Breakouts led the week while sweet spot setups stayed reliable
Backtest patterns, sweet spot performance, and missed-trade analysis for September 07 to September 11.
The Edge This Week
This week showed a clear split between symbol-level edge and scanner-level edge. Backtest results remained strong across the broader universe, but live setup conversion was weak, which suggests conditions favored select names with persistent directional follow-through rather than broad breakout participation.
The most important pattern was a collapse in low-pressure, short-duration setup quality. Historically favorable 3 to 5 bar compressions with pressure below 60 underperformed badly, indicating that textbook-looking structures were less reliable than symbol-specific strength and sector concentration.
BY THE NUMBERS
By The Numbers
The broad backtest sample remained healthy, with 189 symbols producing at least three trades and an aggregate win rate above 64%. That says the underlying model still finds repeatable behavior across many names, even though the live scanner had difficulty converting setups into confirmed breakouts.
Only 53 of 369 scanner setups broke out, while 117 failed. That imbalance matters because it points to a market where participation existed, but follow-through was scarce. In practical terms, traders needed stronger confirmation and more selectivity around see live setups in the scanner rather than assuming compression alone would lead to expansion.
The key divergence this week was strong historical edge paired with weak real-time setup conversion. That usually means pattern quality depended more on context than on structure alone.
SWEET SPOT REPORT
Sweet Spot Report
The sweet spot failed decisively this week. Setups defined by pressure below 60 and 3 to 5 bars produced just a 25.0% win rate, far below the 56.9% historical baseline across 313 trades.
This kind of underperformance usually signals that compression quality was being overstated by the raw scan. In stronger conditions, low-pressure consolidations often act as launchpads. This week, they behaved more like hesitation zones, where price looked orderly but lacked the sponsorship needed for clean continuation.
For traders, the lesson is not that the sweet spot stopped working permanently. It is that the setup likely required an extra layer of filtering, such as sector leadership, stronger relative strength, or a tighter link to symbols already showing robust backtest behavior. When a pattern trades at less than half its historical hit rate, the issue is rarely the definition alone. It is usually the surrounding environment.
When a historically reliable setup drops from 56.9% to 25.0%, treat the pattern as context-dependent until proven otherwise.
SYMBOL SPOTLIGHT
Symbol Spotlight
NOW stands out not just because of its 11 trades at a 100.0% win rate, but because its average result was still essentially a full +1R per trade. That combination suggests a clean, repeatable trend response rather than one or two outsized wins masking uneven execution. In a week where many scanner setups failed, $NOW behaved like a symbol with persistent institutional sponsorship.
GDX is interesting for a different reason. Its 95.45% win rate across 22 trades is one of the larger sample sizes among top performers, which gives the result more research value. Alongside GDXJ at 88.89% over 18 trades, it points to a clear pattern: metals-related vehicles provided cleaner continuation than many conventional growth or consumer names.
COST illustrates the other side of the tape. Despite being one of the most active scanner names with four setups, it produced just a 16.67% win rate over 12 backtest trades and an average of -0.6669R. That is a useful reminder that setup frequency is not the same as setup quality. Activity without follow-through often becomes a trap, especially in names that repeatedly attract attention but fail to expand.
WHAT THE BOTS MISSED
What The Bots Missed
There were no missed trades this week, no unrealized R left on the table, and no TP3 runners blocked by filters. That is an important data point because it means performance drag did not come from excessive filtering or poor bot participation.
Instead, the weakness came from the opportunity set itself. When the bots are capturing available trades but conversion remains soft, the research question shifts away from execution and toward market structure. In short, the system did not miss edge. The edge was simply narrower than the headline number of setups suggested. Traders can watch the bots in the Edge Lab to see how that distinction plays out in real time.
No missed trades means the process was efficient. The limitation this week was opportunity quality, not opportunity capture.
SECTOR HEAT MAP
Sector Heat Map
Breakouts clustered most heavily in ETFs, Financials, and Technology, with 9, 8, and 8 breakouts respectively. Healthcare followed with 7. Even with some duplicate casing in the source data for technology and consumer categories, the message is still clear: leadership was concentrated rather than broad.
Financials are especially notable because several of the most active scanner symbols came from the group, including AXP, BAC, C, and JPM. That suggests capital was willing to revisit liquid, institutionally followed names, but only selectively. Technology also remained active, though not uniformly productive, which fits the broader pattern of strong symbol dispersion inside an otherwise weak breakout tape.
By contrast, Industrials and Consumer-related groups saw less breakout concentration. Lower counts there imply either weaker trend alignment or less urgency behind new directional moves. For traders, the takeaway is straightforward: focus on where breakouts are clustering, but only after confirming that those sectors are also producing acceptable follow-through.
RESEARCH NOTE
Research Note
This week raises a useful research question: when live breakout rate falls to 14.4% while backtest win rates remain strong, which context filter best restores alignment between scan quality and realized outcomes?
The evidence suggests the answer may lie in symbol and sector persistence rather than in setup geometry. Names like LLY, NOW, and GDX delivered strong repeated outcomes, while the average low-pressure compression did not. That implies a possible ranking improvement for future testing: prioritize symbols with established positive expectancy over the prior sample window, then apply setup filters second.
In other words, the better research model may be symbol-first, setup-second. This week did not reward every clean chart. It rewarded clean charts in the right names.