What technical analysis claims, and what the evidence supports

Technical analysis is the study of price and volume history in order to form expectations about future price. That is the whole claim, stripped of decoration. It does not look at earnings, interest rates, or the quality of a management team. It looks at what the market has already done and asks whether that record contains information about what the market will do next.

Most arguments about technical analysis fail because the two sides argue about different things. One side defends a specific ritual — a head-and-shoulders pattern, a Fibonacci retracement — and the other side attacks the general proposition that past prices can carry information. Those are separate questions with separate answers. The general proposition is defensible and has serious academic support behind parts of it. Many of the specific rituals are not. This lesson separates them, so that when you learn the tools in the rest of this course you know which drawer each one belongs in.

The three assumptions underneath every chart

Charles Dow, whose editorials in the 1890s became the ancestor of the field, and the later codifiers of "Dow Theory" rested the discipline on three assumptions. They are worth stating plainly because every indicator you will ever meet inherits them.

First, the market discounts everything. Price is a summary statistic of all participants' beliefs, constraints and information. If a hedge fund learns something and acts on it, the trace of that action appears in price and volume before it appears in a press release. So the chart is a lagging record of information, but it may lead the public disclosure of that information.

Second, prices move in trends. Once a move is underway it is more likely to continue than to reverse at any given moment. This is the load-bearing assumption. Without it, technical analysis reduces to pattern-matching noise.

Third, history repeats, because behaviour repeats. The same incentive structures — the leveraged holder who must sell, the underweight manager who must buy before quarter-end — recur, and so do their price signatures.

Assumption one is close to standard finance. Assumption three is a hypothesis about human institutions, plausible but hard to test. Assumption two is the one that has actually been measured, repeatedly, for decades.

What the evidence supports: trend and momentum

The strongest empirical support for anything in the technical family is momentum: assets that have outperformed over an intermediate horizon have tended, on average, to keep outperforming over the following months, and vice versa.

Jegadeesh and Titman documented this in cross-sectional form for US stocks in a 1993 Journal of Finance paper, ranking stocks on their prior three-to-twelve-month returns and finding that the winners continued to beat the losers over the next several months. The finding has since been reproduced across international equity markets, and Moskowitz, Ooi and Pedersen extended it in 2012 to "time series momentum" — an asset's own past return predicting its own future return — across dozens of futures markets in equities, bonds, currencies and commodities.

This matters because time series momentum is, formally, a moving-average crossover rule wearing an academic suit. A rule that says "hold the asset if its twelve-month return is positive, otherwise stand aside" is a trend-following rule. It has been tested on out-of-sample data across asset classes and long histories, and it has not vanished on contact with scrutiny in the way most chart patterns do.

The mechanisms proposed for it are also concrete rather than mystical. Underreaction: information diffuses slowly through a fragmented investor base, so price adjusts in steps rather than in one jump. Institutional flow: pension funds and index providers rebalance on calendars, and mandates force buying into strength. Risk transfer: in futures markets, hedgers pay speculators to absorb their directional risk, and that compensation shows up as a persistent drift. Herding and extrapolative expectations from retail participants add a further push.

Note the honesty required here. Momentum is a statistical tendency measured across hundreds of assets and decades. It says nothing about whether your particular chart will continue tomorrow. It also has a well-documented failure mode — momentum crashes, where sharp market reversals savage the trend-following portfolio in a matter of weeks. A tendency that pays in eight years out of ten and takes a violent loss in the ninth is a real edge and a real risk simultaneously.

What the evidence does not support: most of the pattern zoo

Now the uncomfortable half. Brock, Lakonishok and LeBaron published a widely cited 1992 study finding that simple moving-average and trading-range-breakout rules had predictive power on the Dow Jones Industrial Average over roughly a century of data. It was the most rigorous defence technical analysis had received.

