Live Trading Center | ← 返回 factors 首页
Rank29 已停止实盘并降为 P0 archived。 这页不再把它当“主运营页”,而是把相关证据收口成一页审计:future leak 污染有多重、strict-causal 后为什么失效、旧 live / shadow 该怎么留档。
页面结构已压缩:shadow_dashboard 保留为研究/历史口径页,gate_live 保留为 retired debug/raw 细页。任何旧的 Rank29 收益曲线,都只能按“历史污染样本”来读,不能再当作已验证有效的策略证据。
下面用 SOL-USD 在 2026-04-29 23:45 UTC 的案例举例。现在这页默认把 causal replay 当主基准:只有当时真实可见的信息能给出信号,才算可交易;若某个标记只会在事后整段历史重算后才出现,就会被归类成 hindsight-only,不再当成真实可交易星号。
这里用 橙色实线 画“该案例视角下真正参与判定的那条 confirmed line”,灰色虚线只保留作参考轨迹。蓝色竖线 = signal bar;绿色 = entry bar;红色 = exit bar。
当前页面展示的 trend_score 实际就是 tbn_composite_trend。它不是一个连续回归分数,而是三个时间框架趋势状态的整数求和:
tbn_composite_trend = tbn_long_trend + tbn_medium_trend + tbn_short_trend
其中每个 timeframe 的 trend 只取三种值:+1(多头)、0(中性 / 尚未定向)、-1(空头)。本例我同时把 causal 和 hindsight 两个视角摆出来:
causal:
long_trend = 1
medium_trend = 1
short_trend = -1
=> trend_score = 1
hindsight:
long_trend = 1
medium_trend = 1
short_trend = 1
=> trend_score = 3
做多准入要求:trend_score ≥ 2;做空准入要求:trend_score ≤ -2。
以本例的 medium 线为例,配置参数是:swing_medium=30、swing_right=1、min_pivot_gap=5。
确认 pivot 的规则:
pivot high at center c:
high[c] > max(high[c-L : c-1])
and high[c] > max(high[c+1 : c+R])
confirmed on bar c+R
pivot low at center c:
low[c] < min(low[c-L : c-1])
and low[c] < min(low[c+1 : c+R])
confirmed on bar c+R
一旦趋势建立后,线本体用两点式表示:
line_value(i) = anchor_price + line_slope × (i - anchor_origin)
line_slope = (pivot_price - anchor_price) / (pivot_origin - anchor_origin)
本例 medium 线的可见参数是:
bullish breakout condition: close[bar] > line_value[bar]
bearish breakout condition: close[bar] < line_value[bar]
causal view:
anchor_origin = -1
anchor_time = -
pivot_origin = -1
pivot_time = -
line_slope = -
line_value = -
hindsight view:
anchor_origin = -1
anchor_time = -
pivot_origin = -1
pivot_time = -
line_slope = -
line_value = -
signal_bar_close = 83.04
close_minus_line = -
只有当 breakout 落在 line_is_provisional = 0 的 confirmed line 上时,这条信号才允许进入 rank29 live / shadow。若图里某段线只有 A 没有 P,那不是漏画,而是它还停留在 provisional line:第 2 个定义点尚未确认,所以只能暂时把它当成“候选结构”,不能当完整两点趋势线来读。若 causal 与 hindsight 的 pivot / slope / trend_score 不同,就说明这根 bar 被未来确认信息改写过。
low_trend_high_noise = 1 时,这条线会缩到约 25% 仓位,而不是完全停做。你刚才盯到的问题是对的:每一条趋势线的生命周期内都应该是一条直线,而 breakout 也应该和它实际突破的那条线发生可见交汇。所以这张总览图现在进一步改成:同一生命周期继续画成一条直线,若该段后来触发 breakout,就把通向 breakout bar 的最后一段改成虚线延长;同时把 A=anchor、P=pivot 两个定义点直接标出来。若只有 A 没有 P,就表示这段仍是 provisional,第二个点还没确认。
口径一致性:图里的 breakout 圆点现在画在 line_value 上,方块画在当根 close 上,中间用竖向虚线连接;这样你能同时看到“突破发生在什么价位的线”与“收盘最终冲到了哪里”。黄色实心星表示 causal 真信号,粉色空心星表示 hindsight-only:事后回看会出现,但当时并不能诚实算出来。
同一根 bar 里,breakout 点是“满足突破+分数”的候选;这里额外列出 close_minus_line,方便直接看收盘价到底越线了多少。signal_kind=causal 才是当前基准认可的真信号;hindsight_only 表示这根 bar 只有事后重算才会被打星。
| breakout_bar | tf | side | trend_score | line_value | close | close_minus_line | is_provisional | signal_kind |
|---|---|---|---|---|---|---|---|---|
