β Home
Bracket-exit study Β· Published 2026-04-25
Bilbo Box Γ HTF PO with real exits.
This study tests SPY Bilbo Box breaks against higher-timeframe Phase Oscillator context using a trade-path bracket model: R = box height / price range, T1 = 0.5R, T2 = 1R, T3 = 2R, stop = opposite side of the box, and a 15-bar time stop. If a stop and target are both inside the same candle, resolution waits for that candle close.
01 β Study method
Why use trade-path exits: the question is whether a target is reached before a stop, not where price marks after a fixed number of bars.
Box risk unit.
Box βwidthβ is time. The price risk unit is box_high - box_low, i.e. box height.
Same-bar resolution.
If stop and target touch in the same candle, wait for the candle close: close beyond target = target, close beyond stop = stop, otherwise continue.
- Entry: immediate Bilbo boundary break.
- Stop: opposite side of the box.
- Targets: T1 = 0.5 box height, T2 = 1 box height, T3 = 2 box heights.
- Time stop: 15 bars, same RTH session only.
- Same-bar conflicts observed: 274 of 18,067 3m bull baseline trades (1.5%) and 76 of 5,224 10m bear baseline trades (1.5%).
- Lookahead guard: HTF PO timestamps shifted forward by one full HTF bar before
merge_asof.
02 β Headline bracket results
18,067
3m bull breaks Γ 10m PO sample
38.7%
3m baseline T2 hit rate
5,224
10m bear breaks Γ 1h PO sample
36.4%
10m baseline T2 hit rate
| Cohort | N | T1 0.5H | T2 1H | T3 2H | Stop | Timeout | Med bars |
|---|
| 3m bull baseline | 18,067 | 62.4% | 38.7% | 14.0% | 37.6% | 48.4% | 12 |
| 3m bull Γ 10m PO compression | 8,752 | 61.9% | 38.0% | 13.1% | 36.7% | 50.2% | 12 |
| 3m bull Γ 10m PO high zone | 1,213 | 62.8% | 41.9% | 17.4% | 35.5% | 47.1% | 11 |
| 3m bull Γ 10m PO low zone | 612 | 60.3% | 34.6% | 11.6% | 43.5% | 44.9% | 11 |
| 10m bear baseline | 5,224 | 59.3% | 36.4% | 14.6% | 29.8% | 55.6% | 8 |
| 10m bear Γ 1h PO compression | 2,304 | 60.5% | 38.1% | 16.3% | 29.3% | 54.4% | 8 |
| 10m bear Γ 1h compression + falling | 1,399 | 62.3% | 38.5% | 16.9% | 26.8% | 56.3% | 8 |
| 10m bear Γ 1h bull_exp + rising | 411 | 53.0% | 32.1% | 11.9% | 32.1% | 56.0% | 7 |
Read this as bracket probability. T2 means price resolved one full box height in favor within 15 bars under the candle-close same-bar rule.
03 β What survives?
| Filter | N | T2 hit | Baseline T2 | Ξ T2 | Stop | Baseline stop | Ξ stop |
|---|
| 3m bull Γ 10m compression | 8,752 | 38.0% | 38.7% | -0.6pp | 36.7% | 37.6% | -0.9pp |
| 10m bear Γ 1h compression | 2,304 | 38.1% | 36.4% | +1.7pp | 29.3% | 29.8% | -0.6pp |
| 10m bear Γ 1h compression + falling | 1,399 | 38.5% | 36.4% | +2.1pp | 26.8% | 29.8% | -3.0pp |
Conclusion: 10m bear Bilbo breaks with 1h PO compression show a modest improvement: T2 38.1% vs baseline 36.4%, stop 29.3% vs 29.8%. Compression + falling is cleaner: T2 38.5%, stop 26.8%.
3m bull compression remains weak as a green-light. It lowers stop rate (36.7% vs 37.6%) but also lowers T2/T3 hit rates. Low-zone 10m PO remains the better anti-filter: T2 34.6%, stop 43.5%.
04 β Same-direction PO confirmation: 3-contract P&L strategy
Strategy tested: take only same-direction PO-confirmed breaks: 3m bull + 10m bull_exp+rising, and 10m bear + 1h bear_exp+falling. Position size = 3 SPY contracts. Take one contract at T1, one at T2, one at T3. The stop stays at the opposite side of the box after T1; after T2, the final-contract stop moves to the break-side edge of the box β box top for bull breaks, box bottom for bear breaks. Unclosed contracts exit after 15 bars.
| Strategy bucket | Trades | T1 hits | T2 hits | T3 hits | Stops | Timeouts | Win% | Total P&L pts | Avg/trade |
|---|
| 3m bull Γ 10m bull_exp+rising | 2,419 | 1,525 / 63.0% | 951 / 39.3% | 327 / 13.5% | 1,031 / 42.6% | 1,061 / 43.9% | 57.1% | +161.8 | +0.067 |
| 10m bear Γ 1h bear_exp+falling | 715 | 428 / 59.9% | 251 / 35.1% | 80 / 11.2% | 252 / 35.2% | 383 / 53.6% | 55.4% | +51.4 | +0.072 |
P&L read: values are SPY points summed across the 3-contract position, not option premium dollars. The less aggressive stop model improves the 3m bull bucket to +161.8 pts across 2,419 trades; the 10m bear bucket remains positive at +51.4 pts across 715 trades.
Important: stops are more frequent here than in the simple bracket table because managed stops can close remaining contracts after partial profits. That is trade management, not a contradiction.
05 β Trading takeaway
Use Bilbo as texture, not a standalone system. A half-box target hits often, but the full-box target is only ~36β39% baseline and two-box continuation is ~14β16%. Most trades stop or time out inside 15 bars.
- Best candidate: 10m bear Bilbo break when 1h PO is compressed, especially compressed + falling.
- 3m bull: avoid low/accumulation 10m PO; do not treat compression alone as a green light.
- Next test: time-of-day buckets and matched random breakouts with the exact same bracket model.
06 β Caveats
- 1h PO caveat: hourly Phase Oscillator has known accuracy drift versus TradingView; treat 1h PO zones directionally.
- Same-bar ambiguity: candle-close resolution is better than stop-first for these ambiguous bars, but OHLC still cannot know intrabar order.
- Gap-through entries: if entry occurs beyond the boundary, actual stop distance can exceed one box height; the script tracks
actual_stop_R.
- Not an exploit: the useful findings are modest filters/anti-filters, not an automatic trading strategy.