Strategies

Jesse Community IFR2: Faithful RSI2 Pullback Port on ADA Binance Spot

Outcome: Abandoned

JesseIFR2AdaSpot

Outcome Summary

After three iterations, JesseIFR2AdaSpot reached backtest review as a frozen adaptation of the Jesse community IFR2 rules. Across 261 closed trades, it returned 15.82% net with Sharpe 0.272, but maximum drawdown reached 71.15%. Review abandoned the run because its 0.24923% average net trade return narrowly missed the spot floor, modeled impact was material, and no tuning was authorized. The decision concerned this ADA four-hour adaptation; later evaluation stages were not reached.

Strategy report

Hypothesis

Translate the supplied IFR2 strategy without changing its trading rules. Credit: Jesse community; the source identifies gabrielweich/jesse-strategies. Use ADAUSDT.BINANCE_SPOT on 4-HOUR bars, CASH account, long-only, leverage 1.0. Instrument and interval are platform research choices: neither is specified by the source. Preserve RSI period 2, entry threshold 10, Ichimoku periods 20/30/120 and displacement 60, Hilbert Transform trend-mode filtering, entire-balance fee-adjusted sizing, unconditional cancellation of pending entries, and the two-prior-high exit. There is no source stop-loss, profit target, maximum holding period, pyramiding rule, or signal-dependent sizing. Performance remains unverified. [Pinned IFR2 source](https://github.com/jesse-ai/example-strategies/blob/7c91e0a37bf62165790120d730442e4f6eb00364/IFR2/__init__.py).

The user's explicit faithful-reproduction request takes precedence over changing this into a novel mechanism. Spot is appropriate because the code never shorts or requests leverage and is underrepresented at 10.9% versus the 15% target. ADA avoids another BTC experiment; corpus_concentration is empty, so its actual research share cannot be verified. The single-symbol, long-only design adds to already overrepresented buckets, justified only as a requested reproduction. Four-hour sampling avoids the fastest fee-sensitive implementations while leaving the source rules intact. This remains a classical pullback strategy with additional filters, not a demonstrated structural edge; assign low research priority and do not claim novelty or profitability.

Decline optimization of this frozen ADA four-hour adaptation: a substantial executed sample delivers average net trade return of 0.24923% against the 0.25% spot review floor, PF 1.015, and material modeled impact drag. No parameters are authorized for tuning. This rejects this platform adaptation's measured economics, not the Jesse community source strategy generally; net account PnL is positive, so this is economic marginality rather than demonstrated negative net expectancy.

Implementation

Faithful Stratmill translation of the Jesse community IFR2 (RSI2 pullback) strategy on ADAUSDT Binance Spot 4-hour bars. It goes long when RSI(2) is below 10, the close is above both legacy-Jesse Ichimoku spans (20/30/120, displacement 60) and TA-Lib HT_TRENDMODE equals 1. It exits when the close is above the highs of both of the two previous candles. Sizing uses the entire balance with fees reserved, and pending BUY entries are always cancelled.

The iteration-3 brief had previous_code set to SEE_OUTBOX and no outbox file existed, so the Layer-1 'no FactoryStrategy class' failure came from a submission with no class body. I took the strategy code from the iteration-2 brief, which had passed Layer 1, and changed it as little as possible. To fix the earlier Layer 1.5 mismatch ('optimization_plan fixes leverage at 1.0'), every fixed value in optimization_plan now appears exactly in parameters: leverage 1.0, allow_short false, rsi_period 2, rsi_entry_threshold 10, Ichimoku 20/30/120/60, required_trend_mode 1, exit_high_lookback 2, balance_fraction 1.0 and cancel_pending_entry true. The code reads leverage and allow_short. Leverage now multiplies the position size, but because it is 1.0 the size does not change. The trading rules are the same as before. Platform adaptations: market orders at the bar close replace Jesse's limit order at the current price, and the real Binance spot taker fee is used for size_to_qty's three-fee reservation. Performance has not been verified. This is a classic pullback strategy with no claimed structural edge.

Verification Results

The sandbox's 71.42% drawdown exceeds the research contract's 40% constraint, and modeled impact materially affects economics. These are advisory findings for the analyst, not correctness failures.

Backtest Review

Sharpe: 0.27
Total return: 15.82%
Max drawdown: 71.15%
Trades: 261
Win rate: 62.5%
Profit factor: 1.02

261 closed trades over nearly eight years provide a functioning sample; all trades are long, consistent with the spot hypothesis.

Closed-trade PnL reconciles to the positive 15.823% account return; no ending unrealized gain drives the result.

Average net trade return is 0.24923%, narrowly below the supplied 0.25% spot review floor; PF is only 1.015. Positive net expectancy exists, but the measured economic margin is weak.

Modeled impact consumes 90.19% of gross PnL. QA's concern about impact and severe drawdown is borne out, although stressed-regime performance is positive overall.

All strategy parameters are frozen, leaving no authorized tuning path. Drawdown is 71.15% versus the declared 40% constraint; this supports the concern but is not the sole rejection reason.

Buy-and-hold returned approximately 15.70%, with 91.82% drawdown on the sampled benchmark curve. Strategy information ratio is -0.435.

Reported average position utilization and trade leverage conflict with engine peak leverage below 1; the reporting discrepancy needs separate verification, not an assumed accounting root cause.

Iteration History

Verification failed (Layer 1.5 — hypothesis/config consistency) [class=hypothesis_mismatch]: - optimization_plan fixes leverage at 1.0

Verification failed (Layer 1 — static analysis): - No class extending FactoryStrategy found (line None)

Outcome Summary

Positive net account returns can coexist with marginal trade economics and severe drawdown, so source fidelity alone does not establish viability for a chosen instrument and timeframe.

Backtest review rejected optimization because average net trade return narrowly missed the 0.25% spot review floor, economic margins were weak, and all parameters were frozen. Optimization, post-optimization analysis, and risk review were not reached.

Faithfully port the Jesse community IFR2 RSI2 pullback strategy to long-only ADA Binance Spot four-hour bars, preserving Ichimoku and Hilbert filters, full-balance sizing, and the two-prior-high exit.

The backtest produced 261 closed trades, a 15.82% net return, Sharpe 0.272, and 71.15% maximum drawdown. Average net trade return was 0.24923%, profit factor was 1.015, and modeled impact consumed 90.19% of gross PnL.

Analysis

IFR2

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