| 2026-04-29 01:45:00+00:00 | long | long | 3 | 83.04428571428572 | 84.1 | 1.0557142857142736 | 0 | hindsight_only |
| 2026-04-29 23:45:00+00:00 | short | long | 3 | 82.253125 | 83.04 | 0.7868750000000091 | 0 | hindsight_only |
每一段对应一条直线生命周期;当 anchor/pivot/slope/trend 任一变化,就结束当前段并开始下一段。现在表里也直接给出 anchor/pivot 的价格与索引;如果 pivot_idx 仍为空,说明这段还只是 provisional。
| tf | segment_start | segment_end | bars | trend | is_provisional | anchor_idx | anchor_px | pivot_idx | pivot_px | anchor_ts | pivot_ts | slope | end_reason |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| short | 2026-04-28 14:30:00+00:00 | 2026-04-29 01:15:00+00:00 | 44 | 1 | 0 | 510 | 82.98 | 539 | 83.62 | 2026-04-28 14:30 UTC | 2026-04-28 21:45 UTC | 0.0220689655172414 | line_missing |
| short | 2026-04-29 04:30:00+00:00 | 2026-04-29 18:15:00+00:00 | 56 | -1 | 0 | 566 | 85.06 | 635 | 83.22 | 2026-04-29 04:30 UTC | 2026-04-29 21:45 UTC | -0.026666666666666717 | lifecycle_change |
| short | 2026-04-29 18:30:00+00:00 | 2026-04-30 00:30:00+00:00 | 25 | 1 | 0 | 622 | 81.4 | 654 | 82.7 | 2026-04-29 18:30 UTC | 2026-04-30 02:30 UTC | 0.04062499999999991 | lifecycle_change |
| short | 2026-04-30 00:45:00+00:00 | 2026-04-30 06:30:00+00:00 | 24 | -1 | 1 | 647 | 84.01 | -1 | nan | 2026-04-30 00:45 UTC | - | 0.0 | window_end |
| medium | 2026-04-28 14:30:00+00:00 | 2026-04-29 16:45:00+00:00 | 106 | 1 | 0 | 510 | 82.98 | 608 | 83.12 | 2026-04-28 14:30 UTC | 2026-04-29 15:00 UTC | 0.0014285714285714344 | line_missing |
| long | 2026-04-28 11:45:00+00:00 | 2026-04-28 14:15:00+00:00 | 11 | -1 | 0 | 367 | 88.08 | 548 | 84.19 | 2026-04-27 02:45 UTC | 2026-04-29 00:00 UTC | -0.021491712707182323 | lifecycle_change |
| long | 2026-04-28 14:30:00+00:00 | 2026-04-29 16:45:00+00:00 | 106 | 1 | 0 | 510 | 82.98 | 608 | 83.12 | 2026-04-28 14:30 UTC | 2026-04-29 15:00 UTC | 0.0014285714285714344 | line_missing |
这张表就是上面图里的原始数值:你可以直接看到哪一根 bar 的 medium_bull_breakout 从 0 变成 1,以及 short / medium / long 三个趋势分数是如何相加成 trend_score 的。
| bar_time | open | high | low | close | short_trend | medium_trend | long_trend | trend_score | medium_bull_breakout | medium_line_is_provisional |
|---|---|---|---|---|---|---|---|---|---|---|
| 2026-04-29 19:45 UTC | 82.08 | 82.29 | 82.06 | 82.19 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 20:00 UTC | 82.2 | 82.53 | 82.17 | 82.26 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 20:15 UTC | 82.27 | 82.53 | 82.23 | 82.53 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 20:30 UTC | 82.53 | 82.72 | 82.47 | 82.6 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 20:45 UTC | 82.6 | 82.81 | 82.58 | 82.62 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 21:00 UTC | 82.62 | 82.95 | 82.52 | 82.84 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 21:15 