Then Sullivan, Timmermann and White re-examined the same rules in 1999 using a bootstrap procedure designed to correct for data snooping — the statistical inflation you get when many rules are tried on one dataset and only the survivors are reported. If you test 10,000 rules on a single price history at a 5% significance threshold, roughly 500 will look "significant" from chance alone. Their conclusion was that the original results were far weaker once this was accounted for, and that the rules' apparent edge did not carry forward into subsequent data.

Park and Irwin's 2007 survey in the Journal of Economic Surveys reviewed the modern literature and found a split picture: rules tested in currency and futures markets showed more evidence of profitability than rules tested in equity markets, and evidence of profitability in developed equity markets weakened markedly from around the 1990s onward, as transaction costs fell, electronic execution spread, and quantitative funds began arbitraging the more obvious regularities.

Meanwhile Lo, Mamaysky and Wang, in a 2000 Journal of Finance paper, did the most sympathetic thing anyone has done for chart patterns: they defined head-and-shoulders, double-tops and similar formations algorithmically using kernel smoothing, then tested whether the return distributions conditional on those patterns differed from unconditional distributions. They found statistically detectable differences for several patterns in US stocks. Read carefully, though, this is a modest result — it says the patterns carry some information, not that trading them net of costs makes money.

The pattern here is consistent. Simple, mechanical, trend-based rules survive testing better than complex, discretionary, shape-based ones. The more degrees of freedom a method has, the more likely its apparent success is an artefact of the fitting.

Why the false positives are so convincing

Three cognitive mechanisms manufacture confidence out of nothing, and understanding them is more useful than any indicator.

Survivorship in your own memory. You remember the head-and-shoulders that preceded a 20% decline. You do not remember the eleven that resolved into nothing, because at the time they resolved into nothing you were not looking. There is no ledger, so there is no correction.

Ambiguity of definition. A "support level" is not falsifiable unless you define in advance how close a touch counts and how deep a break counts. Without that pre-commitment, every chart supports the theory after the fact.

Self-fulfilment is real but bounded. If enough participants place stop orders just below an obvious round number, a move through it triggers cascading sell orders and the level does "work" — for minutes. That is a genuine microstructure effect, not superstition. But it operates on the timescale of the resting orders, not on the timescale of a weekly chart, and it dissipates as soon as the queued orders are consumed.

The Swiss dimension: who is selling you the method, and in what status you trade it

An edge is a gross number. Two Swiss facts sit between that number and what you keep: the rules that govern anyone selling you a method, and the tax status under which you trade it. Costs are the third — spread, commission and, for securities transferred through a Swiss dealer, the federal transfer stamp duty — and the relative strength lesson works that arithmetic in full.

Start with the seller. Chart-based signal services, model portfolios and trading courses are marketed in Switzerland as freely as anywhere else, but the surrounding activities are regulated and it is worth knowing which duties actually attach. Under the Financial Services Act (FinSA/FIDLEG), a provider of financial services must classify clients as retail, professional or institutional, and the retail protections — an appropriateness or suitability assessment, disclosure of risks and costs, a record of what was recommended and on what basis — attach to services such as investment advice, portfolio management and execution. They do not attach to general market commentary, which is precisely how most chart-based content is published. Under the Financial Institutions Act (FinIA/FINIG), an independent portfolio manager requires a FINMA licence and ongoing oversight by a supervisory organisation; client advisers acting for providers that are not prudentially supervised, including foreign ones, must be entered in an advisers' register before serving Swiss clients. The practical test is one question: is the person showing you the equity curve licensed and supervised for what they are actually offering, or publishing commentary that carries no duty toward you at all? The evidence standard in this lesson is unchanged either way. Your recourse is not.

