Technical Analysis
Technical Analysis of Stock Trends
This is the book every chart pattern in every other technical book was copied out of, and after seventy-eight years the definitions still hold. The problem is that the pattern half is now common knowledge, while the half almost nobody reads, the stop and basing-point rules in Part Two, is where the money actually is.
The one idea
Edwards and Magee argue that you do not need to know why a stock is moving. Price is the only place where every buyer's hope and every seller's fear gets settled into one number, and that number already contains everything anyone knows or guesses. So instead of forecasting earnings, you read the tape. Their word for it is supply and demand: at every price level there is a quantity of stock people want to unload and a quantity people want to own, and the chart is a picture of which side is winning.
The second half of the idea is the part that made the book famous. When that balance flips, it does not flip instantly. A group that owns twenty thousand shares cannot dump them in a morning without killing the price they are dumping into, so they feed the stock out in waves, supporting the price on the dips and selling into the rallies. That process leaves a shape on the chart. Distribution over three pushes with shrinking volume draws a head-and-shoulders. Distribution against a fixed ceiling draws a descending triangle. The shapes are not magic. They are the visible residue of a large position changing hands slowly.
And because the shape is made by a real transfer of stock, its size tells you something about the size of the move that follows. That is why every pattern in this book comes with a measuring rule and a minimum target, and why the book insists on volume as a second signature: the price picture and the volume picture have to agree, or the pattern does not count.
What it actually teaches
Edwards and Magee run a strict order of operations, and skipping a step is how people lose money with charts.
READ THE PRIMARY TREND FIRST
Before you look at a single stock, decide which way the whole market is going, using the twelve Dow tenets. A Primary trend usually lasts more than a year and moves the averages more than 20%. Secondary corrections run 3 weeks to many months and retrace one third to two thirds of the preceding primary swing, most often close to 50%. Minor swings are usually under 6 days and are ignored outright. The trend is assumed to continue until reversal is definitely signaled. The signal is only valid when both averages confirm (originally Industrials and Rails, now Transports), on closing prices only, and any penetration counts, even 0.01. Intraday highs do not exist for this purpose.
WAIT FOR THE PATTERN TO COMPLETE
A head-and-shoulders top needs four things, all of them: a high-volume rally and pullback (left shoulder), a higher high on heavy volume that reacts back below the left shoulder top (head), a third rally on decidedly lighter volume that fails to reach the head (right shoulder), and finally the break of the neckline. Miss one and it is not a pattern. A symmetrical triangle cannot even be drawn until four minor reversals exist: top, bottom, lower top, higher bottom. Flags and pennants have to form after a straight-line move with volume drying up continuously throughout.
DEMAND THE 3% CLOSE AND THE RIGHT VOLUME
The same confirmation rule runs through the whole book: price must close beyond the pattern boundary or trendline by roughly 3% of the stock's price. It does not have to happen in one day. Upside breaks require a marked increase in volume or you do not trust them. Downside breaks do not require any volume increase at all. One counterintuitive warning: a downside break out of a symmetrical triangle on conspicuously heavy volume, especially deep into the apex, is more often a two or three day shakeout that reverses than a real decline. Triangles work best when the break comes between half and three quarters of the horizontal distance from the base to the apex. Past three quarters, the pattern loses its force and Edwards tells you to go look at a different chart.
MEASURE THE MINIMUM MOVE
Head-and-shoulders: measure the vertical distance from the top of the head down to the neckline, then project that same distance down from the point where price broke the neckline. In the Teledyne 1984 example that was 44 points from a 264 neckline, giving 220. Critical qualifier: there are two minimums, the formula target and the size of the advance that preceded the pattern, and the smaller one applies. Triangles: draw a line from the top of the first rally parallel to the lower boundary, and expect price to reach it at roughly the same angle it was travelling before the pattern. Flags and pennants are half-mast patterns: measure the pole from where the prior move broke away to where the flag began, then project that same distance from the flag breakout, measuring chart distance on a semilog scale rather than counting points. The triangle formula is explicitly less reliable than the head-and-shoulders formula. There is no maximum rule for any of them.
