← 返回首页

Scout Seat · Rank 3：third-touch + EMA/MACD confluence · first verdict

生成时间：2026-03-16 14:40:12 UTC ｜ clean_room spec 已存在，当前这页把 Rank 3 从 spec-only 推到 first verdict。

hard verdict

hard verdict：Rank 3 `third_touch_plus_ema_macd` 已补齐最小 Light Stability Pack 后，仍显示样本与邻域都过薄；当前最诚实的 desk 读法应是 `park`，保留为 structure-filter evidence，不进入 paper candidate pool，也不争夺 Live Seat。

variant aggregate

variantassets_testedpositive_assetsmean_total_returnmedian_total_returnmean_max_drawdownmean_false_break_ratiomean_persistence_pass_ratemean_time_to_failure_barsmean_tradesmean_win_ratepositive_asset_ratio
third_touch_plus_ema_macd3.001.000.78%0.78%0.00%0.00%100.00%9.000.33100.00%33.33%
third_touch_plus_ema3.001.000.01%0.01%-0.23%75.00%100.00%4.001.0025.00%33.33%
third_touch_only3.002.00-0.01%0.33%-0.32%66.67%100.00%3.672.0033.33%66.67%
raw_breakout3.000.00-45.14%-45.00%-45.84%49.17%100.00%4.83499.3332.76%0.00%

artifact：reports/artifacts/scout_third_touch_ema_macd_15m/variant_aggregate.csv

轻量 friction recheck

friction recheck：`third_touch_plus_ema_macd` 在 10/15/20bps per side 下仍守住正的跨资产平均收益；当前更像值得继续做轻量 forward 复核的窄门 structure-confirmation challenger。

variantcost_bps_per_sidemean_total_returnmean_false_break_ratiopositive_asset_ratiomean_tradesmean_time_to_failure_bars
raw_breakout6-45.14%49.17%0.00%499.334.83
third_touch_only6-0.01%66.67%66.67%2.003.67
third_touch_plus_ema60.01%75.00%33.33%1.004.00
third_touch_plus_ema_macd60.78%0.00%33.33%0.339.00
raw_breakout10-63.19%49.17%0.00%499.334.83
third_touch_only10-0.17%66.67%66.67%2.003.67
third_touch_plus_ema10-0.11%75.00%33.33%1.004.00
third_touch_plus_ema_macd100.70%0.00%33.33%0.339.00
raw_breakout15-77.64%49.17%0.00%499.334.83
third_touch_only15-0.37%66.67%0.00%2.003.67
third_touch_plus_ema15-0.26%75.00%0.00%1.004.00
third_touch_plus_ema_macd150.60%0.00%33.33%0.339.00
raw_breakout20-86.42%49.17%0.00%499.334.83
third_touch_only20-0.57%66.67%0.00%2.003.67
third_touch_plus_ema20-0.41%75.00%0.00%1.004.00
third_touch_plus_ema_macd200.50%0.00%33.33%0.339.00

artifact：reports/artifacts/scout_third_touch_ema_macd_15m/friction_ladder.csv ｜ 这里只做成本敏感性快检，不引入新 bar，也不把它误写成 forward 证据。

trade-count honesty / cadence dry-check

trade-count honesty：`third_touch_plus_ema_macd` 的有效交易过于稀疏，当前只能保留为 keep-narrow guard 证据，还不够资格往 paper candidate / live challenger 方向写。

gatestatusactualthresholdwhy_it_matters
total_trade_floorfailtotal trades = 1>= 9 total trades for 3-way time split至少要有最小样本量，才配谈 3 段时间稳定性；否则只能如实承认这是极稀疏 pocket。
asset_coverage_floorfailactive assets = 1/3all 3 assets should print at least one trade如果只有单一币种留下信号，它更像局部结构 pocket，不够支撑 desk 级候选。
min_asset_trade_floorfailmin asset trades = 1>= 2 trades on every active asset至少每个活跃资产要不止一笔，否则 trade-count honesty 过于脆弱。
calendar_breadth_floorfailmin active months per asset = 1>= 2 active months on every active asset至少跨过不止一个月，避免所有信号都挤在单段局部行情。
asset_concentration_watchwatch100.00%<= 60% of trades from one asset若几乎所有交易都来自一个币种，就不该把 headline 误写成跨资产稳定。
idle_gap_guardpasssingle-print / no gap yet<= 30d max gap between trades per asset若交易之间长期空窗，就更像极稀疏结构警报，而不是可连续观察的候选。

artifact：reports/artifacts/scout_third_touch_ema_macd_15m/trade_count_honesty.csv ｜ 这张卡不追求好看，只回答当前最优版本的交易分布够不够诚实。

time stability dry-check

time stability：`third_touch_plus_ema_macd` 当前样本薄到还不配做诚实的 early/mid/late 三段切片，因此只能继续停在 keep-narrow / one-more-light-check。

