{
  "audit_date": "2026-04-27",
  "scope": "Additional audit angles (A-O) missed in first audit asym_v2_HONEST_AUDIT.json. Strategy = ASYM_MOO_1100_V1 organic_pump SHORT (sub5_pump). Empirical anchor = 26 April 2026 raw signals.",
  "first_audit_reference": {
    "verdict": "INFLATED 2.5x",
    "phase1_realistic_mid": 20000,
    "remaining_red_issues_before_deep": ["universe survivorship", "MOO routing broken", "Sharpe 7.58 violates own cap"]
  },
  "summary": {
    "additional_red_findings": 4,
    "additional_yellow_findings": 5,
    "additional_green_findings": 2,
    "cumulative_inflation_estimate": "8x to 16x (vs first audit's 2.5x). Combined inflation vs ORIGINAL $96K = ~20-40x",
    "true_realistic_year1_phase1_dollar": "$1,200 to $4,000 net of all DEEP factors (was $20K mid in first audit)",
    "biggest_new_killers": [
      "G_spread (-50%): 18/26 SLs hit on tickers with 0.5-2% bid-ask spread = real SL is -2% not -1%",
      "F_fees (-30%): GPUS at $0.17 = 88K shares = $899 round-trip fee on $1050 gross PnL = 86% fee/PnL",
      "C_concurrent_cap (-25%): max=2 cap drops 12 of 26 raw April signals (Apr 8 cluster: 9 sigs -> 2 taken)",
      "J_multiple_comparisons (-35%): v3 picked from {v1..v5} -> Bonferroni-like Sharpe haircut"
    ]
  },
  "findings": [
    {
      "id": "A",
      "topic": "Borrow / locate cost (HTB risk on small-cap pumps)",
      "severity": "YELLOW",
      "evidence": "3 of 26 April trades fit HTB-proxy heuristic (prev_close<$2 AND gap>30%): QNCX (+105%), BFRG (+33%), MYSE (+283% gap = clear news catalyst). For typical px<$5 pump tickers borrow fee assumption = 30% annual avg, with HTB spikes to 100%+. At $15K position * 1.5h hold = $0.77/trade base, $2.57/trade HTB. Across 280 trades/yr = ~$216-720 borrow cost. ALSO: HTB unavailable = trade simply does not happen = ~11.5% trade count haircut.",
      "quantified_impact": "-$216 to -$720/yr borrow drag + 11.5% trade count haircut. Combined factor 0.876x.",
      "fix_or_mitigation": "Pre-trade locate check via IBKR API. If shares not available or fee >50%, skip ticker. Build borrow_fee column in daily snapshot. Most realistic to assume QNCX/BFRG/MYSE class trades will frequently fail locate."
    },
    {
      "id": "B",
      "topic": "Gap-through-stop execution (1% SL becomes 3-8%)",
      "severity": "RED",
      "evidence": "8 of 18 SL exits had high_day > open*1.05 (gap-through risk zone). Mean worst excursion +23.7% above entry, max +77.8% (MAXN Apr 6: open $0.76, high $1.35). Stop-market on SHORT covers at high tick when price gaps through. Top offenders: MAXN +77.8%, SNBR +45.1%, EVTL +17.9%, QNCX +15.1%, SPCE +14.0%. With 1% stop limit triggered but stop-market executing 5-15% above stop_target, real SL fill = -3% to -10% on 44% of SL hits. Realistic mean SL = -2.3% (not -1%).",
      "quantified_impact": "-$1,932/yr extra loss vs assumed (at 280 trades, $750 size, 70% SL rate, +0.92% extra cost per trade). Factor 0.903x.",
      "fix_or_mitigation": "Use stop-LIMIT not stop-market on SL (accepts some unfilled stops as catastrophic-loss prevention). Reduce position size on prev_close<$2 tickers where intraday +50% pumps occur regularly. Add hard $-loss cap = -2% portfolio risk per trade not -1% nominal stop."