UTC | 82.83 | 82.95 | 82.75 | 82.92 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 21:30 UTC | 82.92 | 83.09 | 82.85 | 83.05 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 21:45 UTC | 83.05 | 83.22 | 82.89 | 82.91 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 22:00 UTC | 82.91 | 82.96 | 82.81 | 82.96 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 22:15 UTC | 82.96 | 83.03 | 82.85 | 82.88 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 22:30 UTC | 82.88 | 83.05 | 82.87 | 83.03 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 22:45 UTC | 83.03 | 83.13 | 83.0 | 83.07 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 23:00 UTC | 83.07 | 83.08 | 82.97 | 83.08 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 23:15 UTC | 83.07 | 83.08 | 82.82 | 83.02 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 23:30 UTC | 83.02 | 83.09 | 82.94 | 83.01 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-29 23:45 UTC | 83.02 | 83.11 | 82.94 | 83.04 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-30 00:00 UTC | 83.05 | 83.3 | 83.05 | 83.2 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-30 00:15 UTC | 83.21 | 83.35 | 83.1 | 83.27 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-30 00:30 UTC | 83.27 | 83.89 | 83.27 | 83.61 | 1 | 1 | 1 | 3 | 0 | 0 |
| 2026-04-30 00:45 UTC | 83.6 | 84.01 | 83.6 | 83.81 | -1 | 1 | 1 | 1 | 0 | 0 |
| 2026-04-30 01:00 UTC | 83.8 | 83.94 | 83.66 | 83.94 | -1 | 1 | 1 | 1 | 0 | 0 |
| 2026-04-30 01:15 UTC | 83.93 | 83.98 | 83.76 | 83.87 | -1 | 1 | 1 | 1 | 0 | 0 |
| 2026-04-30 01:30 UTC | 83.87 | 84.0 | 83.7 | 83.82 | -1 | 1 | 1 | 1 | 0 | 0 |
| 2026-04-30 01:45 UTC | 83.82 | 83.96 | 83.73 | 83.75 | -1 | 1 | 1 | 1 | 0 | 0 |
| 2026-04-30 02:00 UTC | 83.75 | 83.96 | 83.75 | 83.88 | -1 | 1 | 1 | 1 | 0 | 0 |
| 2026-04-30 02:15 UTC | 83.87 | 83.9 | 83.52 | 83.55 | -1 | 1 | 1 | 1 | 0 | 0 |
| 2026-04-30 02:30 UTC | 83.55 | 83.56 | 82.7 | 82.88 | -1 | 1 | 1 | 1 | 0 | 0 |
| 2026-04-30 02:45 UTC | 82.89 | 83.07 | 82.71 | 82.98 | -1 | 1 | 1 | 1 | 0 | 0 |
这里专门回答你刚才问的核心问题:旧口径里的星号,有多少其实是被未来确认的 pivot / slope 回填误导出来的? 当前统计口径是:先用 confirmed_line_only 扫一遍旧信号,再用 causal_replay 重放同一窗口;凡是旧信号里有、但 causal 重放里没有的,都记作 hindsight-only。
如果某个资产的 misleading_pct 特别高,说明它在最近样本里特别依赖事后确认结构,实盘可信度就更低。
| asset | hindsight_signals | causal_signals | hindsight_only | misleading_pct | latest_hindsight_only |
|---|---|---|---|---|---|
| BTC-USD | 8 | 0 | 8 | 1.0 | 2026-04-29 13:30:00+00:00 |
| ETH-USD | 6 | 0 | 6 | 1.0 | 2026-04-29 13:15:00+00:00 |
| SOL-USD | 6 | 0 | 6 | 1.0 | 2026-04-29 23:45:00+00:00 |
这些就是“图上会亮星,但当时其实算不出来”的代表案例。
| asset | event_ts | direction | trigger_tf | entry_ts |
|---|---|---|---|---|
| SOL-USD | 2026-04-29 23:45:00+00:00 | long | short | 2026-04-30 00:00 UTC |
