Technical Analysis
A Short Course in Technical Trading
Kaufman tests four different ways to find a trend across every speed from 5 to 200 days and shows they all produce nearly the same profit curve. That single chart is the point of the book: the method you pick barely matters, the speed and the position size decide everything.
The one idea
Almost every technical analysis book sells you a way to find the trend. Kaufman built four of them, moving average, exponential smoothing, regression slope, and N-day breakout, ran all four across calculation periods from 5 to 200 days on the same data, and plotted the net profits on one chart. The four lines sit on top of each other. They all peak around 15 days, all collapse around 35 days, and all climb steadily after that. His conclusion is blunt: if a market trends, any of the four will make money, and the choice of method should not turn a loss into a profit.
That kills the question most traders spend years on. It also raises the one that actually pays. If the four methods land in the same place, the difference between them is style, not edge. Breakouts and regression slopes hold trades longer, take fewer of them, and make five times more per trade, but they let price swing further against you first. Moving averages and smoothing cut losses faster and trade far more often, because a moving average always whipsaws at a turn. You are not picking a better system. You are picking how you want your losses shaped.
The rest of the book follows from that. If the entry signal is close to a commodity, then the money is made in speed selection (slow beats fast, and he shows fast losing across all six methods on Eurodollars), in stop placement (at the point where the reason for the trade disappears, never at a dollar figure you can afford), and in sizing (equalize the daily dollar volatility of every position so no single stock decides your year). Those three chapters, 4, 8, and 17, are the book. The other fifteen are context.
What it actually teaches
Kaufman runs the reader through a full trading process in order, from finding direction to sizing the bet.
PICK ONE TREND METHOD AND MAKE IT SLOW
Any of his four calculations works. The moving average trend is up when today's average is above yesterday's average, not when price is above the average. The breakout version is his preferred rule 3: buy on an N-day high only if today's close is above yesterday's close, sell on an N-day low only if today's close is below yesterday's. Speed matters far more than method. He calls 25 to 60 days medium, warns that high-volume index markets carry the most noise and need the longest view, and notes a trading month is 21 days. Fast trend trading fails across every method he tests.
CONFIRM THE SPEED IS DEFENSIBLE, NOT DISCOVERED
If breakouts make money on your stock but the moving average and smoothing both lose, do not trade it. He says that pattern means you found an odd combination of past prices, not a trend. All methods should be roughly profitable or the market has no clear trend to follow. He also warns against picking the calculation period by grid search, and suggests using periods with a real-world reason: a week, a month (21 days), a quarter, because that is the rhythm of earnings reports, government data, and pension flows.
PLACE THE ORDER BEFORE THE CROWD, ON THE CLOSE
Calculate the trendlines shortly before the close and place the order for the close, not for the next open. He gives a trick for anticipating a moving average turn: drop the oldest day and average the remaining N minus 1 days, and that value is the exact price at which the trend flips today. On breakouts, put your order slightly ahead of the level everyone else is watching, so the gap through resistance works for you rather than against you. He calls that burst of movement free exposure. If you are wrong and the close goes the other way, exit on the next open and take the small cost.
SET THE STOP WHERE THE REASON FOR THE TRADE DISAPPEARS
Not at an amount you can afford to lose. The best exit for a long is the natural point where you would go short. His hard floor: a stop must never sit closer than 1.5 times the current high to low range, and if you want it closer you should just close the trade. When trading purely off volatility with no chart, he uses a profit target of 2 times the daily range and a stop of 3 times the daily range, so the target is closer than the stop. Raise stops to each new support level as the trade develops, never lower them, and use stop close only orders so an intraday spike does not take you out.
SIZE BY VOLATILITY, NOT BY DOLLARS
Equal share counts are the worst allocation and equal dollar amounts only work if volatility is exactly proportional to price, which it is not. His worked example: BIG trades at $100 with $3.00 average daily range, NEW trades at $25 with $1.00. For $10,000 you solve 10,000 = 100B + 25N with 3B = 1N, giving 57 shares of BIG and 171 of NEW, so each position risks $171 a day. For a bigger list he builds a table at an arbitrary target risk (his example uses $5,000 a day per stock), then scales the whole thing down by ten or a hundred to fit real capital. Measure volatility over the number of days you expect to hold, not over ten years.
DIVERSIFY, THEN STAY OUT OF THE MARKET MOST OF THE TIME
Three or four uncorrelated markets, and past four the benefit fades. Correlation above 0.20 is moderate, above 0.50 is strong, and if the average pairwise correlation of your book is over 0.20 your only real option is to trade less capital. He is clear that none of this saves you in a shock, because on days like September 11 everything correlates. His two defenses against that are structural: hold a position in any one stock no more than 30 percent of the time, and keep the position small. Being flat 70 percent of the time is what makes a price shock survivable.
What it looks like Monday morning
Two things change immediately. First, stop placement. If you have been setting stops at a round dollar loss you are comfortable with, measure the stock's current high to low range and check that your stop is at least 1.5 times that away, sitting under a real support level. If it is not, either widen it or cut the position size until it can be. Kaufman's point is that a tight stop and a trend system are fighting each other, and the stop always wins, which means you get taken out of the exact trades that were supposed to pay for the year.