gatestatusactualthresholdwhy_it_matters
three_bucket_sample_floorfailtotal trades = 1>= 9 trades before attempting early/mid/late split没有最小样本量，就不该假装自己做出了时间稳定性结论。
bucket_asset_coveragefailassets with trades = 1/3all 3 assets should appear before 3-way time split如果时间切片前就只剩少数资产，时间稳定性读法天然不可信。
positive_bucket_floorfailnot attempted: sample too sparse for honest split>= 2 positive buckets out of 3当前更重要的是先如实承认样本太薄，而不是捏一个看似完整的时间切片。
bucket_trade_floorfailnot attempted: sample too sparse for honest split>= 2 trades in every time bucket若还没到能切桶的样本量，时间稳定性默认视作未通过。

artifact：reports/artifacts/scout_third_touch_ema_macd_15m/time_stability_drycheck.csv ｜ 若样本量连三段时间切片都不配做，就如实显示 fail，而不是伪造稳定性。

parameter stability dry-check

parameter stability：`third_touch_plus_ema_macd` 的本地参数邻域仍明显偏薄，当前不足以把单点正收益写成可升级的 paper candidate / live challenger。

gatestatusactualthresholdwhy_it_matters
positive_neighbor_floorpass7/7 configs positive>= 4 positive configs across local parameter neighborhood参数轻微变动后若多数近邻立刻翻负，说明这条线更像单点 lucky pocket，不像稳定 guard。
cross_asset_neighbor_floorfail0/7 configs keep >=2/3 positive assets>= 2 configs keep cross-asset floor如果参数一改就只剩单币种存活，就不该把它写成 desk 级候选。
trade_count_neighbor_floorfail0/7 configs keep >=1 mean trades / asset>= 4 configs keep >=1 mean trades / asset若邻域内大多配置交易数都塌到接近 0，就说明这条线的可观察性本身不足。
false_break_neighbor_guardpass0.00%<= 10% max false-break ratio across neighbor configs即便样本稀疏，也要确认参数轻微扰动后不会立刻把假突破率放大。
worst_neighbor_return_watchpasstol0.04_win24_slope3 mean_total_return=0.78%; positive_assets=1/3worst neighbor ideally > -0.50% mean total return记录最弱近邻，防止 base 配置只是在邻域边缘侥幸为正。
best_neighbor_snapshotinfotol0.04_win24_slope3 mean_total_return=0.78%; false_break_ratio=0.00%; mean_trades=0.3reference only保留邻域最强快照，方便 desk 判断是否只是一个参数偶然出彩。

artifact：reports/artifacts/scout_third_touch_ema_macd_15m/parameter_stability_drycheck.csv ｜ 这里只做本地小参数邻域的诚实干检，不做大规模 optimizer。

per-asset summary

assetvarianttradestotal_returnwin_ratefalse_break_ratiopersistence_pass_ratemean_time_to_failure_barsmax_drawdown
BTC-USDraw_breakout511-45.00%31.31%49.71%100.00%4.81-45.47%
BTC-USDthird_touch_only2-0.68%0.00%100.00%100.00%2.00-0.21%
BTC-USDthird_touch_plus_ema0------
BTC-USDthird_touch_plus_ema_macd0------
ETH-USDraw_breakout467-41.43%34.48%46.47%100.00%5.01-42.30%
ETH-USDthird_touch_only20.33%50.00%50.00%100.00%4.00-0.30%
ETH-USDthird_touch_plus_ema1-0.30%0.00%100.00%100.00%3.000.00%
ETH-USDthird_touch_plus_ema_macd0------
SOL-USDraw_breakout520-48.99%32.50%51.35%100.00%4.68-49.75%
SOL-USDthird_touch_only20.33%50.00%50.00%100.00%5.00-0.45%
SOL-USDthird_touch_plus_ema20.33%50.00%50.00%100.00%5.00-0.45%
SOL-USDthird_touch_plus_ema_macd10.78%100.00%0.00%100.00%9.000.00%

