# AI Trader — six-portfolio trial report **Provisional report: final accounting and operational review are still required.** Common window: 2026-09-15T15:26:53.536947+00:00 to 2026-09-29T15:26:53.536947+00:00. State: terminal. Each portfolio began with USD 10,000.00 simulated cash. All six remain in this comparison, including halted portfolios. | Portfolio | Cutoff USD | USD gain | SGD gain | Peak SGD drawdown | Risk coverage | Decision coverage | | --- | ---: | ---: | ---: | ---: | ---: | ---: | | trend_following:qwen | 10,000.00 | 0.00 | 48.00 | 0.29% | 98.38% | 9.09% | | trend_following:astra | 10,000.00 | 0.00 | 48.00 | 0.29% | 98.97% | 9.09% | | mean_reversion:qwen | 9,909.51 | -90.49 | -67.58 | 2.05% | 99.28% | 9.09% | | mean_reversion:astra | 9,909.99 | -90.01 | -66.97 | 2.05% | 97.26% | 9.09% | | defensive_rotation:qwen | 10,021.71 | 21.71 | 75.73 | 0.50% | 99.13% | 9.09% | | defensive_rotation:astra | 10,021.95 | 21.95 | 76.04 | 0.50% | 99.56% | 9.09% | ## trend_following:qwen Recorded fees: USD 0.00. Fills: 0; decisions: 3. Mean sampled exposure: 0.00% (3894 samples). SGD gain split: trading 0.00; FX 48.00. Terminal cash: USD 10,000.00; change after cutoff: USD 0.00. This change includes liquidation and later cash adjustments, not the measured trial return. Additional execution-cost sensitivity (USD net gain, per executed notional): 0 bps: 0.00, 5 bps: 0.00, 10 bps: 0.00, 25 bps: 0.00. USD cash benchmark return: 0.00% USD / 0.38% SGD. SPY benchmark: 0.93% USD / 1.31% SGD; accrued dividend income USD 24.55. Assessment: inconclusive pending final review. Risk halt: none recorded. Coverage thresholds met: False. ## trend_following:astra Recorded fees: USD 0.00. Fills: 0; decisions: 1. Mean sampled exposure: 0.00% (3894 samples). SGD gain split: trading 0.00; FX 48.00. Terminal cash: USD 10,000.00; change after cutoff: USD 0.00. This change includes liquidation and later cash adjustments, not the measured trial return. Additional execution-cost sensitivity (USD net gain, per executed notional): 0 bps: 0.00, 5 bps: 0.00, 10 bps: 0.00, 25 bps: 0.00. USD cash benchmark return: 0.00% USD / 0.38% SGD. SPY benchmark: 0.93% USD / 1.31% SGD; accrued dividend income USD 24.55. Assessment: inconclusive pending final review. Risk halt: none recorded. Coverage thresholds met: False. ## mean_reversion:qwen Recorded fees: USD 0.01. Fills: 2; decisions: 3. Mean sampled exposure: 61.80% (3894 samples). SGD gain split: trading -115.15; FX 47.57. Terminal cash: USD 9,922.76; change after cutoff: USD 13.25. This change includes liquidation and later cash adjustments, not the measured trial return. Additional execution-cost sensitivity (USD net gain, per executed notional): 0 bps: -90.49, 5 bps: -93.58, 10 bps: -96.66, 25 bps: -105.92. USD cash benchmark return: 0.00% USD / 0.38% SGD. SPY benchmark: 0.93% USD / 1.31% SGD; accrued dividend income USD 24.55. Assessment: inconclusive pending final review. Risk halt: none recorded. Coverage thresholds met: False. ## mean_reversion:astra Recorded fees: USD 0.01. Fills: 2; decisions: 1. Mean sampled exposure: 61.79% (3895 samples). SGD gain split: trading -114.54; FX 47.57. Terminal cash: USD 9,923.56; change after cutoff: USD 13.57. This change includes liquidation and later cash adjustments, not the measured trial return. Additional execution-cost sensitivity (USD net gain, per executed notional): 0 bps: -90.01, 5 bps: -93.10, 10 bps: -96.18, 25 bps: -105.44. USD cash benchmark return: 0.00% USD / 0.38% SGD. SPY benchmark: 0.93% USD / 1.31% SGD; accrued dividend income USD 24.55. Assessment: inconclusive pending final review. Risk halt: none recorded. Coverage thresholds met: False. ## defensive_rotation:qwen Recorded fees: USD 0.01. Fills: 1; decisions: 2. Mean sampled exposure: 30.49% (3895 samples). SGD gain split: trading 27.63; FX 48.10. Terminal cash: USD 10,022.54; change after cutoff: USD 0.83. This change includes liquidation and later cash adjustments, not the measured trial return. Additional execution-cost sensitivity (USD net gain, per executed notional): 0 bps: 21.71, 5 bps: 20.20, 10 bps: 18.68, 25 bps: 14.14. USD cash benchmark return: 0.00% USD / 0.38% SGD. SPY benchmark: 0.93% USD / 1.31% SGD; accrued dividend income USD 24.55. Assessment: inconclusive pending final review. Risk halt: none recorded. Coverage thresholds met: False. ## defensive_rotation:astra Recorded fees: USD 0.01. Fills: 1; decisions: 1. Mean sampled exposure: 30.49% (3895 samples). SGD gain split: trading 27.93; FX 48.11. Terminal cash: USD 10,022.59; change after cutoff: USD 0.64. This change includes liquidation and later cash adjustments, not the measured trial return. Additional execution-cost sensitivity (USD net gain, per executed notional): 0 bps: 21.95, 5 bps: 20.44, 10 bps: 18.92, 25 bps: 14.38. USD cash benchmark return: 0.00% USD / 0.38% SGD. SPY benchmark: 0.93% USD / 1.31% SGD; accrued dividend income USD 24.55. Assessment: inconclusive pending final review. Risk halt: none recorded. Coverage thresholds met: False. ## Interpretation and limitations The SGD 500/month reference corresponds to about SGD 233 over 14/30 of a month. Observed target pace is descriptive, not a forecast or evidence of dependable income. No trial is automatically extended or promoted to real money. The SPY benchmark can be nearly fully invested while each strategy caps any ETF target at 35%. Midpoint marks and simulated fills do not reproduce real liquidity, slippage, taxes or all fees. Exposure is averaged over available samples, not imputed across missing intervals. Recorded broker cash and fills, marked at IEX midpoint; unposted receivables and simulator omissions require separate final review. Cost sensitivity is additional cost per executed notional, not a replacement for recorded fees. Final review remains: - Reconcile unposted portfolio dividend receivables, withholding, corporate actions and delayed fees; compare with benchmark accrual accounting. - Verify provider corporate-action completeness and all native terminal account snapshots. - Review risk-policy, paper-boundary and reconciliation events; coverage thresholds alone do not prove operational validity. - Attribute missed opportunities and model availability, review exposure differences and costs, and record actual user attention. - Compare Qwen/Astra within each strategy and strategies within each model; assign the final three-way assessment without tuning to these outcomes. The JSON export retains cutoff quotes, FX, benchmarks, missing risk-minute slots, cash adjustments and cost sensitivity. The CSV provides one row per portfolio.