Does Technical Analysis Work? What the Evidence and Markets Say

Not pseudoscience. Not the holy grail. Something more complicated — and more useful — than either camp admits.

The Wrong Question

The debate around technical analysis has always been framed badly. On one side: traders who swear by their indicators, draw trendlines with religious conviction, and post screenshots of every winning trade. On the other: quants and academics who dismiss the entire discipline as astrology with better fonts.

Both sides are arguing about the wrong thing.

The question is not whether technical analysis works. The question is: under what conditions, in which markets, and for how long? That reframing changes everything. It moves the conversation from belief to evidence — and from a binary answer to a conditional one.

My experience running systematic strategies across US equities and crypto futures has led me to a simple conclusion: some indicators carry a genuine edge in specific market conditions, none of them work all the time, and the traders who do well with TA are the ones who understand the difference.

What Technical Analysis Actually Is

Before evaluating whether TA works, it’s worth being precise about what it actually does — because most of the confusion comes from conflating two very different uses of the same tools.

Descriptive TA — using price and volume data to describe what has happened and what is currently happening in a market. Where is price relative to recent highs? Is volume confirming or contradicting the move? Are buyers or sellers in control? This use of TA is uncontroversial. It’s just reading the tape.

Predictive TA — using the same tools to forecast what will happen next. This is where most of the debate lives, and where most of the failure occurs. The assumption is that past price patterns reliably predict future price behavior. The evidence for this, as a general claim, is weak.

The distinction matters because a lot of what traders do with TA is actually descriptive — they’re using it to understand current market structure, not to predict the future — and then mistaking that for predictive power. An indicator that tells you “the market is currently trending” is useful. The same indicator telling you “the market will continue trending” is a different — and much stronger — claim.

What the Academic Evidence Actually Says

The academic literature on technical analysis is larger and more nuanced than most traders realize. The honest summary: the evidence is mixed, condition-dependent, and has shifted significantly as markets have evolved.

The early efficient market hypothesis literature, starting with Fama’s 1970 paper, argued that prices already reflect all available information — making technical analysis theoretically useless. If price fully reflects past data, studying past data cannot give you an edge. This view dominated academic finance for decades.

Then the anomaly literature arrived. Studies in the 1990s and 2000s documented systematic patterns that shouldn’t exist if markets were fully efficient: momentum effects, mean reversion at specific timeframes, and the persistence of certain support and resistance levels. Jegadeesh and Titman’s 1993 paper on momentum strategies found that stocks that performed well over the past 3-12 months tended to continue outperforming — a result replicated across dozens of markets and timeframes.

More directly relevant to TA, Brock, Lakonishok, and LeBaron (1992) tested simple moving average crossover rules and found statistically significant returns on Dow Jones data from 1897 to 1986. Lo, Mamaysky, and Wang (2000) found that several classic chart patterns — head and shoulders, double tops and bottoms — had statistically significant predictive power in US equity data.

The problem: many of these edges have eroded or disappeared as they became widely known and as algorithmic trading increased market efficiency. A 2004 paper by Ready found that the moving average results in Brock et al. largely disappeared in out-of-sample data from 1987 to 1996. The market learned.

Why Indicators Stop Working

This is the part that most TA education skips entirely — and it’s the most important part.

Markets exist in regimes. A regime is a persistent state defined by a specific combination of volatility, trend, and correlation structure. A trending, low-volatility regime behaves fundamentally differently from a ranging, high-volatility regime. The indicators that work well in one will fail in the other — not because the indicator is wrong, but because it was designed for different conditions.

Consider a simple moving average crossover. In a sustained trend, it works well: it keeps you on the right side of the move and exits you when momentum fades. In a ranging market, it generates constant false signals — price oscillates around the moving average, triggering entries and exits that go nowhere. The indicator hasn’t changed. The market has.

This is the concept of non-stationarity — the statistical properties of market data change over time. A strategy that has positive expected value in one regime may have negative expected value in another. The tragedy is that most traders backtest over long historical periods that contain multiple regimes, see a positive overall result, and assume the strategy will work going forward — not realizing they’re averaging across conditions that may never coexist the same way again.

There’s a second mechanism that kills indicators: alpha decay through crowding. When an indicator or pattern becomes widely known and widely traded, the edge it once contained gets arbitraged away. The strategy works until everyone is doing it — at which point the market adjusts to price in the pattern before it fully develops. This is precisely what happened to many of the moving average strategies documented in 1990s academic papers.

What Has a Real Edge — and Why

Not all TA is equally supported by evidence. Some tools have structural reasons to work that go beyond pattern recognition. Here’s what I consider genuinely useful and why:

Support and resistance — the most durable concept in TA, and the one with the clearest structural explanation. Support and resistance levels work because they are points where a large number of market participants have positions, stop losses, or limit orders clustered. This creates genuine liquidity at those levels — not a prediction, but a measurable reality. The self-fulfilling prophecy element is real: if enough traders treat a level as significant, it becomes significant because of their collective reaction to it.

