Technical Analysis
The Art and Science of Technical Analysis
Grimes tested the things other chart books assert, and reported the ones that failed. He hid the price bars on a chart, drew lines at random, put the bars back, and the fake levels held, broke, and got retested exactly like the real ones. Everything else in the book is built to survive that finding.
The one idea
Markets are random most of the time. That is the starting position, not a hedge. Prices sit in equilibrium, buyers and sellers roughly agree, and the wiggles mean nothing. In that state no pattern, indicator, or level can make money, because there is nothing there to be right about. The trader's entire job is to sit out that condition and act only in the small windows where it breaks.
What breaks it is an imbalance of buying and selling pressure. Someone with size has to move a position and cannot do it all at once, so they buy, wait for the market to absorb it, and buy again. That leaves a footprint: a push, a pause, another push. Grimes says every edge a technical trader has traces back to that, and nothing else. You are not trading the flag or the triangle. You are trading the order flow that happened to draw one.
The consequence is a standard the rest of the field ducks. If a pattern is real, it has to beat a random baseline, and Grimes shows you how easily it fails to. His random-lines exercise is the cleanest demolition of chart-reading in print, and he wrote it into his own book. He also throws out volume analysis (he could not substantiate the claims), throws out the idea that a level gets stronger each time it is tested (it gets weaker), and throws out Bollinger's textbook containment numbers. What is left is smaller than what most chart books sell, which is the point.
What it actually teaches
The book is a sequence: qualify the setting, read the structure, name the trade, then let the risk math size it and score it.
TEST THE LEVEL AGAINST A RANDOM LINE
Before trading any support or resistance, prove it is not noise. Hide the price bars in your charting package, draw lines anywhere, restore the bars, then study every touch. The random lines hold, break, and hold again on the retest, exactly like the levels you drew on purpose. Grimes posted video of himself doing this in December 2011. What survives his filter is short: levels obvious to every participant, meaning multiply tested highs and lows, the extremes of large spikes, and the previous day's high and low. And note the reversal of conventional wisdom: a level tested three or more times is more likely to break, not less.
READ THE THREE-LEG PATTERN
Every trend on every time frame is impulse, retracement, impulse, labeled A-B-C-D. The retracement usually ends near 50 percent of the setup leg, with a working range of 25 to 75 percent and no surprise outside it. The next leg (C to D) tends to run about as far as the first (A to B). He calls that the measured move objective and treats it as a zone, not a price. Complex pullbacks, two countertrend legs instead of one, are common in mature trends and are exactly what stops out traders who only planned for the simple version.
SET UP TWO INDICATORS AND STOP THERE
Keltner channels at 2.25 times average true range around a 20-period exponential moving average, plus a modified MACD: 3-period simple average minus 10-period simple average, with a signal line that is a 16-period average of that, no histogram. The channel marks emotional extremes, not barriers. A new momentum extreme on the MACD says the next pullback is worth buying. A new price high without a matching MACD high says it is not. He also checked the standard textbook claim that 96 percent of closes sit inside two-standard-deviation Bollinger bands and found it is closer to 88 percent.
NAME WHICH OF FOUR TRADES YOU ARE MAKING
Trend continuation, trend termination, support or resistance holding, support or resistance breaking. Every technical trade fits one of the four. The label sets your expectations before you enter. Continuation trades win often for modest payoffs. Termination trades lose often and pay large when they work, and a win means the trend merely stopped, not reversed. Breakouts mostly fail, and the reward-to-risk has to carry them.
PICK FROM EIGHT TEMPLATES, NOT A LIBRARY
Three primary patterns: the pullback, the failure test (a probe past a level that reverses on the same or next bar, which is Wyckoff's spring and upthrust and Sperandeo's 2B), and the breakout. Five derived: the Anti (first pullback after a trend break), buying support inside a pullback, paying a lower-time-frame breakout inside a pullback, entering the base before a breakout, and trading the failed breakout. Notice what is deliberately missing: no trading inside ranges, no plain buying at support, no fading a market for being extended. He says those were not profitable for him over a very large sample.
PLACE THE STOP BEFORE YOU PLACE THE ENTRY
The one rule he says cannot be broken. Stops go where the trade is proven wrong, not at a distance that feels comfortable. Rarely closer than two average true ranges, sometimes more than four. Never tighter than one average bar's range, or you are trading inside the noise. Add a few ticks of random jitter so your stop is not sitting on the same price as everyone else's. Fixed percentage stops fail because volatility varies: as of April 2011, across 500 active stocks the average daily range was a bit under 3 percent of price, but plenty ran 9 percent or more, so a single 10 percent stop means two completely different things.
SIZE OFF THE STOP, NOT OFF CONVICTION
Trade size equals the dollars you will risk divided by the per-share distance to the stop. Risking 1 percent of a $100,000 account on a $50 stock with a $47.50 stop gives 400 shares. Under 1 percent per trade is conservative, 3 percent or more is extremely aggressive. His Monte Carlo makes the case: 1,000 traders, 250 trades each, running a genuine edge (win 1.2 times risk, lose 1.0 times risk, 50/50 odds). Risking 2 percent, nobody went broke, but one path fell from $100,000 to $12,400, an 87.5 percent drawdown on a winning system. At 10 percent per trade, 17.6 percent of accounts went to zero. At 25 percent, 47.7 percent did.
