Does technical analysis work? Indicators, Fibonacci levels, pivot points and what research shows

The short answer: research does not support technical analysis as a way to make steady money on short-term trades. Simple rules such as moving averages were profitable on historical data up to the 1990s, but once you correct for «fitting» to past data and for trading costs, that profit mostly disappears, and fewer than 1% of day traders make money consistently. The only thing that survives strict testing is slow trend following in diversified futures portfolios, and that has nothing in common with what courses on indicators, Fibonacci levels and pivot points teach.

Andrii Kolisnyk November 13, 2026 13 min read

What technical analysis is and what it rests on

Technical analysis is an attempt to predict price movements from past prices and trading volumes alone, without analysing the company, the economy or the news. Its supporters rely on three claims: the price already reflects all information, prices move in trends, and history repeats itself because people behave the same way in markets.

In practice, technical analysis consists of three large groups of tools:

indicators - formulas that turn past prices into a new line or histogram: moving averages, RSI, MACD, stochastic, Bollinger Bands, volume indicators;
levels - prices at which a reversal is supposedly likely: support and resistance, Fibonacci levels, pivot points;
chart and candlestick patterns - «head and shoulders», «triangles», pin bar, engulfing, flat.

The question is not whether you can draw a line on a chart from which the price once bounced. You always can. The question is whether it gives you an edge in advance, after all costs and on data the analyst has not yet seen. That is what research tests.

What the research says

Economists have been testing technical analysis since the 1960s. Below are the most important studies that today's debate builds on. All of them were published in peer-reviewed journals.

Study	What was tested	Result
Brock, Lakonishok, LeBaron, 1992	26 simple rules (moving averages, range breakouts) on the Dow Jones index for 1897-1986	The rules looked profitable. For a long time this study was the main argument in favour of technical analysis.
Sullivan, Timmermann, White, 1999	The same and thousands of similar rules, corrected for «fitting» to history, plus 10 new years of data	The best rule on the old data did not beat simply holding the index on the new 1987-1996 data.
Park, Irwin, 2007	A review of 95 modern studies	56 found profits, 20 found losses, 19 had mixed results. But most studies have methodological flaws: rules chosen after the fact, costs and risk estimated imprecisely.
Kuang, Schröder, Wang	25,988 rules on 10 emerging-market currencies	Without correcting for fitting, thousands of «profitable» rules turned up. With the correction, profitability disappeared in most cases.
Bajgrowicz, Scaillet, 2012	Rules on the Dow Jones for 1897-2011, testing whether the best rule could be picked in advance	An investor could never have chosen the future best rules in advance. Even on historical data, low trading costs wipe out the profit completely.
Moskowitz, Ooi, Pedersen, 2012	«Time series momentum» on 58 liquid futures: indices, currencies, commodities, bonds	A trend over the past 1-12 months tended to continue. A diversified portfolio of such strategies delivered notable excess returns.

The overall picture is this. Early positive results for simple rules mostly did not survive stricter tests correcting for fitting and costs. Only slow trend following has a robust positive result: trades once every few weeks or months, dozens of markets at once, the low costs of a large fund. Minute charts, indicators from courses, Fibonacci levels and pivot points do not belong to this category.

What happens to real traders

Data on real people with their real accounts is even more convincing. There is no fitting of rules to history here: only the money people made or lost.

Taiwan. Researchers analysed all day traders on the Taiwan stock market for 1992-2006. Fewer than 1% of them were able to earn predictably and reliably after fees.
Brazil. Among everyone who started day trading index futures in 2013-2015 and kept going for more than 300 days, 97% lost money. Only 1.1% earned more than the Brazilian minimum wage. The authors specifically note that this contradicts what course providers claim.
Europe. According to the European regulator ESMA, 74 to 89% of retail client accounts lose money on CFDs.

The Brazilian study shows another pattern: the longer people traded, the smaller the share that stayed profitable. Among those who traded for one day, 29.8% made a profit; among those who traded more than 300 days, only 3%. Experience did not turn into skill: the sample simply grew and the effect of random luck faded.

Why everything «works» on a chart: fitting to history

The main reason technical analysis looks effective is a statistical trap researchers call data snooping. If you try enough rules on the same historical prices, some will inevitably show excellent results purely by chance.

That is why serious studies test rules on data that was not used to select them and correct for the number of variants tried. Courses, blogs and screenshots of «trades» never do this. On top of that, screenshots show winning trades, while losing ones simply are not published.

Why we see patterns where there are none

The human brain is wired to look for patterns: in clouds, in noise, in charts. On a random chart generated by tossing a coin it is easy to «find» trends, support levels and even «head and shoulders». It is easy to check for yourself: build a «price» that rises or falls each day depending on how a coin lands, and you will find every pattern in the textbook on that chart.

Two more traps add to this. The first is confirmation bias: trades where a level «worked» are remembered, and those where it did not are explained by a «wrongly drawn level». The second is the illusion of control: the more complex the method and the more lines on the chart, the stronger the feeling that you are managing risk. In fact, the number of lines has nothing to do with forecast accuracy.

Indicators: what they really show

Any indicator is a formula based on the same past prices you already see on the chart. It adds no new information and only presents it differently. That is why most indicators lag: the signal appears when a large part of the move has already happened.