    },
    {
      "id": "C",
      "topic": "Concurrent position cap (max=2) destroys breadth",
      "severity": "RED",
      "evidence": "Day-of-cluster counts: Apr 2: 3 sigs, Apr 8: 9 sigs, Apr 13: 4 sigs, Apr 16: 4 sigs. After max_concurrent=2 cap: 14 of 26 raw signals taken (54%). Apr 8 dropped 7 of 9 candidates by random/score selection. The 12 dropped signals contained 4 TP wins (USEG, GPUS, BATL_13, BFRG, SLDP) by replay outcome. With max=2 cap and selection randomness, ~half of TP wins are lost. Trade count drops 46% on top of any other filter haircut.",
      "quantified_impact": "-25% to -45% on annual $/yr at same per-trade WR (because dropped signals had higher TP-rate than survivors). Factor 0.55-0.75x.",
      "fix_or_mitigation": "Either lift max_concurrent to 4-5 (with proportional position sizing reduction so total exposure stays $10K), OR build a deterministic ranking (highest gap_pct first) so survivor selection is reproducible. Currently 'random pick when over-cap' cannot be backtested honestly."
    },
    {
      "id": "D",
      "topic": "Capital scaling: share count infeasibility on sub-$1 tickers",
      "severity": "YELLOW",
      "evidence": "$15K position at sub-$1 prices = absurd share counts: GPUS @ $0.17 = 88,235 sh, FFAI @ $0.37 = 41,040 sh, DGNX @ $0.63 = 23,719 sh, MAXN @ $0.76 = 19,758 sh. For sub-$1 small-cap, available short borrow is typically 10-50K shares max per account. 4 of 26 trades (15%) need >20K shares. Realistic fill = 30-50% of intended size. Compounding effect: position size capped = $/yr capped on the very tickers (GPUS, BFRG) that actually WON in April.",
      "quantified_impact": "-10% on annual $/yr. Factor 0.90x. Worse: small-priced tickers happen to be 4/7 of TP wins -> capping these = capping the upside disproportionately.",
      "fix_or_mitigation": "Add min_price = $0.50 filter (drops GPUS-class). Or cap position size = min($150/SL_pct, $5/share_count, 5% of pm_dvol). YAML currently has no share-count cap."
    },
    {
      "id": "E",
      "topic": "Days-zero-trades ignored in compounding & Sharpe model",
      "severity": "YELLOW",
      "evidence": "April 2026: 18 trading days, signals on only 8 days (Apr 2, 6, 7, 8, 9, 13, 14, 16) = 56% zero-trade days. Compounding model in 12month_forecast.html assumes daily PnL; zero days = no growth. Sharpe computed only over active days inflates daily-Sh by sqrt(252/N_active) ~ sqrt(252/112) = 1.5x. So reported daily Sh 7.58 over active-days = ~5.0 over all-days (matches strategy's own daily_sh_realistic=5.0).",
      "quantified_impact": "Sharpe overstated 1.5x if computed on active-days-only. $/yr impact zero (already counts active-days actual PnL). But annualization formula sensitive: 660 trades/yr / 112 active days = 5.9 trades/active day vs strategy yaml says max=2. Inconsistency suggests trade count claim itself uses active-days basis incorrectly.",
      "fix_or_mitigation": "Always annualize over 252 trading days, not active-trade days. Recompute annualized Sharpe as daily_PnL_series_with_zeros.std()*sqrt(252)."
    },
    {
      "id": "F",
      "topic": "Fee/commission drag (IBKR Pro $0.005/sh min $1, sub-$1 tickers catastrophic)",
      "severity": "RED",
      "evidence": "Computed actual round-trip IBKR Pro fees on April trades: GPUS 88,235 sh = $899 fee, FFAI 41K sh = $418, DGNX 24K sh = $237, MAXN 20K sh = $201, BFRG 13K sh = $130. Total fees on 26 trades = $3,088. Total gross PnL = $5,206. **Fee/PnL ratio = 59%**. The GPUS TP win of $1,050 gross becomes $151 net after $899 fees. Strategy wins on small-priced TPs are MOSTLY EATEN by per-share commissions.",
      "quantified_impact": "-30% to -60% on net $/yr. At realistic 280 trades/yr with mean 30K shares per low-priced trade, fee drag = $5K-$15K/yr. Factor 0.40-0.70x.",
      "fix_or_mitigation": "Switch to IBKR Lite (zero commission) but pay payment-for-order-flow drag (5-10 bps) which is BETTER for high-share-count low-priced trades. Or cap share count at 5,000 per trade. Or hard min_price = $1.00."