SCORE EVERY TRADE IN R, THEN RUN A T-TEST
Tag each trade with the setup that produced it, then convert the result to a multiple of what you risked. His worked example: 68 trades, $3,158 net on $100,000, p equals 0.263, meaning you would see that by luck alone about a quarter of the time. Convert the same trades to R multiples and they average 0.3R with p equals 0.000, a real edge that his own inconsistent sizing was hiding. Rerun at a flat 1 percent risk and those trades make $22,935 instead of $3,158. One system in the set had an 81.3 percent win rate and almost no edge, because the average loss was three times the average win.
What it looks like Monday morning
You go back through your own trade log and retag every trade by the setup that produced it, then convert each result to a multiple of what you risked instead of leaving it in dollars. That one change tells you whether your profits come from one pattern while another quietly gives them back, and it separates whatever edge you have from your position sizing decisions. Most traders find what his example trader found: the dollar column was lying, in one direction or the other.
Then you go to the charts and try to break your own levels. Hide the bars, draw lines at random, restore the bars, and study how those lines behave. If the levels you actually trade do not clearly beat the fake ones, you stop trading them. What tends to survive is a short list: previous day highs and lows, the extremes of large spikes, and prices so obvious that every participant is staring at the same number.
| Test group | Range inside the bands | Bars fully outside |
|---|---|---|
| Large-cap stocks (496) | 87.7% | 3.4% |
| Small-cap stocks (500) | 85.4% | 4.3% |
| Futures (16 contracts) | 85.9% | 3.8% |
| Forex (9 pairs) | 89.8% | 2.3% |
| Randomly generated data (7 series) | 86.8% | 3.8% |
Random data lands in the middle of the real markets, which is worth sitting with before you treat a band touch as a signal.
Almost any random line drawn in the market will function as believable support or resistance.Adam Grimes, The Art and Science of Technical Analysis
Where it fails
- The eight setups are never tested. Chapter 6 calls them concrete expressions of quantifiable, directional edges. Then it gives no win rate, no expectancy, no sample size, and no test period for any of the eight. The single real test table in the book measures how often price stays inside a Keltner channel, which is not a claim about money. Grimes raises the bar for everyone else with the random-lines demonstration and does not clear it himself. He tells you in Chapter 9 that if you do not know what your edge is, you do not have one, and then hands you eight patterns whose edges he has not shown.
- His own headline table cannot tell real markets from noise. Table 7.1 runs 2,403,774 bars across 496 large-cap stocks, 500 small-caps, 16 futures contracts, 9 forex pairs, and 7 randomly generated series. The randomly generated data comes in at 86.8 percent of range inside the bands, sitting comfortably in the middle of the real markets, which run 85.4 to 89.8 percent. He presents this as confirmation the channel is set correctly. Read it the other way and it says the band alone tells you nothing about whether a market is behaving nonrandomly.
- Some load-bearing rules are too wide to be wrong. The retracement guideline is 50 percent, plus or minus a very large margin of error, with a working range of 25 to 75 percent and a note not to be surprised outside it. A significant new momentum reading means significant relative to recent history, defined nowhere. Climax means a swing two to three times the average swing, sometimes. He is right to refuse false precision, and that honesty is why the book is good. But a discretionary trader cannot test what a rule like that would have done, and neither can you.
- It quietly requires more than most readers have. Three to five years to reach the stage where trading pays you. Two years to reach the stage where it merely covers commissions. Risk capital above $25,000 that you have already accepted you will lose. Screen time during the session, and a scanning habit he describes as 400 to 500 charts a day. Grimes puts the odds of substantial percentage gains at fewer than one in a thousand, and possibly much worse. He is being honest. He is also describing a reader who does not have a day job.
- The data stops in 2011 and there has never been a second edition. The copyright page reads 2012, the preface is dated September 2011, and the charts run through mid-2011. So nothing in the book has been revisited against the 2012 to 2019 low-volatility grind, zero-commission retail brokerage, the 2020 to 2021 retail options wave, 0DTE, or crypto. His pitch for the three-day to two-week swing (step up out of the noise the HFTs create) was written when high-frequency trading was the topical threat. Every number cited here, from the 2.25 ATR channel to the 88 percent Bollinger figure to the 3 percent average daily range across 500 stocks, comes from that one edition.
- The trade examples are curated, and some were never traded. Chapter 10 is better than the usual parade of winners, because it spends real space on failures and on how patterns break. But Grimes states outright that pedagogical concerns won and he included patterns he did not actually trade, alongside ones from his Waverly Advisors report. These are teaching illustrations chosen after the fact, not a track record. The book contains no equity curve of his own and no audited results.
- It is roughly twice as long as its argument. The book itself says the essential elements of market structure and trade management could fit on a single piece of paper. It then runs 480 pages. Appendix A explains bid-ask spreads and order types to an audience of self-directed traders who are already placing trades. Chapter 11 on psychology covers cognitive biases, flow, and intuition at a level any general book does better. The material that justifies the book is Chapters 3 through 9 and 12.
Who it is for
Buy it if
You already read charts fluently and you are stuck: your patterns work sometimes and you cannot say why. This is the only book on the shelf that gives you a way to separate a real edge from a good story, plus the risk math to stay solvent while you find out which you have. Best for a swing trader on daily bars with a funded account and a trade log worth analyzing.
Skip it if
You want setups you can run Monday with evidence that they pay. The book never publishes a win rate or an expectancy for a single one of its eight patterns, and it will send you off to build and test your own. Skip it also if you are trading a small account you need the income from, because Grimes tells you plainly that will not work, and paying 480 pages to hear it is a poor trade.
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.