SolFourHourKeltnerChannelBreakoutTrailLongShort

SOL Keltner Channel Breakout with ATR Trail

A trend-following breakout strategy on SOLUSDT futures using Keltner Channels to identify volatility expansion. When price closes decisively above the upper Keltner Channel, the strategy enters long, anticipating continuation of the breakout move. When price closes below the lower channel, it enters short. Positions are managed with an ATR-based trailing stop that locks in profits as the trend develops. SOL is chosen for its high beta and strong trending behavior among large-cap alts. Uses 4-hour bars for a balance between signal quality and responsiveness. This is a simple, robust trend-capture system with only 3-4 core parameters, reducing overfitting risk.

Implements the Keltner channel volatility-expansion breakout with an ATR trailing stop, a simple robust trend-capture system with only a few core parameters. The Keltner band (EMA +/- mult x ATR) captures breakouts beyond price's own volatility envelope, distinct from a Donchian high/low channel. calculate_signal returns the continuous channel-position score so the verifier always sees varying values, while the discrete band-breakout booleans gate should_enter and position_size, avoiding the gated-to-0.0 anti-pattern. The trailing stop is path-dependent state: it is re-initialised when a new position opens (detected via entry-price change) and only ratchets in the trade's favour (max for longs, min for shorts), with the initial 2x ATR stop defining the risk used for capital-relative position sizing. Venue is BINANCE USD-M futures because the strategy goes both long and short and needs a margin account; leverage is left at 1.0 (no amplification used) so the leverage_set_but_unused gate does not trigger. SOL's high beta and strong 4H trends suit channel-breakout trend capture, and a wide ATR trail lets winners run well beyond the ~0.10% round-trip futures fee.

Fat-tail-carried trend edge that fails multiple-testing deflation and walk-forward overfitting, exactly the kurtosis-207 risk flagged at pre-optimization. The optimizer improved the headline (Sharpe 1.25, PF 2.83, Sortino 8.59, capacity $3.8B) and passed sensitivity (0 cliffs) and the holdout (ratio 1.571), but the decisive gates reject it: deflated_sharpe 0.6518 (<0.95), is_significant FALSE, sharpe_ci_low -0.514 (CI straddles 0), pbo 0.6468 (>0.5), and the selected Sharpe 1.25 is only marginally above the expected-max luck bar 1.05 over 225 trials; walk-forward is_overfitted=TRUE with avg IS 1.534 collapsing to avg OOS 0.488 and OOS windows [0.38, 1.83, -0.74] (one negative, propped by one window). Root cause: the result is outlier-dependent -- optimized return_kurtosis 124 (base 207), return_skew 10.66, tail_ratio 0.0448 on just 34 trades, with the entire Sharpe resting on ~3 pump days (2021-09-08 +42.8%, 2023-02-01 +20.7%, 2023-12-11 +38.3%, largest_win $64K). DSR's skew/kurtosis adjustment correctly discounts this to ~0.65, and the edge is decaying (optimized 2025 -10.1%, recent rolling Sharpe negative). The high probabilistic_sharpe (0.99) is the trap: PSR ignores trial count, whereas DSR (the gate that corrects for best-of-225 selection) rejects. The passing sensitivity/holdout measure in-sample smoothness and one forward window -- neither can override a CI straddling 0, PBO 0.65, is_overfitted=TRUE, and a Sharpe built on 3 tail days (a holdout pass on a fat-tail strategy is itself fragile to one tail flip). Not iterate (attempt 1 of 2): sensitivity is flat ~0.7 everywhere with no high-Sharpe plateau to tune toward, the edge is structurally outlier-dependent, and the optimizer already over-thinned to 34 trades -- a second sweep re-selects another tail-carried, sub-significant config. Not revise_hypothesis: this is the fat-tail arm of the path-dependent trend family, not a proven mechanism stranded on a dead target. FAILURE PATTERN: single-asset long-short Keltner/channel-breakout trend-followers on a high-beta crypto major (SOL) produce a headline carried by a handful of fat-tail pump days (kurtosis >100, skew >10, tail_ratio <0.05); after best-of-225 selection they fail deflated Sharpe (0.65, barely above the 1.05 luck bar) with a CI straddling 0, PBO 0.65, and is_overfitted=TRUE -- the same outlier-luck signature as the abandoned XRP Chandelier (DSR 0.01) and NEAR Supertrend (kurtosis 175). A passing holdout/sensitivity does not rescue an edge resting on 3 trades.

Long-short Keltner channel breakout trend-following on SOLUSDT.BINANCE USD-M futures, 4H bars, pure OHLCV single feed. The Keltner channel is a 20-EMA midline +/- 2x ATR(14); a close above the upper band enters LONG, a close below the lower band enters SHORT. Positions are managed with an ATR trailing stop initialised at entry +/- 2x ATR and trailed by 3x ATR, plus a 300-bar safety time stop. The signal is the continuous channel position (close-mid)/(2x ATR) recomputed every bar; the discrete band breakout is the entry gate. Sized at 20% equity notional, capped so initial-ATR-stop risk <= 1.5% of equity. leverage 1.0.

