Gold Edge 15-Min
Kalshi lists a fresh gold contract every fifteen minutes. Once the window opens the strike is locked — so we can tell you what it is actually worth, and what it costs to act.
Once a window opens, its reference price is locked. From that moment the contract's fair value is arithmetic rather than a forecast:
- S — the live gold price: a public interbank spot quote our engine reads and scores against every Kalshi settlement (see the feed note below), refreshed on this page every ten seconds. The price itself stays inside the engine; the page shows Kalshi's own reference.
- K — the reference price locked at the window's open (for example $4,241.34).
- τ — time left in the window.
- σ — how much gold has actually been moving, measured over the last fifteen minutes.
There is no drift term. We are not claiming to know which way gold goes next — at this horizon nobody credibly does, and our own hourly model tested worse than the market's own pricing. Setting drift to zero is the honest choice, and it is what makes the number above a measurement rather than an opinion.
FREE · LIVE EVERY 15 MINUTES
Live fair value and the true cost of trading — KXGOLD15M, every 15 minutes
What is this?
Kalshi lists a brand-new gold contract every fifteen minutes — 96 of them a day on weekdays, dark on weekends. Each one asks a single question: is gold higher at the end of this window than it was at the start? It resolves YES if the Pyth 1-minute candle close at the window's close is at or above the candle close at the window's open. The moment the window opens, that opening price is locked as the reference (for example $4,241.34 — Kalshi rounds gold to 2 decimals).
That lock is what makes this tool possible. Before the window opens there is no strike, so there is no fair value — only a coin flip. After it opens, fair value stops being a forecast and becomes arithmetic: how far is the live price from the locked reference, how much time is left, and how much has gold actually been moving? We compute it with zero drift — no directional call, because at a fifteen-minute horizon nobody credibly has one and our own hourly model tested worse than the market's own pricing. The tool then draws the round-trip fee band on top, so you can see whether any gap is even big enough to act on.
Real-World Example
→ A Live Window
A window opens and locks its reference at $4,241.34. Six minutes in, gold is trading at $4,243.10 — a hair above. Nine minutes left. Realized volatility over the trailing fifteen minutes annualizes to about 18%.
Run those four numbers through the model and fair value for YES comes out at roughly 71¢. The Kalshi book is quoting a mid of 65¢. That is a +6pp gap — YES looks cheap.
Now the part everyone skips. Kalshi's taker fee on these series is 0.07 × P × (1 − P) per contract — about 1.5¢ each way at these prices, so ~3pp for a round trip, plus roughly 2pp of spread. The dead zone is about 5pp wide. Your +6pp gap clears it — barely. Most windows do not, which is exactly why the tool draws the band instead of manufacturing a signal every fifteen minutes.
✅Action: Read the gauge, not the gut. If the book sits inside the shaded fee band — which is where it sits most of the time — the honest answer is no trade, and churning 96 windows a day compounds fees faster than almost anyone expects. Act only when the book drifts clearly outside the band with time still on the clock. The page also publishes the hour-by-hour base rate of gold windows finishing up, which is the number to check before you assume any hour is special.
Bottom line: Free, no account, live every fifteen minutes. We are logging our own flags in the open to see whether they beat the market — and we will publish that answer either way. Until it clears a real bar, nothing here is a pick.
Full guide →Beta — recalibrating in public
Two things changed underneath this tool in September. Kalshi replaced the weekly metals contracts with daily ones, and Pyth moved its historical price bars behind a paywall — which killed the tooling we used to check our feed against Kalshi's settlement print. So we are rebuilding the evidence from zero, in the open.
Where it stands, measured across 2,086 settled windows per metal: our number does not beat the Kalshi book. The book is fractionally sharper both at window open and five minutes in, and the side we flag wins about 55% of the time while still losing money — roughly a third of a cent per contract on gold, over a cent on silver, and that is measured at the mid, which nobody can actually trade at.
We have now started recording the price you would really pay rather than the mid, because that is the number that decides whether any of this is real. The tool is live and free and the maths is the same maths. We are simply not going to tell you to trade it until the record says you should — and when it does, or does not, that record will be on this page.
Quick Answer
Gold Edge 15-Min prices Kalshi's KXGOLD15M markets — one up-or-down contract per 15-minute window, settling on Pyth 1-minute candle closes. Once a window opens its reference price is locked, so fair value becomes arithmetic: we compute it live from the distance to that strike, the time left, and realized volatility, with no directional call. We also draw the round-trip fee band, because near 50¢ it costs about 3.5¢ to get in and out and most windows offer less edge than that. Window open.
Current Gold window
Closes in 6:09
MODEL READS UP
The model puts 67% on gold closing above $4,042.07, Kalshi's reference — it is above the line with 6:09 left. Kalshi has YES at 61¢.
A read of the model, not a call: following this arrow lost 1.39¢ a contract after fees across 2,925 windows (at the mid, through Sep 23, 2026).
Closes below → NO | YES ← closes above← NO | YES →
Does Kalshi’s price agree with the math?
