Execution & cost · Measured against the official print
We don't estimate execution cost. We measure it — against 291 M official auction prints.
The book executes at the market-on-open / market-on-close auction cross, and every fill is reconciled field-by-field against the official cross price from a 291.4 M-print TAQ archive. The result: auction execution is near-costless — 1,105 events reconcile to a 0.0 bp median (AAL 0.0 bp, PLTR +3.5 bp). The cost model isn't an assumption; it's validated against ground truth. What's left to manage is capacity at scale, not per-trade cost.
The cost model, validated — measured vs naive
measured effective cost at the auction cross → book runs at 1.36 excess Sharpe
median slippage vs the official cross across 1,105 reconciled events
the full continuous-market spread — what a naive execution pays, and what the auction avoids
How we know — reconciliation, not assumption
291.4 M cross-prints as ground truth
Every trading day's official opening and closing auction cross (2003–2026) is ingested and field-validated. Each real fill is diffed against the official print for its symbol and session — so execution cost is a measurement, not a modeling choice.
0.0 bp median across 1,105 events
Auction execution lands on the official cross: 0.0 bp median slippage over 1,105 reconciled events (AAL 0.0 bp, PLTR +3.5 bp). The cost model was validated to the measured reality, which is why the book sits at the top of its cost range, not the bottom.
The +116 bp defect — caught, not incurred
A pre-open MKT/DAY proxy order type (not the auction) showed +116 bp
— flagged by the TAQ reconciliation, within 65 s of the open so drift is
excluded. It's a one-line routing fix (MKT/OPG), not a strategy cost. Catching
it is the point of measuring against the print.
The real remaining constraint — capacity, measured from the same data
flagship long-only book, from auction-clearing depth
event-book peak (a band, $150k–$500k)
closing-auction depth vs continuous market — why MOC scales