One share. Two prices.
The exchange is shut and the oracle is holding its last value, but 25 pools are still quoting NVIDIA to whoever turns up — and they do not agree with each other.
On a $25,000 order of NVDA, the wrong pool costs $618 more for exactly the same tokens.
basis · is it priced right against the real world?
+105 bps · DARK · the real NVIDIA last traded at the close
What the wrong pool costs
One token, a dozen pools, a dozen prices. Far more people lose money landing in the wrong one than will ever make money on the gap.
Buying $25,000 of NVDA gets you 109.847 tokens at the best pool. The same order routed into NVDA/ETH costs $25,618 — $618 more for exactly the same thing. The real NVIDIA is $229.01.
This order is 15% of the best pool's depth. The quoted price is the top of book, not your fill.
What you are holding, and where the numbers come from
Three prices, none of them ours, so a reading can be re-derived without trusting us. Hover a source to see which figures it feeds.
A Stock Token is an ERC-20 on Robinhood Chain giving economic exposure to a listed share — tokenised debt securities issued by Robinhood Assets (Jersey) Limited, without legal or beneficial rights in the underlying.
They never pay a cash dividend and never rebase. A split or a dividend moves an on-chain multiplier instead — the ERC-8056 standard. Your raw balance never changes; what each token represents does.
underlying_shares = raw_balance × uiMultiplier() / 1e18
The pools trade all 168 hours of the week. The shares trade about 32.
Exchange bid and ask from the issuer's keyless API, times the multiplier. Frozen when the market shuts.
Chainlink total-return feed on chain 4663. Holds its last value off-hours with no heartbeat.
Every Uniswap pool holding the token, ranked by depth. Never authoritative — it is what gets measured.
spread = ( dearest ÷ cheapest − 1 ) × 10 000 // do the pools even agree with each other
The readout
Every tracked Stock Token against the real market. Sort by any column; open a row for its pool distribution and the commands that reproduce the number.
The underlying market is closed and the reference is frozen, so every row reads DARK. The gap is unmeasurable, not zero — the pools are still trading against a price that stopped moving.
| Token | Reference | Deepest pool | Basis ↓ | Flag |
|---|---|---|---|---|
| GME | $34.44 | $19.28 | -4402 | DARK |
| AMC | $2.66 | $2.57 | -338 | DARK |
| AAPL | $310.23×1.001 | $320.40 | +328 | DARK |
| MSFT | $512.22 | $500.31 | -233 | DARK |
| CRWD | $857.76×4.000 | $843.98 | -161 | DARK |
| NVDA | $229.01 | $231.42 | +105 | DARK |
Reference and pool in USD · basis and spread in bps · depth is the deepest pool's TVL, which is not executable size · 12 of ~194 tokens tracked in this build
Watch your own tokens, ask what a reading means in plain English, and see which corporate actions are about to pause an oracle while the pools keep trading.
When the market sleeps
The pools trade 168 hours a week, the underlying about 32. Chainlink on these feeds: they “do not have heartbeats during off-hours.”
“With the equity market closed and no market makers on the other side, the on-chain price ran 35 times past the real one.”
A Robinhood Chain launchpad, 31 August 2026, on what happened to AMC that Saturday — and announcing “risk labels so traders see a premium before they buy rather than after.”
A dividend or split moves the ERC-8056 multiplier, and the issuer pauses that token's oracle while it processes. The pools keep trading throughout.
How you would catch us lying
A measurement is only worth reading if someone who distrusts the publisher can check it. Two of these work today; two do not exist yet and say so.
Three commands, none of them ours
Every row expands to the issuer's price and multiplier endpoints and the pool index. If basis lies, one curl catches it.
# QQQ · 16:35 → 17:21 UTC reference 708.77 → 707.84 pool 724.12 → 719.98 basis +217 → +172 bps # the gap closed while we watched
The biggest pool is often a trap
A pool can hold millions and trade nothing, and its stale quote drags a naive reading anywhere. We rank on the deepest pool that turned over a tenth of its own depth in 24h.
# TSLA · 2026-09-03
largest pool $418,744 held · $8,835 traded · 2%
quotes 356.24
active pools 382-384 reference 383.29
ranked on size -706 bps a dislocation that isn't there
ranked on trade -28 bps TIGHTA hash-chained archive — each sample storing the hash of the one before, the head anchored on chain daily, every missed minute logged with its cause. Then a weekend board: the frozen reference posted each Friday, settled each Monday, so a claim made before the outcome can be checked after it. Both need a host with a database. This build keeps nothing between requests, so neither runs and no board exists. A feed claiming perfect uptime is one you should not read — so is a feed claiming an archive it does not have.
What is live today: /api/v1/status
The token, and what it is for
The readout is free and stays free. $BASIS is a claim on three engines running off the same data — one live from the first block, two gated on published conditions. No price on this page.
The launchpad's locker holds the pool and there is no withdraw function to call · no pre-mine and no team allocation is structurally possible · the only way the creator holds tokens is a disclosed buy at launch price
Buyback & burn
Every dollar the project earns is split in half — one half buys the token off the market and destroys it, the other funds the treasury. Two sources: the creator share of the pool's swap fees, and what launchpads pay to put the premium warning beside their own buy button.
The basis treasury
The instrument finds mispricings. The treasury trades them. Pool-to-pool, weekend reversion, and corporate-action windows where the oracle pauses and the pools do not. The rule that keeps the feed honest: publish first, then trade the published number, fills timestamped beside the alerts so the ordering is checkable. It costs the treasury its best fills; that is the price of the feed being worth reading.
Oracle staking
basis moves on chain and the token becomes the thing you stake to run it. A contract other protocols read — a peg guard, a collateral haircut, a router avoiding a dislocated pool. The second gate is the real one: with no protocol wanting to read it, building this would be infrastructure cosplay. The roadmap says may-never-ship for that reason.