FabORB — Fabio Valentini ORB on Crypto Perps

−$4,234

Aggregate net (6mo × 50 syms)

Only 9/50 symbols profitable

15/50

Walk-forward OOS positive

The 30-day edge was a low-vol regime artifact

15×

Edge concentration (2-factor screen)

Trend + breakout follow-through — but not profitable alone

Unit tests

stdlib-only research package

Verdict

The ORB edge does NOT transfer to crypto perps at scale. Over 6 months × 50 symbols: aggregate net −$4,234, only 9/50 symbols profitable, walk-forward OOS positive on just 15/50. The apparent 30-day edge was a low-volatility regime artifact.

A two-factor symbol screen (trend + breakout follow-through) concentrates ~15× of the available edge but does not create profitability alone. The research protocol itself (Bybit v5 data layer, ORB engine, confluence filters, vol-target sizing, walk-forward validation, 54 unit tests) is the reusable asset — the honest verdict is that this strategy family needs a different mechanism on crypto perps.

01 · Strategy

The strategy being tested

Opening range

First 30 minutes of the New York session (09:30–10:00 ET).

Entry

First 5-minute candle that closes above the opening-range high (long-only).

Stop / target / time

Stop = opening-range low · target = 1× risk (1R) · time exit = 15:00 ET. One trade per day, long only.

Delta filter

Fabio's delta (>200 contracts) filter was tested and added nothing (coin-toss across halves) — removed, matching the video.

02 · Results

Headline numbers

Measure	Result	Read
6mo × 50 symbols aggregate net	−$4,234	9/50 symbols profitable
Walk-forward OOS positive	15/50 symbols	Regime artifact: the 30-day edge was low-vol
2-factor symbol screen (trend + follow-through)	~15× concentration	Concentrates edge but not profitable alone
Regime-gate A/B sweep	no rescue	Vol-target sizing + regime gate don't create the edge

Takeaway: the Nasdaq ORB is a session-anchored, institutional-flow phenomenon. Crypto perps trade 24/7 with different microstructure — the ORB event has no standalone edge here. This is a clean negative result, published as loudly as a positive one.

03 · Methodology

How it was tested

Data

Rate-limited Bybit v5 client, top-N by turnover, disk cache, retry-with-backoff; NY-time conversion; 90/180-day windows.

Confluence filters

Volume confirmation (breakout > 1.5× avg), range tightness (<4%), 4H-EMA trend — all tested and reported honestly.

Validation

Chronological train/val/test split + rolling walk-forward OOS; regime analysis; winners-vs-losers feature investigation; proper OOS symbol-selection validation.

Engineering

stdlib-only Python package, 54 unit tests, gitignored regenerable cache (~15 min to refetch 50 symbols × 180 days).

04 · Reproduction

Run it yourself

pip install requests
python -m unittest discover -s orb_research/tests   # 54 tests
python scripts/test_enhanced_orb_sample.py --top 20 --days 90 --wf --cache orb_research/cache
python scripts/sweep_regime.py --symbols BTCUSDT,ETHUSDT,SOLUSDT --days 90
python scripts/test_walkforward_symbol_selection.py --days 180