RBOT EVIDENCE LAB
公开 AI 交易测评

交易报告分析:Vigorous EA · MT4 EURUSDi M15 Backtest

The Vigorous EA backtest on EURUSDi M15 shows a modest +3.37% return over 276 days but is driven by a handful of outlier trades, carries a 19.51% max drawdown, and rests on Control-point modelling quality that the tester itself says must not be considered.

分析时间:2026-10-12 23:39

本页分析用户提交的 MT4/MT5 报告,不认证账户归属、不重跑独立回测,也不预测未来收益。

0–100 AI 评分是基于证据的评审判断,不是盈利概率;应结合评分依据与未知事项阅读。

阅读指标公式、报告类型对比与证据边界
回测

Vigorous EA · MT4 EURUSDi M15 Backtest

0–100 AI 评分是基于证据的评审判断,不是盈利概率;应结合评分依据与未知事项阅读。

38/100策略安全评分
3.4%收益
19.51%最大回撤
1.29PF
505交易数

数据透明度

本页分析用户提交的 MT4/MT5 报告,不认证账户归属、不重跑独立回测,也不预测未来收益。

数据来源用户上传的 MT4 报告
经纪商TradersWay-Demo
品种EURUSDi
周期M15
交易区间2026-01-06 — 2026-10-09
报告类型回测
报告交易数505
处理方式解析器提取与支持的核算 + AI 风险解读;非账户认证

核心指标

下列指标来自导入记录与支持的核算,比较前应核对资金基数、费用、出入金和交易期间。

收益3.4%
最大回撤19.51%
PF1.29
胜率76.83%
净利润336.97
单笔期望0.67
平均盈亏比0.39

增长

以导入报告的起始资金归一化为 100。该图按已保存的逐笔结果采样,不等同于经纪商实时权益曲线。

以导入报告的起始资金归一化为 100。该图按已保存的逐笔结果采样,不等同于经纪商实时权益曲线。

详细计算结果

下列指标来自导入记录与支持的核算,比较前应核对资金基数、费用、出入金和交易期间。

10,000.00初始资金
336.97净利润
76.83%胜率
1,488.87总盈利
1,151.90总亏损(绝对值)
3.85平均盈利单
-9.85平均亏损单
0.39平均盈亏比
0.67单笔期望
334.41最大盈利单
-139.93最大亏损单
31最长连胜
13最长连亏
0.03平均手数
276.4 天样本跨度
1.83日均交易

收益率 = 净利润 ÷ 初始资金 × 100%;PF = 总盈利 ÷ 总亏损绝对值;单笔期望 = 净利润 ÷ 已平仓交易数;平均盈亏比 = 平均盈利单 ÷ 平均亏损单绝对值;最大回撤采用报告输出的最大回撤值。若报告含入金或出金,收益率应结合现金流另行解释。

策略安全评分依据

Verified: 505 closed rows, 2026-01-06 to 2026-10-09 (276.4 days), net profit 336.97 on a 10,000 base, profit factor 1.29, win rate 76.83%.

  • Verified: top 1% of trades (6 rows) account for 64.35% of gross profit; the remaining 499 rows net -621.22 with profit factor 0.46.
  • Verified: top 5% (26 rows) account for 75.08% of gross profit; the rest nets -780.86 with profit factor 0.32.
  • Verified: max drawdown of 19.51% against total net return of only +3.37% — a poor return-to-drawdown ratio.
  • Verified: a single clustered close on 2026-06-30 accounts for roughly +450 swing in balance across many rows, an extreme concentration event.
  • Verified: modelling is Control points, modelling quality is n/a, 4 mismatched charts errors, and the tester explicitly warns results must not be considered.
  • Verified: StopLoss is configured as 0 in report_context.
  • Pending: no Demo or Real forward record is supplied; only a Strategy Tester report exists.

关键风险信号

Profit concentration: 6 of 505 rows (top 1%) supply 64.35% of gross profit; without them the book is net negative.

