# Mini Dragon Hunt v1 — SPXW 0DTE research strategy

## Objective and current verdict

Identify inexpensive call/put reversals earlier than Dragon v1 while reporting losses and missed 10× moves. It cannot predict all 10× gains. The frozen September 1–October 9, 2026 retrospective backtest failed: it is not approved for live buy alerts.

## Changes from Dragon v1

1. Early reversal may act before MACD crosses zero: three improving histogram closes, a recent support/resistance sweep and reclaim, close back across EMA9 and the previous candle's high/low. Support/resistance is the already-completed SPY rolling range, not a retrospectively drawn SPX pivot.
2. Trend continuation uses an EMA9 retest and reclaim with aligned EMA20, MACD line and histogram. Calls require bullish alignment, puts the exact bearish mirror.
3. Option confirmation uses four contiguous traded minute bars and a break over the preceding two-minute high, instead of eight bars and five-minute confirmation. Both call and put premiums rise when their directional setup works.
4. Watch first; verify the exact live bid/ask before treating an observation as executable. Current research remains watch-only because its backtest failed.

## Frozen entry rules

- SPY supplies directional context; its price is never multiplied to price SPXW.
- At least 34 contiguous completed RTH SPY minute bars; SPY minute volume at least 0.8× its preceding 20-minute average.
- Exact OCC-listed SPXW call/put, expiry equal to trading date, one-minute TRADES bars.
- Premium $0.20–$3; signal-minute volume at least 20 contracts; four-minute volume at least 100; current minute volume at least 1.25× preceding three-minute average.
- Premium at most 2.5× the low already observed in the preceding 15 traded minutes; never use the future day low.
- Rank on observed volume/evidence; at most one new contract per side every five minutes, one selection per contract/day. No new entry in the last hour.
- Live observations additionally require current exact bid/ask, spread ≤10%, size ≥5 each side. Missing/delayed tape or quotes block entry. Current bridge coverage is a watched sample, not an all-chain scan.

## Frozen paper execution and exits

- Entry starts a full minute after confirmation; use the next eligible trade-bar open plus 10% and $0.05 assumed cost. No fill if that bar does not arrive within two minutes.
- Initial stop at 60% of assumed entry premium (40% loss). A completed SPY close beyond the signal's recent swing invalidation exits at the next option-bar open. Stop wins ambiguous stop/target bars.
- Sell half at 2×; remaining half moves to break-even for subsequent bars. After a completed bar reaches 3×, trail the runner at 65% of its already-seen peak. Target the runner at 10×; flat 30 minutes before exchange close.
- Half at 2× plus half at 10× produces **6× whole-position proceeds**, not 10×. A later option peak after a stop is not realized profit.
- Missing minute paths or ambiguous ordering when a partial target and newly activated stop occur in the same bar remain unresolved. Historical trade highs do not verify executable bid/ask or actual fills.

## Backtest and promotion criteria

Replay each minute using only its prefix and retain every failed selection. Publish day-low-to-later-high winners separately from after-entry and before-exit results. Report contract coverage, gaps, losing trades, unresolved paths and mean resolved proceeds; do not call overlapping research selections portfolio returns.

The rules were written after seeing October 1–7 failures, so this is retrospective research, not a prospective holdout. Complete point-in-time option-chain and quote data, independent future sessions and positive performance after costs are needed before any live trading promotion. No parameter was relaxed after this run to manufacture a 10× success. Increasing risk or holding through every stop merely to match historical peaks is not evidence of prediction.

## Observed validation results

All 28 requested trading dates have minute SPY coverage; the saved SPXW sample covers 2,238 distinct contract-days and 98 observed day-low-to-later-high >10× winners. The strategy selected 382 candidates, including 27 winner contracts before their observed peak. It captured zero managed 10× runners. There were 51 profitable, 314 losing and 17 unresolved paths. The mean resolved whole-position proceeds were 0.7963× assumed entry cost. Seventeen selections eventually reached 10× relative to the assumed entry before the time limit, including after stops; those peaks are not captured returns.

The early reversal trigger selected 119 candidates (17 profitable, 97 losing, 5 unresolved); the trend pullback selected 263 (34 profitable, 217 losing, 12 unresolved). Both resolved averages were below entry cost. Keep research observations separate from buy alerts.