CHECK WHAT IS OVERHEAD BEFORE YOU BUY
Old tops become support and old bottoms become resistance, which is the reverse of what most people in a boardroom will tell you. Estimate the strength of a level with three criteria: how much volume traded there originally, how far price subsequently fell below it (for stocks between 20 and 35, expect little supply unless price dropped more than 10% under the level), and how long ago it formed. A level that has already been attacked once has spent some of its supply, and a third attack usually gets through. The standing instruction is to overestimate resistance, never underestimate it.
SET THE STOP OFF A BASING POINT
A minor low only becomes a basing point once price has moved three days away from it: three full days whose entire range sits above the high of the day that made the low. Then the stop goes a fixed percentage below that low, taken from Magee's table, from 5% for a conservative stock over 40 up to 15% for a speculative stock under 5. Never tighter than 5%, ever. The underlying formula is the stock's normal range for its price divided by 15.5, multiplied by its sensitivity index (what we now call beta), multiplied by 5%. The point of the padding is that everyone can see the obvious low, so a stop sitting right on it gets run.
RATCHET THE STOP, DO NOT PREDICT THE TOP
Stops on longs are never moved down, only up (and never up on shorts). You raise the stop each time a new basing point confirms, and each time a new high closes more than 3% above the previous high. Then the one exception: after a big run, when a day prints exceptional volume and it is not the day of a fresh breakout to new highs, cancel the protective stop and place a day-only stop one tick under that day's close. Repeat it every day until it fills. Magee's own worked example risked one eighth of a point and captured an extra two and five eighths. This is how the book takes profits: it never uses a limit order at the measured target, because the moves that overshoot are exactly the moves you cannot afford to cut short.
What it looks like Monday morning
You stop looking for the pattern and start looking for the confirmation. Most people who use charts see a head-and-shoulders forming and act on it. Edwards and Magee will not let you: the pattern does not exist until price closes 3% through the neckline, and about one in five of them get saved before that happens. That single rule kills the majority of the trades a chart reader would otherwise take, which is the whole point. On Monday you go back through your open positions and ask which ones you entered on a shape rather than a confirmed break.
Then you go find the last basing point under every long you hold, count three days away from it, and put a real stop there with 5% to 15% of padding depending on how volatile the name is. Not a mental stop, not a level you are watching. And you write down the measured target for each position and accept that it is a minimum, not an exit, because the exit is the ratcheting stop. That is the actual working system in this book, and it lives 435 pages in, which is why almost nobody who cites Edwards and Magee is using it.
| Stock price | Conservative (beta under 0.75) | Median (beta 0.75 to 1.25) | Speculative (beta over 1.25) |
|---|---|---|---|
| Over 100 | 5% | 5% | 5% |
| 40 to 100 | 5% | 5% | 6% |
| 20 to 40 | 5% | 5% | 8% |
| 10 to 20 | 5% | 6% | 10% |
| 5 to 10 | 5% | 7% | 12% |
| Under 5 | 5% | 10% | 15% |
The stop goes this far below the last confirmed basing point, never closer than 5%, and once set on a long position it is only ever moved up.
In brief, a Reversal Pattern has to have something to reverse.Edwards and Magee, Technical Analysis of Stock Trends
Where it fails
- The headline evidence is a curated table, and the book half admits it. The proof offered for Dow Theory is a compounding record: $100 in 1897 becomes $11,236.65 by 1956 in the original, updated to $345,781.94 by the end of 2005 against $39,685.03 for buying the low and selling the high. Read the footnotes. The record excludes commissions, transfer taxes, and dividends. The signal dates come from one source (Jack Schannep), and the editor concedes that 'not all theorists are in 100% agreement as to the exact date or nature of the signals' and that the theory 'is not a 100% objective algorithm.' A backtest where the entry dates are a matter of opinion is not a backtest. The original table also quietly shows the cost: 15 round trips, one outright loss, and three occasions where the reinvestment happened at a higher price than the preceding sale.
- The specific rules were calibrated to a market that no longer exists. The progressive stop is stated as one eighth of a point under the close, in a market that decimalized in 2001 (the 8th edition editor flagged this as an open question and it was never resolved). The famous 30-degree trendline angle is an artifact of the publisher's own TEKNIPLAT chart paper, and the book says so outright: 'It is pure happenstance that the TEKNIPLAT sheets tend to produce the 30-degree ascending line.' The volume rules assume the NYSE consolidated tape, where one venue printed nearly every share; today volume splits across dozens of exchanges and dark pools, so 'notably heavier than recent activity' is a much noisier reading. And the pattern examples come from issues trading a few hundred round lots a day, where a single fund's accumulation genuinely did leave a visible three-wave signature on a chart.