artifact：reports/artifacts/scout_third_touch_ema_macd_15m/asset_summary.csv

sample events

assetvariantsidesignal_time_utcentry_time_utcboundary_levelentry_priceexit_time_utcexit_pricenet_returnmax_drawdownfalse_breaktime_to_failure_barspersistence_pass
BTC-USDraw_breakoutshort2025-11-16 15:45 UTC2025-11-16 16:00 UTC95117.480094573.46002025-11-16 16:00 UTC94894.5979-0.46%-0.39%0.00%9.001
BTC-USDraw_breakoutshort2025-11-16 16:15 UTC2025-11-16 16:30 UTC94368.240094304.42002025-11-16 18:00 UTC94653.1929-0.49%-1.32%0.00%4.001
BTC-USDraw_breakoutshort2025-11-16 21:30 UTC2025-11-16 21:45 UTC93589.020093495.95002025-11-16 22:45 UTC93973.2914-0.63%-1.06%100.00%3.001
BTC-USDraw_breakoutlong2025-11-17 07:30 UTC2025-11-17 07:45 UTC95638.290095751.56002025-11-17 09:00 UTC95447.3436-0.44%-0.46%0.00%4.001
BTC-USDraw_breakoutshort2025-11-17 13:30 UTC2025-11-17 13:45 UTC94500.000093855.90002025-11-17 14:00 UTC94185.7364-0.47%-0.35%0.00%4.001
BTC-USDraw_breakoutshort2025-11-17 16:00 UTC2025-11-17 16:15 UTC93571.300093176.54002025-11-17 16:30 UTC93933.3971-0.93%-0.94%100.00%2.001
BTC-USDraw_breakoutshort2025-11-17 17:30 UTC2025-11-17 17:45 UTC93051.710092852.38002025-11-17 19:45 UTC91678.93001.16%-0.40%0.00%9.001
BTC-USDraw_breakoutshort2025-11-17 18:15 UTC2025-11-17 18:30 UTC92602.500092462.98002025-11-17 20:30 UTC91634.79000.78%-0.65%100.00%1.001
BTC-USDraw_breakoutshort2025-11-17 19:00 UTC2025-11-17 19:15 UTC92413.310092311.37002025-11-17 21:15 UTC91783.31000.46%-0.17%0.00%9.001
BTC-USDraw_breakoutshort2025-11-17 19:30 UTC2025-11-17 19:45 UTC92103.260091864.25002025-11-17 21:45 UTC91959.7100-0.22%-0.42%0.00%6.001
BTC-USDraw_breakoutshort2025-11-18 03:00 UTC2025-11-18 03:15 UTC90687.420090385.58002025-11-18 04:45 UTC89264.63001.14%-0.60%100.00%3.001
BTC-USDraw_breakoutlong2025-11-18 15:15 UTC2025-11-18 15:30 UTC92450.230092801.63002025-11-18 17:30 UTC93325.99000.45%-0.43%0.00%9.001
BTC-USDraw_breakoutlong2025-11-18 16:00 UTC2025-11-18 16:15 UTC93049.910093146.28002025-11-18 18:15 UTC93231.9900-0.03%-0.35%0.00%5.001
BTC-USDraw_breakoutshort2025-11-19 03:30 UTC2025-11-19 03:45 UTC91929.860091892.53002025-11-19 05:00 UTC91060.26570.79%-0.12%0.00%9.001
BTC-USDraw_breakoutshort2025-11-19 04:30 UTC2025-11-19 04:45 UTC91597.680091450.02002025-11-19 05:30 UTC90715.32710.69%-0.05%0.00%9.001
BTC-USDraw_breakoutshort2025-11-19 15:30 UTC2025-11-19 15:45 UTC91037.500090696.88002025-11-19 16:15 UTC89730.02140.96%-0.45%0.00%9.001
BTC-USDraw_breakoutshort2025-11-19 16:15 UTC2025-11-19 16:30 UTC89880.000089696.36002025-11-19 18:30 UTC88906.92000.77%-0.48%100.00%2.001
BTC-USDraw_breakoutshort2025-11-19 17:00 UTC2025-11-19 17:15 UTC89453.880089360.01002025-11-19 19:15 UTC88954.20000.34%-0.43%100.00%1.001
BTC-USDraw_breakoutshort2025-11-19 19:30 UTC2025-11-19 19:45 UTC88792.000088723.33002025-11-19 20:15 UTC89233.1850-0.69%-0.75%100.00%1.001
BTC-USDraw_breakoutlong2025-11-20 00:15 UTC2025-11-20 00:30 UTC91594.000091855.18002025-11-20 02:00 UTC92784.10140.89%-0.04%0.00%9.001
BTC-USDraw_breakoutlong2025-11-20 01:30 UTC2025-11-20 01:45 UTC92500.000092568.21002025-11-20 03:45 UTC92440.3200-0.26%-0.27%100.00%3.001
BTC-USDraw_breakoutlong2025-11-20 04:45 UTC2025-11-20 05:00 UTC92798.140092962.84002025-11-20 05:30 UTC92640.6014-0.47%-0.42%100.00%2.001
BTC-USDraw_breakoutshort2025-11-20 14:00 UTC2025-11-20 14:15 UTC91440.550090705.37002025-11-20 14:15 UTC91092.4636-0.54%-0.86%100.00%1.001
BTC-USDraw_breakoutshort2025-11-20 15:45 UTC2025-11-20 16:00 UTC90213.150089869.88002025-11-20 16:00 UTC88625.30861.28%-0.05%0.00%9.001

artifact：reports/artifacts/scout_third_touch_ema_macd_15m/event_sample.csv

怎么读这页

下一步

默认 park；只有当未来来自更宽历史样本或更诚实 repo/paper 复现出现明显更厚的 trade-count / cross-asset 证据时，才值得重开。

如需回看冻结口径，可参考既有 spec artifact：reports/artifacts/scout_third_touch_ema_macd_15m/clean_room_spec_v1.csv