Your private oracle (TOSDB-native AHMA) vs the product path (Alpaca + Python PAA),
compared tick-by-tick during market hours. This page is the gate: TOSDB stays your personal feed
until the numbers below prove the Python path is a 1:1 stand-in.
NO DATAWaiting for samples — the harness records while the bridge streams during RTH.
Direction κ
—
Cohen's kappa, long/short/flat
Tier κ (weighted)
—
quadratic, ordinal −3…+3
Direction agree
—
raw observed agreement
Tier within ±1
—
tier off by at most one step
Tier MAE
—
mean absolute tier error
⚖️ Switch criterion: direction κ and weighted tier κ both ≥ 0.90 ("excellent"),
sustained across multiple full sessions, with no systematic lag and disagreements clustered at inflections —
then Alpaca+Python is proven and TOSDB becomes optional, even for you.
Per symbol
0 valid TOS samples.
Parity only records while TOSDB is the active source — it compares the TOS-native AHMA against
Python. So first check the engine source: if the bridge has failed over to Alpaca (normal when
TOSDB drops), there's no TOS side to compare and this page stays empty by design until TOSDB
reconnects. Only if the source says TOSDB and it's still empty during RTH is it worth confirming
your kh_custom1_2m_hull / kh_custom2_5m_tier studies are streaming live values
(2m▼ / 5m-2.00 in the
watchlist, not "loading"). Fills within minutes of valid TOSDB ticks.