    },
    {
      "id": "G",
      "topic": "Stop-loss bid-ask spread cost",
      "severity": "RED",
      "evidence": "Pump tickers in 09:30-11:00 window have wide spreads: typical 0.5% for sub-$5 names with 1-3M ADV, 1-2% for sub-$2 names. Stop-market = pay full spread. For a 1% SL on SHORT, real fill = stop_target + (spread/2 + impact) = +1% nominal stop becomes -1.5% to -3% realized. 18 SLs in April month * +1% extra cost on $750 size = $135 extra loss. Annualized 280 trades * 70% SL rate * +1% extra = -$1,470/yr.",
      "quantified_impact": "-50% on annual $/yr (most aggressive of all factors). Mean PnL per trade drops from +0.45% to -0.25% = strategy goes negative if combined with H/I/J. Factor 0.50x.",
      "fix_or_mitigation": "Use marketable LIMIT orders for SL (limit at stop_target * 1.005 = accept 0.5% extra slip cap, refuse worse). Some stops won't fill = catastrophic risk if price keeps gapping. Or widen SL to -2% nominal absorbing spread cost into design."
    },
    {
      "id": "H",
      "topic": "Tax drag (US short-term capital gains on intraday SHORT)",
      "severity": "YELLOW",
      "evidence": "All trades held <1.5h = short-term gains taxed as ordinary income. Federal 24-37% (depends on bracket) + state (NY 6.85% / CA 9.3% / TX 0%). Assume 30% blended. $20K Year 1 gross -> $14K net. Affects compounding rate (less reinvested capital each cycle).",
      "quantified_impact": "-30% on annual NET $/yr. Factor 0.70x. Compounding to Phase 3 takes longer.",
      "fix_or_mitigation": "Trade in retirement account (Roth IRA limited margin not great for SHORT). Or accept tax drag and forecast NET-of-tax explicitly in dashboard."
    },
    {
      "id": "I",
      "topic": "Regime artifact (2023-2026 high-vol pump era)",
      "severity": "YELLOW",
      "evidence": "Backtest period coincides with peak retail-mania regime: AI bubble (2023-2024 NVDA/SMCI), crypto resurgence (2024-2025 MSTR/CLSK), GME/AMC echoes, SPAC dregs (SPCE/JBLU/EVTL all backtest tickers). Pump-fade strategies historically deliver Sh 0.5-1.5 in low-vol regimes (2017-2019). 2026-2027 may see vol compression -> small-cap pumps fewer/smaller.",
      "quantified_impact": "-30% on forward $/yr if regime mean-reverts. Factor 0.70x. No way to time this; structural assumption.",
      "fix_or_mitigation": "Build regime detector: VIX 20d MA, small-cap RVOL (RUT/SPY ratio), # daily gaps >10%. Pause strategy when 30d-rolling-pump-count drops below 15 (1 per day)."
    },
    {
      "id": "J",
      "topic": "Multiple-comparisons selection bias (v3 picked from v1..v5)",
      "severity": "RED",
      "evidence": "Memory `strategy_moo1100_asym_FINAL.md` confirms v1, v2, v3, v4, v5 were tested. v3 survives audit; v4/v5 failed permutation tests. Selecting best of N=5 versions inflates apparent edge by ~sqrt(N) = 2.24x in Sharpe terms (Bonferroni-like). Per-trade Sh 14 reported at one point in v5 research = clearly N-overfit. True OOS Sharpe of selected v3 likely 35% lower than reported.",
      "quantified_impact": "-35% on $/yr. Factor 0.65x. This is on top of already-honest first-audit estimate.",
      "fix_or_mitigation": "Apply White's Reality Check or SPA test on v1-v5 family. Report adjusted-Sharpe explicitly. Reserve untouched OOS window (e.g. 2026-Q3+) for final validation BEFORE capital scaling."
