Strategies

AVAX Daily Ichimoku Cloud Bullish Breakout Long-Only (Single-Asset AVAXUSDT.BINANCE Daily, Hosoda 1969 Multi-Component Trend System, Bars-Only)

Outcome: Abandoned

AVAXDailyIchimokuCloudBreakoutLong

Outcome Summary

AVAXDailyIchimokuCloudBreakoutLong introduced a genuinely new mechanism to the portfolio — a multi-component Ichimoku trend filter requiring simultaneous confirmation across three bars-only sub-signals — thoughtfully engineered to avoid the supp_data, cross-venue, and concentration failures of prior attempts. The construction was clean and look-ahead-safe, but the asset and sizing choices doomed it: the entire +151% return came from 2024's +313% AVAX pump, with four of six years negative and a catastrophic risk profile (75% drawdown, 312% annualized vol, 128% average position from winner-balloon through parabolas, a +318% single day). The reviewer abandoned it as single-regime alt-beta with ~zero alpha, unpromotable on risk alone, and explicitly not iterable — the extreme position sizing is by-design high-beta-alt risk, not a correctable bug masking an edge. The lesson logged: a novel signal cannot salvage a full-size long-only trend-breakout on a high-beta alt whose performance is one pump year wrapped in ruinous volatility.

Strategy report

Hypothesis

Single-asset long-only strategy using the ICHIMOKU KINKO HYO system (Hosoda 1969) as a multi-component trend confirmation signal — a GENUINELY NEW mechanism class never tested in the portfolio. This proposal addresses EVERY documented failure mode from the analyst's accumulated post-mortems: (a) bars-only — NO supp_data dependency (the analyst has now confirmed 7+ failures from supp_data fallbacks: premium_index, L/S ratio, taker_ratio, HL funding all confirmed unavailable in production), (b) single-asset AVAX — diversifies asset focus from the BTC/ETH/SOL concentration in current pipeline (AVAX is a major L1 with 4+ years of Binance USD-M history but NO direct single-asset strategy currently), (c) single-venue Binance USD-M (sane backtest engine, no cross-venue execution gap, no multi-instrument backtest accounting bug), (d) long-only (no short-leg alt-season blowup), (e) MID-LONG horizon (daily bars with 2-8 week holds — NOT macro horizon which the analyst confirmed is structurally incompatible with factory walk-forward), (f) genuinely novel mechanism — Ichimoku combines THREE independent bars-only sub-signals into a single composite (Tenkan/Kijun cross + price above cloud + cloud color) which is fundamentally different from single-signal EMA crossovers, single-indicator KER/Aroon/ADX, or volume-based breakouts. The classic 'three-line bullish breakout' requires: (1) Tenkan-sen (9-day high-low midpoint) crosses above Kijun-sen (26-day high-low midpoint) — short-term momentum turns up, (2) current close is above the Ichimoku Cloud (Senkou Span A + B from 26 periods ago) — long-term trend is up, (3) cloud itself is green (Senkou Span A > Senkou Span B) — forward-looking trend bias is bullish. ALL THREE conditions together is a high-conviction filter. Hosoda's research and 50+ years of replication across equities, FX, and commodities (Patel 2010 'Trading with Ichimoku Clouds', Murphy 1999) consistently shows multi-component Ichimoku signals outperform single-line crossovers because they require simultaneous confirmation across multiple horizons. AVAX has had multi-month uptrends (2021 Q3-Q4, 2024 Q1, 2024 Q4) that Ichimoku-filtered systems would capture cleanly. Bars-only, NO supp_data dependency, NO fallback path. 3 parameters total (tenkan_period, kijun_period, senkou_period — Hosoda's standard 9/26/52 defaults with narrow optimization range).

Iteration-2 fix for the Layer-2 synthetic failure: the only defect was that `_midpoint` was a `@staticmethod`, and the synthetic verification proxy rebinds class methods onto its proxy instance, which stripped the staticmethod nature and injected `self` as a 5th positional argument (`takes 4 positional arguments but 5 were given`). The smallest safe change is to move `_midpoint` to a module-level function (no `self` binding is ever possible) and update the five call sites from `self._midpoint(...)` to `_midpoint(...)`. All passing logic — the three-component Ichimoku scoring, entry/exit conditions, sizing, imports, and parameters — is unchanged, so the earlier static layer stays green. Verified locally: static analysis passes and the synthetic scenarios (steady_uptrend, steady_downtrend, flat_ranging, volatility_spike, zero_volume, price_gap) now run without error and the synthetic layer passes. The strategy remains bars-only with no supp-data dependency, single-asset/single-venue AVAX daily, long-only, leverage 1.0.

Single-year-dependent with catastrophic risk and no generalizable edge — not optimizable or iterable into viability. The entire +151 total return is one year (2024 +313%); the strategy is NEGATIVE in 4 of 6 years (2022 -37%, 2023 -17%, 2025 -15%, 2026 -11%), so it is a 2024-alt-pump bet, not a robust trend signal. Risk is ruinous: max_drawdown 75.2%, annualized_volatility 312%, avg_position_pct 128% (winner-balloon from holding a 95%-notional position through AVAX's parabolic runs), a +318% single day (2024-01-03), kurtosis 216, skew 13.7, sharpe_ci_low -2.54. The sample is also thin (42 trades, ~7/yr) with a 23.8% win rate, and alpha is ~0 (+0.0116, beta 0.455, corr 0.69) — the Sharpe 0.73 is flattered 2024 AVAX beta. NOT optimize: a single-regime strategy with 75% drawdown / 312% vol / 42 trades will fail DSR and is undeployable at this risk regardless of parameter tuning. NOT iterate: the avg_position 128% / +318% day is winner-appreciation through the 2024 parabola at full 95% sizing (by-design high-beta-alt risk, not a correctable sizing bug masking a hidden edge), and the core problems — single-year dependence, catastrophic drawdown, thin sample, ~zero alpha — are fundamental, not a density/parameter fix. FAILURE PATTERN: a long-only Ichimoku (or any full-size trend-breakout) on a high-beta alt like AVAX produces a flashy headline driven entirely by one alt-pump year (2024 +313%) with negative returns in every other year, a 75% drawdown, 300%+ annualized vol from 95%-notional sizing through parabolas, and a thin low-win-rate sample — alt-beta in one regime, not a generalizable edge, and unpromotable on risk alone.

