Hi thereBacktests
Historical results and validation studies behind the strategy ensemble. Each study below summarizes a real research report from the backtest archive — verdicts, key figures, and the artifacts they were derived from. Expand any study for its findings.
Validation Studies
Completed studies with written reports. Status: green ✓ = validated / actioned, amber ! = engine ready or archived data.
| Study | Scope | Date | Key result | Status |
|---|---|---|---|---|
| SVStrategy Validation 2026 | SPX 0DTE + 1DTE | 2026-05-13 | No production changes; walk-forward passes | ✓ Final |
| BB1DTE BB Violation v2 | SPX 1DTE (bull) | 2026-05-16 | Sharpe ridge 4.7–5.7; edge structural | ✓ Validated |
| 1D1DTE vs 0DTE EOD-Hold | SPX 1DTE | 2026-05-06 | Sharpe 6.53 · 80.0% win · deployable candidate | ✓ Validated |
| 1M1-min vs 15-min Entry | SPX 0DTE | 2026-05-04 | Replace 15-min entry; Sharpe gap +1.141 | ✓ Actioned |
| ETEntry Timing / Dual Bias | SPX 0DTE | 2026-05-04 | Chart TZ fix; Sharpe 1.96 → deployable | ✓ Actioned |
| PTPT Threshold Analysis | SPX 1DTE EOD | 2026-05 | PT=90% best on Sharpe, DD, profit factor | ✓ Final |
| RRRSI/RSI Validation | SPX multi-day | 2026-05-04 | DEPLOYABLE · Sharpe 1.843 · 72.4% win | ✓ Live |
| ECETF Credit-Spread Engine | 25 sector ETFs | 2026 | Engine complete, 18/18 selftests; run queued | ! Queued |
Study details
Strategy Validation 2026 — SuperTrend mult + 1DTE exit matrix ›
- Question A: does SuperTrend mult=3.0 beat production mult=3.9 for the 0DTE strategy? No — mult=3.9 wins by 11.7% on Sharpe; production kept at 3.9.
- Question B: does any 1DTE exit cell beat the current hold-to-EOD? No — top cell (PT_90) only +3.4% on Sharpe, within noise.
- Walk-forward checks pass on both winners; no overfit flags fire.
- Proxy fidelity 98.26% (SPY 15m bars as SPX proxy; documented caveat — results indicative, not definitive).
- Data spend this iteration: $4.68 of a $75 cap.
Artifacts:
backtests/strategy_validation_2026/BACKTEST_REPORT.md · walk-forward + mult-comparison CSVsSPX 1DTE BB Violation — Diagnostics v2 ›
- 4×4 structural grid (ITM offset × width): the Sharpe surface is smoothly monotonic as the long leg moves from deep-ITM to OTM — decisive evidence the v1 result was a ridge, not a curve-fit peak.
- Best cell: long 25-OTM, Sharpe 5.71; neighbors 4.7–5.7. Deep-ITM cells strongly negative (worst −7.78).
- Bull side only — bear side disqualified in v1 and not re-tested.
Artifacts:
backtests/spx_1dte_bb/REPORT.md, REPORT_v2.md · bootstrap CIs · diagnostics.jsonSPX 1DTE vs 0DTE — EOD-Hold Comparison ›
- Sell SPXW credit spread day T 8:45 CT (T+1 expiry), close T+1 14:30 CT; ST+VIX bias filter; sell-at-bid / buy-at-ask fills, $2.60 round-trip commission.
- 1DTE ITM20 W10 (EOD-hold): N=225, total P&L $35,060, Sharpe 6.53, 80.0% win, max DD −$2,426 — deployable candidate.
- Simulator integrity fixes applied (exit debit capped at width; NaN bids treated as worthless expiry).
Artifacts:
backtests/spx_1dte/COMPARISON_1DTE.md · 1dte_trades.csv · SPXW chains DBSPX 0DTE — 1-min vs 15-min Entry Detection ›
- Validated with real OPRA fill data.
- Verdict: replace 15-min entry with 1-min — Sharpe gap +1.141 exceeds the +0.3 action threshold.
Artifacts:
backtests/spx_0dte_1min_vs_15min/COMPARISON.md · trade-level CSVSPX 0DTE — Entry Timing / Dual Bias ›
- Validated with real OPRA fill data. Timing discrepancy traced to chart timezone configuration, not Pine logic.
- Fix: set the TradingView chart to Central Time. With chart in CT, entry fires 09:44 CT → Sharpe 1.9565 ≥ 1.5 → deployable after fix.
Artifacts:
backtests/spx_0dte_dual_bias/COMPARISON_TIMING.md · phase CSVsSPX 1DTE EOD — Profit-Target Threshold Analysis ›
- Thresholds tested: 50–90% across the 230-day live signal universe (2023-05-02 → 2026-04-27).
- Recommendation: PT = 90% — highest Sharpe, lowest max drawdown, highest profit factor; ranking monotonic across every robustness split.
Artifacts:
backtests/spx_1dte_eod/pt_threshold_analysis.md · PT comparison CSVsRSI/RSI Strategy Validation ›
- DEPLOYABLE — Sharpe 1.843 (clears the 1.5 bar) with the $30 server-side debit stop enforced.
- 72.4% win rate, $14,125 over 2.9 years; max drawdown reduced from −$4,110 to −$1,705 by the stop.
- Phase 2B live: server-side polling enforces the stop every 10 minutes during RTH.
Artifacts:
backtests/spx_rsi_rsi/COWORK_REPORT.md · P&L CSVs · chains DBETF Credit-Spread Backtest — Engine ›
- Validates RSI / Bollinger credit-spread setups on the sector-ETF universe against real OPRA chains, with American-model Greeks and pessimistic (short@bid / long@ask) fills.
- Engine complete and unit-tested offline — 18/18 selftests pass. Full backtest run not yet executed (queued pending data budget).
- Safety rails: live cost estimate before every pull, hard stop at $50 data budget, 20% out-of-sample holdout, combos under 30 trades flagged inconclusive.
Artifacts:
etf_spread_backtest/README.md · config.py (single source of truth) · selftest.pyArchive — data-only studies (no written report)
spx_0dte_rsi— RSI signal/chain/backtest pipeline with trade log CSVs.spx_supertrend_long— SuperTrend long P&L phase run, 2023 trade set.spx_supertrend_v2— v2 puller + simulator validation scripts.
Important: All figures are derived from historical backtests and simulations, not live trading. Some are real-fill backtests over a single market regime (2023–2026); others are model simulations and may run optimistic. Backtested and simulated performance does not guarantee future results. These strategies are in validation; a verified live track record is being established.
Restricted access · victorngoma.com