Study design and coverage
The study follows a fixed sample of 15,577 distinct Cup and Handle formations through breakout, measuring subsequent price behavior under the same rules for every formation.
The detector evaluated 290,334,077 rolling windows across 59 Binance USDT-margined perpetual markets and all 15 configured timeframes, with candles from January 2021 through August 2026. It produced 15,787 raw observations. After validation and earliest-observation family selection, the analytical sample contains 15,577 Cup and Handle formations.
Each formation enters the sample when its handle has returned to the neckline, before any breakout. Of the 15,577 formations, 11,390 closed above the neckline within the handle-scaled deadline (73.12%), 2,252 closed below the handle bottom first, and 1,935 did neither. The main 5% rise estimate uses 11,387 evaluable upward breakouts; other endpoints report their own denominators.
15,577 distinct formations
Every formation enters the sample once. The earliest valid observation represents each geometry, and the selection rule never reads a later outcome, so successful patterns are not preferentially selected.
Claim boundary
The headline rates describe what happened after an upward breakout. They are chart outcomes, not net trading returns, and they do not include entries, stops, fees, slippage, funding or position sizing.
Sample coverage by timeframe
| Timeframe | Evaluated rolling windows | Formations | Upward breakouts |
|---|---|---|---|
| 1m | 173,878,091 | 9,562 | 6,982 |
| 3m | 57,959,331 | 3,265 | 2,419 |
| 5m | 34,775,579 | 1,793 | 1,302 |
| 15m | 11,591,827 | 457 | 320 |
| 30m | 5,795,889 | 231 | 166 |
| 1h | 2,897,870 | 151 | 115 |
| 2h | 1,448,936 | 61 | 46 |
| 4h | 724,435 | 24 | 14 |
| 6h | 482,924 | 12 | 9 |
| 8h | 362,159 | 7 | 7 |
| 12h | 241,390 | 7 | 5 |
| 1d | 120,581 | 6 | 5 |
| 3d | 37,545 | 1 | 0 |
| 1w | 16,037 | 0 | 0 |
| 1M | 1,483 | 0 | 0 |
| Total | 290,334,077 | 15,577 | 11,390 |
One window is one eligible scan endpoint under the study's fixed inclusion rules. The 1w and 1M searches completed with full coverage and found no formations.
Why lower timeframes contain more observations
Lower timeframes contain more completed candles over the same calendar span, which creates more opportunities for patterns to appear. The 1m, 3m and 5m strata supply 94.0% of all breakouts, so the dedicated 5m+ and 15m+ views are essential context for the all-timeframe headline.
Detection atlas and timeframe sample sheets
One hundred visually distinct detector outputs, cropped and de-identified for public comparison. Statistical results use the complete analytical cohort, not these selected examples.
Outcome definitions and statistical method
Post-breakout outcomes use the same measurement rules as the Bull Flag and Bear Flag studies, with Cup and Handle geometry defining the target and failure levels.
Favorable rise
The rise from the open of the candle after the breakout close to the highest close before the first qualifying reversal, a close below the handle bottom, or the time cap. The primary Bulkowski-style threshold is 5%.
Structural hold
Whether an upward breakout avoids a close below the handle bottom during the predeclared width-scaled observation window. This is not a profit measure.
Full cup-depth target
The classical measure rule: cup depth added to the neckline. A hit uses the candle high and must occur before the move resolves.
Uncertainty
Binary outcomes use two-sided 95% Wilson intervals plus a symbol-cluster block bootstrap. Continuous moves report the mean, median and quartiles because the distribution is right-skewed.
Cup anchors, handle, neckline and handle bottom were fixed when the handle returned to the neckline, before any later candle was read. A formation could not be redrawn after its result was known. A breakout is the first later close above the neckline, before a close below the handle bottom and within max(3, handle length) candles.
The 5% threshold is identical across timeframes. The move ends at a close-to-close reversal of 12 × ATR(14)%, bounded by timeframe-specific limits (for example 1% to 4% at 1m and 20% to 30% at 1d), at a close below the handle bottom, or at a cap of max(500, 5 × formation width) candles. This is the same stopping rule used in the Bull Flag and Bear Flag studies, and at 1d it approximates Bulkowski's 20% trend-change rule.
Structural hold uses a window of 3 × formation width, bounded to 30 through 500 candles. A throwback is a return to within max(0.5 × ATR%, 0.3%) of the neckline within 30 candles. Every result prints its own evaluable denominator; censored rows are excluded, never counted as misses.
Data verification
Each formation was checked against Binance historical price data before post-breakout outcomes were calculated. The checks confirmed the recorded measurements for 15,575 of 15,577 formations, including breakout timing, invalidation, price moves and targets. The remaining 2, both 1m formations, had a candle-data discrepancy. Their original breakout classification is retained in the formation counts, but they are excluded from all post-breakout results. An independent recalculation of the detection results and a visual review of 120 charts also passed.
