The three principles of technical analysis, and their limits
The whole discipline rests on three premises, set out in this form by John Murphy in Technical Analysis of the Financial Markets, one of the profession's standard textbooks. Each holds some truth and a limit worth knowing before you draw your first line.
Price discounts everything. Earnings, rates, rumours, fear: whatever is known is supposed to be in the price already, which is the licence to study price alone. The limit sits in the word “known”. An announcement that has not happened yet is in no chart, and it is the one that produces the most violent moves.
Prices move in trends. The premise is that a move under way is more likely to continue than to reverse. The limit: a market also spends long stretches with no trend at all, and every trend-following method racks up false signals there.
History repeats itself. The same setups come back because the same human reactions come back, the fear of missing out as much as the fear of losing. The limit: a setup everybody knows ends up being anticipated, and so it wears out. The opening lesson explains why a pattern works, and why it stops working.
Technical analysis vs fundamental analysis
Fundamental analysis asks what an asset is worth: a company's earnings, a central bank's policy, supply and demand for a commodity. Technical analysis asks when to act and at what price, by reading the outcome of all those causes in the price itself. One studies the causes, the other their effect.
The two are not rivals, they work on different horizons. The standard survey of the foreign exchange market, published by Lukas Menkhoff and Mark Taylor in the Journal of Economic Literature in 2007, finds that almost all professionals use technical analysis, and that its weight in the decision rises as the horizon shortens. The faster you decide, the more the chart counts. The further out you aim, the more fundamentals take over.
Timeframe, trend and structure
Two choices come before any tool. Timeframe first: the same session is a collapse on the five-minute chart and a mere pause on the daily, so a level or a signal only exists on the timeframe it was read on. Pick a decision timeframe, judge your positions on it, and step up one notch only to place the context, before you enter.
Trend next, which has needed no indicator since Charles Dow: an uptrend is a run of higher highs and higher lows, a downtrend the reverse, and everything else is a range. It lasts until that run is broken. The lesson on market structure covers that breaking point in detail. It is the first reading to master, because the same pattern does not carry the same weight with the trend as against it.
Support and resistance: zones, not lines
Support is a price zone where a fall has already stopped, resistance a zone where a rise has already stalled. There is nothing magical about either: they are prices a lot of people saw and remembered, and therefore prices where orders gather. Two rules prevent most mistakes. You draw a band, never a line to the penny, because nobody places orders at exactly the same price. And a level that breaks changes role: broken resistance often turns into support, which is known as polarity.
The trap is drawing too many, since a screen covered in lines agrees with any scenario. The price action lesson gives three criteria for qualifying a support or resistance zone, and teaches you to keep only three on screen. Some levels are not drawn but calculated: the lesson on pivot points gives the formulas, and says why two platforms do not show the same ones.
Candlesticks and chart patterns: shape and location
A candlestick sums up a period in four prices: open, high, low, close. A chart pattern is a shape drawn by several candles, and textbooks sort them into two groups: reversal patterns, such as the double bottom or the head and shoulders, and continuation patterns, such as the triangle or the flag.
Two precautions apply to all of them. A pattern is confirmed only by the close that breaks its threshold, not while it is still forming. And location matters more than shape: the same hammer does not carry the same weight on a support everybody watches as in the middle of nowhere. The lesson on candlestick charts sets out both rules, and the course on chart patterns takes each pattern in turn with its exact threshold and its measured performance.
Technical indicators, family by family
An indicator is a calculation applied to price, sometimes to volume. It therefore sees nothing the chart does not already contain, it simply shows it differently. There are hundreds of them, and they fall into four families according to the question they ask.
Trend: which way is the market going, and since when? These are moving averages, the EMA clouds that shade the gap between two of them, Gerald Appel's MACD, and Welles Wilder's ADX and Parabolic SAR. They smooth price, so they react late, and that lag is the cost of filtering out noise.
Momentum: is the move speeding up or running out of steam? The best known is the RSI, published by Welles Wilder in 1978, which swings between 0 and 100 with the usual thresholds at 70 and 30. The stochastic oscillator asks the same question. Their shared flaw: in a strong trend they can sit in overbought territory for a very long time while nothing reverses. The barest of the family is momentum, or the Rate of Change (ROC): the close compared with the close n candles earlier, with no bounds and no fixed threshold. John Ehlers' Relative Vigor Index looks inside the candle: the close against the open, relative to the range, smoothed and centred on zero.
Volatility: how far does price usually move? Bollinger Bands plot, by default, two standard deviations either side of a twenty-period average. The ATR measures the average range of recent candles. Neither says anything about direction: they are for placing a stop at a sensible distance or spotting a squeeze.
Volume: how many people are behind the move? OBV, Chaikin Money Flow, VWAP with its standard deviation bands, and the volume profile answer that, provided the volume on your screen is real. Reading volume candle by candle, with no indicator, is the subject of the lesson on the Wyckoff method and VSA.