| BTC-USD | 2026-04-29 13:30:00+00:00 | short | medium | 2026-04-29 13:45 UTC |
| ETH-USD | 2026-04-29 13:15:00+00:00 | short | short | 2026-04-29 13:30 UTC |
| ETH-USD | 2026-04-29 03:15:00+00:00 | long | short | 2026-04-29 03:30 UTC |
| BTC-USD | 2026-04-29 03:00:00+00:00 | long | short | 2026-04-29 03:15 UTC |
| SOL-USD | 2026-04-29 01:45:00+00:00 | long | long | 2026-04-29 02:00 UTC |
| BTC-USD | 2026-04-28 17:30:00+00:00 | long | short | 2026-04-28 17:45 UTC |
| ETH-USD | 2026-04-27 15:00:00+00:00 | short | short | 2026-04-27 15:15 UTC |
| ETH-USD | 2026-04-27 07:45:00+00:00 | short | long | 2026-04-27 08:00 UTC |
| ETH-USD | 2026-04-27 05:15:00+00:00 | short | medium | 2026-04-27 05:30 UTC |
| SOL-USD | 2026-04-27 05:15:00+00:00 | short | short | 2026-04-27 05:30 UTC |
| ETH-USD | 2026-04-26 05:00:00+00:00 | long | short | 2026-04-26 05:15 UTC |
| BTC-USD | 2026-04-26 04:45:00+00:00 | long | short | 2026-04-26 05:00 UTC |
| SOL-USD | 2026-04-26 02:00:00+00:00 | long | medium | 2026-04-26 02:15 UTC |
| BTC-USD | 2026-04-25 21:45:00+00:00 | long | short | 2026-04-25 22:00 UTC |
| SOL-USD | 2026-04-25 15:45:00+00:00 | short | short | 2026-04-25 16:00 UTC |
| BTC-USD | 2026-04-25 15:45:00+00:00 | short | short | 2026-04-25 16:00 UTC |
| BTC-USD | 2026-04-24 16:15:00+00:00 | short | short | 2026-04-24 16:30 UTC |
| SOL-USD | 2026-04-24 03:15:00+00:00 | short | short | 2026-04-24 03:30 UTC |
| BTC-USD | 2026-04-24 03:00:00+00:00 | short | short | 2026-04-24 03:15 UTC |
这里把 rank29 live、gate shadow、baseline shadow 放在同一张图上。看盘时先看这张:如果 live 明显开始背离 gate shadow,就说明要优先检查执行质量或优先级冲突,而不是先怀疑策略本体。
向上 = 赢,向下 = 亏。live 如果还没成交,这里会为空。
这是 matching shadow 的逐笔参考,用来判断最近 shadow 本身是在顺风还是逆风环境。
这里直接看真钱 closed trades 和 gate shadow proxy 的同窗差值。当前如果还没很多真钱成交,这块会先比较稀疏,是正常的。
暂无数据。
暂无数据。
暂无数据。
暂无数据。
| timestamp | message |
|---|---|
| 2026-04-04T08:36:41Z | exchange reports external busy symbol on higher-priority universe |
| 2026-04-04T08:35:41Z | exchange reports external busy symbol on higher-priority universe |
| 2026-04-04T08:34:41Z | exchange reports external busy symbol on higher-priority universe |
| 2026-04-04T08:33:41Z | exchange reports external busy symbol on higher-priority universe |
| 2026-04-04T08:32:42Z | exchange reports external busy symbol on higher-priority universe |
| 2026-04-04T08:31:41Z | exchange reports external busy symbol on higher-priority universe |
| 2026-04-04T08:30:41Z | exchange reports external busy symbol on higher-priority universe |
| 2026-04-04T08:29:40Z | exchange reports external busy symbol on higher-priority universe |
| 2026-04-04T08:28:41Z | exchange reports external busy symbol on higher-priority universe |
| 2026-04-04T08:27:41Z | exchange reports external busy symbol on higher-priority universe |
暂无数据。
这些页面还保留,但默认不再是主入口。后面排障或回看研究时再点进去。
这块只做 sanity check,不再占首页 headline。