Pooled outcomes across all measured timeframes
These results combine evaluable upward breakouts across all represented timeframes using the same outcome definitions.
| Endpoint | Numerator / denominator | Estimate | 95% Wilson interval |
|---|---|---|---|
| Structural hold | 3,424 / 11,387 | 30.07% | 29.23% to 30.92% |
| Favorable rise ≥5% | 1,322 / 11,387 | 11.61% | 11.03% to 12.21% |
| Full cup-depth target | 3,893 / 11,387 | 34.19% | 33.32% to 35.06% |
| Neckline-proximity throwback (30 candles) | 11,274 / 11,388 | 99.00% | 98.80% to 99.17% |
Symbol-cluster block bootstrap intervals (95%): rise ≥5% 10.71% to 12.55%, structural hold 28.74% to 31.43%, full target 32.79% to 35.64%. Structural hold is a chart-validity outcome, not a profit rate.
How to read these rates: the formations are small
The median cup is 1.63% deep and the median handle bottom sits only 0.54% below the neckline (6.91% and 2.21% at 15m+). The full cup-depth target is therefore often a smaller move than 5%, which is why it is reached more often than the 5% threshold. A close below that shallow handle bottom anywhere in the median 210-candle hold window ends structural hold, which is why hold is lower than in the flag studies. The throwback rule counts any low within max(0.5 × ATR%, 0.3%) of the neckline; because breakout closes sit just above the neckline, 96.9% of throwbacks occur on the first candle. It is a boundary-proximity measure, not Bulkowski's rise-then-return throwback, and is not used in the Bulkowski comparison.
Practical timeframe aggregates
The 5m+ row removes the 1m and 3m strata. The 15m+ row isolates slower intraday and swing-style observations. Both are shown so the all-timeframe estimate is not mistaken for the result on the timeframes most traders use.
| Cohort | Breakouts | Structural hold | Rise ≥5% | Full cup-depth target | Median rise |
|---|---|---|---|---|---|
| 5m+ | 1,989 | 622/1,988 (31.29%) | 533/1,988 (26.81%) | 731/1,988 (36.77%) | 1.49% |
| 15m+ | 687 | 219/686 (31.92%) | 273/686 (39.80%) | 260/686 (37.90%) | 3.07% |
Wilson intervals for the 5m+ rise ≥5% and full-target rates are 24.91% to 28.80% and 34.68% to 38.91%. Corresponding 15m+ intervals are 36.20% to 43.51% and 34.35% to 41.59%.
10.35%
15m+ mean rise
3.07%
15m+ median rise
0.64%
15m+ 25th percentile
11.73%
15m+ 75th percentile
All-timeframe rise distribution
2.35%
Mean rise
0.71%
Median rise
0.15%
25th percentile
2.36%
75th percentile
Across all 11,387 resolved breakouts, the mean rise is about 3.3 times the median. A small number of large moves lift the average above what the typical formation experiences, so the median is the better description of a typical outcome.
Results for all 15 configured timeframes
Every configured timeframe is shown, including zero-detection strata. Sparse higher-timeframe rows are descriptive, not stable rankings.
| Timeframe | Breakouts | Structural hold | Rise ≥5% | Full cup-depth target |
|---|---|---|---|---|
| 1m | 6,982 | 2,136/6,980 (30.60%) | 447/6,980 (6.40%) | 2,355/6,980 (33.74%) |
| 3m | 2,419 | 666/2,419 (27.53%) | 342/2,419 (14.14%) | 807/2,419 (33.36%) |
| 5m | 1,302 | 403/1,302 (30.95%) | 260/1,302 (19.97%) | 471/1,302 (36.18%) |
| 15m | 320 | 90/320 (28.13%) | 99/320 (30.94%) | 111/320 (34.69%) |
| 30m | 166 | 64/166 (38.55%) | 73/166 (43.98%) | 71/166 (42.77%) |
| 1h | 115 | 39/115 (33.91%) | 54/115 (46.96%) | 45/115 (39.13%) |
| 2h | 46 | 20/46 (43.48%) | 26/46 (56.52%) | 24/46 (52.17%) |
| 4h | 14sparse | 3/14 (21.43%) | 7/14 (50.00%) | 5/14 (35.71%) |
| 6h | 9sparse | 3/9 (33.33%) | 5/9 (55.56%) | 3/9 (33.33%) |
| 8h | 7sparse | 0/6 (0.00%) | 1/6 (16.67%) | 0/6 (0.00%) |
| 12h | 5sparse | 0/5 (0.00%) | 3/5 (60.00%) | 0/5 (0.00%) |
| 1d | 5sparse | 0/5 (0.00%) | 5/5 (100.00%) | 1/5 (20.00%) |
| 3d | 0 | N/A | N/A | N/A |
| 1w | 0 | N/A | N/A | N/A |
| 1M | 0 | N/A | N/A | N/A |
Timeframes with fewer than 30 breakouts are marked sparse; their percentages are descriptive and should not be treated as precise rankings. No formation was found at 1w or 1M. The denominator in each fraction is the number of breakouts for which that endpoint was fully observed.
