Jesse Community Donchian: Faithful 20-Bar Breakout with SMA200 Filter on ATOM Spot

Hypothesis

Faithful fixed-rule translation of the Jesse community strategy, credited to the supplied author and its referenced gabrielweich/jesse-strategies repository. Trade ATOMUSDT.BINANCE_SPOT using 4-HOUR bars, long-only, without leverage. The source does not specify an instrument or interval; these are platform experiment choices. Preserve its full-balance sizing and channel exit. Performance is unverified. Source: [pinned Donchian strategy](https://github.com/jesse-ai/example-strategies/blob/7c91e0a37bf62165790120d730442e4f6eb00364/Donchian/__init__.py).

The explicit faithful-port request takes precedence over introducing a new mechanism. Spot is appropriate for an unleveraged long-only strategy and is below its portfolio target. ATOM avoids another BTC experiment; corpus_concentration is empty, so low recent ATOM concentration cannot be asserted. This remains a saturated channel-trend family, with relevant turtle failures, so priority is low. Its SMA200 entry filter and symmetric 20-bar channels distinguish the exact rules from the reported 20/10 turtle port, but do not establish novelty or profitability. Do not optimize defaults or reinterpret a failed test as support for a modified strategy.

negative_expectancy: reject this fixed ATOM spot 4-hour experiment on measured net losses across 113 closed trades, PF below 1 and negative dollar expectancy. This judges this instrument/timeframe translation, not the source strategy on other markets. Drawdown alone and historical family failures are not the basis for rejection.

Implementation

Stratmill translation of the Jesse community Donchian strategy (gabrielweich/jesse-strategies) on ATOMUSDT Binance spot 4h bars. It goes long when the close breaks above the prior 20-bar Donchian upper band while the close is above the SMA200. It exits when the close falls below the prior 20-bar lower band. Sizing uses the full balance, following Jesse's size_to_qty: balance*(1-3*fee)/price, floored to 3 decimals.

Iteration 2 fixes only the Layer 1.5 failure. The optimization plan fixes `leverage` at 1.0, so `leverage: 1.0` (float) is now in parameters. It mirrors config.leverage and multiplies the sizing notional, which does nothing at 1.0 on the CASH spot venue. Every other fixed plan parameter is present with a matching type. Entry, exit and sizing logic are unchanged from the faithful port. The one bar-level adaptation: Jesse evaluates on candle close and fills at market, and here a market order is submitted on the bar close.

Verification Results

The sandbox reports -68.70% total return and 92.23% drawdown. Positive average percentage trade return alongside negative average dollar PnL warrants reviewing trade weighting and equity evolution before drawing economic conclusions. These results do not establish a code defect.

Backtest Review

Sharpe

0.13

Total return

-69.64%

Max drawdown

92.73%

Trades

113

Win rate

30.1%

Profit factor

0.90

113 closed trades over nearly seven years provide a meaningful initial economic test. All recorded trades are long, consistent with the stated direction; the code implements prior-channel breakout entries, SMA200 filtering and lower-channel exits.

Average trade return is 0.709%, above the 0.25% spot fee floor; this is not a below-fee-floor rejection.

Net closed-trade PnL totals -69,635.84 USDT, reconciling to ending equity of 30,364.16 USDT. Profit factor is 0.905 and dollar expectancy is -616.25 USDT per trade. QA's economic concern is borne out.

Positive arithmetic average percentage trade return does not establish profitable compounding: compounding the recorded trade percentages is also negative. Trade weighting and capital evolution matter.

Buy-and-hold returned -69.10%, versus the strategy's -69.64%; sampled benchmark drawdown was approximately 96.02%, versus 92.73% for the strategy.

All rules and sizing values are frozen, with no tunable parameters. No verified code defect supports iteration or spending compute on optimization. Reported position percentages conflict with engine leverage diagnostics; their cause remains unverified.

Iteration History

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

Outcome Summary

A positive average percentage trade return above the fee floor does not establish profitable compounding; dollar expectancy and capital evolution also matter.

Backtest review returned abandon because measured net losses, profit factor below 1 and negative dollar expectancy rejected this fixed instrument/timeframe experiment. With no tunable parameters or verified code defect supporting iteration, optimization and subsequent review stages were not reached.

AtomSpotJesseDonchian tested a fixed, long-only ATOM spot strategy on 4-hour bars, buying prior 20-bar Donchian breakouts above SMA200 with full-balance sizing and exiting below the lower channel.

The backtest recorded 113 closed trades, a net return of -69.64%, Sharpe of 0.132 and maximum drawdown of 92.73%. Profit factor was 0.905 and dollar expectancy was -616.25 USDT per trade, despite an average percentage trade return of 0.709%.

Analysis

Donchian