Technical Analysis: Reading Charts to Predict Prices, and Why It Usually Doesn't Work
A chart pattern feels like it is telling you something a company's balance sheet never could: exactly when to act. That feeling is the entire appeal of technical analysis, and it is also its central vulnerability, because the human eye is extraordinarily good at finding patterns in random noise whether or not those patterns predict anything. Understanding what technical analysis actually is, and what the evidence says about it, matters as much for what it teaches about your own instincts as for the strategy itself.
The core principle
Technical analysis is the practice of forecasting future price movement from historical price and volume data, chart patterns, and momentum indicators, rather than from a company's earnings, balance sheet, or competitive position, which is the domain of fundamental analysis instead. Practitioners study shapes on a price chart, such as head-and-shoulders formations, support and resistance levels, and moving average crossovers, on the premise that price action encodes the collective behavior of every market participant and that certain patterns of that behavior tend to repeat.
The theoretical objection to technical analysis comes from the efficient market hypothesis, which holds that if a pattern in historical prices reliably predicted future prices, enough traders would exploit it that the exploitation itself would erase the pattern's edge. In its weaker forms this does not rule out short-lived, small edges; it does suggest that any edge robust and large enough to survive real-world trading costs, taxes, and competition from well-capitalized quantitative funds is rare, and that most patterns visible to a retail chartist on a screen have already been arbitraged down to something close to zero, or were never real to begin with.
None of this means price data is worthless. Momentum, the tendency of recent winners to keep outperforming over intermediate horizons of several months to about a year, is one of the most replicated patterns in the academic finance literature and does appear to be a genuine, if inconsistent and crash-prone, phenomenon. The distinction worth holding onto is between disciplined, statistically tested, systematically applied price-based strategies, which a minority of quantitative funds run with real risk controls, and the retail practice of eyeballing a chart and declaring that a pattern is forming, which is closer to reading shapes in clouds.
How the math works
Example 1: transaction costs erase small edges quickly. Suppose a chart-based trading system generates 100 round-trip trades in a year, and each round trip carries a combined cost of 0.15% from the bid-ask spread and any slippage on execution, even with a commission-free broker. That cost alone drags annual return down by 100 x 0.15% = 15 percentage points. For context, the long-run average annual nominal return of the U.S. stock market has historically been in the neighborhood of 10%. A strategy trading that frequently needs a raw, pre-cost edge larger than the market's entire average annual return just to break even after costs, before even accounting for the extra tax drag of short-term gains being taxed at ordinary income rates rather than the lower long-term capital gains rate.
Example 2: a "68% win rate" can be statistical noise, not skill. A trader backtests a chart pattern over 50 historical trade signals and finds it was profitable 68% of the time, which sounds compelling. But with a sample of only 50 trades, the standard error of that observed win rate, using the formula for the standard error of a proportion, SE = sqrt(p x (1-p) / n), works out to sqrt(0.5 x 0.5 / 50) ≈ 7.1%. A 68% observed win rate sitting roughly 2.5 standard errors above a coin-flip baseline of 50% looks statistically meaningful at first glance, but 50 trades is a small sample by the standards of rigorous strategy testing, and patterns discovered by testing many variations of a strategy against the same historical data, a practice called overfitting, routinely produce results this strong purely by chance, only to fail once traded on new, out-of-sample data.
How it shows up in real portfolios
The most common retail application is a self-directed trader using indicators like the relative strength index or moving average crossovers to time entries and exits in individual stocks, often trading frequently enough that transaction costs and short-term capital gains taxes eat heavily into any raw edge, exactly as the math above illustrates. Academic studies tracking real retail day traders using brokerage account data have consistently found that the large majority lose money net of costs over multi-year periods, with any persistent skill confined to a small minority of accounts.
At the institutional end, systematic trend-following and momentum-based quantitative funds, sometimes called managed futures or CTAs, do run technical, price-based strategies professionally, but with meaningful differences from retail chart reading: rigorous statistical backtesting across decades and many markets, strict position sizing and stop-loss discipline, diversification across dozens of uncorrelated markets simultaneously, and institutional-scale trading costs far below what a retail investor pays. Their live, audited track records show these strategies can work as a small, uncorrelated sleeve in a diversified portfolio, though performance is famously inconsistent and prone to multi-year stretches of underperformance.
A high-earning professional with concentrated employer stock, timing the sale of vested shares around a perceived chart pattern rather than a plan tied to tax and diversification goals, is a common and costly version of this mistake. Deciding to hold shares an extra month because a chart "looks like it wants to break out," while ignoring that the position represents both a large piece of net worth and the same paycheck-generating employer, layers speculative timing risk on top of an already concentrated position, for a benefit the evidence suggests is unlikely to be real.
Published track records of successful technical trading strategies also suffer from a well-documented statistical distortion known as survivorship bias: the strategies and traders whose results get publicized, written about, or sold as courses are disproportionately the ones who happened to succeed, while the far larger number who tried similar approaches and failed simply disappear from view without ever publishing anything. A newsletter or course showcasing an impressive multi-year track record of chart-based calls says very little, on its own, about the base rate of success among everyone who attempted the same approach, since only the winners tend to stick around long enough, and feel motivated enough, to advertise their results.
Actionable breakdown
- Questions to ask before trusting any chart-based signal:
- Was it tested out of sample, on data not used to design it?
- Does the edge survive realistic trading costs and taxes?
- How many trades support the claimed win rate?
- Lower-risk ways to engage with price-based investing:
- A small, diversified sleeve in a systematic trend fund.
- Simple rules like rebalancing bands, not chart reading.
- Momentum exposure via a low-cost factor fund, not timing.
- Red flags in any technical trading pitch:
- A backtest with no mention of transaction costs.
- A short track record shown without a drawdown history.
- Claims of a pattern that "always" works.
Common pitfalls
- Confirmation bias: remembering the trades where the pattern worked and mentally discounting the ones where it failed, which inflates a trader's sense of the strategy's real win rate over time.
- Overfitting a backtest by testing dozens of pattern variations against the same historical data until one looks good, then treating that one as a discovered edge rather than a statistical accident.
- Ignoring transaction costs and short-term capital gains taxes when evaluating a strategy's historical returns, both of which fall disproportionately on frequent, chart-driven trading.
- Mistaking a compelling narrative, such as a well-drawn trendline or a named pattern with an authoritative-sounding label, for actual predictive evidence.
Related concepts
The theoretical challenge to technical analysis is the efficient market hypothesis. The one price-based pattern with real academic support is momentum, while reversion to the mean describes the opposing tendency of extreme results to normalize. For the discipline of testing a strategy properly, see backtesting, and for the retail practice most closely tied to chart-based trading, see day trading. For a broader framework, see the guide on stock analysis.
The bottom line
Chart patterns can feel like insight, but the rigorous evidence supporting most of them is thin once trading costs, taxes, and the human talent for seeing patterns in noise are honestly accounted for.