The answer changes with the win threshold
The complete 11,387-breakout favorable-rise curve is more informative than one headline percentage.
Volatility-normalized outcomes
ATR thresholds use the 14-period average true range at the breakout. Across all timeframes, a 1% rise occurred in 43.07% of resolved breakouts, a 5% rise in 11.61% and a 10% rise in 4.30%. The curve is descriptive and was not used to choose a preferred threshold. This is why “the Cup and Handle win rate” is incomplete unless the outcome threshold and cohort are stated.
Results by BTC market regime
Breakout-day regime splits are descriptive secondary analyses and were not used to select or redraw patterns.
| BTC market regime | Breakouts | Structural hold | Rise ≥5% |
|---|---|---|---|
| Bear market | 5,296 | 1,607/5,293 (30.36%) | 646/5,293 (12.20%) |
| Bull market | 6,094 | 1,817/6,094 (29.82%) | 676/6,094 (11.09%) |
Regime uses the previous completed BTCUSDT daily close relative to its trailing 200-day moving average on the breakout day; missing history is unclassified. These unadjusted splits reflect different symbol and timeframe composition and are not causal regime effects.
Comparison with Bulkowski's stock reference
The closest aligned comparison is the share of breakouts that rise at least 5%, Bulkowski's break-even convention.
| Metric | Bulkowski cup with handle (stocks) | All crypto timeframes | Crypto 15m+ |
|---|---|---|---|
| Average rise | 54% | 2.35% | 10.35% |
| Median rise | Not reported | 0.71% | 3.07% |
| Failure below 5% | 5% | 88.39% | 60.20% |
| Success at or above 5% | 95% | 11.61% | 39.80% |
| Full measure-rule target | 61% | 34.19% | 37.90% |
The difference is descriptive, not proof that one market is easier or harder to trade. Bulkowski's 913 “perfect trades” are daily stock cups lasting 7 to 65 weeks, held to the ultimate high before a 20% decline. This crypto cohort is dominated by intraday formations lasting hours, and its move ends at a volatility-scaled reversal. The pattern construction and trading rules are documented in our cup and handle pattern guide.
Does a shorter handle perform better?
Bulkowski reports that handles shorter than the median (22 days in his stock sample) showed superior post-breakout performance. The crypto equivalent compares handles shorter than the median handle of the same timeframe with the rest.
| Cohort | Breakouts | Structural hold | Rise ≥5% | Full cup-depth target | Median rise |
|---|---|---|---|---|---|
| Shorter than timeframe median | 5,401 | 1,660/5,400 (30.74%) | 565/5,400 (10.46%) | 1,883/5,400 (34.87%) | 0.62% |
| Median or longer | 5,989 | 1,764/5,987 (29.46%) | 757/5,987 (12.64%) | 2,010/5,987 (33.57%) | 0.79% |
In this crypto sample, shorter handles rose at least 5% less often than longer ones (10.46% versus 12.64%), so Bulkowski's stock finding does not carry over as a simple rule. This is an unadjusted descriptive split; handle length also correlates with cup size and timeframe mix, so it is not a causal test.
Limitations and claim boundary
- 1
Conditional post-breakout denominator
Of 11,390 upward breakouts from 15,577 formations, 11,387 are evaluable for the main 5% rise estimate. Other endpoints report their own denominators. The 73.12% breakout conversion is reported separately and is not a win rate.
- 2
Timeframe imbalance
1m, 3m and 5m contribute 94.0% of breakouts. Dedicated 5m+ and 15m+ aggregates prevent that mix from being mistaken for a universal timeframe result.
- 3
Timeframe-dependent stopping rule
The 5% threshold is fixed, but the reversal rule uses ATR and timeframe bounds. Timeframe rows are descriptive cohort results, not a controlled experiment on timeframe alone.
- 4
Dependence
Symbols can share the same market-wide shock. Wilson intervals treat observations as independent; the symbol-cluster block bootstrap is reported for the headline endpoints as a dependence-aware check.
- 5
Sparse higher timeframes
4h and above have small samples, and 1w and 1M have none. Those strata cannot support rankings.
- 6
No execution model
Entries, stops, fees, slippage, funding, liquidity, latency and position sizing are outside this descriptive study. Outcomes are not net strategy returns.
