| 每輪取前幾名 | 分數上限 | s1觸發%(權益) | s1回落%(權益) | s1賣出% | s2觸發%(權益) | s2回落%(權益) | 單筆等效複利率 | 軌道獲利率 | 截斷率 |
|---|---|---|---|---|---|---|---|---|---|
| 5 | 5.0% | 0.000% | 1.565% | 15.8% | 71.493% | 1.220% | +6.547% | 100.0% | 14.4% |
| 5 | 5.1% | 0.000% | 1.750% | 12.2% | 66.277% | 1.278% | +6.547% | 100.0% | 13.1% |
| 5 | 5.0% | 0.000% | 1.298% | 18.3% | 72.632% | 1.008% | +6.444% | 100.0% | 14.7% |
| 5 | 5.1% | 0.000% | 1.232% | 10.6% | 59.617% | 1.062% | +6.367% | 100.0% | 11.6% |
| 5 | 5.0% | 0.000% | 1.936% | 18.2% | 68.904% | 1.080% | +6.265% | 100.0% | 13.8% |
| 5 | 5.0% | 0.000% | 1.011% | 19.6% | 66.958% | 1.169% | +6.110% | 100.0% | 13.4% |
| 5 | 5.4% | 0.000% | 1.528% | 16.2% | 71.829% | 0.621% | +6.093% | 100.0% | 13.0% |
| 5 | 5.0% | 0.000% | 1.770% | 16.0% | 59.017% | 0.927% | +6.046% | 100.0% | 11.7% |
| 5 | 5.0% | 0.000% | 1.701% | 16.4% | 65.174% | 1.881% | +5.975% | 100.0% | 12.9% |
| 5 | 5.1% | 0.000% | 1.444% | 10.2% | 52.323% | 0.936% | +5.952% | 100.0% | 10.0% |
| 5 | 5.0% | 0.000% | 1.880% | 16.7% | 61.077% | 1.692% | +5.910% | 100.0% | 11.9% |
| 5 | 5.1% | 0.000% | 1.897% | 14.9% | 58.922% | 1.495% | +5.863% | 100.0% | 11.6% |
| 5 | 5.2% | 0.000% | 1.535% | 10.4% | 52.984% | 1.165% | +5.859% | 100.0% | 9.9% |
| 5 | 5.6% | 0.000% | 1.632% | 15.7% | 70.060% | 0.590% | +5.798% | 100.0% | 12.1% |
| 5 | 5.7% | 0.000% | 1.727% | 11.7% | 69.226% | 1.725% | +5.787% | 100.0% | 11.7% |
| 5 | 5.0% | 0.000% | 1.594% | 21.6% | 59.428% | 0.867% | +5.695% | 100.0% | 11.6% |
| 5 | 5.5% | 0.000% | 1.638% | 17.7% | 69.624% | 0.797% | +5.688% | 100.0% | 12.2% |
| 5 | 5.0% | 0.000% | 1.649% | 21.4% | 59.246% | 0.992% | +5.673% | 100.0% | 11.6% |
| 5 | 5.0% | 0.000% | 1.732% | 19.4% | 62.227% | 1.414% | +5.650% | 100.0% | 12.2% |
| 5 | 5.0% | 0.000% | 1.850% | 28.7% | 72.848% | 0.933% | +5.559% | 100.0% | 14.7% |
綠色標示為最佳試驗。兩段式：第一段觸及s1觸發%後開始追蹤最高點，回落s1回落%就賣出s1賣出% 比例的倉位(先落袋一部分)；第二段觸及s2觸發%(通常設得比第一段高很多，確認是真正的大行情)後 改用s2回落%追蹤剩餘倉位，回落就全部賣出。2026-09-08實測發現：s1設得寬鬆、只賣一小部分, 真正的保護留給門檻夠高的s2用緊的回落%鎖利，比單階段版本(optimize_simulation_params) 在同樣修正後的模型下表現好很多。「截斷率」是交易到資料結束都還沒真正觸發出場、只能用最後 一天收盤價估報酬的比例——這是用還沒兌現的帳面數字冒充已實現獲利，搜尋範圍放寬時優化器容易 鑽這個漏洞(截斷率愈高、分數愈好看但愈不可信)，這個版本已經對截斷率設了上限(見上方meta)過濾掉 不可信的組合。僅供參考，非投資建議。