Objective: sustained account growth after costs. We will judge return, drawdown and loss size alongside entry frequency and win rate. Profitable months are not guaranteed. The rules stay frozen while fresh evidence accumulates.

1 prospective decision record saved. Latest signal session: 2026-09-04; recorded 2026-09-07T14:58:32.186452+00:00. Its next-session opening deadline is 2026-09-08T09:30:00-04:00.

No forward trades or profit results yet. The first future entry session has not been recorded. Historical returns shown below are not forward earnings.

Recorded research plan

For the next opening reference, the frozen account selected XOM, MA. Entry prices and actual units remain unknown. No orders were submitted; unresolved data and execution checks prevent activation.

Capture status: the first record was made manually. The daily step is ready for the existing 08:30 New York research workflow; continuous unattended collection has not been verified. The next required completed signal session is 2026-09-08. A missed deadline must remain a visible gap.

Price reliability: 31 intraday checks completed

Each stock’s largest remaining vendor disagreement, plus SPY, was checked against fresh daily and minute data. The disputed daily values persisted. Two prominent conflicts remain:

These fields remain disputed; neither vendor was silently substituted. Some other extreme prints appear in minute data too, which does not establish that the trade was valid or executable. Source correction and the opening-order/protective-stop procedure remain acceptance work.

Next evidence needed: fresh outcomes, verified source prices and IBKR execution observations. Paper results must be assessed separately from historical reference fills. No live trading, paper orders or new paid data subscriptions are enabled.

The calculations passed. The strategy is not ready for live trading. A separate calculation engine reproduced the original trades and account values. A second price provider reproduced the latest year’s entry count and win rate. Delayed entries and combined execution stress materially weakened the outcome.

What happens when execution is less favourable?

A high win rate is not enough: the tighter stop won 73.8% of its trades but lost 2.1% of the account. Combining higher costs, worse fills and late entries lost 16.5%, despite a 52.7% win rate. These are disclosed stress assumptions, not estimates of actual IBKR costs.

Comparison with SPY and dependence on individual stocks

Removing each of the 30 stocks separately left every latest-year account result positive. The weakest was +43.38% without GOOGL. Skipping the baseline’s largest cash-profit trade and replaying gave +47.65%. These are hindsight dependence checks, not rules for excluding stocks.

The continuous account made 485 entries, about 50.2 per year, with 29.3% annualized account growth in this simulation. It carries positions and cash across years. SPY and the strategy both exclude dividends and taxes; open holdings are marked at the final close.

Data correction and independent price check

We excluded 238 early META rows belonging to another instrument or the period before Meta’s ticker change. Meta’s comparable history now starts on 9 June 2022, with fresh indicator warmup. Original archives and earlier experiment results are preserved. No FB predecessor history was joined.

Nasdaq ticker-change notice · Roundhill’s former META ticker

Some daily fields still disagree substantially between providers, including SPY’s 2 February 2026 low and NVDA’s 10 June 2024 high. Matching recent performance does not certify every price field. No vendor value was silently substituted.

Next: observe fresh signals under the frozen rules

Resolve the material price discrepancies and confirm the exact opening-order and stop procedure. Then record new signals before outcomes are known, including skipped trades and real spreads. After IBKR is unlocked and paper execution is authorized, compare simulated entries with paper fills. Judge account return, drawdown and loss size alongside frequency and win rate. Any rule change starts a new version; do not retune the current test to improve its score.

Research audit complete; forward validation and execution readiness remain incomplete. No broker orders, operational watchlist expansion or strategy activation were performed.

The audit above supersedes these earlier-history conclusions with the META identity correction and stress tests. These original numbers remain visible for comparison; the latest-year baseline is unchanged.

2025-09-05 to 2026-09-04: 55 entries, 75.9% winning closed trades. 41 wins and 13 losses across 54 closed positions; 1 still open. Results below include an assumed 0.25% round-trip cost per trade.