Genuine deployable trend edge: PF 1.51, Sharpe 0.71, total +150%, Sortino 2.33, max_drawdown 19.6%, expectancy +$637/trade; healthy sample (208 trades, both directions), positive in 5 of 7 years

Excellent capacity (~$238M, impact 2.05%) and low fee drag (1.89% of gross) — scalable, not a toy-scale edge

Implemented correctly and trades as specified; proven Keltner/ATR-trail trend mechanism

Extreme fat-tail dependence: return_kurtosis 207, skew 12.4, tail_ratio 0.38, win_rate 38% — the +150% is carried by a few enormous trend days (largest_win $45,646; +30%/+27%/+22.5% single days); the deflated-Sharpe penalty will heavily discount it

Sharpe CI lower edge -0.035 (just below 0); recent softening (2026 +1.8% YTD, rolling Sharpe negative in early 2026) — holdout risk

Headline concentrated in 2021 (+40.6%) and 2023 (+54.9%) bull-trend years

Clean implementation, adequate base sample (208 trades), sensitivity passed with 0 cliffs and a flat ~0.7 Sharpe surface (no fragile parameter), holdout passed (ratio 1.571, holdout_sharpe 0.767)

Healthy headline risk-adjusted metrics in the optimized config (profit_factor 2.83, Sortino 8.59, recovery_factor 10.25, capacity $3.8B) and positive alpha 0.078

Fails multiple-testing deflation: deflated_sharpe 0.6518 (<0.95), is_significant FALSE, sharpe_ci_low -0.514 (CI straddles 0), pbo 0.6468 (>0.5); optimized Sharpe 1.25 barely above the expected-max luck bar 1.05 over 225 trials

Walk-forward is_overfitted=TRUE: avg IS 1.534 collapses to avg OOS 0.488 with OOS windows [0.38, 1.83, -0.74] (one negative, average propped by one window)

Fat-tail / outlier-carried: optimized return_kurtosis 124 (base 207), return_skew 10.66, tail_ratio 0.0448 on only 34 trades; headline rests on ~3 pump days (2021-09-08 +42.8%, 2023-02-01 +20.7%, 2023-12-11 +38.3%, largest_win $64K)

Edge decaying: optimized 2025 -10.1%, recent rolling Sharpe negative in 2026; high probabilistic_sharpe (0.99) is misleading because PSR ignores trial count

This strategy traded SOL Keltner channel breakouts both directions with an ATR trail, and the base looked genuinely strong and deployable — +150%, Sharpe 0.71, ~$238M capacity, positive in 5 of 7 years — earning an 'optimize' verdict, though flagged for extreme fat-tail dependence (kurtosis 207). Optimization polished the headline (PF 2.83, Sharpe 1.25, Sortino 8.59) and even passed sensitivity and the holdout, but the multiple-testing gates rejected it: deflated Sharpe 0.65 barely above the luck bar, PBO 0.65, a CI straddling zero, and an overfit walk-forward, all because the result rested on ~3 explosive trend days across only 34 trades and was decaying recently. The analyst abandoned it on iteration 1, judging the edge structurally outlier-dependent with no high-Sharpe plateau to tune toward — a passing holdout on a strategy built on three tail days is itself fragile to a single tail flip.

A passing holdout and flat sensitivity surface cannot rescue an edge built on a handful of fat-tail pump days — single-asset long-short channel-breakout trend-followers on a high-beta major (kurtosis >100, skew >10, tail_ratio <0.05) fail deflated Sharpe after best-of-N selection with a CI straddling zero and is_overfitted=True, the same outlier-luck signature as the abandoned XRP Chandelier and NEAR Supertrend; PSR (0.99 here) is a trap because it ignores trial count where DSR corrects for it.

The post-optimization analyst abandoned it despite passing sensitivity (0 cliffs) and the holdout (ratio 1.571): the decisive multiple-testing gates failed — deflated Sharpe 0.6518 (<0.95, not significant) with optimized Sharpe 1.25 barely above the 1.05 expected-max luck bar over 225 trials, PBO 0.6468, sharpe_ci_low -0.514 (CI straddles 0), and walk-forward is_overfitted=True (avg IS 1.534 → avg OOS 0.488, one window -0.74) — because the entire edge rests on ~3 pump days (optimized kurtosis 124, tail_ratio 0.0448 on 34 trades) and is decaying (2025 -10.1%).

A long-short Keltner channel breakout trend-follower on SOLUSDT 4H perpetual futures (pure OHLCV) — going long when price closed decisively above the upper Keltner band (EMA + multiple×ATR) and short below the lower band, managed by an ATR trailing stop that ratchets in the trade's favor, on the thesis that SOL's high beta produces strong, capturable trends.

The base backtest was a genuine, scalable edge — 208 trades (112 long, 96 short), +150% total, Sharpe 0.71, profit factor 1.51, Sortino 2.33, expectancy +$637/trade, 19.6% max drawdown, ~$238M capacity at 2.05% impact, positive in 5 of 7 years — but fat-tail dependent (return kurtosis 207, skew 12.4, win rate 38%, with the headline carried by a few +30%/+27%/+22.5% trend days); optimization produced PF 2.83, Sharpe 1.25, Sortino 8.59 on just 34 trades.