This demo replays 12 curated 30-60 minute windows of real Binance USD-M futures market data (BTCUSDT and ETHUSDT), each drawn from the study window behind the project's memo. Two things are on screen for every window. The real-data layer shows the book's 12 percentage-band depth profile evolving snapshot by snapshot, next to the trade tape, and lets you walk a hypothetical order against that real band data. The engine layer is a synthetic playground: the project's own C++ order book, compiled to WebAssembly, seeded from the window's opening band totals only -- not from real order-level data.
Pricing conventions used throughout this demo: (1) the mid price used to price the bands is the last trade price at or before the relevant moment -- the project's trade-price proxy for the true midprice (D-008), valid at the coarse, window/snapshot scale used here. (2) Liquidity within a band is treated as uniform and priced at that band's own percentage-offset price (its "band midpoint").
The square-root cost curve plotted against the real walk-the-book result is
cost(Q) = Y × day_sigma × sqrt(Q / day_volume_base), with
day_sigma and day_volume_base read from the selected window's
own day and Y = 1, the practitioner "order of unity" convention from
Toth, Lemperière, Deremble, de Lataillade, Kockelkoren & Bouchaud (2011),
"Anomalous price impact and the critical nature of liquidity in financial markets",
Physical Review X 1, 021006 (arXiv:1105.1694). This is a literature reference
curve, not the memo's own fitted model. The memo's own finding is a concave
impact exponent gamma ≈ 0.76 -- that difference from the square-root exponent
(0.5) is exactly why the two curves on screen diverge. Nothing on this page makes a
fine-scale impact claim (see D-010): the curves shown are at window/snapshot scale only.
This demo makes no performance claims, native or WebAssembly, about the engine.