Median holding period: 5 sessions. Average closed stock outcome: +1.93%; median: +1.62%. Gains are compounded through the simulated account; individual trade percentages are not added to calculate account return.

The change that increased activity

Buy a pullback in an established uptrend, then exit on recovery to the 20-session average. Allow two positions, reserving up to half the account for each new purchase. This keeps the recovery rule and lets the account take another stock while the first is held.

Earlier periods

The entry family had to pass both earlier periods before the two-position test. These periods were used for research selection; they are not untouched validation.

Next evidence needed: forward paper results under frozen rules. The operational watchlist, approved strategy and broker order settings have not been changed by this research.

How often did the latest year trade?

2025-09-05 to 2026-09-04: 18–20 simulated entries across the existing 10-stock shortlist. Start with no positions and require fresh setups inside this year. Earlier prices still warm the indicators. Unavailable and rejected cases do not count as entries.

Your preferred pace is monthly to biweekly. Compare the yearly total and the monthly distribution separately.

Every tested approach had no new entries in 2026-03, 2026-04, 2026-07. This is not a consistent monthly trading schedule.

Opportunity count is not your account trade count. Depending on the rule, maximum simultaneous holdings ranged from 5 to 7 stocks. A 100% stock allocation cannot take those simultaneous entries. Win rates and return summaries use closed positions only.

Known-history research; not unseen validation. Existing entry and exit rules remain unchanged. Costs, slippage, earnings exclusions, dividends, historical gamma and account allocation are not modelled. No orders enabled.

Frozen history through 2026-09-04. Each stock has at most one holding at a time and needs a fresh pullback after exit. The two new rules are compared with the original baseline and the earlier immediate-reclaim rule.

Neither new rule improves win frequency while retaining the original average winning gain. Keep the original strategy unchanged. A higher mean outcome alone is not proof of better account performance.

The mean includes both winners and losers. The median is the middle completed case, not the account return. Different stocks can still overlap, and no allocation or account compounding is simulated.

Waiting audit: 28 setups entered a waiting state; 18 later qualified for an entry-price evaluation. 10 cancellations were recorded, and 0 remain pending at the cutoff. A qualified price evaluation can still be rejected or unavailable before entry.

Before waiting or entry: 0 origin confirmations had unavailable reference-stop inputs and 0 produced a rejected reference stop. These remain separate from entered outcomes.

Waiting, cancelled and unavailable decisions are not losses or wins. Open or interrupted holdings block further entries in that stock. Unknown earnings, costs, slippage, dividends and executable quotes remain outside these gross research outcomes.

Research status: these fixed experiments use a selected universe and historical prices already examined. No rule was tuned after seeing this round’s results. The original strategy and broker execution settings are unchanged.

The aim is to win more often while retaining large winning gains. Both tests use the existing pullback dates. One requires the stock and SPY 50-session averages to rise from pullback start to confirmation; the other requires a closing price above every earlier daily high within the pullback.

Same frozen history through 2026-09-04. These rules were defined after the original results had been examined. This is research on known history, not an unseen validation.

Compare the same original signals

Each accepted case keeps its original entry, initial stop and exit. A failed filter consumes that setup; it does not wait for a later breakout. The rules are tested separately, never combined.

Unverified filters are separate from known failures. Neither missing nor unfinished outcomes count as wins or losses. These original cases can overlap, including repeated entries into the same large trend.

Allow one holding per stock

This second test regenerates entries after each exit. A new pullback must start strictly after that exit; setups forming during a holding are ignored. An open or interrupted holding blocks further entries.

Result: neither added rule improves both measures in the stock sequences. Keep the original rules as the reference. These experiments do not justify a strategy change.

These are separate stock sequences. Different stocks can still be held at the same time. This is not a simulation of choosing one stock for your concentrated account, and the returns cannot be compounded into an account result.

No candidate is automatically adopted. Costs, slippage, dividends, executable quote checks and your earnings-exit policy remain unmodelled. A better historical percentage alone does not establish a profitable improvement.