Volume — volume is not a prediction tool, but it is a confirmation tool with real informational content. High volume on a breakout indicates institutional participation; low volume on the same breakout suggests retail-driven movement that may not sustain. Volume reveals conviction. A price move without volume is a weaker signal than the same move with volume — and this holds up in the data.

Momentum across specific timeframes — the academic evidence for momentum is among the strongest in the TA literature. Stocks that outperform over a 3-12 month lookback period tend to continue outperforming in the near term. This is not about moving average crossovers on a 5-minute chart — it’s a medium-term behavioral phenomenon driven by underreaction to information and institutional herding. The edge exists, has been replicated across dozens of markets, and has a plausible behavioral explanation.

VWAP as institutional reference — the Volume Weighted Average Price is not a prediction tool. It is the benchmark price that institutional traders use for execution quality measurement. Knowing that institutions buy dips below VWAP and sell rallies above it gives you real information about where institutional orders are likely to be clustered — which creates genuine support and resistance dynamics around it during the regular session.

The Most Popular Indicators: A Brief Overview

The following indicators are among the most widely used in technical analysis. This section covers what each one actually measures and the conditions where it tends to have the most relevance — not as a definitive ranking, but as a reference. Each deserves its own dedicated treatment, which we cover in individual articles linked below.

Indicator	What it measures	Works best in	Weakest in
Moving Averages (SMA / EMA)	Trend direction and smoothed price momentum	Trending markets with sustained directional moves	Ranging, choppy markets — generates constant false signals
RSI	Relative speed of recent price changes (overbought/oversold)	Ranging markets and mean-reversion setups	Strong trends — stays overbought/oversold for extended periods
MACD	Momentum via difference between two EMAs	Trending markets with clear momentum shifts	Low-volatility ranges — produces lagging, unreliable crossovers
Bollinger Bands	Price volatility and deviation from a moving average	Mean-reversion setups in ranging markets	Trending markets — price rides the band, not a reversal signal
VWAP	Volume-weighted average price — institutional execution benchmark	Intraday trading on liquid assets with clear session structure	Multi-day or swing setups — resets daily, loses context
Support & Resistance	Price levels where buying or selling pressure has historically concentrated	Most market conditions — most durable TA concept	Extremely high-volatility breakout environments

We break down each of these indicators in depth — how they’re calculated, where the edge comes from (or doesn’t), and how to apply them in practice:

• → Moving Averages: SMA vs EMA — what they actually tell you: coming soon
• → RSI: how to use it without misreading it: coming soon
• → MACD: the momentum indicator traders misapply most: coming soon
• → Bollinger Bands: volatility, not prediction: coming soon
• → VWAP: why institutions use it and how to read it: coming soon
• → Support and Resistance: the most durable concept in TA: coming soon

What Is Mostly Noise

The honest answer here requires naming things that a significant portion of the trading community treats as gospel. The evidence does not support that treatment.

Candlestick patterns with exotic names — doji, hammer, evening star, three black crows. The academic literature on candlestick patterns is largely negative. Caginalp and Laurent (1998) tested 8 candlestick patterns on 349 stocks and found that only a small fraction showed statistically significant predictive power, and the results were inconsistent across different samples. The problem is selection bias: traders remember the patterns that worked and forget the ones that didn’t. In a systematic backtest across all pattern occurrences, the edge largely disappears.

Indicators layered on indicators — a stochastic of stochastic, a moving average of ATR applied to Bollinger Band width, triple crossover systems built on top of other crossover systems. These constructions create the illusion of analytical depth while actually reducing information. Every additional derivation removes you further from the raw price and volume data that contains whatever signal exists. More inputs on a chart do not equal more information — they typically equal more noise with the original signal buried underneath.

Subjective trendlines — lines drawn by hand across price action are subject to confirmation bias in a way that systematic indicators are not. Two traders looking at the same chart will draw different trendlines, and both will find evidence that their line was the correct one. If a pattern requires subjective interpretation to identify, it cannot be backtested systematically — which means you have no way to know whether you’re seeing a real edge or a post-hoc rationalization.

Elliott Wave and Fibonacci retracements as prediction tools — these frameworks are flexible enough to be fitted to almost any price action after the fact. That flexibility is their fatal flaw: when a theory can explain anything, it predicts nothing. Fibonacci retracement levels at 38.2%, 50%, and 61.8% are popular enough that they create self-fulfilling clusters — but that’s a different claim from saying they have intrinsic predictive value.

How to Use TA Without Fooling Yourself

The conclusion is not “TA is useless.” It’s “TA requires a more rigorous framework than most traders apply to it.”

