BTC SCALPER IA[BTCUSD][H1][notandor.cn]… AI 测评｜QuantTester
BTC SCALPER IA[BTCUSD][H1][notandor 报告：收益 20,969.9%、回撤 60.43%、PF 6.03、488 笔交易。AI 风险解读。
策略安全信用评级: B (60 / 100) · 常态中性 (Moderate)
交易风险分析结果
Strategy Type: Strict Lot Discipline. Classified as Single-Order Trend (high-frequency intraday scalper): a fixed 0.1-lot, always-one-position BTCUSD system with a hard 1,950 USD clustered stop discipline and 4.71 realized payoff drove PF 6.03 and net profit of 2,096,994 USD, but a 60.43% maximum drawdown against a 10,000 USD base confines the profile to speculative, not allocatable, capital.
Darwinex 集中度压力测试 (Top 5% Windfall Dependence Test)
剔除排名前 25 笔（5%）极端盈利单后：毛利占比 34.58%，剩余净利润 $1,227,542.00，剩余 PF 3.94。
数据透明度与环境说明
- MT4/MT5: MetaTrader 5
- 品种: BTCUSD
- 周期: M5 (2025.01.03 - 202
- 总净利润: $2,096,994.00
- 总收益率: 20,969.9%
- 最大回撤: 60.43%
- 总交易笔数: 488
- 盈利因子 (PF): 6.03
AI 识别的核心高危漏洞清单
- [Rs Risk Stability / Drawdown Depth] Peak-to-trough equity impairment reached 60.43% against a 10,000 USD starting base while the single-trade risk unit is fixed at 1,950 USD (19.5% of initial capital); a modest streak of only 6 consecutive losses (longest_loss_streak) would erase ~11,700 USD, demonstrating that survival depends on the win rate never reverting and exposes a live transmission mechanism of compounding balance erosion during a regime flip.
- [La Loss Aversion / Payoff Asymmetry] Win rate is 56.15% with a 4.71 payoff; the arithmetic is robust, but the realised profit distribution is bimodal (a repeated cluster of exactly -1,950 USD stops alongside 5,850 USD TP exits), so any compression of large trend runs toward the 1,950 USD stop level narrows the payoff ratio and rapidly degrades expectancy toward the break-even boundary.
- [Execution Feasibility / Friction Sensitivity] Average holding time is 0.4 minutes across 488 trades (1.15 trades/day) on an M5 chart with M1-scale entries; at this turnover, spread and slippage on BTCUSD—unquantified in report_context—are incurred on every position, and a widened spread of even a fraction of the 19,500-point SL distance represents a recurring proportional drag that the backtest does not price.
- [Momentum Decay / Equity Path] Balance advanced from 10,000 USD to 2,106,994 USD, with peak balance growth concentrated in trend phases (e.g. 2025-02 to 2025-09); the 60.43% drawdown shows the curve is capable of a deep, prolonged unhedged excursion, so a stalled trend regime would leave the account underwater for an extended period without any averaging-down recovery mechanism.
客观数学打分明细
- [Base Score] All valid statements start at 50 points neutral benchmark (Grade B-)
- +15 pts [Lot Discipline]: Fixed/proportional lot execution, no adverse counter-trend doubling
- -25 pts [No Hard SL]: No disciplined hard stop-loss; holding unhedged drawdown to severe depths
- +15 pts [Payoff Ratio]: Win/loss ratio reached 4.71 (> 1.8), strong asymmetric payoff
- +5 pts [Expectancy]: Overall Profit Factor reached PF 6.03 with solid positive mathematical expectancy
- +10 pts [Organic Profit]: Removing top 5% windfall deals leaves PF 3.94 >= 1.2; robust and non-luck-driven
- -20 pts [Extreme Drawdown]: Max historical drawdown reached 60.4% (> 35%), weak defensive execution
- +10 pts [Sample Horizon]: Sample duration spans 425 days (>= 180 days), covering diverse regimes
- Final Strategy Safety Score: 60 / 100 (Grade B).
下一步行动与优化建议
- Re-run the backtest with explicit commission and realistic BTCUSD spread (including widened-spread scenarios) to test whether PF 6.03 survives friction given the 0.4-minute average holding period.
- Stress-test the equity curve under spread widening of 2x and 3x the assumed value to estimate the break-even cost threshold implied by the 4.71 payoff ratio.
- Reconstruct paired entry and exit tickets for same-timestamp rows to confirm single-position discipline is genuine and not a reporting artefact.
- Extend the sample window backward and forward to include a low-volatility BTC regime and verify that the 1,950 USD clustered stop remains the termination level.
- Run a walk-forward parameter sensitivity sweep around InpStopLoss 19,500 and InpTakeProfit 58,500 to test for parameter plateau versus isolated peak.
关于本报告的关键质询与答疑 (Q&A About This Report)
The backtest shows a 20,969% return and PF 6.03 — why does the strategy score only Grade B?
The rating is driven by internal downside risk, not by return. The strategy's maximum drawdown reached 60.43% of a 10,000 USD base, and the audit facts show no configured hard stop-loss was set (sl_preset 0); protective termination is inferred solely from the clustered -1,950 USD loss distribution. Those two factors weigh heavily against the 4.71 payoff ratio and PF 6.03, producing a Grade B rather than a higher rating.
Are the repeated -1,950 USD losses evidence of a fixed stop, and why does the ledger still record 'No Hard SL'?
The stop_loss_evidence pattern is clustered_fixed_stop: all 214 losing trades terminate at exactly -1,950 USD with max_over_median of 1.0 and a 100% clustered share, which is consistent with an executed protective stop. The deduction reflects that no stop was configured in the inputs and that a 60.43% equity drawdown occurred in the curve, so the configured protection does not correspond to a tightly bounded capital-at-risk envelope. The two statements together show that trade-level stop discipline exists but account-level risk control was loose.
Is the profit dependent on a few lucky trades?
Only partly. The top 1% of trades (5 rows) account for 11.73% of gross profit and the top 5% (25 rows) for 34.58%; stripping the top 5% still leaves 1,227,542 USD net profit at PF 3.94, so the edge survives outlier removal and is not purely luck-driven. The largest single win (91,168 USD) is still large relative to the drawdown scale, so path dependence on expansion moves remains material.
The 0.4-minute average holding time seems extreme — what needs to be verified before this is tradable?
The 488-trade sample is executed over 424.7 days at 1.15 trades/day, so the 0.4-minute figure is a mean across a small number of very long trend holds and many short scalp-like exits; it implies meaningful per-trade cost exposure. Because report_context contains no commission, swap or spread figures, the net-of-cost expectancy is unverified and should be the first item tested before any consideration of capital.
How is QuantTester different from asking a general-purpose AI assistant myself?
QuantTester parses MT4/MT5 reports into structured metrics and trade records before AI review. Supported calculations can be checked against the original export, guaranteeing consistent audit facts.
Which broker is recommended for trading and running MT4/MT5 strategies?
For algorithmic trading, EA execution and reliable backtesting, we strongly recommend IC Markets (Raw Spread Account) with 0.0 pip spreads.