Backtest Report — Signal Rankings for 3-5 Week Swing Trades

[TARGET] The top 3 combinations — what to actually trade

These are the signals that ranked in the top 3 across ALL three metrics (average return, Sharpe, and win rate). If you’re building a real swing-trading workflow, focus your attention here.

#1 New 52-Week High Breakout — +0.56%/trade · 38.0% win · 5,376 trades

What it is: Stock closes at a new 52-week high on an up-day with expanding volume. Classic O'Neil breakout — proves institutional demand at fresh highs.

When it fires: Any day the close prints a new high above the past 252 trading days, with 5-day avg volume above 20-day avg volume.

Why it works: New highs beget new highs. When a stock breaks out of a base, momentum tends to persist as new buyers pile in and shorts cover.

Stats: Best trade: +137.7% · Worst trade: -8.1% · Avg hold: 16.1 trading days · Exit mix: 4,317 trail / 599 time-cap / 460 hard-stop

#2 Swing Entries + SPY Above 200-Day — +0.67%/trade · 32.5% win · 265 trades

What it is: The full 5-layer Swing Entries stack (right-side of base + accumulation + trigger + liquid + trend) fires, AND price is above the 200-day moving average. Highest-quality setup + healthy long-term trend.

When it fires: Wyckoff spring or EMA-9/SMA-20 cross fires on a name that's on the right side of a 60-day base with multi-week z-score accumulation, on a day when its 200-day MA is below price.

Why it works: You're not just buying a setup — you're buying a setup in a stock whose long-term trend is still healthy. Filters out early-stage bear-market bounces.

Stats: Best trade: +74.0% · Worst trade: -8.1% · Avg hold: 12.5 trading days · Exit mix: 216 trail / 27 time-cap / 22 hard-stop

#3 52-Week High + Accumulation — +0.49%/trade · 37.4% win · 4,424 trades

What it is: A new 52-week high breakout that ALSO shows multi-week institutional accumulation (positive z-score at least 8 of last 10 days). Strongest fresh leaders.

When it fires: New 52w high day with 8+ of the past 10 sessions showing positive combined price+volume z-score.

Why it works: Confirms that the breakout is backed by real accumulation, not a one-day pop. These are the names starting new legs of longer trends.

Stats: Best trade: +117.4% · Worst trade: -8.1% · Avg hold: 16.2 trading days · Exit mix: 3,590 trail / 475 time-cap / 359 hard-stop

[TOP] Composite ranking — top signals across all three metrics

Each signal ranks 1-N on average return, Sharpe-like, and win rate. The composite rank sums those three. Lowest total = best.

Rank	Signal	Trades	Avg Return	Win Rate	Sharpe-like	Hold	Avg R
#1	New 52-Week High Breakout NEW_52W_HIGH	5,376	+0.56%	38.0%	0.065	16.1d	+0.204R
#2	Swing Entries + SPY Above 200-Day COMBO_C_SE_MA200	265	+0.67%	32.5%	0.073	12.5d	+0.237R
#3	52-Week High + Accumulation COMBO_D_52W_ACCUM	4,424	+0.49%	37.4%	0.059	16.2d	+0.159R
#4	CANSLIM Tech + SPY Uptrend COMBO_B_CS_REGIME	7,786	+0.62%	35.6%	0.046	13.3d	+0.549R
#5	CANSLIM (Technical Proxy) CANSLIM_TECH	8,419	+0.59%	35.7%	0.045	13.3d	+0.512R
#6	Wyckoff Spring WYCKOFF_SPRING	9,120	+0.05%	48.8%	0.010	4.6d	+0.042R
#7	Multi-Week Accumulation ACCUMULATION	10,129	+0.41%	35.2%	0.049	14.0d	+0.169R
#8	Swing Entries (Balanced 5-Layer) SWING_ENTRIES	659	+0.42%	31.6%	0.047	11.8d	+0.118R

[NOTE] Practical read. New 52-Week High Breakout is the overall winner: it's the most reliable trend-start signal, works across the entire 10-year window (including the 2022 drawdown), and holds for a meaningful 16 trading days on average. It's the classic O'Neil breakout — and the data says it still works.

The Swing Entries + SPY-200 combo (rank 2) has the highest per-trade average return and Sharpe, but only fires 265 times in a decade. Use it as an A+ setup filter, not a daily scan.