The rule of use follows from the classification: one tool per question, no more. Three momentum oscillators flashing together are not three confirmations, they are the same calculation run three times on the same price.
Volume: the only data that does not come from price
Everything above is computed from prices. Volume adds a separate piece of information, the number of shares or contracts traded during the period. A breakout on heavier volume than usual drew a crowd. The same breakout in an empty market committed almost nobody, and nothing holds it up if price comes back.
That number still has to be real. On stocks and futures, trades are reported and the published volume is that of actual transactions. Spot forex has no central market: your platform shows tick volume, meaning the number of price changes received from your broker, not the amounts traded. In crypto, each exchange publishes only its own trades. A volume indicator is never worth more than the data it is fed.
How to start with technical analysis, step by step
1. One market, one timeframe. Take a single liquid asset and a single decision timeframe, and stay with them for several weeks. Switching market after every disappointment stops you from seeing anything repeat.
2. A naked chart. Start with no indicator: mark the trend by its highs and lows, then three support or resistance zones you can each justify in one sentence.
3. One indicator per question, at most. Add one only if it answers a question the naked chart leaves open, and keep the default settings: they are the ones everybody else is watching.
4. The scenario in writing, beforehand. What you expect, the level that proves you wrong, the deadline. Without those three, any reading ends up having been right.
5. Measure before money. Replay your rule over history, count the cases where it fails as carefully as the ones where it works, and risk nothing until that record exists. In Tradoshi, backtesting is included from the Pro plan, with ten strategies, and unlimited on Premium; replay covers forex, gold, silver, six indices, oil, bitcoin, ether and litecoin.
6. Risk on its own track. Position size comes out of the calculation in the risk lesson, never out of how confident the chart makes you feel.
What the research says
Whether it works has been tested seriously, and the answer comes in three stages. In 1992, William Brock, Josef Lakonishok and Blake LeBaron ran 26 simple rules, moving averages and trading-range breaks, on the Dow Jones from 1897 to 1986: in the Journal of Finance, they show that buy signals are followed by higher returns than sell signals, which a random walk does not explain.
In 1999, Ryan Sullivan, Allan Timmermann and Halbert White raised the objection that matters: try enough rules on the same data and one of them will work by chance. They widened the test to 7,846 rules over a hundred years of the Dow Jones and corrected for that data-snooping bias. Their finding: the best rule holds up over the historical period, but its performance is not repeated over the next ten years, 1987 to 1996.
The most cited review, by Cheol-Ho Park and Scott Irwin in the Journal of Economic Surveys in 2007, counts 95 modern studies: 56 report positive results, 20 negative, 19 mixed. The authors themselves warn that most suffer from data snooping, rules picked after the fact and poorly estimated transaction costs. And when Pierre Bajgrowicz and Olivier Scaillet went back over the Dow Jones from 1897 to 2011 with those costs included, the edge disappeared.
What you can take from this without misquoting anyone: regularities did exist, they are small, they fade as they become known, and no rule read in a book spares you from measuring it on your market, your timeframe and your costs. The biases that distort that measurement, hindsight first among them, are the subject of the lesson on what a chart can tell you and what it never will.
Frequently asked questions
Is technical analysis a self-fulfilling prophecy?
Partly, and that is its mechanism rather than its flaw. A level triggers a reaction because many people are watching it and have placed their orders there, not because the line holds any power. The practical consequence: the most widely followed landmarks are the most useful ones, and an exotic setting nobody else uses coordinates nobody.
Does technical analysis work on forex, stocks and crypto?
Its tools apply to any market liquid enough to produce a continuous price history, and they are used on all three. What changes is the data. On a market that trades around the clock, such as forex during the week or bitcoin, the daily candle has no official close: it depends on the cut-off hour your platform chose, and two brokers can show two different candles for the same day. And on a thinly traded instrument, a level watched by ten people coordinates nothing.
What is the difference between technical analysis and price action?
Price action is a branch of technical analysis: the one that reads price alone, candles, levels and structure, with no computed indicator. It does not oppose it, it strips one layer off. Since nearly every indicator is derived from price, many traders start with that bare reading and add an indicator only for a specific question.
Which book should I read to learn technical analysis?
Three references cover the essentials. John Murphy's Technical Analysis of the Financial Markets (1999) for the full overview. Technical Analysis of Stock Trends by Robert Edwards and John Magee (1948) for chart patterns, whose catalogue it fixed. Thomas Bulkowski's Encyclopedia of Chart Patterns for pattern-by-pattern statistics. None of them replaces your own record: a book tells you what a pattern did elsewhere, not what it does on your market.
Is reading a chart well enough to be profitable?
No. A reading gives you a scenario, not a result. An account is decided as much by position size, costs and discipline in execution as by the reading itself: a method that is right six times out of ten loses money if its losses are twice the size of its wins. That is why a reading is judged over a series of recorded trades, never on the last one.