    },
    {
      "id": "K",
      "topic": "EOD daily-bar exit vs strategy 11:00 forced exit",
      "severity": "GREEN",
      "evidence": "Verified all 7 April TP wins have close BELOW the SHORT TP target (target = open*0.93). Specifically: USEG close $0.854 vs target $0.914 (-6.5%), GPUS close $0.142 vs target $0.158 (-10.2%), all TPs show close < target. This means low_day was reached and price KEPT FALLING beyond TP -> 11:00 forced exit would have realized AT LEAST the TP gain. NO inflation from this angle. Backtest's use of day-low for TP detection is conservative-friendly here.",
      "quantified_impact": "0% (clean). Factor 1.00x.",
      "fix_or_mitigation": "None needed. Counter-intuitively this is one of the few audit angles that does NOT inflate the forecast."
    },
    {
      "id": "L",
      "topic": "Margin/cash drag (Reg-T)",
      "severity": "GREEN",
      "evidence": "Strategy is intraday (exit 11:00, never holds overnight under normal conditions) so 4:1 margin applies. $10K equity = $40K BP intraday. With max 2 concurrent at $15K each = $30K, well within 4:1 cap. No margin call risk. Overnight risk only on halt-and-cannot-exit scenario (low probability for sub-$5 names which freely trade).",
      "quantified_impact": "0% under normal ops. Factor 1.00x.",
      "fix_or_mitigation": "Reserve $5K cash buffer for halt-overnight scenario. Already standard practice."
    },
    {
      "id": "M",
      "topic": "SEC fee + FINRA TAF on cover (sell-side)",
      "severity": "YELLOW",
      "evidence": "SEC fee $0.0000278/share + FINRA TAF $0.000166/share on the SHORT cover (sell-side equivalent). On GPUS 88K sh: $1.66 + $14.65 = $16.31. On 280 trades/yr avg 30K shares: $1,628/yr. Small relative to commissions but not zero.",
      "quantified_impact": "-$1,500 to -$2,000/yr drag. Already partially folded into F. Factor 0.97x marginal.",
      "fix_or_mitigation": "Already unavoidable; same fix as F (cap share count or hard min_price)."
    },
    {
      "id": "N",
      "topic": "Idle capital opportunity cost",
      "severity": "GREEN",
      "evidence": "56% zero-trade days * $10K equity * 5% risk-free annual / 252 = $11/day idle * 142 zero days = $1,560/yr foregone. Strategy uses cash sweep at IBKR earning ~4.5% on idle, so this cost is mostly recovered. Real opportunity cost vs SPY buy-and-hold = much larger (~$700/yr expected SPY return on $10K) but that is comparison to passive, not strategy cost.",
      "quantified_impact": "-$0 to -$200/yr (covered by IBKR cash sweep). Factor 0.99x.",
      "fix_or_mitigation": "Ensure cash sweep enabled. Already standard."
    },
    {
      "id": "O",
      "topic": "Hidden correlation on cluster days inflates Sharpe",
      "severity": "YELLOW",
      "evidence": "Apr 8: 9 simultaneous signals all on retail-mania day, market-wide pump regime. Correlation across same-day positions ~0.4-0.6 (not the assumed 0). Effective N for variance calc = N/(1+rho*(N-1)) = 9/(1+0.5*8) = 1.8 not 9. Daily PnL stdev computed as if independent = understated by sqrt(5) = 2.24x. Reported daily Sh inflated 1.5-2x ON TOP OF first-audit's per-trade-vs-daily inflation.",
      "quantified_impact": "0% on $/yr (mean PnL unchanged). Sharpe overstated additional 1.5-2x. Combined with first-audit per-trade inflation, true daily Sh = Sh_reported / 3 = 7.58 / 3 = ~2.5. Factor 1.00x for $/yr but Sharpe 2.5 not 5-7.",
      "fix_or_mitigation": "Compute Sharpe on daily PnL series with zero-fill, NOT per-trade. Cluster cap at max=2 already partially addresses this."