- 7
Detector scope
The detection specification defines the instrument. A future update to geometry rules, filters, market universe or timing can produce different estimates.
- 8
Price data
Post-breakout outcomes use Binance USD-M futures candlestick data. The recorded measurements were verified for 15,575 formations; 2 were excluded from post-breakout results because the price data differed from the original observations. Later follow-up candles were not part of the initial observations.
Questions this study answers
What is the Cup and Handle win rate in crypto?
Using a favorable post-breakout rise of at least 5% as the win definition, 1,322 of 11,387 resolved breakouts succeeded: 11.61%. The rate was 26.81% at 5m and above and 39.80% at 15m and above. Different endpoints and timeframe mixes produce different rates.
How many Cup and Handle patterns are in the study?
The sample contains 15,577 distinct formations, including 11,390 upward breakouts. The main 5% rise estimate uses 11,387 evaluable breakouts; other endpoints report their own denominators.
How often does a Cup and Handle break out?
11,390 of 15,577 formations closed above the neckline within the handle-scaled deadline (73.12%). A breakout is not a win; the rise and target endpoints measure what happened next.
Were all timeframes tested?
Yes. The run covered 1m, 3m, 5m, 15m, 30m, 1h, 2h, 4h, 6h, 8h, 12h, 1d, 3d, 1w, 1M. No formations were found at 1w or 1M.
What is the 5m+ result?
Among 1,988 resolved 5m+ breakouts, 26.81% rose at least 5%, 31.29% held structurally, and 36.77% reached the full cup-depth target.
What is the 15m+ result?
Among 686 resolved 15m+ breakouts, 39.80% rose at least 5%, 31.92% held structurally, and 37.90% reached the full cup-depth target.
How often does price reach the full cup-depth target?
3,893 of 11,387 resolved breakouts reached the neckline plus one cup depth before the move ended: 34.19%.
Does the 30.07% structural hold rate mean the strategy is profitable?
No. Structural hold describes chart validity through a fixed observation window. It does not include entries, exits, fees, slippage, funding, liquidity or position sizing.
Conclusion
The study contains 15,577 distinct Cup and Handle formations across 1m through 3d, including 11,390 upward breakouts. The most defensible all-timeframe headline is endpoint-specific: 11.61% rose at least 5%, 34.19% reached the classical cup-depth target, and 30.07% held structurally.
At 5m and above, 26.81% of 1,988 resolved breakouts rose at least 5%. For the 686-breakout 15m+ subset, the rate was 39.80%. None of these descriptive outcomes is a strategy return. A Cup and Handle has no single win rate until the endpoint, denominator, timeframe exposure and sampling design are stated.
Sources and references
Primary source. The dataset and outcome statistics in this article are original ChartScout research. Pattern identification, family selection and post-breakout measurement were designed and run in-house. When citing these figures, attribute them to ChartScout Research (2026).
- ChartScout Research. Cup and handle win rate in crypto: a study of 15,577 pattern formations. ChartScout, 2026.
Binance-only study across 59 pairs and 15 configured timeframes, followed by the fixed post-breakout outcome method of the ChartScout flag studies. - Bulkowski, Thomas N. Cup with handle: chart pattern statistics. ThePatternSite.com.
Stock-market baseline for average rise, break-even failure and measure-rule target. thepatternsite.com/cup.html. - Bulkowski, Thomas N. Measurement and pattern-statistics glossary. ThePatternSite.com.
Definitions for break-even failure, average rise, ultimate high and throwback. thepatternsite.com/glossary.html. - O'Neil, William J. How to Make Money in Stocks. McGraw Hill.
The original cup-with-handle framework. ChartScout's operational geometry is defined by this study rather than assumed equivalent. - Lo, Andrew W., Harry Mamaysky, and Jiang Wang. Foundations of Technical Analysis: Computational Algorithms, Statistical Inference, and Empirical Implementation. Journal of Finance 55, no. 4 (2000).
Methodological precedent for automated, rule-based pattern detection and formal statistical evaluation. nber.org/papers/w7613. - Sullivan, Ryan, Allan Timmermann, and Halbert White. Data-Snooping, Technical Trading Rule Performance, and the Bootstrap. Journal of Finance 54, no. 5 (1999).
Basis for the study's cautions on multiple-threshold reporting and dependence-aware inference. doi.org/10.1111/0022-1082.00163.
Stjepan Ivanović
Founder of ChartScout · Crypto Trader Since 2013
Trading crypto since 2013 with his first Bitcoin bought at ~$200. Four complete bull/bear market cycles, traded on early exchanges like Mt.Gox and BTC-e, on-chain trading on IDEX and EtherDelta, and ~70 crypto project investments. Built ChartScout after 23+ months of development to automate what no trader can do manually. Watch hundreds of charts 24/7.