Most writeups on the cup and handle pattern teach you how to find it. Almost none tell you how often it fails once you do. That gap is the reason this piece exists. Across our sample, a plain "it looks like a cup and handle" read is close to a coin flip. The three handle-level traits that separate winners from failures are measurable. They are not a matter of feel.
What we actually counted, and how
We built a dataset of 412 candidate cup and handle formations. The sample covered liquid US large-cap and index names from 2019 through 2025. Each candidate was first screened on daily bars inside Quantzee's AI Adaptive Quant Toolkit, then hand-checked against price and volume. Our data set used a stated, testable definition before any formation was counted. This was not a loose "looks like a U" eyeball call:
- Cup depth: 12% to 35% retracement from the pre-cup high. Shallower cups behaved like noise. Deeper ones behaved like broader downtrends, not bases.
- Cup duration: 7 to 65 weeks, rounded, measured from the left lip to the right lip.
- Handle retracement: no more than 50% of the cup's total depth, measured from the right-lip high.
- Minimum bars: at least 5 trading days inside the handle. A single down day was never counted as a completed handle.
According to this method, a pattern only entered the count once the handle closed. We did not retroactively count formations that broke down mid-handle and never completed. That choice is a source of survivorship bias worth naming up front. We return to it in the methodology note near the end.
The headline number: a 46% success rate, not the “70-85%” you’ve seen quoted
Of the 412 completed cup and handle formations, 190 reached our measured-move target within 12 weeks. The target is the cup's depth, projected upward from the breakout point. To count as a clean win, price also could not close 8% below the breakout level first. That gives a 46% hit rate for a full win.
Another 99 formations, or 24% of the sample, produced a positive move that stalled short of the full target. We call that a partial win, not a clean failure, since real money was still made on most of these. The remaining 123 formations, or 30% of the sample, either failed to clear the breakout level on volume, or reversed back through the handle low within four weeks of breaking out.
We tracked this result against the "70% to 85% success rate" figure that circulates on many pattern-identification sites. We found no published counted sample behind that figure in the sources we checked. It traces back to William O'Neil's original CAN SLIM framework. That framework was built on growth-stock leadership names, during strong market regimes, not a general population of chart formations. When we restricted our own 412-formation sample to names that were also in a confirmed market uptrend at breakout, the hit rate rose to 61%. Regime matters a great deal here. Most retail write-ups strip regime out entirely.
Trait one: handle retracement depth is the strongest single predictor
We split the sample into three retracement buckets. Each bucket is measured as a percentage of the cup's total depth:
| Handle retracement (% of cup depth) | Formations | Full measured-move hit rate |
|---|---|---|
| Under 20% | 121 | 58% |
| 20–35% | 186 | 47% |
| 35–50% | 105 | 29% |
A shallow handle, under 20% retracement, signals that sellers barely showed up after the cup's recovery. In our experience running this screen live across several quarters, this is the trait that most separates a genuine accumulation base from a pattern that merely resembles one on a static chart. A handle that gives back more than a third of the cup's recovery is, more often than not, a stock still finding a floor. It is not coiling for a breakout yet.
We also checked whether retracement depth interacts with cup depth itself. A shallow 12-18% cup paired with a shallow handle hit the full target 60% of the time in our sample, close to the shallow-handle average on its own. A deep 28-35% cup paired with a shallow handle still hit 52%, only modestly lower. That told us handle depth carries most of the predictive weight here, with cup depth playing a smaller supporting role.
Trait two: handle duration relative to the cup
We computed handle duration as a ratio of handle bars to cup bars. Patterns with a short, tight handle, under 15% of the cup's own duration, outperformed patterns with a long, drawn-out handle:
| Handle duration (% of cup duration) | Formations | Full measured-move hit rate |
|---|---|---|
| Under 15% | 147 | 55% |
| 15–30% | 178 | 44% |
| Over 30% | 87 | 31% |
A handle that drags on past roughly a third of the cup's own duration tends to lose the momentum it needed for a clean breakout. We found that the market tends to "forget" a base the longer the handle lingers. New sellers arrive who were not part of the original decline. The clean supply-and-demand picture that made the cup worth watching starts to erode once the handle overstays.
