Chart Patterns
Encyclopedia of Chart Patterns
The only book that puts a measured base rate on every chart pattern instead of an adjective. Bulkowski hand-measured more than 38,500 patterns in US stocks between 1991 and 2004 and published the failure rate, average move, and throwback rate for each one. Every number assumes a perfect trade.
The one idea
Before this book, technical analysis ran on adjectives. Books said a head-and-shoulders top was reliable, that triangles usually broke out in the direction of the prior trend, that volume should confirm. Nobody said how often. Bulkowski, an engineer who got laid off at 36 and retired into the market, went to the library, found the same vague claims in every book, and decided to count. That is the whole idea: chart patterns have base rates, the base rates are knowable, and they differ enormously from one pattern to the next.
So each of the 63 chapters (53 chart patterns, 10 event patterns) opens with a Results Snapshot giving the same handful of measured numbers. Break-even failure rate, meaning how often price fails to move even 5% after the breakout. Average rise or decline, measured from the breakout price to the ultimate high or low, which he defines as the extreme reached before price reverses 20%. Throwback or pullback rate, meaning how often price comes back to the breakout price within 30 days. And percentage meeting the price target from the measure rule. Each is split into a bull market column and a bear market column.
The payoff is the spread between patterns. In a bull market a high and tight flag rose 69% on average with a 0% break-even failure rate. A descending scallop with an upward breakout rose 22% and failed to move 5% a fifth of the time. Same chart, same screen time, wildly different expected value. The spread inside a single pattern is nearly as wide: head-and-shoulders tops that were tall and narrow fell 26% on average, while short and wide ones fell 17%. Reading this book properly means using it to decide which setups you are allowed to take, not learning what a triangle looks like.
What it actually teaches
This is a lookup manual, so the method is the six-step loop you run at the chart before a trade, not a curriculum.
START AT THE PATTERN INDEX
The back of the book carries an Index of Chart and Event Patterns: thumbnail drawings of all 63 patterns with page numbers. Match the shape on your screen to a thumbnail, turn to that chapter, and check the Identification Characteristics table before you assume you have it right. For a head-and-shoulders top that means three peaks with the center highest, shoulders at roughly the same price and roughly the same distance from the head, and a neckline drawn across the two troughs between them. The neckline slopes up about 52% of the time and down about 45%.
READ THE SNAPSHOT, PICK THE RIGHT COLUMN
Every chapter opens with a one-page box: performance rank, break-even failure rate, average rise or decline, change after trend ends, throwback or pullback rate, and percentage meeting the price target. Two columns, bull and bear. His bull market is every date outside March 24, 2000 to October 10, 2002. Head-and-shoulders tops in a bull market rank 1 of 21 bearish patterns: 4% break-even failure, 22% average decline, 50% pullbacks, 55% meeting target.
IGNORE THE HEADLINE FAILURE RATE
Break-even failure only counts patterns that never moved 5%, on the theory that 5% covers trading costs. It is the friendliest possible definition. Each chapter also prints a ten-row failure table at 5%, 10%, 15% and up. Bull-market head-and-shoulders tops: 4% fail to fall 5%, 15% fail to fall 10%, 35% fail to fall 15%, 54% fail to fall 20%, 68% fail to fall 25%. Find the row that matches your actual target and use that number, not the one on the cover of the chapter.
APPLY THE FOR BEST PERFORMANCE LIST
Each chapter ends with a checklist of conditions that measurably improved the result, keyed to the table each came from. For head-and-shoulders tops: tall and narrow beat short and wide by 9 points (26% versus 17% average decline), where tall means height above the median 17.27% of breakout price. Rising volume through the pattern, U-shaped volume, volume highest on the head, a horizontal or up-sloping neckline, a higher left shoulder, and a breakout in the middle third of the 12-month range each added one to four points. This list, not the headline stat, is where the book earns its price.
SET THE TARGET, THEN CHECK ITS HIT RATE
The measure rule is almost always pattern height projected from the breakout price. For a head-and-shoulders top, take the highest high in the head minus the neckline price directly below it, then subtract that from the breakout. His worked example: head at 51, neckline at 47.38, height 3.62, breakout high 48.50, target 44.88. Then look up how often the target is actually reached. Head-and-shoulders tops: 55% in a bull market. He also gives an alternative measurement, height from the head down to the low of the higher trough, projected down from that trough, which raises the hit rate to 62%.
PLAN FOR THE RETRACE BEFORE ENTRY
A throwback (upward breakout) or pullback (downward breakout) is price returning to the breakout price within 30 days, with clear white space between the retrace and the breakout level. Half of head-and-shoulders tops pull back and the average round trip takes 12 days. It costs you: patterns with a pullback fell 20%, patterns without fell 24%. Before entering, look at what support or resistance sits within a few percent of the breakout, and decide in advance whether the retrace is your add or your exit.
What it looks like Monday morning
You stop treating chart patterns as one category. Before taking a breakout trade you open the relevant chapter, read the Results Snapshot, and score your specific setup against the For Best Performance list. A short, wide head-and-shoulders top with a down-sloping neckline and obvious support 3% below the neckline is not the same trade as a tall, narrow one with a horizontal neckline breaking out in the middle of its yearly range, and the book quantifies the gap in average outcome. Some patterns come off your list entirely: symmetrical triangles fail to move 5% roughly one time in ten in both directions, and their downward breakouts hit the projected target less than half the time.