Then there is tax status. Private individuals resident in Switzerland do not pay income tax on capital gains from private wealth — a genuine advantage over most of Europe. But that treatment is not unconditional. The Federal Tax Administration distinguishes private investors from professional securities dealers (gewerbsmässiger Wertschriftenhandel), and its circular guidance sets out safe-harbour criteria covering, among other things, how long positions are held, how large turnover is relative to the portfolio, whether borrowed money finances the positions, and whether derivatives are used for more than hedging. Fail the safe harbour and you may be reclassified, at which point trading gains become taxable income and social-security contributions can apply. Losses become deductible too, which is small consolation. A high-frequency chart-based strategy is precisely the profile that attracts this question, so the tax consequence should be modelled before the strategy is scaled, not after. This is general information about how the system is structured, not advice about your situation; the assessment is cantonal and fact-specific, and a Swiss tax adviser is the right person to answer it.

How to hold the discipline honestly

The defensible position is this. Price history contains information. The information is weak, it is statistical, and it is concentrated in simple persistence effects rather than in ornate geometry. Technical tools are best understood as a language for describing market state — trending or ranging, expanding or contracting volatility, orderly or disorderly participation — and as a framework for defining risk, which is where they earn most of their keep. A chart tells you where a position stops being right long before it tells you where it will be profitable.

Everything in the remainder of this course should be judged against one test: could this rule be written down in advance, applied mechanically to data it was not designed on, and still show an edge after realistic costs? If yes, it is worth your time. If it requires your judgement to identify the pattern after the fact, you are looking at a story, not a signal.

What to take away

• The general claim that past prices carry information is supported; momentum and time-series trend effects have been documented across markets and decades by Jegadeesh and Titman, and by Moskowitz, Ooi and Pedersen, with plausible mechanisms in slow information diffusion, mandated institutional flows and risk transfer from hedgers.
• Data snooping is the central methodological danger: Sullivan, Timmermann and White showed that the classic Brock–Lakonishok–LeBaron moving-average results shrink substantially once you correct for the number of rules searched, so treat any backtest whose rule was chosen by looking at the same data as unproven.
• Simple rules with few parameters survive out-of-sample testing far better than discretionary chart patterns, because every additional parameter is another opportunity to fit noise.
• The FinSA conduct duties — client classification, suitability, cost and risk disclosure — attach to advice and portfolio management, not to published market commentary, so establish what a signal seller is actually licensed and supervised to provide before you weigh their track record.
• Frequent, leveraged, derivative-heavy trading can move you from tax-free private capital gains into professional securities dealer status under the Federal Tax Administration's criteria, which changes the after-tax arithmetic of the whole strategy; check the position with a Swiss tax adviser before scaling.

Sources

• Narasimhan Jegadeesh and Sheridan Titman, "Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency", Journal of Finance, 1993.
• Tobias J. Moskowitz, Yao Hua Ooi and Lasse Heje Pedersen, "Time Series Momentum", Journal of Financial Economics, 2012.
• William Brock, Josef Lakonishok and Blake LeBaron, "Simple Technical Trading Rules and the Stochastic Properties of Stock Returns", Journal of Finance, 1992.
• Ryan Sullivan, Allan Timmermann and Halbert White, "Data-Snooping, Technical Trading Rule Performance, and the Bootstrap", Journal of Finance, 1999.
• Andrew W. Lo, Harry Mamaysky and Jiang Wang, "Foundations of Technical Analysis: Computational Algorithms, Statistical Inference, and Empirical Implementation", Journal of Finance, 2000.
• Cheol-Ho Park and Scott H. Irwin, "What Do We Know About the Profitability of Technical Analysis?", Journal of Economic Surveys, 2007.
• Burton G. Malkiel, A Random Walk Down Wall Street — the standard statement of the sceptical case.
• Swiss Financial Services Act (FinSA/FIDLEG) and Financial Institutions Act (FinIA/FINIG), and FINMA's published information on licensing, supervisory organisations and the client adviser register.
• Swiss Federal Tax Administration (Eidgenössische Steuerverwaltung), guidance on the distinction between private investors and professional securities dealers (gewerbsmässiger Wertschriftenhandel).