Do chart patterns predict stock returns?
By Mehmet Saruhan, Founder·Opulence Alpha Research · Published · Data through
Not on their own. Across 1,354,468 weekly-sampled detections of 15 chart patterns on US stocks since 1995, a stock showing a pattern went the way it pointed against the median stock over the next month 47 to 51 times in 100 — the range a coin flip produces. How often a pattern reaches its drawn target depends mostly on how far away the target is.
Keywords: chart patterns, do chart patterns work, technical analysis evidence, bull flag success rate, head and shoulders, double bottom·JEL classification: G11, G12, G14, C12, C58·Concepts: Multiple testing, Point-in-time data, Backtest overfitting
- Universe
- The same for every study: 1,767 US common stocks in the 11 GICS sectors, 422 of them since delisted; S&P 500 members since 1996 plus large and mid-sized companies outside the index
- Period
- Every Wednesday from 4 Jan 1995 to 19 Aug 2026: 1,550 Wednesdays across 7,959 trading sessions
- Sample
- n = 1,764 stocks of the 1,767-stock universe; 1,361,444 formation detections, a median of 906 per Wednesday
- Outcome
- Return over the next 5, 21 and 63 trading sessions against the same day's median stock; whether the drawn target was reached
- Inference
- Date-clustered intervals; t on non-overlapping dates; proven only when |t| ≥ 3
1Introduction
Flags, triangles, head and shoulders, double tops and bottoms: chart patterns are among the most taught tools in technical analysis, and handbooks such as Bulkowski's list how often each one reaches its target. Academic tests are more guarded: Lo, Mamaysky and Wang found that some patterns carry a little information about the returns that follow, without testing whether a stock showing one goes on to beat other stocks. This study asks that plain question on one universe: does a stock showing a pattern go on to beat the same day's median stock, and does reaching the drawn target say more than how far away the target was?
2Data and method
2.1Sample design
Population and frame. The population is US common stocks listed on the NYSE and Nasdaq; funds, ETFs, trusts, preferred shares, warrants and units are excluded. The sampling frame is a fixed universe of 1,767 companies, drawn once when the platform was built and not re-sampled since, in two strata. Stratum 1 is a census of the S&P 500: every company in the index at any time since 1996 whose price history could be recovered, 1,027 companies of which 413 have since delisted; it holds 76% of the index's members in 1996 and at least 96% in every year from 2010. Stratum 2 is 740 large and mid-sized companies outside the index, selected in proportion to the market's sector weights from the stocks that passed a minimum share price of $15 and a minimum average daily trading value of $25 million; 9 of them have since delisted.
Sample. The unit of observation is a formation detection: one stock on one Wednesday on which the pattern detector recorded a formation. Weeks the platform's regime model flags as a likely change of market regime are left out, as are bars flagged as bad data. This study's sample is n = 1,764 stocks of the 1,767: 1,361,444 detections on 1,550 Wednesdays from 4 Jan 1995 to 19 Aug 2026, a median of 906 stocks with a formation per Wednesday. 217,945 detections (16%) come from the 420 companies that have since delisted. Every detection and every Wednesday carries equal weight.
Representativeness. Table 1 gives the sample by GICS sector beside the S&P Composite 1500: 7.9% of companies would have to change sector for the two to match exactly. By latest market value, 55% of the listed companies with a recorded market value are large (at least $10bn), 38% mid ($2–10bn) and 6% small. Because Stratum 2 was chosen from companies listed at construction, its history carries survivorship bias. Section 4 reads the result by market volatility.
Companies in this study's sample, of them those since delisted, the size of the listed companies with a recorded market value, the sample's share of stock-weeks and the sector's share of the S&P Composite 1500.
