Martial Systems
LLC
Loading feed…
Martial Systems
LLC
Personal non-commercial research summary for KXBTC15M-style short-dated event contracts. Public figures are high-level aggregates only - no tickers, strikes, settle spots, entry prices, or per-trade/window dumps.
Post-reset research sample (unique markets, last write wins; path $ from equity series only). Pre-reset archive PDF remains for method notes.
Download pre-correction research baseline (PDF) - estimator error, correction, and pre-correction aggregates (no trade tapes)
Parallel microstructure study (Level-2 snapshots → VWAP penetration → fill/footprint models → multi-window schedule → Path P&L under size caps). Complements 1-contract shadow Path P&L for edge discovery; does not scale that path linearly. Frozen shadow exit policies are unchanged.
Capacity methodology note (PDF) : aggregate methods only; no market identifiers or trade tapes
Operational gates (GraphForge product-law checks: hard rails, freeze discipline, capacity phase order, HTS isolation): verification board · GraphForge overview
Provisional: treat win rates and hybrid deltas as research evidence. Exit policies stay shadow until a dated re-eval. Freeze gate: 20 to 30 win65 first_hot unique tickers (one first_hot row per market, strictest gate).
High-level mid-window forecast quality on unique settled markets only. Research-only; not a claim the model is ready for live trading.
Shadow WR = entry current_exit on unique markets (not window-path +rate). Model research-post-reset. Path $ = equity-series sum (unique last-write windows). Not audited · not a solicitation to trade.
Unique markets, last write wins · simulated cooldown path · anonymized window order. Closed trades from outcomes log are never included.
1-contract cooldown sim path. Flats (no trade) count against “incl. flats” only.
Share of the last N non-flat (active) windows that were positive. Flats are omitted from this series. Not Shadow WR.
Aggregate mean shadow P&L per entry when the confidence bar for an entry is raised or lowered (cooldown, 1 contract). Counts and win rates only - no individual signals.
Same first_hot markets under alternate exit rules only (1-ct, est. fees). Primary counter: cumulative Δ total $ for the leading policy vs current exits (lead can change as the sample grows). One row per market ticker · anonymized entry order.
Solid lines: cumulative shadow P&L by exit policy. Thick dashed line: cumulative Δ$ for the leading policy (shadow − current). Hover a line to focus it; Reset view (KPI row) restores all series. Not live trading.
Running average Δ¢ per entry for the leading policy (positive = shadow beating current so far).
Isolated paper path. Enter only in the first minute when a side is well below even money. Brain/model does not decide whether to go in. Hold a short time and stay only while still green. Reconstructs the pre-automation operator style. Not Shadow WR. Not live trading.
Cumulative 1-contract paper P&L in anonymized close order. Own book: does not enter first_hot Shadow WR or live calibration.
Same 1-ct size. Shadow line is the equal-weight mean of Shadow exit comparison policies on the first N unique first_hot closes, where N is the live scalp sample. The scalp path is not in that mean. Different entry book: not the same trades. Hover a line to focus it. Reset view restores both series.
Venue (context only). Kalshi is a CFTC-regulated U.S. exchange for event contracts. KXBTC15M-style markets are short-dated binaries on BTC vs a strike over a fixed interval. This page does not republish live books, quotes, or market-by-market rows.
Research process. Internally we estimate win probability and apply thresholds and risk rails. Public charts show anonymized shadow-path aggregates (cumulative P&L, rolling win rate, threshold bars, shadow exit comparison) plus closed-trade side counts - not downloadable trade tapes or window dumps.
Publication scope. Personal non-commercial research summary. Speculative risk remains; capital can be lost. Not investment advice.
Martial Systems LLC - quantitative research on short-dated event contracts.
Investor inquiries: martialsys@gmail.com
Exchange-listed contract that pays on a defined outcome at a fixed time (here: BTC vs strike after 15 minutes).
Contract sides on binary event contracts. Public stats show aggregate side counts and win rates only.
Model estimate that the chosen side wins by settlement. Thresholds such as p ≥ 70% filter weaker signals.
Minimum wait after an entry before another entry is counted on the baseline path.
Share of unique first_hot markets with positive current_exit P&L (1-ct). Not path +rate (which includes no-trade windows).
Equity-series sum of the 1-contract cooldown sim path on unique last-write windows. Excludes trade_outcomes closes.
Path is simulated research only. Live P&L is closed outcomes log aggregates. No per-trade tape is shown.
No tickers, strikes, settle spots, entry prices, signal counts per window, or CSV downloads on this site.
Phased diligence layer: L2 book snapshots → VWAP pot-ladder EV → footprint gates → multi-window capital → liquidity-capped Path P&L. Proves scale without linear 1-ct × N math. Legacy shadow exits remain frozen.
Volume available at each price tick on the bid book (Kalshi). Public site shows snapshot counts only; full ladders stay on research hosts (disk-only).