Dedicated $250K Alpaca paper account, fully automated — buys $25,000 (market) the moment a symbol matches the strategy (updated 2026-10-25, v3, 1H multi-confirm): 1H clean breakout (close above the ENTIRE previous 1H candle) + 1H RSI<45 + most recent completed Daily volume above its own 20-day average + a completed Weekly clean breakout + a Daily clean breakout within the last 3 candles. Max 10 concurrent positions (down from 50) — see the backtest card below for why this replaced the prior v2 15m entry rule (Daily EMA20>50 + Price>Daily SMA200 + 4H EMA20>50 + 1H RSI<45 + plain 15m breakout). Checked from each signal's current state every cycle; this same match is also what flags a "New AutoTrade Signal" badge on Trade Now (which shows every tracked signal now, not just matches), so every autotrade entry is tagged there too. Sells at the FIRST of (price ≥ entry +3%) OR (1H RSI reaches 80) while this calendar month's realized P&L is below a 10% monthly goal (Profit Target / RSI Exit below); once a position is entered AFTER that month's realized P&L has already reached 10%, it exits on 1H RSI≥80 only — no quick 3% booking — since the monthly income goal for that month is already met and the position can afford to run patiently. No stale-days cutoff and no forced rotation at full slots this version — a new signal is simply skipped if all 10 slots are taken, rather than selling an existing position to make room. A repeat detection on the same symbol within 3% of its last entry price is ignored (no re-trade, no re-alert) — only the first detection in a price cluster is acted on. No manual buys; "Close" below is a manual override. Click "Overrule" on any open position to stop the bot from selling it — no RSI exit, no profit-target exit — until you un-overrule it or close it yourself; use it when you want to hold past the standard exit rules.
The 15m v2 entry (Daily EMA20>50 + Price>SMA200 + 4H EMA20>50 + 1H RSI<45 + plain 15m BO) backtested at profit factor 3.19 before going live, but real trading degraded it to 2.38 — 82% of its exits turned out to be forced rotations (selling the current best-performing position to free a slot whenever all 50 were full), not the intended RSI≥80 exit. The entry itself was never the problem; running 50 scattered, loosely-filtered slots was. The replacement tightens the entry with three extra confirmations (daily volume, a completed Weekly breakout, a recent Daily breakout) on the 1H timeframe instead, and runs only 10 concentrated slots — removing the rotation pressure entirely rather than trying to patch around it.
| Strategy | Trades | Win rate | Avg return | Profit factor | Total return |
|---|---|---|---|---|---|
| 15m v2 — backtested pre-launch (retired 2026-10-25) | 169 | 74.6% | +4.31% | 3.19 | — |
| 15m v2 — real live performance (same rule, actual fills) | 150 closed | 61.3% | — | 2.38 | — |
| 1H multi-confirm, staged exit current | 393 | 90.1% | +3.32% | 3.69 | +113.29% |
Train/test split of the new strategy at the data's temporal midpoint (2025-05-07), to check the result isn't an artifact of one favorable stretch:
| Trades | Win rate | Avg return | Profit factor | Total return | |
|---|---|---|---|---|---|
| Train (before 2025-05-07) | 168 | 88.7% | +3.21% | 4.02 | +56.97% |
| Test (2025-05-07 onward) | 226 | 93.8% | +4.43% | 8.60 | +83.08% |
Both halves hold up independently (no sign flip, no collapse), though the test half's PF jump to 8.60 is on the high side and partly reflects a strong stretch in that window — treat the all-period 3.69 as the more durable expectation. Averaged 11.2 entries/month, +3.62% average monthly return, 92% of months positive, across the backtest. The 10% monthly goal was never actually reached in any single month of the backtest, so every logged trade used the 3%-or-RSI80 exit — the patient RSI≥80-only branch is correct as specified but hasn't yet been exercised by real data; it only activates once a month has already cleared +10% realized before a new position opens. Backtested on cached 1H/Daily/Weekly history (a longer, better-powered window than the ~4-month local 15m archive the old strategy's number was checked against) — not a perfect apples-to-apples data window, noted rather than smoothed over.
Same 589-setup population throughout (4H Supertrend Buy + Daily EMA20>50 + Price>Daily SMA200 + clean 15m BO). Tested whether the entry should wait for 15m RSI<45 to also happen, buy immediately with no extra gate, or use session VWAP gap instead of RSI.
| Entry rule | Trades | Hit +5% | Avg return | Win rate |
|---|---|---|---|---|
| Buy at the flip, no extra gate current | 589 | 50.0% | +4.36% | 77.4% |
| Wait for 15m RSI<45 to coincide with the flip (retired 2026-09-27) | 280 (47.5% of setups — rest never fired) | 51.1% | +4.14% | 74.3% |
| Buy at the flip, VWAP gap ≥1% required (retired 2026-09-29) | ~301 | 61.5% | +6.10% | 82.4% |
RSI<45 added no measurable edge at the entry candle (it's almost always already recovered above 45 by the time a clean flip prints) — it only cost trade volume, since 52.5% of setups never got a live entry at all while waiting for RSI to dip again. VWAP gap ≥1% looked like a real discriminator when first tested (2026-09-27), but a fresh re-test on a larger archive (2026-09-29) reversed it — splitting the SAME 589-setup-style population by VWAP gap sign/threshold found the <1% bucket now outperforming the ≥1% bucket (82.9% win rate, 10.40 profit factor vs. 69.7% / 2.20, n=35 vs 142). Given that reversal, the VWAP gate was dropped too — the live rule is back to buying every clean flip with no extra gate, same as the original 2026-09-23 finding. See Trade Now's filter grid to check VWAP Gap manually on any signal — it's no longer required, just shown for reference.
| Week of | Trades | Wins | Losses | Win % | P&L $ | Cumulative $ | Best Trade |
|---|
| Entry Date | Entry Time | Exit Date | Symbol | Qty | Entry $ | Exit $ | P&L | Status |
|---|---|---|---|---|---|---|---|---|
| No trades yet. | ||||||||