Duration and depth also compound. Our best-performing bucket combined a short handle (under 15% of cup duration) with a shallow handle (under 20% retracement): 64 formations, 67% full-target hit rate. Our worst bucket combined a long handle with a deep retracement: 41 formations, 22% hit rate. Those two extremes show how much the combination of traits matters, not any single number read in isolation.
Trait three: volume behavior inside the handle
This trait had the widest spread of the three. We measured average daily volume inside the handle against the cup's trailing 50-day average volume:
| Handle volume vs. 50-day average | Formations | Full measured-move hit rate |
|---|---|---|
| Drying up (under 70%) | 159 | 59% |
| Roughly average (70–110%) | 163 | 42% |
| Elevated (over 110%) | 90 | 27% |
Volume drying up through the handle, followed by a sharp expansion on breakout day, was the cleanest single tell we measured across all three traits. Per our breakout-day volume check alone, at least 1.5 times the 50-day average, formations that also had a drying-up handle hit the full target 64% of the time. That is meaningfully above any single trait measured on its own. Elevated volume during the handle usually means distribution is still happening, not accumulation, even though the price shape still looks like a textbook handle on the screen.
We also looked at where inside the handle the volume dried up. When the lightest volume day fell in the final third of the handle, right before breakout, the full-target hit rate climbed to 61%. When the lightest volume day fell early in the handle and volume crept back up before breakout, the hit rate dropped to 38%, even though the average volume for the whole handle looked similar in both cases. Where the dry-up happens matters almost as much as whether it happens at all.
Two traps that inflate how good this pattern looks on a chart
Trap one: judging a half-formed cup. It is easy to look at the left half of a developing cup, see the rounded shape starting to form, and assume the right half will complete to plan. We tested this directly. We walked forward through 85 formations that were later disqualified. In each case, the cup rounded out roughly as expected, but the handle either never retraced into our 50% bound, or never ran the minimum five-bar window needed to call it finished. A pattern is not tradeable until the handle has actually formed and the breakout has cleared on volume. A forming cup is a chart to watch, not a chart to act on.
Trap two: timeframe shopping. We reran the same 412 formations on the 4-hour chart, not just the daily. 38% of them did not reproduce. The cup shape that looked clean on the daily became choppy and ambiguous on the 4-hour. In other cases, a 4-hour "cup and handle" resolved into noise once zoomed out to daily. If a pattern only exists on one specific timeframe you had to search several charts to find, treat that as a sign the setup is weaker than it looks. It is not confirmation that you found something others missed.
Measured-move target versus what actually happened
The textbook measured-move target projects the cup's depth upward from the breakout price. Among the 190 formations that reached this target, the median time to target was 7 weeks. The median overshoot beyond the target, before the move eventually stalled, was an additional 11%.
Among the 99 partial-win formations, the median stall point was 58% of the way to the full measured-move target. That number is useful if you manage a position with partial exits rather than expecting an all-or-nothing outcome. Treat the measured-move target as a reasonable planning level for where to consider scaling out. It is not a promise of where price will stop.
What volume confirmation adds — and what it does not
Breakout-day volume at or above 1.5 times the 50-day average was present in 71% of the 190 full-target winners. The same condition appeared in only 34% of the 123 clean failures. That is a meaningful split, and it is why we treat volume confirmation as a real filter, not decoration.
What volume confirmation does not do is turn a weak setup into a strong one. A formation with a deep, long handle and average breakout volume still only hit the full target 33% of the time in our sample. Volume confirms the shape underneath it. It does not override a weak shape.
How the AI Adaptive Quant Toolkit scores this instead of just drawing it
The AI Adaptive Quant Toolkit on Quantzee's AI Adaptive Quant Toolkit does not simply outline a cup and handle shape on the chart. It scores a detected formation against the handle-depth, handle-duration and volume-dryup ranges described above. These are the same three traits this study isolated. The toolkit flags non-repainting signals only once the handle has closed and breakout volume has confirmed.
That turns "this looks like a cup and handle" into a graded read you can weigh against the rest of your plan, rather than a binary yes-or-no call off a static chart. A formation that scores well on all three traits is a different proposition from one that merely has the right outline.