Kalshi is asking (YES)
61¢
60–62¢ · book as of 0s ago
Our math says (YES)
67¢
fair value · drift 0
Kalshi is asking 5.9¢ less than the math says YES is worth — outside the 4.1¢ round-trip cost band. That is worth a look, not a trade signal: we are logging these to see whether they beat the book over time, and we will publish the answer either way.
Go Pro · 26% off your first year
Every edge engine on the site, live — $150/yr
Pro opens every engine on the site — the full silver, gold, oil and bitcoin strike grids that price commodity contracts in real time, Thee Oracle, Theta Edge, and the full cross-venue price-gap board.
$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 →
This window
- YES bid / ask
- 60¢ / 62¢
- Contracts traded
- 63,915
- Open interest
- 29,147
- Round-trip fee at fair
- 3.10¢
Next window
Opens at 6:45 AM ET. Its reference price locks at that moment — until then there is no strike to price against, which is why the card above only exists once a window is live.
Trade gold windows on KalshiRelated Tools
The fee is the whole story
Kalshi charges a quadratic taker fee on these markets — 0.07 × P × (1 − P) per contract. That works out to about 1.75¢ at 50¢, 1.47¢ at 70¢, and 0.63¢ at 90¢. Near the middle, where these contracts spend most of their lives, a round trip costs roughly 3.5¢ before the spread. Add a typical few-cent spread and the market has to be around two percentage points wrong just for you to break even.
That is why the gauge above draws a dead zone instead of a signal. Ninety-six windows a dayis a lot of opportunities to pay a fee, and churning every window compounds the cost far faster than most people expect. If you want to see what a given entry actually costs, the Kalshi fee calculator does the arithmetic.
Why we can price these honestly
These contracts settle on Pyth's 1-minute gold candles. That index is gated — no free feed can read it — so the engine prices on a public interbank spot quote and we score that quote against every settlement Kalshi publishes (the 15-minute bitcoin markets settle through CF Benchmarks instead). The page refreshes from our engine every ten seconds, so the gauge keeps moving between server refreshes; the order book comes with it and is labelled with its own age so you always know how fresh each half is.
On the feed, because nobody else states it. Kalshi's settlement source for this series is listed as Metal.Index.GOLD/USD, a gated Pyth feed that is not in Pyth's free catalogue. What we price on is a public interbank spot quote for gold.
So rather than assume those agree, we measured. Kalshi publishes the actual settlement print on every settled window; before switching to this feed on September 23, 2026 we compared interbank spot against 244 of them and counted how often it called the same up-or-down result: 100% of the time. That is the number that decides a contract. Every settlement since is scored automatically, and an alert fires if agreement falls below 95% — scripts/check-metals-spot-proxy.mjs in our repo does exactly this.
Gold 15-minute base rates
Measured across 5,631 settled windows. A window resolves YES when the closing price is at or above the reference price locked at its open — so the up-rate is close to, but not exactly, a coin flip, and the average move tells you how far these windows typically travel.
Windows settled
5,631
Finished up
49.1%
Avg move per window
$4.07
| Hour (ET) | Windows | Up rate | Avg move |
|---|---|---|---|
| 7 PM | 248 | 50.0% | $4.35 |
| 8 PM | 247 | 48.6% | $5.48 |
| 9 PM | 248 | 53.6% | $4.80 |
| 10 PM | 248 | 45.2% | $4.07 |
| 11 PM | 248 | 44.0% | $3.10 |
| 12 AM | 252 | 48.8% | $4.19 |
| 1 AM | 252 | 48.8% | $4.23 |
| 2 AM | 221 | 53.8% | $3.55 |
| 3 AM | 212 | 48.6% | $3.98 |
| 4 AM | 241 | 50.6% | $3.26 |
| 5 AM | 252 | 50.4% | $3.12 |
| 6 AM | 251 | 49.4% | $3.41 |
| 7 AM | 248 | 54.0% | $5.95 |
| 8 AM | 247 | 48.2% | $6.58 |
| 9 AM | 244 | 51.6% | $6.43 |
| 10 AM | 242 | 46.3% | $5.12 |
| 11 AM | 243 | 46.9% | $4.09 |
| 12 PM | 243 | 45.3% | $3.17 |
| 1 PM | 243 | 49.0% | $3.65 |
| 2 PM | 248 | 52.8% | $3.35 |
| 3 PM | 248 | 40.3% | $2.57 |
| 4 PM | 105 | 79.0% | $2.58 |
| 5 PM | 158 | 43.7% | $2.60 |
| 6 PM | 242 | 45.0% | $2.67 |
The shadow record — not a track record
UnpromotedWhen the market drifts outside the cost band mid-window, we log it and grade it against what actually settled. Nothing here is a pick. No position was taken, no alert was sent, and none of it is scored in our track record. It exists so that the question “is there really an edge here?” gets answered with evidence instead of a claim — and we publish the answer either way.
Graded windows
5,631
Our Brier
0.2390
Market's Brier
0.2332
We beat the market by
-0.0057
Brier score is mean squared error on a probability — lower is better. We score our fair value at the moment the gap first opened, against the market's own mid at that same instant. A positive difference means we were closer to the truth than the price was; a negative one means the market was. Right now, it is the market — which is the result we expected and are reporting anyway. A negative delta is a failed promotion gate, not a near miss.
