BTC OKX 4H Trend-Bar Continuation: Typical-Price SMA/EMA Trend Filter + Directional Expansion-Bar Trigger, Long-Short, Fixed % Stop, Reverse-on-Opposite-Signal (BTCUSDT.OKX 4H, pure OHLCV)

Distilled from the '日内交易策略' section of 善's FMZ digest 'Six simple strategies for beginners' (https://www.fmz.com/digest-topic/4050, the My-language dual-MA script, fmz.com/strategy/129079). Following lesson L167 it is restated from first principles, not ported. It keeps the economic core: take a position only when a bar that EXPANDS in the trend direction appears inside an established trend. That bar closes beyond the previous close, makes a new high or low beyond the previous bar, and has a body in the trend direction. The source's intraday cadence and pyramiding (BK add-on) are dropped because the 0.10% round trip kills them. It runs on 4H bars. The other five sections were screened and rejected. HODL/DCA has no timing edge to test. Bollinger scalping on 1-5m is the dead sub-fee microstructure class. RSI 30/70 is generic mean reversion, 0 survivors in 203 runs. 'Avoid pump groups' is advice, not a rule. The swing MACD+MA script was pre-tested with 0.12% RT cost: LTC COIN-M 4H lost in every cell (PF 0.68-0.87). BNB and DOGE 12H 'worked' only through the 2021 mania year and lost in most other years. BTC OKX 4H came out at PF 1.16-1.28 with a weaker year profile than this rule. Venue choice: OKX is 0.1% of the corpus against a >=5% quota, and OKX history in the catalog only runs to multiple years for BTCUSDT.OKX and ETHUSDT.OKX. The ETH OKX slot is already taken (EthVolumeConfirmedMomentumLS_OKX), so BTC is the only OKX instrument with enough history (14,848 4H bars since 2019-12-16). BTC is above the concentration share, and this rationale is why it is used anyway. Pre-study, point-in-time, next-bar-open fills, 0.12% RT cost, full notional, 2019-12 to 2026-09. At P1=50/P2=20/SL=3%: 274 trades (~41/yr), mean net +1.03%/trade, PF 1.46, win rate 22%, trade-level annualized Sharpe ~0.63, and yearly sums +0.99/+1.15/+0.28/+0.48/-0.15/-0.21/+0.28 for 2020-2026 (5 of 7 years positive). Robustness: every cell in a 12-cell neighbourhood (P1 40-80, P2 10-30, SL 3-4%) had PF 1.19-1.68 and trade Sharpe 0.33-0.67. That is a plateau, not a spike. The best cells (P1 60-80) were NOT chosen. Disclosed selection choices: venue (OKX, quota), instrument (only OKX name with history not already in the pipeline), timeframe 4H (source was intraday; the fee floor forced it up), and P1=50/P2=20 as the middle of the plateau. This is plain trend following, and the honest pre-study Sharpe is ~0.6, below lesson L166's 1.0 comfort bar. It is proposed at low priority as the most viable mechanism in the source, with modest pre-registered numbers. Risk profile: 50% of equity notional per position (leverage 0.5x effective, 1x max), so a 3% stop costs ~1.5% of equity. There is no pyramiding and at most one position.