Whatever tool or method you use to spot the pattern, paper-trade the setup first across a run of signals before sizing real risk behind it. A pattern with a 46% base rate, and a 61% rate in the right regime, is still a probabilistic tool. It is not a certainty. The gap between those two numbers is exactly why regime and trait filtering matter more than pattern recognition alone.
Two formations from the sample, side by side
Two examples from our 412-formation set show how the same rough outline can lead to opposite outcomes. We are not naming the underlying tickers here, since the point is the shape and the numbers, not the stock.
Formation A: a 22-week cup with 19% depth, a 9-day handle retracing 14% of the cup's depth, and handle volume averaging 58% of the 50-day baseline. The lightest volume day inside the handle fell three sessions before breakout. Breakout-day volume came in at 2.1 times the 50-day average. This formation reached the full measured-move target in 5 weeks and overshot it by 9% before stalling. Every trait here sat in our best-performing bucket, and the outcome matched.
Formation B: a 31-week cup with 33% depth, a 19-day handle retracing 44% of the cup's depth, and handle volume averaging 118% of the 50-day baseline. Breakout-day volume was only 1.1 times the 50-day average, below our 1.5x confirmation threshold. Price broke out, stalled for six sessions, then closed 9% below the breakout level twelve weeks later. Every trait here sat in our worst-performing bucket, and the outcome matched there too.
Neither formation would have looked obviously different to someone scanning a chart for the outline alone. Both have a rounded cup and a pullback that technically qualifies as a handle. The difference only shows up once you measure the three traits against the ranges in this study, which is exactly the gap a shape-only read misses.
Applying this the next time you spot a cup and handle
When you see a candidate on your own charts, run through the three traits in order. Check handle retracement first, since it carried the widest spread in our data. Anything retracing more than a third of the cup's depth deserves real skepticism, even if the overall outline looks textbook. Check handle duration next. A handle that has already run longer than a third of the cup's own duration is losing the momentum that made the setup worth watching in the first place.
Check volume last, and check it twice: once for the dry-up pattern through the handle, and once for the expansion on the breakout bar itself. Our data showed that where the dry-up occurs inside the handle matters almost as much as whether it occurs. A late dry-up, just before breakout, was a stronger tell than an early one. Finally, ask whether the broader market or sector is in a confirmed uptrend. That single regime filter moved our hit rate from 46% to 61% on its own, more than any individual handle trait did in isolation.
Frequently Asked Questions
A note on methodology and its limits
How we tested: 412 formations were screened on daily bars inside the AI Adaptive Quant Toolkit's pattern detector, then manually reviewed against OHLCV data. Each formation had to meet the stated depth, duration and handle bounds before being added to the sample.
Our methodology carries a known limitation worth stating plainly. Formations that broke down before a handle could complete were excluded from the base count. That means the 46% figure describes "completed formations," not "every cup-shaped setup that ever appeared." That is a narrower and more honest claim than most pattern guides make, and it is worth remembering before comparing our number directly against a looser definition elsewhere.
For broader background on reading price patterns as part of a research process, the SEC's Investor.gov glossary is a useful starting reference. For whether technical patterns carry statistically measurable information, and not just visual appeal, the academic literature is summarized in Lo, Mamaysky and Wang's NBER study on the foundations of technical analysis. That study found that several classical patterns do carry incremental information content when measured against a non-pattern baseline, which lines up with the gap we measured between regime-filtered and unfiltered hit rates.
If you want to go deeper on validating a pattern-based idea before trusting it with size, our guide on how to backtest a TradingView indicator walks through the same counted, out-of-sample approach used here. It pairs naturally with reading support and resistance levels, since the cup's prior high is itself a resistance level that the handle and breakout both have to clear. Before trusting any backtest or live result built on a pattern like this, it is also worth checking our breakdown of common backtest-vs-live trading mismatch causes, since slippage and fill assumptions can quietly move a 46% base rate in either direction.
Quantzee's indicators are analytical software built to score and flag chart conditions. They are not investment advice and do not predict outcomes with certainty. Paper-trade any pattern-based approach across a meaningful run of signals before committing real capital.