  • Return/drawdown asymmetry: +3.37% return against a 19.51% max drawdown — the payoff is far below the risk taken.
  • A single 2026-06-30 close cluster moves balance from ~10,224 to ~11,188 then back to ~10,238, showing extreme single-event dependency.
  • Control points modelling with 4 mismatched charts errors means the reported P/L, PF, and win rate cannot be relied upon as the strategy's true behaviour.
  • StopLoss set to 0 in report_context — protective-stop behaviour is not evidenced by the supplied configuration.
  • Close-only rows without paired tickets prevent verification of whether the many same-timestamp closes are independent trades, layered entries, or position management.

现有数据无法判断

Whether same-timestamp closes (e.g., 2026-01-06 01:43, 2026-02-02 15:21, 2026-06-30 17:42) are independent trades, layered entries, or position management — no entry tickets accompany the rows.

  • Whether the 0.93-lot, 0.62-lot, and 0.41-lot sizes on 2026-06-30 follow a documented sizing rule — the rule is not provided.
  • True spread, commission, swap, and slippage impact on net P/L (slippage 10 is a tester setting, not a cost model).
  • Whether the strategy has any protective stop in live operation given StopLoss: 0.
  • Behaviour under tick-level or real-tick modelling, since only Control points was used.

下一步行动

Rerun the backtest on the correct EURUSDi chart history to clear the mismatched-charts errors, then report the resulting modelling-quality value.

  • Rerun using a higher-quality modelling mode (e.g., every tick or real ticks) to test whether the outlier-driven P/L structure survives.
  • Export paired entry/exit tickets to verify whether same-timestamp closes are separate trades, layered entries, or position management.
  • Run sensitivity analysis on spread, commission, and slippage to test whether the thin net edge (336.97 on 10,000) survives realistic costs.
  • Forward-validate on a Demo account for a comparable duration before drawing any live-performance conclusion.

常见问题

0–100 AI 评分是基于证据的评审判断,不是盈利概率;应结合评分依据与未知事项阅读。

Q. Is the +3.37% return broad-based or dependent on a few trades?

A. It is heavily concentrated. The top 1% of closed rows (6 trades) contribute 64.35% of gross profit, and the remaining 499 rows net -621.22 with a profit factor of 0.46. The top 5% (26 trades) contribute 75.08% of gross profit, with the remainder netting -780.86. The headline return therefore rests on a small minority of trades.

Q. What does the 19.51% max drawdown mean relative to the 3.37% return?

A. The account gave up 19.51% at its worst point to earn a net +3.37% over 276.4 days. That is a weak payoff for the risk taken and means the equity curve carried materially more risk than its final return suggests.

Q. How reliable is this backtest given the tester warnings?

A. Reliability is limited. The report uses Control points modelling, reports modelling quality as n/a, lists 4 mismatched charts errors, and the tester itself explicitly warns the results must not be considered. Under those conditions the reported profit factor of 1.29, 76.83% win rate, and 19.51% drawdown cannot be trusted as representative of live behaviour.

Q. Why do many trades close at the same timestamp, such as 2026-06-30 17:42?

A. The supplied rows are close-only events and do not include paired entry tickets, so the same-timestamp closes cannot be classified as independent trades, layered entries, or position management from this evidence. This needs confirmation from paired ticket data; the 2026-06-30 cluster alone swings balance from roughly 10,224 up to 11,188 and back to about 10,238, which is why it should be verified before drawing conclusions.

Q. How is RBOT different from asking a general-purpose AI assistant (such as ChatGPT, Gemini, Claude or DeepSeek) to analyse my report myself?

A. RBOT parses MT4/MT5 reports into structured metrics and trade records before AI review. Supported calculations can be checked against the original export; AI wording and scores can vary and are not guaranteed to be deterministic or comparable across reports. General AI assistants may also use calculation tools. Missing actual slippage, ownership or forward validation still requires other evidence.