Session-aware oracle · live on Robinhood Chain
Price any equity. Publish the error bar with it.
Tokenised equities trade around the clock. The shares behind them price for six and a half hours a day. Every oracle in production returns one number and hides which of those two regimes it came from — so a contract cannot tell a live print from a weekend estimate. hoodoracle returns the price, its provenance, and a confidence interval fitted on two years of realised gaps.
US market session
—
loading
Next session
—
until the tape reopens
Feeds modelled
—
all live prints
Widest band
—
across all feeds
Live feeds
| Instrument | Last | 95% confidence band | Provenance | Session | Src | Drift | Age |
|---|---|---|---|---|---|---|---|
| loading feeds… | |||||||
Live on-chain
deployedContract
0x65cf45524407a5e700188a8a8178d5d5c0c38d30
Robinhood Chain · chainId 4663
Verify it yourself
# a liquidation path, tape shut cast call 0x65cf4552… \ "getPriceIfTraded(string,uint64)" \ "HOOD" 50 --rpc-url rpc.mainnet.chain.robinhood.com → reverted: not a live print
Why the bands are believable
Fitted on 3,990 real gaps, then checked against them.
Every beta and every interval comes from regressing two years of realised close-to-open moves. Coverage is the test that matters: replay each historical gap and count how many landed inside the published band. A 95% interval should catch about 95% of them.
93.4—96.4%
observed coverage, all 8 instruments
k = 0.107
measured time exponent, not the assumed 0.50
3,990
close-to-open gaps in the fitting window