Settled trades
899
since 22 Jul 2026
What is left after the exchange fee — +1.23¢ on a contract that pays a dollar — is the whole business. Most people selling you a trading bot will not print that number, because printing it invites you to do the arithmetic. We publish it, we publish every rail that lost, and we publish the ideas we killed before they ever touched money.
What this is
Railyard trades Kalshi binary contracts — short-dated yes/no markets on where
Bitcoin’s index settles. A contract costs somewhere between 1¢ and 99¢ and
pays exactly $1 if it resolves your way, nothing if it does not. Six rails read those markets
differently and trade them on your own Kalshi account.
Fifteen rails have run. Six are on sale, one is parked and eight are retired. This is all of them — every rail, every trading day, winners and losers together. There is no selected subset, no excluded “test” period, and no rail quietly dropped for having gone wrong.
| Rail | Trades | Win % | Mean entry | Net | Status |
|---|---|---|---|---|---|
| Rail 1 Continuous | 169 | 66.9% | — | +$157.89 | Beta |
| Rail 2 Late Entry | 119 | 66.4% | 61.8¢ | +$217.63 | Beta |
| Rail 3 Confirmed | 248 | 79.0% | — | +$534.04 | Graduated |
| Rail 4 Tier | 25 | 88.0% | — | +$69.45 | Beta |
| Rail 5 Weekend | 10 | 90.0% | 77.3¢ | +$18.67 | Beta |
| Rail 6 Hourly | 5 | 60.0% | 60.6¢ | +$6.20 | Beta |
| The six on sale | 576 | 73.3% | — | +$1003.88 | |
| Cross-asset port | 23 | 73.9% | — | −$34.52 | Parked |
| PARALLAX | 41 | 46.3% | — | −$101.40 | Retired |
| CROSSTIE | 62 | 66.1% | — | +$41.25 | Retired |
| TOWPATH | 73 | 87.7% | — | −$85.26 | Retired |
| CINDER | 16 | 50.0% | — | −$60.91 | Retired |
| DRAYAGE | 98 | 75.5% | — | −$103.47 | Retired |
| GANTRY | 2 | 100% | — | +$23.91 | Retired |
| DERRICK | 8 | 50.0% | — | −$13.31 | Retired |
| SOUNDING | 0 | — | — | $0.00 | Never fired |
| Every rail ever run | 899 | 72.4% | — | +$670.17 |
They all trade the same instrument. What separates them is when they look, how sure they insist on being, and how often that lets them fire. A rail that trades dozens of times a day and one that trades twice a week are not the same bet with a different dial — they fail in different weather, which is the point of running several.
Works the 15-minute market continuously and takes almost every window where our valuation beats the price by more than the cost of trading.
Different because it is the high-frequency one — dozens of positions a day, a thin edge on each, and the one whose results settle fastest into something you can read.
Waits until most of a window’s outcome is already determined, then trades the part the price has not caught up with yet.
Different because it enters last — in the final minute rather than ten minutes out — and it is the only rail that trades several crypto markets rather than Bitcoin alone.
Requires two independent readings to agree before it will act. When they disagree it does nothing, however attractive either one looks alone.
Different because it is the workhorse — much the longest record in the yard, and so far the only rail to have cleared all three graduation bars and left beta. The setups it insists on are rarely cheap, so it fires less often than it could.
Only the strongest single setups clear its bar. Days pass without a position, and that is the design rather than a fault.
Different because it is the most selective thing in the yard: it passes on setups every other rail would take, which is why its record is still small enough to read as luck.
Read this carefully. 22–3 on 25 settled trades is a small sample, and a small sample flatters. The record is real and exactly reconciled; it is not yet evidence of an 88.0% rail, and it will not stay where it is.
Runs Saturday and Sunday, when the book thins out and prices behave differently from a weekday.
Different because the weekday rails are dormant while it works. It exists because the program measured its weekday edge and its weekend edge separately and found they were not the same thing.
Hour-long windows instead of quarter-hours, and it commits fifteen minutes before settlement rather than ten.
Different because it is the only rail on a different clock. It has the thinnest record in the yard — 5 settled trades — which is nowhere near enough to justify anyone’s money.
Positive in dollars, negative per contract. The $6.20 is positive only because the winning positions were the larger ones. On the per-contract measure that graduation actually uses it is −1.87¢. Of everything on this page, it is the rail to trust least.
Railyard never holds your funds and never touches your exchange balance. It places orders through your own Kalshi API key, which lives encrypted in your settings and is never visible to us once you have entered it.
Email and password, then two-factor from an authenticator app. We ask for nothing else — no phone number, no identity documents.
Step-by-step instructions in your settings, then a test-fire button that places and immediately cancels a one-cent order so you know it works before anything real happens.
Turn each rail on or off. Set your balance and pick low, moderate or high per rail — every stop and position size is derived from that and shown to you before you arm anything.
It is not active today, and we would rather open with that than let you assume otherwise. It switches on when two rails graduate out of beta. One has so far — Rail 3 (Confirmed).
Follow only graduated rails for a full quarter, and if the published record for those rails finishes that quarter negative, your next quarter is free.
A free quarter, not money back. The guarantee is paid in access, not in cash — we do not send you a cheque, and we would rather say so here than let you discover it at the point of claiming. If you would rather have the money than another quarter, this guarantee is not worth anything to you, and that is a fair reason not to buy.
Ours. Not yours.
