QSL

Quant Signal Lab

Live Strategy Performance
APEX Book 1 (SHADOW)
MARKET CLOSED API marks ⋯ reconcile… Updated: loading...
API unreachable — showing last cached data
Portfolio Value
Program Total Return
$100k NAV · dollar-neutral L/S
Long P&L
— longs
Short P&L
— shorts
APEX v1 — Dollar-neutral long/short equity, accumulating weekly tranches  ·  Capital recycles week to week (compounding book)  ·  Returns measured on deployed capital  ·  Equal-weight, 5 bps execution slippage
What you are looking at: APEX v1 is the alpha signal on a minimal-risk shell — weekly entries, equal-weight, fixed 10-trading-day hold (Mon open → 2nd Fri close), no stop-loss and no intraweek risk controls. Volatility and drawdowns on this board are expected for an unmanaged book; they are not the end-state product. The live window is a signal-verification phase — early tranches on simulated capital, Alpaca paper from tranche 3 onward — while the Sentinel overlay (in development) is validated out-of-sample before it touches production.

Why allocators should read these numbers carefully

  • Sample size. Live track is only a handful of trading days — far below the 12–24 months most allocators require. Annualized Sharpe/vol here are preliminary (see Probabilistic Sharpe + Path Inspector below). The research thesis rests on 2016–2025 out-of-sample backtest evidence in Signal Research, not this live window alone.
  • Execution model. Tranches 1–2 use Polygon / AlphaVantage / FMP API marks on simulated capital. From tranche 3 onward the book trades on Alpaca paper with broker fills and position marks on open tranches; closed Alpaca tranches reconcile to recorded exits. Early simulated weeks are a falsification baseline, not proof of live capacity.
  • Volatility is intentionally unmanaged today. Concentrated weekly L/S with no intraperiod controls can produce very high ann. vol and multi-percent weekly drawdowns even when the signal edge is real. That is the problem Sentinel is engineered to compress — not evidence the signal is broken.
  • Dollar-neutral ≠ factor-neutral. Early beta to SPY can look elevated (and noisy at small n) when the book clusters in crowded factors (e.g. AI/semis). Sentinel prices marginal portfolio risk via a factor covariance model, not single-ticker vol alone.

Sentinel overlay — in development (advisory-only until gated)

Sentinel v1 is a variance / tail / drawdown safety rail on top of the APEX book — not a return-raiser. It will not auto-trade until a pre-registered out-of-sample gate passes. Planned capabilities that directly address what you see on this board:
  • Cluster & tail controls — same-day correlated-print cap (limits cluster blowups); conformal CVaR tail monitor (early stress alerts before drawdowns compound).
  • Factor-aware risk pricing — marginal covariance sizing (Σ=BFBᵀ+D), not own-vol; basis-risk monitoring so dollar-neutrality moves toward factor-neutrality.
  • Target-weight engine — CRRA growth objective with CVaR constraint, no-trade band, and event jump caps; resize tiers instead of blind trims.
  • Exit-alpha layer — path-geometry screens, risk-gated trails, and forecast-vol de-risking (Exit-Alpha Lab) to harvest regime-B gains without cutting recoveries.

Status: core engine + cluster cap + tail monitor built; full validation gate (J2) and advisory dashboard (K1) in progress. Until Sentinel ships, APEX v1 numbers reflect signal quality under minimal risk management — the honest baseline we are improving from.

APEX Performance Metrics
Loading metrics…
Stochastic Path Inspector — distribution of outcomes
Your live curve is one realized path drawn from a distribution of futures that share the same edge. This resamples the strategy's own realized daily returns into thousands of forward paths (default: stationary block bootstrap, which preserves serial correlation and fat tails — the path-risk counterpart to the HAC / Newey-West Sharpe above). It surfaces the sequence / path risk a single Sharpe number hides: the drawdown distribution, risk of ruin, and time-to-recovery.
Simulating paths…
Loading path fan chart…
Percentile fan (5–95%) + sample paths · click any path to inspect day-by-day returns
Path inspector Terminal Max drawdown click another path or an archetype to compare
Strategy Breakdown — Combined vs Long vs Short
Metric Combined Long book Short book
Loading...

Portfolio Value Over Time

API vendor marks on early tranches · Alpaca paper from tranche 3 onward

Long vs Short Basket

Long basket vs short basket · equal-weight · dollar-neutral per tranche · position win rate = tickers in profit / tickers in leg (not daily hit rate)
Open & Closed Positions
SymbolSideWeekEntry ExitExit dateP&L $Return % ScoreStatus
Loading...
All Tranches — click a row to view its full metrics
TrancheG(t)G bandExecutionOps alert BookBook closedLong P&LShort P&L Net P&LNet returnStatus
Loading...
Realized Tranche Results — 10 trading-day hold (no stop)
Each tranche enters at the Monday open (5 bps entry slippage) and exits at the 2nd-Friday close (10 trading sessions) — no intraperiod stop. Exit fills include 5 bps adverse slippage (≈10 bps round-trip). Trading days, not calendar days. Book closed = all positions sold at maturity.
TrancheEntry+10td scheduledBook closed Days in bookPositions Long P&LShort P&L Net P&LNet returnExit-slip drag
Loading...
Ch13 G-only regime gate — counterfactual (production is ungated)
Loading…
Signal Research — Forward-Return Decay & Rank IC
Each weekly signal is an independent cohort: forward return of its picks at +5 / +10 / +15 / +20 trading days after the signal Monday, plus rank IC. Only cohorts on/after 2026-06-01 — pre-June telegram weeks are excluded (unreliable decision logging). A horizon shows until that many trading days have passed for that cohort (we started tracking 1 Jun). Equal-weight per ticker. Decay means are not annualisable (overlapping cohorts).
Convention: +Hd is a close-to-close measurement — H trading-day steps after the signal-Monday close (so +10d for the 1 Jun cohort = the next Monday, 15 Jun). This differs from the live tranche, which enters at Monday's open and exits the 2nd-Friday close = 10 sessions of exposure (≈ the +9d close-to-close point).
Decay curve (mean across cohorts)
HorizonMaturedLong-Short %Long leg %Short leg %Rank ICCohorts IC>0Hit rate
Loading...
Per-cohort (long-short return % / rank IC)
Signal dateSourcePicks+5d+10d+15d+20d
Loading...