Does any of this actually work? Each signal is replayed as-of every past monthly anchor using only the data available then, and scored on its realized forward return. The headline number is edge / anchor — how much the signal beat the average card, counting each month once. Cards move together within a month, so months (not card-months) are the real sample size.
Anchors
32
Horizons
1mo · 3mo
Prices
Real trades only
How to read this.n is pooled (printing × anchor) observations — impressive-looking but overstated, since cards are correlated within a month. Edge / anchor fixes that: the signal’s edge over the universe is computed separately at each entry month (months with fewer than 3 signal hits are excluded), and the headline is the average across months with a bootstrap 95% CI from resampling months. If the CI excludes zero, the edge is unlikely to be luck. Months + shows consistency — how many individual months beat the universe. Dropped is the survivorship check: signal entries whose exit price vanished (delisted/illiquid). If that rate is much higher than the universe’s, treat the edge as inflated.
Signal
Horizon
n
Fwd return
Hit %
Base
Pooled edge
Edge / anchor
By month
Dropped
Raw buy verdict
The raw buy rule (score ≥ 76, confidence ≥ 70, no extreme spread) on the Setups board rebuilt as of each month
1mo
184
+8.5%
med +2.9%
65%
+4.9%
+3.7%
+1.2%
95% CI -1.2% … +4.0%
9/16 months +
1.1%
base 1.2%
Raw buy verdict
The raw buy rule (score ≥ 76, confidence ≥ 70, no extreme spread) on the Setups board rebuilt as of each month
3mo
165
+27.6%
med +11.8%
71%
+15.6%
+12.0%
+8.4%
95% CI +0.1% … +16.7%
10/14 months +
1.2%
base 2.0%
Setup
On a proven floor AND recovering (the Setups thesis)
1mo
285
+9.2%
med +3.0%
64%
+4.9%
+4.3%
+2.3%
95% CI -0.3% … +5.2%
11/21 months +
1.4%
base 1.2%
Setup
On a proven floor AND recovering (the Setups thesis)
3mo
265
+23.8%
med +11.8%
71%
+15.6%
+8.2%
+5.0%
95% CI -0.6% … +10.8%
14/19 months +
2.2%
base 2.0%
At Support
Sitting in a tested accumulation zone
1mo
12,512
+6.3%
med +0.4%
53%
+4.9%
+1.4%
+0.5%
95% CI -0.3% … +1.4%
13/22 months +
1.7%
base 1.2%
At Support
Sitting in a tested accumulation zone
3mo
11,500
+20.8%
med +6.8%
71%
+15.6%
+5.1%
+2.0%
95% CI +0.6% … +3.5%
13/20 months +
2.7%
base 2.0%
Recovering
≥25% off sustained peak and turning up
1mo
1,856
+9.5%
med +2.3%
61%
+4.9%
+4.7%
+4.1%
95% CI +2.1% … +6.4%
20/24 months +
1.2%
base 1.2%
Recovering
≥25% off sustained peak and turning up
3mo
1,729
+22.6%
med +11.2%
71%
+15.6%
+6.9%
+6.3%
95% CI +1.9% … +10.7%
17/22 months +
2.3%
base 2.0%
Momentum
Robust monthly gain ≥15% vs trailing median
1mo
56,727
+6.4%
med +2.1%
59%
+4.9%
+1.6%
+0.8%
95% CI +0.4% … +1.2%
22/28 months +
1.2%
base 1.2%
Momentum
Robust monthly gain ≥15% vs trailing median
3mo
47,696
+18.8%
med +11.6%
73%
+15.6%
+3.2%
+0.9%
95% CI +0.0% … +1.6%
16/26 months +
1.7%
base 2.0%
Walk-forward replay over 32 monthly anchors · real-trade entry/exit only · baseline = all eligible printings · edge/anchor = mean of per-month edges (months with ≥3 signal hits), CI via 5,000-draw bootstrap over months · dropped = entries with no real exit price (survivorship check). Strong overall market in this window — even a clean CI is edge over the universe, not a guarantee of profit after fees. Personal, non-commercial prototype — not investment advice.