    }
  ],
  "factor_breakdown_for_dollar_impact": {
    "A_borrow_unavailability": 0.876,
    "B_gap_through_stops": 0.903,
    "C_concurrent_cap": 0.75,
    "D_share_count_infeasibility": 0.90,
    "E_zero_days_compound": 1.00,
    "F_fees_drag": 0.70,
    "G_spread_on_stops": 0.50,
    "H_tax_drag_pretax_excluded": 0.70,
    "I_regime_revert_risk": 0.70,
    "J_multiple_comparisons": 0.65,
    "K_eod_vs_1100": 1.00,
    "L_margin_drag": 1.00,
    "M_sec_taf_partial_in_F": 1.00,
    "N_idle_capital": 1.00,
    "O_correlation_sharpe_only": 1.00,
    "combined_pretax_dollar_impact": "0.876 * 0.903 * 0.75 * 0.90 * 0.70 * 0.50 * 0.65 = 0.122x",
    "combined_posttax_dollar_impact": "0.122 * 0.70 = 0.085x",
    "combined_with_regime_pretax": "0.122 * 0.70 = 0.085x",
    "combined_full_pessimistic": "0.122 * 0.70 * 0.70 = 0.060x"
  },
  "consolidated_realistic_year1": {
    "before_first_audit_forecast": 96000,
    "after_first_audit_phase1_mid": 20000,
    "after_DEEP_audit_phase1_low": 1200,
    "after_DEEP_audit_phase1_mid": 2400,
    "after_DEEP_audit_phase1_high": 4000,
    "calculation": "Phase 1 mid $20K * 0.122 (pretax dollar factors A,B,C,D,F,G,J) = $2,440. Range $1.2K-$4.0K depending on which YELLOW factors materialize.",
    "phase1_likeliest": 2400,
    "phase3_likeliest_pretax_routing_fixed": 9000,
    "phase3_likeliest_posttax_routing_fixed": 6300,
    "biggest_remaining_unknowns": [
      "Borrow availability on QNCX/BFRG/MYSE-class small-cap pumps (no API check pre-trade currently)",
      "True bid-ask spread on stops in 09:30-09:35 window (need TAQ quote data, not available)",
      "Multiple-comparisons true penalty (need White's Reality Check on v1-v5 family)",
      "Regime persistence: will 2026-H2 retain retail-pump frequency at >15/month?",
      "Concurrent-cap selection algorithm: random vs deterministic completely changes WR distribution",
      "Real fill price on stop-market vs stop-limit on sub-$1 tickers (need 1m bar audit, not done)"
    ],
    "honest_one_liner": "After DEEP audit Phase 1 ($10K BP, $15K notional max) realistic Year 1 = $1.2K to $4.0K NET. Total inflation vs original $96K forecast ~ 24x to 80x. Strategy is approximately BREAK-EVEN to slightly-positive at sub-$5 px-floor; becomes consistently profitable only with min_price=$1.00 + IBKR Lite + max_concurrent=4 + stop-limit + locate-API + post-routing-fix."
  },
  "recommended_pre_deploy_actions_priority": [
    "P0: Add min_price=$1.00 hard filter (kills GPUS-class fee disasters; keeps BATL/SPCE/JBLU class)",
    "P0: MOO routing fix (pre-existing P0 from first audit)",
    "P1: Switch IBKR Lite for sub-$5 names (commission-free more important than PFOF cost)",
    "P1: Add pre-trade locate API check + skip if borrow_fee>50% or shares_avail<5K",
    "P1: Use stop-LIMIT not stop-market for SL (cap 0.5% extra slip)",
    "P2: Lift max_concurrent to 4-5 with proportional sizing (preserve cluster-day breadth)",
    "P2: Re-audit Sharpe using daily-PnL with zero-fill across all 252 days",
    "P3: Apply White's Reality Check on v1-v5 family for true OOS Sharpe estimate"
  ]
}