Our house ensemble for crypto — three proven trend components vote on every confirmed close, gated by the 200-week moving average. Built purely from our own backtest evidence.
The Wuguan Master is not a new indicator — it is a curated ensemble of strategies that already run in the Arena, selected purely from our own backtest evidence (26,000+ runs across all markets, evaluated with Deflated Sharpe Ratio checks).
What the data showed (medians across all tested pairs and periods):
The ensemble (v1, frozen):
Gate = close > 200-week SMA
K1 = Supertrend(10, 3.0, Wilder) direction is up
K2 = WMA(15) > WMA(50)
K3 = Ichimoku Tenkan(9) > Kijun(26)
Vote = K1 + K2 + K3
ENTRY: Gate passes AND Vote >= 2
EXIT: Vote <= 1 OR Gate breaks
All component parameters are the unchanged live defaults of the source strategies — nothing was re-fitted. If an asset has less than 200 weeks of history, the gate is inactive (same semantics as our 200-WMA filter, whose evidence this design is based on).
Why a 2-of-3 vote? The vote smooths out whipsaws of any single component: a Supertrend whipsaw in a sideways market gets outvoted by the WMA cross and the TK-cross. The three components use different mechanics — volatility bands (Supertrend), weighted price crosses (WMA), and midpoint crosses (Ichimoku) — so they don't all fail the same way.
Bias protection: Component parameters are frozen in v1 and cannot be tuned — this prevents hindsight bias. Signals form only on confirmed closing data; once a signal is given it is immutable. New data only ever changes future signals, never past ones. Median results mix many pairs, periods and market phases — they are aggregate evidence, not a per-asset guarantee.
The pre-baked mini-backtest is refreshed daily — check back soon or start a live run in the Arena.
Run in Arena →// Components (v1, frozen — not user-tunable)
gate = close > SMA(weekly_closes, 200) // inactive if history < 200 weeks
k1 = supertrend(10, 3.0, wilders).direction == 'up'
k2 = WMA(close, 15) > WMA(close, 50)
k3 = tenkan(9) > kijun(26)
vote = k1 + k2 + k3
// State machine (level-based)
if position.is_flat and gate and vote >= 2:
BUY
if position.is_long and (vote <= 1 or not gate):
SELLDeliberately. The components run with the unchanged default parameters of their source strategies — the same defaults that produced the evidence the selection is based on. The moment we (or you) start tweaking periods until the curve looks better, the strategy stops being evidence-based and starts being curve-fitted. If the evidence ever justifies a change, it ships as a new version (v2) with the old version's results kept intact — never as a silent edit.
Because the data says they hurt this kind of strategy. Across our edge reports, ATR filters cost trend strategies between −15 and −30 percentage points of median CAGR, and the altcoin-season filter cost up to −27pp. The 200-week gate is the one filter that consistently helped — so it is built in, and the harmful ones are locked out.
No. Signals form only on confirmed closing candles — never on a forming bar — and once emitted they are immutable. New data changes future votes, not past ones. There is no repainting: if you backtest the same period twice, you get the same trades.
The ratcheting ATR-based trend follower — one of the most popular indicators on TradingView. Dynamic stop levels that lock in as the trend confirms.
Two weighted moving averages crossing — recent candles weight more, signals fire faster than SMA-based crosses. Validated on BTC weekly.
Goichi Hosoda's Ichimoku Kinko Hyo — TK-Cross with optional Cloud confirmation. Five-line system used by Japanese institutional traders since 1969.
Check out our Strategy Insights Reports — pre-baked deep-dives with historical results, comparisons, and market context.