We never see your balance, your keys or your positions — that is deliberate, and it means a guarantee written on your results would be unverifiable by either of us. So it settles on the same public ledger you can read on any day of the quarter: every settled trade the graduated rails took, at the price they actually filled.
Because a month cannot tell you anything. A rail trades roughly twenty days in a month, and at that sample the noise is several times larger than the edge — a good rail has losing months routinely, and a bad one has winning months just as often.
A quarter is about sixty trading days. It is the shortest window in which the signal is distinguishable from the noise at all. The period is set by the statistics, not by what sounds generous.
Three bars, all three required. They are the same numbers the operator uses to decide whether to keep running a rail with his own money — there is no separate, softer standard for the thing we sell.
Long enough that one good week cannot carry it. Most rails die before this.
Enough that the win rate means something. Below this the error bars swamp the edge.
Not the average — the bottom of the confidence interval, per contract, after fees. It has to be profitable on a pessimistic reading.
Bar 3 does the work. A rail whose lower bound sits above +1.0¢ per contract has an edge that already survived a pessimistic reading of sixty-plus days. Asking it to then lose money across a further quarter is asking for a genuinely unlikely run — on the order of one quarter in ten for a rail that has only just cleared the bar, and rarer for one comfortably above it.
So the expected cost is roughly a tenth of quarterly revenue. That is a real number, it is budgeted, and it is why the promise can be unconditional rather than hedged into meaninglessness.
An annual subscriber has committed to four quarters at once, so a single bad quarter cannot be answered by making the next one free — it is already paid for. The annual term works differently:
If your year finishes negative on the graduated rails’ published record, we add a free quarter on top of your year, and keep adding one quarter at a time — each measured on its own — until a quarter finishes profitable or four free quarters have been added, whichever comes first.
| Term | What it means |
|---|---|
| “Profitable” | Mean per-contract net after fees, above zero, across the graduated rails over that quarter. The same statistic the graduation bar uses — you are never measured on a number that did not let the rail in. |
| Each quarter stands alone | A free quarter that finishes profitable ends the run. It does not have to make up the earlier loss. |
| Capped at four | A full extra year, free. Beyond that the honest answer is not another free quarter — it is that the rails stopped working and you should stop paying us. |
| Still access, not cash | Free quarters extend the subscription. Cancel and there is nothing to refund. |
The quarter closes, the ledger is what it is, and the credit is applied by us. You are not asked to prove anything, because the number the claim turns on is ours.
An email at quarter end with the graduated rails’ net for the period and whether the guarantee triggered — including in the quarters where it did not.
Recorded against your account and applied to the next bill automatically, so you are charged nothing for that period. It is access, not cash: cancel instead of staying and the credit simply goes unused. Nothing is paid out, and nothing is quietly kept either — there was never a cash balance to keep.
One subscription covers the whole yard — every rail, including new ones as they arrive. Cancel any time from your account; access runs to the end of the period you paid for.
Billing runs entirely on Whop. Railyard never sees or stores a card number — the payment form is served by Whop, on Whop’s domain, and we receive only a membership reference.
WHOP* RAILYARD.Our server asks Whop for a checkout session on the plan you chose, with a return address back to your deck, and sends you there. No card data touches our infrastructure, which keeps us out of PCI scope entirely.
An access code does not go through the processor at all. It is checked when you ask for a sign-in link and starts a fixed number of days with nothing on file. There is no card to convert when the days run out, so the access simply ends. A free trial that bills you is not a free trial.
A code is checked when it is offered, and it is the account’s first verified sign-in that consumes it and decrements the seat count — one row, written once, in the same transaction that creates the account. So the cap counts people who can actually see signals, not codes that were typed, and a code that was never redeemed is safe to reissue.
Whop posts membership.activated, payment.succeeded,
payment.failed and membership.deactivated, each signed over
the delivery id, the timestamp and the exact bytes of the body. We store the delivery
id, so a retry cannot book the same renewal twice and hand out a free month.
Every page load checks one thing, in our database: is the subscription
trialing or active, and is the period end in the future.
We deliberately do not ask the processor at request time — an outage,
a freeze or a migration on their side must never lock out somebody who has paid.
Lapse and the rails disarm at the next window; open positions still settle.
Cancel, change plan, update a card, download receipts — all on Whop’s side. We build none of it, and a cancellation is honoured even if our server is down.
| Whop event | What it means | What Railyard does |
|---|---|---|
membership.activated | Subscription started | Mark the account active and set the period end |
| first sign-in | Not a Whop event — ours | Consume the access code, decrement the seat, start the clock |
payment.succeeded | A charge cleared | Extend the period end, record the payment, mark guarantee-eligible |
payment.failed | Card declined | Email, keep access through Whop’s retry window, then disarm |
membership.deactivated | Cancelled or lapsed | Disarm every rail at the next window; positions still settle |
There is no calculator on this page turning your balance into a number. Fifty-one days is not enough history to project from, and a projection built on it would be a decoration, not a forecast.
Nine ideas sit in the graveyard — eight retired and one parked — published with what each one cost, and −$333.71 between them. The most recent lost $101 before it was pulled. You can read all of them before you pay anything.
Binary contracts can settle worthless. 248 of the 899 settled trades on this page are losses, and nine rails have been taken out of service. That is why a rail has to earn its way out of beta on the published record rather than on a story.
Signals are for your account only. Redistributing them ends the subscription immediately and without refund — the one rule with no discretion in it.