| Sector | Sample | Delisted | Large | Mid | Small | Stock-weeks | S&P 1500 |
|---|---|---|---|---|---|---|---|
| Information Technology | 283 | 67 | 122 | 76 | 18 | 13.5% | 12.7% |
| Financials | 255 | 59 | 114 | 74 | 8 | 14.7% | 17.2% |
| Industrials | 255 | 52 | 126 | 67 | 11 | 15.9% | 17.5% |
| Health Care | 237 | 47 | 92 | 81 | 16 | 12.3% | 10.9% |
| Consumer Discretionary | 209 | 39 | 69 | 87 | 12 | 12.4% | 12.9% |
| Energy | 105 | 30 | 39 | 31 | 5 | 6% | 4.7% |
| Consumer Staples | 101 | 38 | 40 | 17 | 5 | 5.9% | 4.9% |
| Materials | 95 | 34 | 35 | 24 | 2 | 5.6% | 5.1% |
| Real Estate | 84 | 10 | 40 | 32 | 2 | 5.9% | 6.9% |
| Communication Services | 80 | 31 | 28 | 13 | 5 | 3.9% | 3.3% |
| Utilities | 60 | 13 | 35 | 12 | 0 | 4.1% | 4% |
| All sectors | 1,764 | 420 | 740 | 514 | 84 | 100% | 100% |
Sample: n = 1,764 of the 1,767 companies. Size by latest market value for the 1,338 listed companies with one recorded: large ≥ $10bn, mid $2–10bn, small < $2bn. S&P 1500 shares count the constituents of the S&P 500, MidCap 400 and SmallCap 600 (1,506 companies, lists read 25 Sep 2026). Sectors are each company's current GICS sector.
2.2Signals
| Signal | Definition | Since | Wednesdays |
|---|---|---|---|
| Bullish formations | Bull flags and pennants, ascending and symmetrical triangles, double and triple bottoms, falling wedges, rounding bottoms and rectangles; a hit beats the same day's median stock | 4 Jan 1995 | 1,550 |
| Bearish formations | Bear flags and pennants, descending and symmetrical triangles, double and triple tops, head and shoulders, rising wedges, rounding tops and rectangles; a hit trails the median stock | 4 Jan 1995 | 1,550 |
| Drawn target and stop | The price the pattern projects and the price that invalidates it; the target counts as reached when it comes before the stop | 4 Jan 1995 | 1,550 |
| Conditions at detection | Volume, confidence, reward-to-risk, distance to target, false-break risk, trend, market volatility, age and stage, recorded when the pattern is found | 4 Jan 1995 | 1,550 |
2.3Measurement
Outcome. The return from the close on the detection's Wednesday to the close 5, 21 and 63 trading sessions later, in excess of the S&P 500, compared with the same day's median stock, so chance is 50 in 100 on every date; and whether the price reached the formation's drawn target before its stop.
Each detection is scored on its Wednesday: a bullish formation is a hit when the stock beat the same day's median stock, a bearish one when it trailed it. The detector also records whether the price reached the formation's drawn target before its stop (21 sessions for bullish formations, 42 for bearish).
Detections are split by the condition recorded at the moment of detection. For each group the study reports how often the drawn target was reached before the stop, and how often the stock beat the same day's median stock over the next 21 sessions.
The t-statistic uses non-overlapping dates only. A result is called proven when |t| is at least 3, it held in at least 60% of years and it is large enough to matter; with 1,470 tests across the studies, a looser bar would pass dozens by luck.
Full data and methods3Results
Each pattern against a coin flip
For every bullish pattern, how often the stock beat the median stock over the next month (21 trading sessions); for every bearish pattern, how often it trailed it. The dashed line is chance. The bar around each square is the 95% interval, clustered by date, because stocks that show a pattern on the same day move together.
How often the drawn target is reached — and why
A pattern draws a target and a stop. Across all bullish patterns, 20 in 100 reached the target before the stop within 21 sessions; the typical hit took 9 sessions. Whether the target is reached depends mostly on how far away it is and how far along the pattern already is — geometry, not a forecast.
Does anything make a pattern work?
We split every pattern by volume confirmation, confidence, reward-to-risk, distance to target, false-break risk, trend alignment, market volatility and age. In none of those splits did the pattern beat the median stock over the next month by a margin that held up across years (every t-statistic below 3).
The one measured effect runs the other way
Over the next week, two patterns went against their own call: 48 in 100 rounding bottoms beat the median stock, and 48 in 100 rounding tops trailed it. Both fit the market's strongest short-term relation — recent winners give some back.
Short-term reversal: the strongest relation we measured4Robustness
Table 3 pools every detection by the market's volatility regime on its date. In every regime, from calm markets to extreme ones, bullish and bearish formations stayed close to a coin flip over the next month: no volatility regime turned chart patterns into a reliable edge.