Second, position size. Build a four-column spreadsheet for every open position: price, average daily range, shares, and daily dollar risk (range times shares). If one line is three times any other, your account is really a bet on that one name. Fix it by solving for share counts that equalize the daily dollar risk. Then check the correlation of your holdings in Excel with the correl function on daily prices. If the average is above 0.20, you are less diversified than the number of tickers suggests, and the honest response is to trade smaller rather than to add more names.
| Days | Net profit | Trades | % profitable | Max drawdown | Profit factor |
|---|---|---|---|---|---|
| 5 | 1.16 | 535 | 30 | -47.92 | 1.00 |
| 15 | 102.90 | 274 | 31 | -22.50 | 1.84 |
| 20 | -25.78 | 246 | 23 | -67.12 | 0.86 |
| 50 | -73.79 | 151 | 21 | -124.12 | 0.55 |
| 100 | -9.71 | 89 | 15 | -80.03 | 0.91 |
| 150 | 116.95 | 41 | 43 | -7.05 | 9.38 |
| 200 | 77.35 | 13 | 23 | -6.11 | 10.91 |
Same method, same stock, same five years: only the day count changes, and the answer swings from a 116.95 profit to a 73.79 loss.
A stop-loss fights with the trend system. The trend system wants to stay long and you want to get out. It just doesn't work.Perry J. Kaufman, A Short Course in Technical Trading
Where it fails
- The chapter that carries the book rests on one stock over five years. Table 4.1 in the 2003 edition tests moving averages on Microsoft alone, five years ending January 2002, and the general conclusion that slower is better comes from that single table plus one Eurodollar chart. The results in the table are wildly unstable: the 150-day period nets 116.95 with a profit factor of 9.38, but the 100-day loses money and the 50-day loses 73.79 with a 124-point drawdown. That spread is the fingerprint of noise, not of a stable relationship. Kaufman warns readers elsewhere in the same book about fitting data, then builds his headline finding on a sample of one.
- The parameter ranges he endorses came out of the largest single-stock bull market on record. He notes that everything from 20 to 55 days lost money on Microsoft and that this is exactly the range trend followers favor, while 150 to 200 days did best. Both halves of that finding are artifacts of Microsoft's 1997 to 2000 run and its 2000 to 2002 collapse. Since publication, the very slow end has been crowded (the 200-day average is the single most watched line in the market) and long-only slow trend following on individual equities has produced nothing like those numbers. Treat the specific day counts as illustration from one dataset, not as settings.
- Kaufman's own famous research is not in this book. There is no adaptive moving average, no efficiency ratio, and no noise-based market classification anywhere in the 326 pages. Search the index and the word adaptive does not appear. If you are buying this because Kaufman is the person who built the AMA, you want Trading Systems and Methods or Smarter Trading instead. This is the Baruch College graduate course he taught in spring 2002, written for students who had never placed an order, and it stays at that level.
- It quietly assumes a six-figure account and a friction-free broker. Trading Game 1 hands you $100,000. Game 2 raises stock positions to 5,000 shares and adds S&P and Treasury note futures, and Kaufman states plainly that there will be no commissions and no slippage. His portfolio arithmetic in chapter 17 produces answers like 552 shares at $100 plus 1,657 shares at $25 plus one crude oil contract at $3,000 margin. None of that scales down to a $5,000 account. Combine a stop at 3 times the daily range with volatility parity across four markets and the minimum viable capital is real money, which the book never says out loud.
- About a quarter of the book is pattern material he then undercuts. Chapters 9 through 13 cover one-day patterns, continuation patterns, tops and bottoms, retracements, Fibonacci, Gann, and volume, roughly 90 pages, with almost no testing. On retracements he writes that if you use all the popular levels the price is bound to stop at one of them, and asks how you could possibly trade that. He makes the same observation about Fibonacci time targets. He is right, and the honest edit would have been to cut the chapters rather than write them and then discount them. Chapter 18 on Dow Theory is openly a summary of a 1978 MTA Journal article and produces no tradeable rule.
- The risk methods are presented as arithmetic, with no evidence they beat the alternative. Volatility parity in chapter 17 is derived algebraically and never backtested against equal dollar weighting. The two moving average system in chapter 8 (long trend of 50 to 200 days, short trend of 5 to 30 days, buy when both point up, exit when the fast one turns down) is supported by exactly one hand-walked Amazon chart from October 1998 to April 1999, where the two-trend approach makes $62 against $78 open profit for the single slow trend, and Kaufman argues the smaller number is better on risk grounds without ever showing the risk figures. The stop-loss chapter also contradicts itself, stating that you never lower a stop and then lowering one two paragraphs later with the note that there is always an exception.
- The plumbing is 2003 and some of it is gone. The examples are Enron, AOL Time Warner, Tyco, and Amazon at $15, and the lead illustration of a clean breakout is Enron. The trading game sends you to BigCharts.com, MoneyCentral.msn.com, INO.com, Eurexchange.com, and Lind-Waldock for quotes, and quotes pit session hours for Treasury notes and the Euro. It tells you the trading game has no uptick rule as if that were a simplification, which is now just true. There is nothing on ETFs beyond a passing mention of Spyders, nothing on decimalization's effect on execution, and nothing on how algorithmic execution reshaped the intraday volume curve he describes.
Who it is for
Buy it if
You already have a signal you trust and keep losing money anyway. This is a book about the machinery around the signal: where the stop belongs, how many shares, how many markets, and how to stop one position from deciding your year. It is also the right first book for someone who has never placed a stop or a limit order, because chapters 5 and 8 explain order types and trade blotters better than most books bother to.
Skip it if
You want a system you can switch on. Kaufman states outright that there are no secrets in this course, and he means it, and the specific parameters in the book came from one stock in one five-year window. Skip it too if you came for the Kaufman adaptive moving average, because it is not here. If you already own Trading Systems and Methods, this is a thinner version of chapters you have.
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.