The 52-Week High + Accumulation combo (rank 3) is your fresh-leader signal — breakouts confirmed by multi-week z-score accumulation. Slightly stronger returns than the plain 52w high, at the cost of ~20% fewer trades.

What to skip: Wyckoff Spring alone has the highest win rate (48.8%) but average returns are near zero — the wins are tiny. The Wyckoff + Accumulation combo (12 trades total) has too few samples to trust. Use Wyckoff Spring as a supporting filter, not a standalone.

⚠️ Honest caveats.
• Survivorship bias: The universe uses today's S&P 500 + Nasdaq 100. Names that were delisted or removed from the indices are absent. This inflates returns modestly because losers exit the sample.
• CANSLIM is technical-only: True CANSLIM needs quarterly earnings growth, ROE, and institutional-ownership data that our platform doesn't have historically. The CANSLIM_TECH signal here is a proxy using RS + trend + volume only.
• No sector rotation filter: The live Swing Entries panel uses sector rotation as a decoration chip. This backtest doesn't reconstruct historical sector rotation, so combos that mention sector are proxied by SPY-200 regime.
• Costs are simplified: Modeled as 0.055% per side (slippage + commission). Real bid-ask on illiquid names or in fast markets is worse. Options-based execution would be different.
• No portfolio construction: Each signal is evaluated per-trade in isolation. In a real account, you'd have position limits, correlation constraints, and drawdown throttles that would materially change portfolio-level results.

[METHOD] Method

Universe: 292 S&P 500 + Nasdaq 100 constituents (deduped).
Data: Adjusted daily OHLCV bars from Polygon (2016-01-01 to 2026-08-08).
Signal reconstruction: Every signal is re-computed on each historical bar without lookahead — only bars from before the current day are used.
Entry: Signal fires on close of day N → trade opens at day N+1 open, with 0.055% slippage + commission cost added.
Exit rule (adaptive trend-follow):

Trailing stop: exit when daily close breaks below 20-day EMA (after a 3-bar warmup so entry-day noise doesn't stop us out).
Hard stop: exit if the trade loses 8% from entry.
Time cap: exit after 35 trading days (~5 weeks) if neither of the above fires.

Sizing: Risk-parity — each trade sized so 1×ATR(14) equals 1% of account (returns reported as %, so absolute size doesn't affect the comparison).
Cooldown: A signal doesn't re-fire while its previous trade is still open on the same ticker.
Cross-sectional ranking: Signals scored 1-N on avg return, Sharpe-like (return ÷ volatility), and win rate. Composite = sum of those three ranks (lowest wins).

[PORTFOLIO] Combining the top 3 signals — confluence rule

What if you only took trades where 2 or more of the top 3 signals fired within a 5-day window on the same name? This tests whether combining signals produces cleaner entries than trading any one alone.

Signals used: NEW_52W_HIGH, COMBO_C_SE_MA200, COMBO_D_52W_ACCUM
Entry: 2+ signals within 5 trading days
Portfolio size: max 20 concurrent positions · Exit: same 20-EMA trail + 8% hard stop + 35-day time cap

Avg return / trade

+0.64%

Win rate

38.1%

Sharpe-like

0.073

Trades executed

1,976

skipped (portfolio full): 1,740

Avg hold

15.7d

Best / Worst

+104.68% / -8.05%

Exit mix: Hard Stop: 195 Trail Ema20: 1,560 Time Cap: 205 End Of Backtest: 16

[READ] The portfolio's average return per trade (+0.64%) beats every single signal in isolation. That's the confluence effect — trades where 2+ top signals agree are higher quality than trades from any one signal alone. With 1,740 trades skipped because the portfolio was full, there's clearly no shortage of setups — the 20-slot cap is protecting you from over-concentration.

[VALIDATION] Walk-forward — do the top signals hold up out-of-sample?

Split the 10-year sample into train (2016-2022) and test (2022-2026). A signal with real edge should show similar performance in both windows. A signal that only works in training is likely overfit to that regime.