The second change is how you set targets. Instead of projecting pattern height and calling it a target, you look up how often that projection actually gets reached, then check the failure table for how often the pattern never travels that far at all. If the measure rule says a 20% decline and 54% of bull-market head-and-shoulders tops never fall 20%, you book profits nearer. You also stop being surprised by week two, because you already know half of these trades come back to your entry price within 12 days.
| Pattern, bull market | Break-even failure | Average move | Throwback or pullback | Hits measure-rule target |
|---|---|---|---|---|
| Flag, high and tight, up breakout | 0% | +69% | 54% | 90% |
| Double bottom, Eve and Eve, up | 4% | +40% | 55% | 67% |
| Head-and-shoulders bottom, up | 3% | +38% | 45% | 74% |
| Ascending triangle, up | 13% | +35% | 57% | 75% |
| Symmetrical triangle, up | 9% | +31% | 37% | 66% |
| Head-and-shoulders top, down | 4% | -22% | 50% | 55% |
| Symmetrical triangle, down | 13% | -17% | 59% | 48% |
Break-even failure only means the pattern never moved 5%, so a 4% failure rate and a 55% target hit rate describe the same pattern without contradicting each other.
Warning: The statistics in this book are based on perfect trades.Thomas N. Bulkowski, Encyclopedia of Chart Patterns
Where it fails
- Every headline number is a perfect trade. Bulkowski buys at the exact breakout price and sells at the exact ultimate high, which he defines as the highest high before price reverses 20%. That top is only identifiable after the fact. No commissions, no slippage, no spread, no bad fills. The 69% average rise on a high and tight flag is the average of 253 such trades. The 38% average rise on a head-and-shoulders bottom is the average of 554 of them. He states in the Glossary that the likelihood of duplicating this in real trading is zero. Treat the tables as a ranking device, never as expected return.
- The famous low failure rates measure almost nothing. Break-even failure rate only counts patterns that failed to move 5%. Head-and-shoulders tops show 4% in a bull market, and that 4% is the number everyone quotes. In the same chapter's failure table, 35% fail to fall 15% and 54% fail to fall 20%. If you hold for the measured move, your real odds are the 54%, not the 4%. The book supplies the honest number and then puts the flattering one on the chapter's first page.
- There is only one bear market in the entire dataset. He defines the bear market as March 24, 2000 to October 10, 2002 and classifies literally everything else in the 1991 to 2004 window as a bull market. So the bear column is the dot-com unwind and nothing else. 2008, 2020, and 2022 are not in this book. Sample counts thin out badly there too: 640 bull-market head-and-shoulders tops against 174 bear-market ones, and the eye-catching 44% average decline for bear-market diamond bottoms rests on 20 samples, which he flags in a footnote most readers skip.
- A lot of the pattern performance is just market beta. Head-and-shoulders tops declined 29% on average in a bear market versus 22% in a bull market, which reads like the pattern works better when it is scary. But the S&P 500 fell 13% during the average bear-market holding period and rose 1% during the average bull-market one. Bulkowski says this himself. The tables are not market-adjusted, so a chunk of what looks like pattern edge is direction you could have captured with an index short.
- The patterns were found by eye and the universe was filtered. He identified patterns manually, merging the first edition's 15,000 with newer hand-found ones. The stock universe required a heartbeat (not too flat), excluded thinly traded or unusually volatile names, and he usually removed stocks that fell below one dollar on the assumption bankruptcy was coming. Cutting the worst outcomes truncates exactly the tail that bullish pattern statistics need to be honest about. No rule in the book is precise enough to hand a computer and reproduce his counts, and he concedes that if you do not follow his methods your results will vary.
- The measure rule is close to a coin flip on some patterns. Symmetrical triangles with downward breakouts reach the projected target 48% of the time in a bull market and 42% in a bear market. Head-and-shoulders tops reach it 55% of the time. Eve and Eve double bottoms, 67%. The book still presents pattern-height projection as a rule in every chapter's Trading Tactics section, and only the snapshot line quietly tells you how often it works.
- The event pattern section has aged worst. The ten chapters on earnings surprises, same-store sales, analyst upgrades and downgrades, and FDA drug approvals are measured on 1991 to 2004 market structure, before decimalization settled, before algorithmic execution, and before most large retailers abandoned monthly comparable-sales reporting in the 2010s. The FDA chapter has no bear-market data at all. These are also the weakest patterns in the book: FDA approvals show a 34% break-even failure rate on upward breakouts and 39% on downward ones, against 20% and 13% average moves.
- It assumes a specific trader, and says so nowhere. Everything is daily bars on individual US stocks with real exchange volume. Nothing here is measured intraday, and futures, forex, and crypto traders lose the volume analysis, the gap analysis, and the event chapters outright. Holding periods are long: head-and-shoulders tops averaged 62 days to the ultimate low in a bull market and 41 in a bear market. And of 1035 pages, the Sample Trade in every chapter is admitted fiction, complete with invented characters, so a meaningful slab of the book illustrates nothing measurable.
Who it is for
Buy it if
You already trade breakouts in US stocks on daily bars and want to know which patterns deserve capital and which conditions actually move the number. Nothing else publishes measured base rates per pattern instead of adjectives, and the For Best Performance checklists are worth more than the headline stats. Buy it as a desk reference you open per trade, not as something you read.
Skip it if
Futures, forex, and crypto traders get almost nothing here, since every number comes from individual US stocks on daily bars and much of the analysis leans on stock-specific volume and gaps. Day traders should skip it too: the average pattern takes 41 to 62 days to reach its ultimate high or low. And if you want a system with entries, stops, and position sizing, this book contains none of that.
As an Amazon Associate I earn from qualifying purchases. It costs you nothing extra, and it does not change what gets recommended or what the verdict says.
Reviewed here is the 2nd edition (2005), which is what most used copies are. The link goes to the current 3rd edition (2021). The pattern statistics were re-measured for it, so the exact numbers quoted above will differ.
Back to all titles