Workspace

StratStack

Stack filters layer-by-layer and watch Sharpe, frequency, win rate, and the 2σ-after-30 win rate evolve with each added probability. Save promising stacks as Strategies.

Layer Builder

4 of 4 active
Universe
Layerstop → applied first
1
RS vs QQQ>90Strength
2
Theme Rank≤3Rank
3
ADR %≥5%Volatility
4
Breakout D+1=trueSetup
Entry strategy

Optional. Layers narrow the universe; entry strategy is the trigger that fires once a name qualifies.

Stop-loss

1.00 × ATR(20) below entry

Save strategy
Baseline
n=9,800
Sharpe0.42
Frequency1 / 1.1d
Win Rate49.0%
WR 2σ-low32.0%after 30 trades
Equity curve (simulated)
+ Layer 1· RS vs QQQ
n=5,162(-47.3%)
Sharpe0.74+76.2%
Frequency1 / 2.1d+90.9%
Win Rate50.0%+2.0%
WR 2σ-low32.9%+2.8%after 30 trades
Equity curve (simulated)
+ Layer 2· Theme Rank
n=1,810(-64.9%)
Sharpe0.87+17.6%
Frequency1 / 6.0d+185.7%
Win Rate51.9%+3.8%
WR 2σ-low34.6%+5.2%after 30 trades
Equity curve (simulated)
+ Layer 3· ADR %
n=835(-53.9%)
Sharpe1.01+16.1%
Frequency1 / 12.9d+115.0%
Win Rate52.7%+1.5%
WR 2σ-low35.3%+2.0%after 30 trades
Equity curve (simulated)
+ Layer 4· Breakout D+1
n=235(-71.9%)
Sharpe1.23+21.8%
Frequency1 / 1.5mo+255.8%
Win Rate53.9%+2.3%
WR 2σ-low36.3%+2.8%after 30 trades
Equity curve (simulated)

Library rank

vs 9 library + 0 saved
Rank by Sharpe#8of 10Sharpe 1.23 · WR 53.9% · n=235
Composite rank#7of 10Sharpe × sample × Wilson 2σ WR
Verdict

Beats 2, loses to 7.

Loses to
  • R1 Rejection · Short1.27n=720
  • 5 SMA Pullback1.34n=1,140
  • R1 Slice Through · Long1.44n=614
Beats
  • Parabolic Short1.21n=198
  • Daily VWAP Reclaim1.18n=1,640

Grid search

Restraint mode
5 × 5 = 25 cells

Restraint mode: 2 dimensions × 5 cells max. Don’t over-search the space — that’s how you find noise.

X: RS vs QQQY: Theme Rank
7
5
3
1
1
58
74
90
100
100
low
high Sharpe
Top 3
  1. 1RS vs QQQ > 58 · Theme Rank ≤ 3
    Sharpe 2.23WR 57.6%n 64Stab 85
  2. 2RS vs QQQ > 74 · Theme Rank ≤ 5
    Sharpe 2.20WR 56.7%n 60Stab 89
  3. 3RS vs QQQ > 90 · Theme Rank ≤ 3
    Sharpe 2.17WR 55.8%n 60Stab 84

Overfit Guard · KNN (K=5)

ML opt-in: KNN here is used only to detect overfit peaks in the grid, not to pick the strategy.

  1. 1RS vs QQQ > 58 · Theme Rank ≤ 3
    Borderline

    Top result Sharpe 2.23; nearest 5 neighbours avg Sharpe 1.95 → mixed signal, retest before sizing up.

    Avg-N Sharpe 1.95Within band 3/5Stability 85/100
  2. 2RS vs QQQ > 74 · Theme Rank ≤ 5
    Real edge — dense cluster

    Top result Sharpe 2.20; 4 of 5 nearest neighbours also Sharpe ≥ 1.80 → robust cluster.

    Avg-N Sharpe 2.02Within band 4/5Stability 89/100
  3. 3RS vs QQQ > 90 · Theme Rank ≤ 3
    Real edge — dense cluster

    Top result Sharpe 2.17; 4 of 5 nearest neighbours also Sharpe ≥ 1.77 → robust cluster.

    Avg-N Sharpe 1.88Within band 4/5Stability 84/100

Walk-forward CV

Current strategy
Sharpe · 10 folds
-0.140.001.542021 H22022 H12022 H22023 H12023 H22024 H12024 H22025 H12025 H22026 H1
Mean
1.18
Std dev
0.15
Min / Max
0.97 / 1.40
CoV
0.13
Pass rate
10 / 10

Each fold is a 6-month walk-forward window. Hover a point for the date range and trade count. Stable lines across folds suggest the edge isn’t period-dependent. CoV and Pass rate above catch the fragile strategies a pretty average can hide.