Flags traded at the ask
3,544
Flagged side won
44.4%
Net after fees
−$87.18
Per contract
−2.5¢
What trading every flag would have paid: one contract on the side the model leaned, bought at the moment the gap first cleared the cost band — at the ask for YES and at 100¢ minus the bid for NO, the prices you could actually fill — less Kalshi's taker fee, held to settlement. It loses money. That is the number that matters, and it is why nothing here is a pick.
Promotion gate — all three must clear
- Beats the market's own pricing (Brier delta > 0) (-0.0057) — not yet cleared
- At least 3 weeks of record (90 / 21 days) — cleared
- At least 60 graded windows (sample-size floor) (5631 / 60 graded) — cleared
Until all three clear, this stays research and the tool stays a fair-value instrument. If the record never clears them, we will say so here and leave it as an educational overlay.
See it in your AI app
The bookHedge Your StackThe 15-minute windows are one chapter. The rest: the gold and silver perps, what they cost to hold, and hedging a stack.Kindle, $2.99 · with a live companion page →Frequently Asked Questions
Does Gold Edge 15-Min actually beat the market?
Not yet, and we would rather say so on the page than let you find out with money. Measured across 2,086 settled KXGOLD15M windows: the Kalshi book is fractionally sharper than our number both at the window open and five minutes in, and the side we flag wins about 55% of the time while still losing money — and that loss is measured at the mid, a price nobody can actually trade at. Two things broke underneath the tool in September: Kalshi replaced its weekly metals contracts with daily ones, and Pyth moved the historical price bars we used for validation behind a paywall. We are rebuilding the evidence in the open, and we have started recording the price you would really pay rather than the mid, because that is what decides whether the edge is real. The maths on this page is sound and the tool is free; the record is what we are still earning.
How do Kalshi's gold 15-minute markets settle?
Each KXGOLD15M contract asks whether the gold price is up over a 15-minute window. It resolves YES if the close of the 1-minute Pyth gold candle at the window's close is at or above the close of the 1-minute candle at the window's open. Pyth is the settlement source Kalshi names in the contract itself, and the reference price is rounded to 2 decimal places. That reference price locks the moment the window opens — before then there is no strike, which is why a fair value only exists once trading is live.
Why isn't fair value 50¢ in the middle of a gold 15-minute window?
Because the strike is already fixed. At the moment the window opens, the contract genuinely is a coin flip. But three minutes in, with the price sitting above the locked reference and eleven minutes left on the clock, the odds are no longer even — they are a specific number you can compute from the distance to the strike, the time remaining, and how much gold has actually been moving. That is what this tool shows: P(YES) = Φ((ln(S/K) − σ²τ/2) / (σ√τ)), with the drift term deliberately set to zero. It is arithmetic on a known strike, not a forecast of where gold is heading.
What does it cost to trade a gold 15-minute contract?
Kalshi charges a quadratic taker fee of 0.07 × P × (1 − P) per contract on these series. At 50¢ that is about 1.75¢ each way, so a round trip near the middle costs roughly 3.5¢ before you pay anything for the spread. On a contract that lives for fifteen minutes and is priced near a coin flip, that fee is the single biggest fact about the product: the market has to be about two percentage points wrong, after the spread, just for a trade to break even. The tool draws that cost as a dead zone around fair value, and most of the time the market sits inside it.
Is there an edge in Kalshi's 15-minute gold markets?
Usually not, and we would rather say so than manufacture a signal every fifteen minutes. We publish fair value and the cost band; when the book drifts outside that band mid-window we flag it as worth a look. We are logging those flags to see whether they actually beat the market's own pricing over time, and we will publish the answer either way. Until that record clears a real bar, nothing here is sold as a pick, alerted on, or scored in our track record.
When do the gold 15-minute markets trade?
Roughly around the clock on weekdays — 96 windows a day, one every fifteen minutes — and dark over the weekend. Those hours are Kalshi's call rather than a stock-exchange calendar, so this page derives them from the windows Kalshi has actually listed instead of assuming a schedule. When nothing is listed, the tool says so and resumes when Kalshi lists the next window.
How is this different from the Gold Edge tool?
Gold Edge covers Kalshi's weekly gold ladder — many strikes on one longer-dated contract, priced against an options-implied view. This tool covers a completely different product: a single up-or-down contract on a fifteen-minute window, where there is no options chain to lean on at that horizon. The weekly tool asks which strike is mispriced; this one asks what the current window is actually worth right now, given a strike that is already locked.
How much volume do these markets have?
Sampled across 200 settled windows in early August 2026, KXGOLD15M averaged ~24,000 contracts per window. For scale, Kalshi's 15-minute bitcoin market runs about 1.8 million per window over the same sample, so metals are a small fraction of that — but ~24,000 contracts in a fifteen-minute window is real retail liquidity, with a two-sided book quoted a few cents wide. These are new products, so treat any volume figure as a snapshot rather than a settled fact.