Source: 发明者量化-小小梦, 'Keltner Channel upgrade: King Keltner strategy', FMZ digest (https://www.fmz.com/digest-topic/4104), sections '升级版的金肯特纳' and '金肯特纳策略My语言'. Credit to the author. The mechanism is restated from first principles, not ported faithfully (L167). Channel: typical price TP=(H+L+C)/3. Midline = EMA(TP, 20), which is Raschke's EMA version from the article's text; the article's My-language code uses an SMA. Band width = SMA(TrueRange, 50), used with a 1.0x multiplier as in the article. Upper = mid + ATR50, lower = mid - ATR50. Direction rules: long when the midline is rising and the close is above the upper band; short when the midline is falling and the close is below the lower band. A long exits on a close below the midline and a short on a close above it. No other gate, filter or supplementary feed (L130). ONE ADAPTATION: the clock moves from the article's 1H to 4H, because my point-in-time pre-study on BYBIT 1H (7 alts/majors, 2021-2026, 0.11% RT) showed the 1H version is fee-dead on every name. Net per trade was -0.14% to 0.00%, with PF 0.82-1.00 on 2,000-2,600 trades. The midline-recross exit cuts trades at about 1 ATR, and on 1H that is too small against the round trip. Instrument selection is disclosed (L122): I screened SOL/XRP/LINK/DOGE/AVAX/ETH/BNB on BYBIT at 4H/12H/1D. I chose ETHUSDT 4H because it is the only cell with (a) >=500 lifetime trades, (b) symmetric long and short capture (net +0.38% long / +0.39% short, so it is not disguised long beta) and (c) positive net in 4 of 6 calendar years with no single-year dominance. Several 1D/12H cells had higher PF (SOL 1D PF 2.6) but only 70-170 trades, and those were concentrated in one year and long-side-driven, so they fail L128/L170. Venue: BYBIT linear perp (0.8% corpus share against a >=5% target). Pure-price trigger, so the L160 supplementary-feed restriction does not bite. Bybit ETH has ~5.5 years of 4H history (2021-03 onward, ~12,100 bars). BTC is avoided (over-concentrated), as is Hyperliquid (L168: short history). The factory's other ETH candidates (OKX SuperTrend, volume-confirmed momentum) use different triggers and exits. This one is a volatility-band breakout with the exit tied to a moving midline, flat ~33% of the time.