Edge over 50 in points at 21 sessions, detections pooled by the volatility regime of the detection date.
| Volatility regime | Bullish | Bearish |
|---|---|---|
| Low | -0.15 · n 245.9K | -0.40 · n 177.8K |
| Normal | 0.14 · n 265.4K | -0.39 · n 237K |
| Elevated | -0.83 · n 113.7K | -0.61 · n 130.8K |
| High | -0.19 · n 55.5K | 0.29 · n 77.8K |
| Extreme | 0.61 · n 16.1K | -0.05 · n 34.4K |
5Limitations
- The formations come from a pattern detector run over the whole history with today's rules, not from charts as readers marked them at the time; another detector, or a human eye, would draw some of them differently.
- Two rare types, bull and bear pennants, had too few detections to read on their own and are left out of the charts by pattern type.
- Before costs. Averages exclude trading costs, taxes and market impact.
- Same-close timing. Returns start at the close the signal is computed from; a real trade would start later.
- Survivors among smaller companies. The non-index names were chosen from companies listed when the universe was built; no delisting returns are added.
6Conclusion
A chart pattern describes the path a price has already taken. Across 15 pattern types since 1995, a stock showing one went the way its pattern pointed against the same day's median stock over the next month 47 to 51 times in 100, the range a coin flip produces, and none of the conditions recorded at detection changed that. The one measured effect, in rounding bottoms and rounding tops over the next week, went against their own call, in line with short-term reversal. How often a drawn target is reached depends mostly on how near it is: geometry, not a forecast.
References
- Lo, A. W., Mamaysky, H. and Wang, J. (2000). Foundations of Technical Analysis. Journal of Finance 55(4).
- Bulkowski, T. N. (2021). Encyclopedia of Chart Patterns, 3rd ed. Wiley.
- Harvey, C. R., Liu, Y. and Zhu, H. (2016). …and the Cross-Section of Expected Returns. Review of Financial Studies 29(1).
- Benjamini, Y. and Hochberg, Y. (1995). Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society, Series B 57(1).
- Grinold, R. C. and Kahn, R. N. (2000). Active Portfolio Management, 2nd ed. McGraw-Hill.
- Harvey, C. R., Liu, Y. and Zhu, H. (2016). … and the Cross-Section of Expected Returns. Review of Financial Studies 29(1).
- Shumway, T. (1997). The Delisting Bias in CRSP Data. Journal of Finance 52(1).
Appendix A. Questions readers ask
- Do chart patterns work?
- Measured against the median stock on the same day, not on their own. None of 15 common patterns beat a coin flip over the next month on US stocks since 1995. How often a pattern reaches its target depends mostly on how far away the target is.
- Do bull flags work?
- 51 in 100 bull flags beat the median stock over the next month, and 17 in 100 reached their drawn target within 21 sessions. That is no measurable edge.
- Does volume confirmation make chart patterns more reliable?
- Not in this data. Bullish patterns on volume at least 1.5 times normal reached their target 16 times in 100, against 23 on below-normal volume, and neither group beat the median stock afterwards.
- How long does a chart pattern take to reach its target?
- When a bullish pattern reached its target, it took 9 sessions at the median, and a quarter took 15 or more. Many never do: 80 in 100 hit the stop or ran out of time first.
- Is a confirmed breakout more reliable?
- Confirmed patterns reached their target 40 times in 100, against 20 for patterns still forming — mostly because the target is closer once price has broken out. Over the next month they did no better than the median stock.
Data availability and citation
Every figure in this paper is quoted from one frozen snapshot, published as JSON with the sample description, the robustness results and the test counts. The same snapshot feeds the research console, so the two cannot disagree. The paper and the snapshot are licensed under CC BY 4.0: share and adapt them with credit to Opulence Alpha and a link to this page.
Mehmet Saruhan (2026). Do chart patterns predict stock returns? Opulence Alpha Research, Opulence Alpha Studies, 25 Sep 2026. https://opulencealpha.ai/studies/chart-patterns
Research, not advice. A measured tendency across hundreds of stocks is a nudge for any one of them, never a forecast of its price.
More evidence
All studiesWhere this is read on the S&P 500 company pages
The company pages show each S&P 500 company's readings from the latest close; a ranking lists every company by one measure.