LongTermTrend_V1478

REJECTEDhybrid

llmReduce overfitting and add regime & volatility sensitivity: 1) simplified entry to require either price above sma_50 with rising sma_50_slope or strong 20d return (>6%) and require vol_pctile < 0.75 to avoid noisy periods. 2) tightened exit: add faster trend-break (close < sma_50) and a combined MACD-hist + falling sma_50_slope exit, and deeper hard stop at -12% ret_20d. 3) added regime gates to only trade in UPTREND and non-HIGH volatility, avoid bubble states (NORMAL only). 4) introduced position sizing: lower size in high vol (vol_pctile>0.7 -> 0.4) and full size when near 52w highs, base_position set to 0.8. These structural changes aim to improve robustness and reduce walk-forward collapse. | rejected: fitness 0.049 below threshold

Fitness

0.0489

Generation

Complexity

Created

2026-10-13

Lineage

Parents (1)

LongTermTrend_V1343S1343 · gen 61ELITEllm

changed cumret_252d threshold -0.1997 -> -0.4102

Children (0)

None

Strategy Logic (pseudocode)

# LongTermTrend_V1478  [hybrid]
require regime.trend in ['UPTREND']
require regime.volatility in ['LOW_VOLATILITY', 'MEDIUM_VOLATILITY']
require regime.bubble in ['NORMAL']
if flat and (((close > sma_50 and sma_50_slope > 0.0) or ret_20d > 0.06) and vol_pctile < 0.75):
    position = 0.8
if long and (ret_20d < -0.12 or close < sma_50 or (macd_hist < 0.0 and sma_50_slope < 0.0)):
    position = 0.0  # exit to cash

Backtest Metrics

Metric	Train	Validation	Hidden Test	Full
CAGR	-2.0%	-11.7%	—	—
Total Return	-15.1%	-31.1%	—	—
Sharpe	-0.10	-0.64	—	—
Sortino	-0.06	-0.31	—	—
Calmar	-0.05	-0.23	—	—
Volatility	12.7%	17.2%	—	—
Max DD	-37.0%	-50.5%	—	—
Win Rate	35.5%	15.4%	—	—
Avg Win	5.2%	19.0%	—	—
Avg Loss	-3.5%	-6.0%	—	—
Payoff	1.50	3.14	—	—
Expectancy	-0.4%	-2.2%	—	—
Profit Factor	0.82	0.57	—	—
Trades	31	13	—	—
Avg Holding Days	11.7	10.0	—	—
Turnover	6.2	6.9	—	—
Exposure	13.2%	12.4%	—	—

Walk-Forward Years

year-by-year stability matters more than any single year

Year	CAGR	Sharpe	Max DD	Trades
No walk-forward results

A robust strategy performs acceptably in most years rather than spectacularly in one — look for consistency across regimes, not a single outlier year.

Regime Performance

Period	Regime	Days	Ann. Return	Sharpe	Exposure
No regime results

Trades (full period)

0 trades

Entry	Exit	Entry Px	Exit Px	Return	Days
No trades recorded