Rolling projection DOE

Find the projection setup whose median is closest to actual outcomes.

Replay completed anchors and rank horizon × sample-size settings by projection error.

Snapshot 2026-09-04

DOE calibrated as-of 2026-09-04

Window 2026-07-02 → 2026-08-31

Anchors 120

Lookback 730D

Confidence low

Computed at 2026年9月8日上午7:46 HKT

Point-in-time Point-in-time safe

Best rolling setting

3D · n=500

DOE results snapshot · Window: 2026-07-02 → 2026-08-31 · anchors: 120 · latest snapshot: 2026-09-04 · computed at: 2026年9月8日上午7:46 HKT

Median abs error2.31%

Mean abs error3.15%

Bias-0.27%

Inside P25/P7548.3%

DOE setting comparison

Same idea as score-vs-outcome analysis: compare each parameter setting by observed actual-minus-median error over rolling completed anchors. Lower median abs error and lower bias win.

Setting Median abs Bias P25/P75 hit n Read

Setting 3D · n=500

Median abs 2.31%

Bias -0.27%

P25/P75 hit 48.3%

n 120

Rank Best setting

Setting 3D · n=320

Median abs 2.38%

Bias -0.25%

P25/P75 hit 46.7%

n 120

Rank #2 Candidate

Setting 3D · n=80

Median abs 2.45%

Bias -0.06%

P25/P75 hit 49.2%

n 120

Rank #3 Candidate

Setting 3D · n=160

Median abs 2.42%

Bias -0.12%

P25/P75 hit 47.5%

n 120

Rank #4 Candidate

Setting 3D · n=240

Median abs 2.42%

Bias -0.22%

P25/P75 hit 48.3%

n 120

Rank #5 Candidate

Setting 3D · n=120

Median abs 2.47%

Bias -0.05%

P25/P75 hit 49.2%

n 120

Rank #6 Candidate

Setting 3D · n=40

Median abs 2.42%

Bias -0.07%

P25/P75 hit 46.7%

n 120

Rank #7 Candidate

Setting 5D · n=240

Median abs 2.76%

Bias +0.04%

P25/P75 hit 50.9%

n 116

Rank #8 Candidate

Setting 5D · n=320

Median abs 2.8%

Bias -0.08%

P25/P75 hit 51.7%

n 116

Rank #9 Candidate

Setting 5D · n=500

Median abs 2.84%

Bias -0.19%

P25/P75 hit 50.9%

n 116

Rank #10 Candidate

Setting 5D · n=120

Median abs 2.87%

Bias +0.21%

P25/P75 hit 51.7%

n 116

Rank #11 Candidate

Setting 5D · n=160

Median abs 2.93%

Bias +0.17%

P25/P75 hit 50.9%

n 116

Rank #12 Candidate

Best setting 3D · n=500: median projection vs actual return

X = Median projection return. Y = actual return. Solid line = linear trend. Orange line = bucket average. Checked: 120 · Pending: 0 · Avg: -0.71%

Best setting 3D · n=500: median projection vs forecast error

X = Median projection return. Y = actual - median. Solid line = linear trend. Orange line = bucket average. Checked: 120 · Pending: 0 · Avg: -0.27%

DOE ranking

Ranked by a composite DOE score: lower median/mean absolute error, lower bias, enough anchors, and better IQR coverage.

Rank	Horizon	n	Median abs	Mean abs	Bias	Inside P25/P75	Anchors	DOE score
#1	3D	500	2.31%	3.15%	-0.27%	48.3%	120	2.177
#2	3D	320	2.38%	3.17%	-0.25%	46.7%	120	2.22
#3	3D	80	2.45%	3.19%	-0.06%	49.2%	120	2.227
#4	3D	160	2.42%	3.21%	-0.12%	47.5%	120	2.233
#5	3D	240	2.42%	3.2%	-0.22%	48.3%	120	2.242
#6	3D	120	2.47%	3.21%	-0.05%	49.2%	120	2.243
#7	3D	40	2.42%	3.29%	-0.07%	46.7%	120	2.255
#8	5D	240	2.76%	3.84%	+0.04%	50.9%	116	2.603
#9	5D	320	2.8%	3.84%	-0.08%	51.7%	116	2.627
#10	5D	500	2.84%	3.82%	-0.19%	50.9%	116	2.658
#11	5D	120	2.87%	3.88%	+0.21%	51.7%	116	2.695
#12	5D	160	2.93%	3.87%	+0.17%	50.9%	116	2.718
#13	5D	80	2.92%	4.01%	+0.39%	50.9%	116	2.795
#14	5D	40	2.73%	4.24%	+0.52%	45.7%	116	2.813
#15	10D	500	3.12%	4.48%	+0.18%	61.6%	99	3.001
#16	10D	320	3.22%	4.57%	+0.43%	60.6%	99	3.123
#17	15D	500	3.37%	4.68%	+0.63%	65.1%	83	3.255
#18	10D	240	3.52%	4.63%	+0.54%	62.6%	99	3.305
#19	10D	160	3.47%	4.77%	+0.68%	60.6%	99	3.355
#20	10D	120	3.54%	4.8%	+0.65%	63.6%	99	3.389
#21	10D	80	3.55%	4.85%	+0.80%	59.6%	99	3.444
#22	10D	40	3.34%	4.99%	+1.11%	53.5%	99	3.449
#23	15D	320	3.85%	4.92%	+1.13%	65.1%	83	3.654
#24	20D	240	3.79%	5.46%	+2.23%	69.6%	79	3.967
#25	15D	240	4.28%	5.16%	+1.43%	66.3%	83	3.995
#26	15D	160	4.2%	5.34%	+1.72%	62.7%	83	4.07
#27	20D	500	4.53%	5.18%	+1.33%	73.4%	79	4.094
#28	15D	120	4.27%	5.39%	+1.89%	63.9%	83	4.145
#29	20D	320	4.45%	5.29%	+1.85%	70.9%	79	4.177
#30	15D	80	4.34%	5.56%	+2.12%	62.7%	83	4.277