Strategies

TASC 2026.03 'One Percent A Week' (Kurczek / PineCoders) — faithful port: TQQQ Monday-open 1% dip limit buy, next-day +1% limit target, break-even after -0.5%, Friday-close flat (TQQQ.USEQ 1H, long-only)

Outcome: Abandoned

TqqqOnePercentAWeekTasc

Outcome Summary

This strategy tried to reproduce the TASC March 2026 'One Percent A Week' snapback trade on TQQQ, a 3x leveraged ETF, on 1-hour bars on the commission-free USEQ venue. It had a statistically weak pre-study edge of +0.37% per trade before costs (t = 1.12) that got worse each year. Over 2 iterations the coding stage timed out three times, and the run was abandoned as 'coding_stuck_3_timeouts'. No verification, backtest, optimization or analyst verdict was ever produced, so whether the edge is real remains untested.

Strategy report

Hypothesis

A faithful translation of the open-source Pine v6 strategy 'TASC 2026.03 One Percent A Week' by PineCoders. It implements the article 'Trading Snapbacks In A Leveraged ETF' by Dion Kurczek (TASC, March 2026), https://www.tradingview.com/script/nVECqIQx-TASC-2026-03-One-Percent-A-Week/, licence MPL-2.0. Credit: the strategy is by the author above. The source is a real, tradeable strategy, not a utility. Instrument: TQQQ.USEQ (ProShares UltraPro QQQ, a 3x Nasdaq-100 ETF). This is exactly the instrument the article names, traded on our commission-free USEQ venue (zero commission, about 0.02-0.05% round-trip spread and impact modeled). Bars: TQQQ.USEQ-1-HOUR-LAST-EXTERNAL. The regular session is 09:30-16:00 ET with 7 bars per day; the 09:30 bar is the session's first bar, and timestamps are bar CLOSE. History runs from about 2024-08-14 to now (about 2.1 years, about 3,700 bars). This is the only intraday history available. Daily bars cannot express the intraday limit and target sequencing, so 1H is required. Sizing is exactly as the source: default_qty_type = percent_of_equity and default_qty_value = 10. Each entry buys whole shares worth 10% of current account equity at the limit price. The backtest starts at $100,000 where the source used $1,000; the scale does not affect returns. No leverage (leverage = 1.0), long-only, pyramiding 1 (the Pine default). Flat state: the strategy is flat every weekend and in every week where the dip never trades. The pre-study shows about 20-30% time in market. Parameters are the source defaults, held fixed and not optimized beyond the ±20% sensitivity check: - dip_pct = 1.0 (entry limit = Monday open * 0.99) - target_pct = 1.0 (target = average price * 1.01) - be_trigger_pct = 0.5 (break-even is armed when low < average price * 0.995) - qty_pct_equity = 10 Pre-study: our own point-in-time bar simulation of these exact rules on TQQQ.USEQ 1H, 2024-08 to 2026-09, before fees and spread, gave 88 trades at a mean of +0.37% per trade, t = 1.12, win rate 49%. - Exits: 64 break-even (mean +0.37%, from opens that gap through the level), 15 target (mean +4.1%, dominated by gap-up opens), 9 Friday close (mean -5.8%). - By year: 2024 +1.04% per trade (18 trades), 2025 +0.51% (39), 2026 -0.19% (31). - The same rules on unleveraged QQQ gave -0.01% per trade on 61 trades. Any edge comes from TQQQ's 3x volatility, as the article claims. The article's claimed performance is treated as unverified. The sample is small and statistically weak, and our backtest decides.

Mirrors the Pine regions 1-5 with the same order sequence and state (wOpen, pt, sendPT, beSent as persisted scalars; fill detection = change in position size, like ta.change(strategy.position_size)). Orders are real venue-resting limits: bar-level fills model gap-through fills at the open, just as Pine does. All order management runs in calculate_signal(), because the base template skips should_enter/should_exit while a non-contingent limit rests. should_enter/should_exit therefore return None/False. Order submission honours the warmup and replay-freeze guards. Platform adaptations: (1) Session first/last bars come from ET close time: the first bar closes at 10:30 and the last closes on the hour (16:00, or 13:00 on half-days). (2) Pine's pending 'Buy' order is replaced each Monday. As in Pine, an unfilled order keeps resting across the weekend until then, and a new one is only placed when flat (pyramiding 1). (3) A break-even limit set while price is already above avg fills immediately at the bar close, where Pine fills at the next open. (4) As in the source, only a Friday session triggers the week-end exit, so on a Good-Friday holiday week the position carries over the weekend. Sizing is equity*10%/limit price, leverage 1. Round-trip spread cost on USEQ is ~0.02-0.05%, against the pre-study's +0.37% gross per trade.

coding_stuck_3_timeouts

Implementation

Faithful port of PineCoders' 'TASC 2026.03 One Percent A Week' (Dion Kurczek) on TQQQ.USEQ 1H, long-only. On Monday's first regular-session bar (09:30-10:30 ET) it records the session open and places a GTC limit buy at open*0.99 sized at 10% of equity (whole shares). After a fill, a +1% limit sell (avg*1.01) goes in on the next session's first bar. If any bar's low trades below avg*0.995, the exit limit is re-priced to break-even (avg price) and the target is never re-armed. Any position still open is closed at market on the Friday session's last bar. calculate_signal returns the % distance of close from the weekly open.

Iteration History

sandbox_timeout: Verification failed (Layer 3 — sandbox backtest): smoke test exceeded the 300s wall-clock limit on a ~20000-bar window (833 days of TQQQ.USEQ-1-HOUR-LAST-EXTERNAL). This almost always means per-bar work that scales with history — e.g. rescanning the full funding/supplementary series, or rebuilding a list and calling min()/sorted() inside calculate_signal()/on_bar() on every bar. Precompute sorted timestamp arrays ONCE in __init__ and use bisect, or cache lookups keyed by timestamp, so per-bar cost is O(log n) not O(n).

Outcome Summary

A port that manages resting limit orders, session-boundary timing and break-even state by hand inside calculate_signal is heavy enough to time out the coding stage, so a future attempt should use a simpler, more incremental implementation that can finish coding and get a real backtest.

The run was abandoned in the coding stage with reason 'coding_stuck_3_timeouts' after 2 iterations. It never reached verification, backtesting, analyst review or risk review.

A faithful port of PineCoders' TASC 2026.03 'One Percent A Week' strategy (Dion Kurczek): on TQQQ.USEQ 1H bars, place a limit buy 1% below the Monday open, set a +1% limit target on the next session, move the exit to break-even after a 0.5% adverse move, and go flat at the Friday close, sized at 10% of equity, long-only, no leverage.

No pipeline backtest or optimization was produced, so no pipeline performance metrics exist. The hypothesis's own pre-study, before fees and spread, found 88 trades averaging +0.37% per trade (t = 1.12, 49% win rate), with the per-trade figure falling each year from +1.04% in 2024 to -0.19% in 2026.

Analysis

TASC 2026.03 One Percent A Week

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