Sky City Network
Every scheduled slot is published. Including the ones that went badly.
SkyTrade posts a scenario at each 4-hour and daily close, on a schedule fixed in advance. Every number is computed by the engine, every slot is written to an append-only ledger, and the day's Merkle root is anchored to Polygon. You are not buying a forecast. You are buying a record you can recount yourself.
We expect activation and hit rates near the historical base rates of the same conditions, and net R near zero after fees. The product is the scored record, not a return.
| Close UTC | Symbol | Slot | Dir | Outcome | Net R |
|---|---|---|---|---|---|
| 09-17 04:00 | BTCUSDT | scenario | ▼ bear | open | — |
| 09-17 04:00 | SOLUSDT | no trade | — | low confluence | — |
| 09-16 20:00 | BTCUSDT | scenario | ▲ bull | objective | +1.62 |
| 09-16 20:00 | ETHUSDT | scenario | ▼ bear | invalidated | −1.08 |
| 09-16 20:00 | XRPUSDT | scenario | ▲ bull | not triggered | — |
| 09-16 20:00 | TONUSDT | data gap | — | OI missing | — |
| 09-16 16:00 | BTCUSDT | scenario | ▲ bull | expired | −0.21 |
| 09-16 16:00 | XAUUSD | outage | — | no row by +300s | — |
anchored on Polygon today, not anchored yet
Four things, in this order.
Nothing here is a judgement call made after the fact.
- The schedule is fixed first.
- 21 symbols × two timeframes, committed as a hash before anything runs. A symbol cannot be quietly dropped on a bad week: the expected slot count is arithmetic, and it is published beside the actual one.
- The engine computes, then the snapshot is hashed.
- Structure, levels, patterns, indicators, derivatives — all deterministic, all from confirmed bar closes only. The snapshot hash is written before any language model is called, so the words can never change the numbers.
- The model may only refer to numbers by id.
- Every figure carries an id (L#3, S#1, I#7). Model output containing a digit in any script is rejected by the verifier and retried; a second failure strips that seat and falls back to a template. Numbers are rendered from the ids, client-side.
- Scoring is pessimistic, and the day is anchored.
- Same-bar stop and objective counts as the loss. Gaps fill at the worse price. Fees, slippage and funding are subtracted, and each row satisfies r_net = r_gross − fees − slippage − funding so you can check the arithmetic. The day's Merkle root goes to Polygon.
A scenario reads like this.
Conditional, with the invalidation named before the objective.
No instruction is given, and no probability is shown. Probability requires the engine's own walk-forward statistics and stays dark until the effective sample reaches 20 — which, on day one, it does not.
What it covers, and what it refuses.
- Now
- 12 crypto pairs on 4H and 1D, spot and perpetual, across Binance, Bybit and OKX. A second venue scores every scenario in parallel; when the two disagree on the outcome, the row says venue disputed rather than picking the flattering one.
- Next
- Gold and seven major FX pairs, once thirty consecutive days of clean data have been logged. Until then those rows carry the deterministic baseline only, and say so.
- Never
- An instruction to buy or sell. A position size for you. A win rate. A percentage without its sample. A claim about what the market will do.
- Not available
- Price levels are redacted for readers in the United States, and the service is geo-blocked in the UAE. Both are precautionary, and both are stated here rather than discovered at checkout.
Access.
The verification pack — daily roots, transaction links, expected against actual slot counts, and the four outcome counts — is public and stays public. Paying opens the scenario detail, the evidence cards and the chat, not the record.
