STATISTICAL ARBITRAGE · THREE EQUITY PAIRS

Trade the gap. Not the noise.

Beta-adjusted overnight gap signals for BAC/JPM, V/MA, and PEP/KO. 59.3% win rate across 1,111 backtested trades.

Open App → How it works
59.3%
Combined Win Rate
3.28
Portfolio Sharpe
1,111
Backtested Trades
3 yr
Backtest Window

Methodology

Signal in three steps.

Every morning at 9:30 AM, two stocks in the same sector have gapped overnight. Quantile finds the ones that moved unusually far apart — and bets on mean reversion by close.

01 — MEASURE
Compute the gap
At market open, record each stock's overnight gap — how much it moved from yesterday's close to today's open.
02 — ADJUST
Isolate the shock
Subtract the target stock's beta-weighted gap from the lead stock's gap. This strips out shared market moves and isolates the pair's divergence.
03 — RANK
Fire the signal
Rank that shock against 3 years of history. Hit the 75th percentile or above — signal fires. Enter at open, exit at close, dollar neutral.
shock = lead_gap − β × target_gap

Short the lead, long the target when shock ≥ 75th percentile. Flip it when shock ≤ 25th percentile. Same-day exit. No overnight risk.

Backtest Results

Three years of data.
Three profitable pairs.

Jan 2023 – Mar 2026 · $1,000/leg · dollar neutral · open-to-close. Transaction costs not modeled.

Pair Trades Win Rate Sharpe Sortino Max DD
BAC / JPM Bank of America · JPMorgan
479
55.3%
1.37
1.63
$153
V / MA Visa · Mastercard
418
63.4%
4.08
5.85
$35
PEP / KO PepsiCo · Coca-Cola
214
59.3%
1.04
0.77
$64
Combined All three pairs
1,111
59.3%
3.28
4.76
$108

Why trust it

Built in the open.

No black box. The logic is public. Losses are shown. You can verify every number yourself.

📐
Verified logic
Python backtest output verified against the original HTML scorecard line by line. Both produce identical signals for the same inputs.
📋
Live trade log
Every signal gets logged with timestamp, entry prices, percentile, and outcome. Win rate updates in real time as trades close.
⚖️
Honest drawdowns
Max drawdown shown for every pair. Transaction costs not modeled — stated clearly. The numbers are what they are, nothing inflated.
🔓
Open source
Full backtest engine, deployment script, and React app on GitHub. Fork it, audit it, run it yourself. Nothing is hidden.

Get started

Free. No account.
Open the app.

Enter today's opening prices, get a signal in seconds. The math does the rest.

Open App → View on GitHub