1
Observed L2
What the simulator can actually see
Waiting for the first two-sided book snapshot.
v0.2.0 · experimental alpha
See exactly when a HIP-4 paper order earns a fill.
A deterministic execution simulator that seeds queue-ahead from observed L2 and advances it only with aggressor trades.
No live orders. No hidden liquidity assumptions. No profitability claim.
One fixture. Every causal step.
This six-event recording is replayed through the actual Python engine. TypeScript only renders the resulting trace.
metadata
Fixture loaded
The deterministic replay fixes the coin, balances, order size, and 50 ms activation latency before any market event is applied.
2
Aggressor trade
The only event allowed to advance queue
No aggressor trade has reached the replay yet.
3
Virtual result
Spot-safe ledger after this event
No fill yetThe ledger is ready; no order has been submitted.
Available USDC10,000
Reserved USDC0
Token inventory1,000
Generated from fixture
2978ac390d3472439311d30d2ed9308f261f64a220b9446e49e851d99a1a1ffdengine v0.2.0 24h mainnet run
A longer run tests the invariants, not the alpha.
From July 13–14, 2026, the recorder followed coin #8220for a full wall-clock day. The aggregate report is public; the raw recording remains outside git and is anchored by its SHA-256.
16,650recorded events
16,087L2 snapshots
562trade messages
38simulated fills
240duplicate IDs ignored
All release invariants passed
Checked after deterministic replay and final cancellation.
balances_non_negativePASStrade_volume_conservedPASSreservations_releasedPASS
Exact reproducibility
The same JSONL replay produces byte-identical JSON reports.
- Coin / quote
- #8220 / USDC
- Source SHA-256
8d155b56c98daaf186eaa12150d2560f71fa2388a7b73776c9b2520ae2249dfc- Window
- 2026-07-13 19:21 UTC → 2026-07-14 19:21 UTC
Show baseline accounting output
Final NAV 10544.669730 · PnL -4.060270. This is a baseline accounting output, not an alpha claim or estimate of live performance.
Model boundaries
What this does not prove.
L2 cannot reveal exact order-level causality. The simulator stays deliberately conservative and publishes where the model stops.
- L2 decreases without trades are treated as cancellations.
- Receive-time gaps include periods of legitimate market inactivity.
- Split, merge, negate, settlement, and fees are outside v0.2.
- The included strategy is a validation baseline, not a profitability claim.
Inspect the assumptions. Challenge the trace.
The useful contribution is a reproducible edge case, not a star.