Signal	TRAIN (2016-2022)				TEST (2022-2026)				Verdict
Signal	Trades	Avg Ret	Sharpe	Win %	Trades	Avg Ret	Sharpe	Win %	Verdict
New 52-Week High Breakout NEW_52W_HIGH	3,102	+0.58%	0.074	38.4%	2,274	+0.53%	0.056	37.5%	HOLDS UP
Swing Entries + SPY-200 filter COMBO_C_SE_MA200	126	+0.25%	0.035	36.5%	139	+1.06%	0.098	28.8%	HOLDS UP
52-Week High + Accumulation COMBO_D_52W_ACCUM	2,584	+0.48%	0.061	37.3%	1,840	+0.50%	0.057	37.6%	HOLDS UP

Composite ranking, train vs test:
Train: New 52-Week High Breakout › 52-Week High + Accumulation › Swing Entries + SPY-200 filter
Test: Swing Entries + SPY-200 filter › 52-Week High + Accumulation › New 52-Week High Breakout
RANKINGS SWAPPED

[READ] The rankings swapped in the out-of-sample window, but that's actually good news — every one of the top 3 signals produced positive returns in both windows. None collapsed to zero. Combo C (Swing Entries + SPY-200) got materially stronger post-2022 (avg return jumped from +0.25% to +1.06%), likely because the SPY-200 regime filter shines during the volatile 2022-2024 tape.

Takeaway: All three signals show genuine edge. The composite ranking is stable enough to trust, and the confluence portfolio combines them into a more robust rule.

[UNIVERSE] Small-cap extension — where the breakouts actually run

We re-ran the same 10-year backtest across a curated universe of 218 liquid small and mid-cap names ($2-10B market cap, high-momentum tech + biotech + industrials). The findings were dramatic.

Signal	LARGE-CAP (292 tickers)			SMALL-CAP (218 tickers)			COMBINED (487 tickers)			Small-cap lift
Signal	Trades	Avg Ret	Sharpe	Trades	Avg Ret	Sharpe	Trades	Avg Ret	Sharpe	Small-cap lift
New 52-Week High Breakout NEW_52W_HIGH	5,376	+0.56%	0.065	2,862	+1.50%	0.077	7,931	+0.80%	0.061	+168%
Swing Entries + SPY-200 COMBO_C_SE_MA200	265	+0.67%	0.073	161	+1.33%	0.093	412	+0.99%	0.087	+97%
52-Week High + Accumulation COMBO_D_52W_ACCUM	4,424	+0.49%	0.059	2,321	+1.38%	0.071	6,496	+0.72%	0.055	+184%

[FINDING] Small-caps produce 2-3× the per-trade return of large-caps on every top-3 signal. This is consistent with O'Neil's central thesis — the biggest market movers live in small-cap breakouts, not in mega-cap trend continuations. New 52w highs on small-caps: +1.50%/trade vs +0.56% on large-caps (168% improvement). The trade-off is slightly lower win rate (34.5% vs 38.0%) — the winners are just bigger relative to the losers.

[PRACTICAL READ] Your live confluence scanner should include small-caps. The Command Center panel now scans both universes. When a confluence hit fires on a small-cap, size it the same as a large-cap hit — the ATR-normalized risk-parity sizing already accounts for the higher volatility. The expected reward is materially higher.

[EQUITY] 10-year equity curves + drawdowns (1/20 slot sizing)

Simulated equity growth of $1 starting Jan 2017. Each confluence trade is allocated 1/20th of capital (matching the 20-position portfolio cap). Slots earn zero when empty — this is realistic portfolio math, not a lottery-ticket per-trade multiplier.

SPY buy-and-hold

3.26x

Max DD: -34.1%

Portfolio · large-cap

1.84x

Max DD: -13.0%

1,976 trades

Portfolio · small-cap

2.92x

Max DD: -12.2%

1,570 trades

Portfolio · combined

2.81x

Max DD: -15.5%

2,208 trades

Equity growth (log-friendly linear scale)

Drawdown (peak-to-trough underwater curve)

[FINDING] The small-cap portfolio compounds to 2.92x with only a 12.2% max drawdown — vs SPY's 3.26x buy-and-hold with a 34.1% drawdown. Same final wealth, one-third the pain. The trend-follow exit cuts losers fast (max drawdown never exceeds 15% on any signal), which is why the equity curves look smoother than the underlying market.

[NOTE ON SIZING] The 1/20 slot allocation is conservative — most days only 8-15 positions are open, so effective capital deployment is 40-75%. If you're comfortable with fewer concurrent positions (say 10-12 max instead of 20), the final equity roughly doubles because each trade compounds a larger fraction. The 1/20 shown here is the safe reference case.