Use TA for timing, not for thesis — the most defensible use of technical analysis is as an entry and exit timing tool on top of a thesis you already hold for other reasons. If you believe COIN is oversold relative to Bitcoin’s fundamentals, TA can help you time the entry at a support level rather than buying into weakness. The TA is not the reason for the trade — it’s the precision layer on top of the reason.

Identify the market regime before applying any indicator — before using any indicator, ask: is this market trending or ranging? High volatility or low? That question should determine which tools you reach for, not habit. A momentum indicator in a ranging market will destroy a strategy that would otherwise have been profitable in the right conditions.

Backtest by regime, not overall — if you’re backtesting an indicator, split the historical data into trending and ranging periods and evaluate performance in each separately. An indicator that shows positive overall results but only works in 30% of conditions is not a robust strategy — it’s a strategy that works in one regime and loses in another, with the wins slightly outweighing the losses in aggregate.

Set a deactivation rule — decide in advance under what conditions you will stop using an indicator. If the profit factor of your strategy drops below 1.0 over 50 consecutive trades, the market has likely shifted regime and the indicator is no longer working. Without a deactivation rule, traders keep using tools that have stopped working because of sunk cost and confirmation bias.

Separate observation from prediction — when you look at a chart, be explicit with yourself about whether you’re describing what has happened or predicting what will happen next. The former is almost always reliable. The latter requires evidence that the specific pattern you’re seeing has historically led to the outcome you’re expecting — and that evidence needs to come from a systematic backtest, not from memory of past trades.

If Foolproof Strategies Existed, No One Would Sell Them

There’s a simple test you can apply to any trading strategy being sold online — a course, a signal service, a proprietary indicator, a “proven system” with a monthly subscription. The test doesn’t require reading reviews or checking credentials. It only requires basic arithmetic.

Suppose a strategy genuinely produced a consistent 2% return per month — not a guarantee of getting rich, just a modest, reliable edge. Applied to $10,000 with moderate leverage and compounded monthly, that’s roughly $10,000 becoming $97,000 in five years. Dial the leverage up slightly and the numbers become obscene. A 3% monthly edge compounded over a decade doesn’t produce wealth — it produces a number that doesn’t fit in a reasonable conversation about money.

Now ask the obvious question: if you had discovered that strategy, what would you do with it?

You wouldn’t sell it. Not for $99, not for $999, not for any price — because selling it creates competition that erodes the edge, and because you don’t need anyone else’s money when compounding is doing the work for you. The moment a genuine edge becomes widely known, arbitrage and crowding destroy it. Selling a real edge is an act of self-destruction.

This isn’t cynicism. It’s the logical implication of what a consistent edge actually means in financial markets. The strategies that get sold are the ones that stopped working, never worked at all, or only worked in backtests on historical data that won’t repeat. The ones that work stay private — inside prop trading firms, hedge funds, and the notebooks of traders who understand that silence is part of the strategy.

None of this means TA is useless or that edges don’t exist. It means that any edge worth having requires work to find, discipline to apply, and the judgment to know when it has stopped working. That work can’t be packaged into a course or outsourced to a signal provider. It’s yours to do — through reading, testing, losing, adjusting, and building your own understanding of how markets actually behave.

The Real Conclusion

Technical analysis is neither the edge that retail traders believe it is, nor the nonsense that academic purists claim. It’s a toolkit with real but conditional utility — and the condition is always the same: regime.

Some indicators have genuine structural reasons to work: they’re based on real liquidity dynamics, institutional behavior, or well-documented behavioral biases. Others are pattern recognition applied to random noise, made convincing by selective memory and confirmation bias. Knowing the difference is not intuitive — it requires systematic testing and honest accounting of when your tools are working and when they’re not.

The traders who use TA effectively aren’t the ones with the most indicators on their charts. They’re the ones who understand exactly what each tool measures, under which conditions it has historically worked, and when to put it down.

That standard — rigorous, conditional, evidence-based — is the only one worth applying.

References

1. Fama, E. F. (1970). “Efficient Capital Markets: A Review of Empirical Work.” Journal of Finance, 25(2), 383–417. Available at: https://www.jstor.org/stable/2325486
2. Jegadeesh, N., & Titman, S. (1993). “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency.” Journal of Finance, 48(1), 65–91. Available at: https://www.jstor.org/stable/2328882
3. Brock, W., Lakonishok, J., & LeBaron, B. (1992). “Simple Technical Trading Rules and the Stochastic Properties of Stock Returns.” Journal of Finance, 47(5), 1731–1764. Available at: https://www.jstor.org/stable/2328994
4. Lo, A., Mamaysky, H., & Wang, J. (2000). “Foundations of Technical Analysis: Computational Algorithms, Statistical Inference, and Empirical Implementation.” Journal of Finance, 55(4), 1705–1765. Available at: https://www.jstor.org/stable/222423