Strategies

XRP COIN-M 4H Trend-Pullback Continuation ('one wave, three folds' band entry): Buy the 3-Day Low Inside a 20-Day Up-Move, Sell the 3-Day High Inside a 20-Day Down-Move, Exit on Pullback Completion / 2.5-ATR Stop / 5-Day Clock (XRPUSD_PERP.BINANCE_CM, pure OHLCV, long-short, flat ~65-70%)

Outcome: Abandoned

XrpCmTrendPullbackContinuationLS

Outcome Summary

XrpCmTrendPullbackContinuationLS turned the 'one wave, three folds' band sub-module of an FMZ composite CTA system into a single-instrument, pure-OHLCV strategy on the XRP COIN-M inverse perpetual. It traded counter-moves inside an established 20-day trend, in the trend's direction, and was expected to be flat about 65-70% of the time. Across 3 iterations the strategy never got through verification: it was abandoned for 'verifying_stuck_3_timeouts', with no backtest, optimization, or analyst verdict. Its trading edge was therefore never tested.

Strategy report

Hypothesis

This hypothesis comes from the 'Band strategy' (波段策略) sub-module of TradeMan's Composite CTA system (FMZ #348225). The idea is that trends do not move in straight lines but in 'one wave, three folds' (一波三折). The system trades the counter-moves inside an established multi-day trend, in the trend's direction. The author runs it on coin-margined contracts, with the same parameters on every market and position size scaled to each signal's volatility. The rest of the description is marketing with no stated mechanism and is NOT used: dozens of undisclosed factor libraries, 'alternative' strategies, equity-curve adaptation, and a claimed 50-150% annualized return. Every return it claims is treated as unverified. This hypothesis is ONE sub-mechanism on ONE leg and ONE instrument: XRPUSD_PERP.BINANCE_CM (Binance COIN-M inverse perpetual), 4H bars, history 2020-09 to 2026-09, about 13,000 bars. The selection choices are stated explicitly (L122). Venue BINANCE_CM: the source names coin-margined contracts, COIN-M is 3.2% of experiments against a 5% target, and it has the lowest taker fee we model (~0.06% round trip). XRP: a top-5-liquidity COIN-M perp with 6 years of history. It is not BTC (62% corpus share) and not ETH (the earlier ETH COIN-M dual-timeframe run died overfit). 4H: the source holds 'days to weeks', and 4H gives about 6 decisions a day without sub-hourly fee drag. No supplementary feed is used, because COIN-M has none (L160). There is no regime gate beyond the trend definition the mechanism itself needs. Declared flat state: the strategy is flat whenever there is no active pullback signal, expected about 65-70% of the time. It is NOT always-in (L150). Sizing: risk 1.5% of equity per trade, with notional = 0.015*equity / (2.5*ATR%), capped at 1.5x equity (leverage 1.5 max). Inverse-contract note for the developer: BINANCE_CM alt contracts are 10 USD face each and PnL is in XRP. Size in contracts = round(target_usd_notional / 10), and skip the trade if that is under 1 contract.

Iteration 3. The reported Layer-1 'syntax error' at line 1 came from a placeholder string that was submitted instead of code. This submission is the full iteration-2 code with only the Layer-2-feedback fix applied: the attribute self._bar_ts, which shadowed the framework method _bar_ts(cls, bar), is renamed to self._cur_bar_ts. The signal, entry, exit and sizing logic are unchanged. The code is pure ASCII.

verifying_stuck_3_timeouts

Implementation

XRP COIN-M 4H trend-pullback continuation. The trend is the vol-normalised 20-day log move (t = ln(C/C[120]) / (ATR% * sqrt(120))), and |t| >= 0.3 counts as a trend. Enter long when an up-trend closes near its 3-day low, and short when a down-trend closes near its 3-day high (depth >= 0.8). Exit on pullback completion (close back in the far 20% of the 3-day range), on a 2.5x entry-ATR close stop, or after a 120-hour calendar clock. Size is 1.5% equity risk / (2.5*ATR%), capped at 1.5x equity, in whole $10 contracts.

Iteration History

Verification failed (Layer 1 — static analysis): - `self._bar_ts = ...` binds a value over framework method `_bar_ts(cls, bar)` — every later call raises "'<type>' object is not callable" (or the framework loses the member). Use a different attribute name. (line 87) - `self._bar_ts = ...` binds a value over framework method `_bar_ts(cls, bar)` — every later call raises "'<type>' object is not callable" (or the framework loses the member). Use a different attribute name. (line 72)

Verification failed (Layer 1 — static analysis): - Syntax error: invalid character '—' (U+2014) (<unknown>, line 1) (line 1)

Outcome Summary

Future iterations should first find out why verification kept timing out on this COIN-M inverse-contract strategy (for example, sandbox runtime or data availability for XRPUSD_PERP.BINANCE_CM) before spending more coding iterations on the signal logic.

The run was abandoned with reason 'verifying_stuck_3_timeouts' after 3 iterations, meaning it repeatedly timed out in the verification stage; backtest review, optimization, analyst, and risk review were never reached.

A long-short 4H trend-pullback continuation strategy on XRPUSD_PERP.BINANCE_CM that buys near the 3-day low inside a vol-normalised 20-day up-move (and sells near the 3-day high inside a down-move), exiting on pullback completion, a 2.5-ATR stop, or a 5-day clock, with 1.5%-of-equity risk sizing capped at 1.5x leverage.

No performance data exists: the backtest report is empty and no optimization report was produced, so there are no return, Sharpe, or trade-count figures.

Analysis

【复合CTA交易系统New】(多因子+多品种+多策略自适应公用版)|【Composite CTA Trading System New 】(Multi-factor + Multi-variety + Multi-strategy Adaptive Public Version)

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