- The reliability numbers are recollections, not measurements. Symmetrical triangles behave 'more than two thirds' of the time, 'lacking an actual statistical count.' Roughly 20% of head-and-shoulders tops get saved before the neckline breaks. About one third show more volume on the left shoulder than the head. No sample size, no period, no method. Every one of those figures is one man's impression of the charts he kept. Later work that actually counted (Bulkowski's pattern encyclopedias, among others) produced materially different breakout-direction and target-fulfillment rates. On top of that, the roughly 200 chart examples were chosen after the outcome was known and nearly all of them worked, with the rare failures explicitly labeled as exceptions.
- Nine editions and three authors argue with each other on the page. The base text is Edwards from 1948, essentially frozen at the 5th edition. Magee revised through the 7th. Bassetti then inserted annotations mid-sentence tagged 'EN' and 'EN9', plus whole interpolated chapters numbered 5.1, 17.1, 18.1, 23 and 28.1. The result contradicts itself. Chapter 3 presents the two-average confirmation rule as the foundation of everything, then a note at the end of the same chapter says Dow Theory 'is no longer adequate to its original purpose' and you now need a composite of the Dow, S&P 500 and NASDAQ, without ever giving a rule for how to combine them. Chapter 11 says SEC regulation ended pool manipulation, and the footnote to the footnote says the skullduggery just changed form. Appendix E is Curtis Faith's Turtle rules, a volatility-sized mechanical futures system with no relationship to chart patterns at all, bolted on at the back.
- The book already documents its own patterns degrading. This is not a criticism from outside. Edwards noted by the 1960s that rectangles and right-angle triangles had become less common and symmetrical triangles 'looser than they were in the 1920s and 1930s, not as clean-cut and conspicuous.' Bassetti's 2005 note goes further: 'frequently patterns are not so neat as they were in the old days. Trend lines, especially horizontal lines seem to be more zones than hard and fast lines and more judgment might be necessary.' Sixty years of drift, admitted in the text, with no adjustment to the 3% rule or the measuring formulas that were built on the sharp version.
- It assumes a daily routine almost nobody now runs. Magee wants 20 to 30 charts as an absolute minimum, 100 if you can, up to 300 if you do this full time. He wants them updated 'every day without fail' at a fixed hour you defend against all other obligations. He wants two years of paper trading before you commit real size, he wants 8 to 10 simultaneous positions for diversification, and he wants at least a probable 15% move before a trade is worth the costs. Computers removed the pencil work but not the requirement: you still have to look at a lot of individual names every day to find the few that are set up. If you hold three ETFs and check them on weekends, the tactical half of this book does not apply to you.
- It quietly does not work for short-term traders or systematic ones. Dow Theory declares minor trends 'meaningless in themselves' and the whole framework works in weeks and months, so there is nothing here for anyone trading intraday. It is also unusable as a mechanical system by design, not accident: the editor states flat out that chart analysis 'is not reducible to an objective algorithm,' the reversal-pattern definitions turn on judgment calls about symmetry and volume relativity, and Robert Colby's 9-day and 39-day channel-breakout approximation of Dow signals is quoted in Chapter 5.1 with the warning that no system should ever be funded until the trader has exhaustively vetted it himself. If you want rules you can code and test, this book will frustrate you for 835 pages.
Who it is for
Buy it if
You already trade individual stocks discretionarily, hold for weeks to months, and keep losing money on breakouts that fail or profits you gave back. Chapters 13, 14, 27, 28 and 28.1 (support and resistance, trendlines, stop orders, basing points) will fix that specific problem better than anything written since. Buy it as a permanent desk reference you open when a pattern is forming, not as a book you read once.
Skip it if
You trade intraday, you trade options, or you want a system you can backtest. You will find a method that refuses to be mechanized, examples from a market where a thousand shares was heavy volume, and stop rules quoted in eighths of a dollar. Also skip it if you already know what a head-and-shoulders is and think that is what this book is for. Buy Bulkowski instead if you want counted statistics on how often the patterns actually pay.
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 9th edition. The link goes to the current 11th edition. Bassetti's editorial additions grew in later printings, so chapter numbering will not line up exactly.
Back to all titles