.
Ninety-six windows a day is a lot of chances to pay that. Trading every window is a reliable way to convert a small edge into a loss, and an excellent way to convert no edge into a large one. This is why both tools draw the cost as a dead zone around fair value rather than printing a signal: most of the time, the honest reading is that there is nothing there. If you want to check a specific entry, the Kalshi fee calculator does the arithmetic.
What the volume actually looks like
Sampling 200 settled windows in early August 2026:
| Series | Avg contracts/window | Median | Max |
|---|---|---|---|
| KXBTC15M | ~1,795,000 | ~1,763,000 | ~3,269,000 |
| KXGOLD15M | ~20,200 | ~19,700 | ~56,700 |
| KXSILVER15M | ~7,200 | ~5,900 | ~25,100 |
Gold runs on the order of one percent of bitcoin's 15-minute volume, and silver less than half of that. But twenty thousand contracts inside a fifteen-minute window is real retail participation, not a ghost market, and the books were quoted two-sided while we sampled.
These are new products, though, and day-three volume can be launch enthusiasm. Treat the table as a snapshot with a date on it rather than a settled fact, and re-check before sizing anything around it. Silver in particular is the thinner book of the two, and thin books are where the difference between a theoretical edge and a filled trade shows up.
How to actually use this
Do not treat a gap between fair value and the market as a trade. Treat it as a question.
The plausible edge in a product like this is not prediction — it is a lagging order book during a fast move in the underlying. That is a microstructure claim, and it is testable. We are logging every window where the book sits outside the cost band, and we will grade those against what actually settled. If they beat the market's own pricing net of fees over a meaningful sample, that becomes a product. If they don't, we will say so on the page and leave the tools as what they already are: an honest live read on what the current window is worth.
Until then, nothing here is a pick. There are no alerts, and none of this is scored in our track record.
If you want the version of these markets with a real strike ladder and an options-implied second opinion, that is the weekly rung — Gold Edge and Silver Edge. Different product, different question, and for most traders a more forgiving place to start than a contract that resolves before you have finished reading about it.
Trade responsibly. These are prediction-market contracts, not investment advice.