Implementation

Long-only single-asset Ichimoku Kinko Hyo bullish-breakout trend system on AVAXUSDT.BINANCE daily bars (pure OHLCV, no supplementary data). calculate_signal returns the count (0–3) of simultaneously-met bullish Ichimoku conditions: Tenkan-sen > Kijun-sen (short-term momentum up), close above the cloud (trend up), and a green cloud Senkou A > Senkou B (bullish bias). Entry requires all three (signal >= entry_threshold=3.0); exit is the classic Ichimoku trend-break (close below Kijun-sen or fully below the cloud) plus a 12% capital-protective stop. Sizes 95% of equity, leverage 1.0.

Verification Results

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['displacement', 'kijun_period', 'min_notional', 'senkou_period', 'stop_loss_pct', 'tenkan_period', 'entry_threshold', 'exit_below_cloud', 'position_fraction'] Check that __init__ sets all attributes from self.parameters.get(). - steady_uptrend: TypeError: AVAXDailyIchimokuCloudBreakoutLong._midpoint() takes 4 positional arguments but 5 were given (bar timestamp: 1735694340000) - steady_downtrend: TypeError: AVAXDailyIchimokuCloudBreakoutLong._midpoint() takes 4 positional arguments but 5 were given (bar timestamp: 1735694340000) - flat_ranging: TypeError: AVAXDailyIchimokuCloudBreakoutLong._midpoint() takes 4 positional arguments but 5 were given (bar timestamp: 1735694340000) - volatility_spike: TypeError: AVAXDailyIchimokuCloudBreakoutLong._midpoint() takes 4 positional arguments but 5 were given (bar timestamp: 1735694340000) - zero_volume: TypeError: AVAXDailyIchimokuCloudBreakoutLong._midpoint() takes 4 positional arguments but 5 were given (bar timestamp: 1735694340000) - price_gap: TypeError: AVAXDailyIchimokuCloudBreakoutLong._midpoint() takes 4 positional arguments but 5 were given (bar timestamp: 1735694340000)

Backtest Review

Sharpe: 0.73
Total return: 151.37%
Max drawdown: 75.23%
Trades: 42
Win rate: 23.8%
Profit factor: 1.19

Genuinely novel mechanism (multi-component Ichimoku), bars-only with no supp_data dependency as designed, clean execution (42/42 submitted, no drops), single-venue

Look-ahead-safe cloud construction; long-only

Single-year-dependent: 2024 +313% is essentially the entire +151 total return; 4 of 6 years negative (2022 -37%, 2023 -17%, 2025 -15%, 2026 -11%)

Catastrophic risk: max_drawdown 75.2%, annualized_vol 312%, avg_position_pct 128% (winner-balloon at 95% sizing through AVAX parabolas), +318% single day (2024-01-03), kurtosis 216, sharpe_ci_low -2.54

Thin sample, low hit rate: 42 trades (~7/yr), win_rate 23.8%; PF 1.19 only via 2024 outliers

No real alpha: alpha +0.0116, beta 0.455, corr 0.69 — Sharpe 0.73 is flattered 2024 alt-beta

only 2024 positive; 4/6 years negative

majority positive

Outcome Summary

A long-only full-size trend-breakout on a high-beta alt like AVAX produces a headline driven entirely by one alt-pump year with negative returns in every other year, a 75% drawdown, and 300%+ volatility from holding 95%-notional positions through parabolas — a genuinely novel mechanism (Ichimoku) cannot rescue an asset/sizing choice that yields single-regime alt-beta and is unpromotable on risk alone.

The analyst abandoned it at the backtest-review gate (on its second iteration) as single-year-dependent with no generalizable edge and ruinous risk: it is a 2024 AVAX-pump bet rather than a robust trend signal, undeployable on risk alone (75% drawdown, 312% vol), with a thin 42-trade sample and ~zero alpha — not fixable by optimization (it would fail DSR and stay undeployable) or iteration (the winner-balloon is by-design high-beta-alt risk, not a hidden-edge sizing bug).

A long-only, single-asset, bars-only daily strategy on AVAXUSDT.BINANCE using the Ichimoku Kinko Hyo system as a multi-component trend filter — entering only when all three sub-signals align (Tenkan>Kijun, close above the cloud, and a green cloud), exiting on the Ichimoku trend-break plus a stop — a genuinely novel mechanism class chosen with no supplementary-data dependency to dodge the factory's documented supp_data and concentration failure modes.

It posted a flashy but fragile +151% total return at Sharpe 0.73 over 42 trades (23.8% win rate, profit factor 1.19) with near-zero alpha (+0.012, beta 0.455) — but the entire return was one year (2024 +313%), 4 of 6 years were negative, and risk was catastrophic: 75.2% max drawdown, 312% annualized volatility, avg_position_pct 128% (winner-balloon), a +318% single day, kurtosis 216, sharpe_ci_low -2.54.

Backtest and paper results are hypothetical. Trading involves risk of loss.