Indicator	What it calculates	Typical promise	What you should know
Moving averages	The average price over N periods	A crossover of averages signals a new trend	The most studied rule. It was profitable on old data, but not after correcting for fitting and costs.
RSI, stochastic	The ratio of rises to falls over a period, on a 0-100 scale	Above 70 is «overbought», below 30 is «oversold»	In a strong trend the indicator stays «overbought» for weeks while the price keeps rising.
MACD	The difference between two moving averages	A signal-line crossover is the entry point	Derived from moving averages, so it has the same limits and lags even more.
Bollinger Bands	A moving average and the deviation from it	Touching a band means a reversal	Shows price volatility. Volatility is easier to predict than direction, but that does not reveal the direction.
Volume indicators	Trading volume together with price change	Volume «confirms» the move	The forex market has no single volume figure: the broker only shows tick counts or volume within its own system.

«Rankings of the best indicators» and collections of «non-repainting indicators» have no research behind them. If a particular indicator consistently gave an edge, funds with billions in turnover and fast algorithms would be the first to use it, and the edge would quickly disappear.

Fibonacci levels: maths instead of magic

Fibonacci levels are built like this: take a noticeable price move, for example a rise from a low to a high, and mark fractions of that move down from the top: 23.6%, 38.2%, 50%, 61.8% and 78.6%. A pullback is supposed to often stop near one of these levels.

Now look at these numbers closely. Five levels are spread across the whole pullback range from 138.2 to 110.7, about 7 points apart on average. Almost any pullback will stop somewhere near one of them, and in hindsight it will look like «the 38.2 level worked» or «the 61.8 level worked». If the price goes through all the levels, people say the move «was not a correction».

The numbers themselves come from the Fibonacci sequence and the «golden ratio», but there is no economic mechanism linking currency or stock prices to them. A self-fulfilling expectation can have some effect: if many traders place orders near a popular level, the price sometimes reacts there. But the strength of that reaction is unpredictable, and we found no convincing independent studies showing that Fibonacci levels are profitable after costs.

Pivot points: yesterday's range in new packaging

Pivot points are calculated from the previous day's high, low and closing price. Floor traders once used them to quickly gauge where the price was yesterday.

Pivot points are arithmetic on yesterday's range, and they know nothing about today's news, central bank decisions or large orders. As with Fibonacci, there are several levels, so the price almost always passes near one of them. Besides, there are several versions of the formula (classic, Camarilla, Woodie, Fibonacci pivots), and each trader has their own levels, which weakens the «everyone watches the same thing» effect.

Candlestick patterns, pin bars and flat markets

Candlestick patterns (pin bar, engulfing, doji) and chart patterns («head and shoulders», «triangle», «flag») have one more problem: their definition is subjective. Two traders looking at the same chart will see different patterns, and there are no precise rules for how long a «wick» has to be to make a candle a pin bar. That makes these methods the hardest to test statistically, and what cannot be tested cannot be refuted either.

A flat market, that is, sideways price movement without a trend, describes what has already happened and forecasts nothing. Indicators that supposedly «detect a flat» detect it with a delay: by the time you see it, the flat may already be over.

The maths of a trade that courses do not show

Even if a method guesses the direction more often than not, that is not yet a profit. The result depends on three numbers: the share of winning trades, the ratio of potential profit to loss, and the cost of each trade (spread, commission, swap).

A high share of winning trades is often achieved exactly like this: profits are taken quickly, while losing trades are allowed to «run». The reverse is also possible: with a 40-pip take profit, a 20-pip stop loss and a 1.5-pip spread, about 36% winning trades is enough to break even. But that is just arithmetic. The key question is whether the win rate stays stable in the future, and that is where the research above says «no»: an edge found in historical data usually does not repeat. How to size a position and why the risk per trade should be small is explained in our article on risk management.

Leverage multiplies these losses: with 1:100 leverage, a 1% price move against you means minus 100% of the money invested. How this ends in a margin call is covered in our article on leverage and stop-out.

So why do professionals use it

Banks and funds do use technical analysis, but mostly not the way courses teach it:

as an execution tool: to split a large order into parts without moving the price against themselves;
as a language for describing the market: «price above the 200-day average» is a handy shorthand in a report;
as part of systematic trend-following strategies: dozens of markets, trades once every few weeks or months, minimal costs, strict risk limits. Strategies of this kind had a positive result in the Moskowitz, Ooi, Pedersen study, and even they go through long losing periods.

What makes money consistently from retail trading is the spread, commissions and swaps. They go to whoever provides market access, regardless of whether you guessed the direction.

How to check a «strategy» someone is selling you

If you are offered a course, signals or a «robot» based on indicators, ask a few questions:

Has it been tested on new data? Results on the history the strategy was fitted to prove nothing. You need real-time trading recorded in advance.
Are the spread, commission and swap included? On short trades, costs decide whether you end up in plus or minus.
How many trades, and what was the maximum drawdown? Ten winning trades are not statistics. What matters is how deep the account fell along the way.
Is there a verified account statement? Screenshots and «monitoring» are easy to fake or to cherry-pick.
Why would the seller share it? If a method really makes money, selling it to everyone makes no sense: the edge would disappear. Selling courses is often more profitable than the trading itself.
Who is the broker, and is it licensed? A strategy is often «bundled» with a particular broker that pays the author for clients. How to check a broker is explained in our separate article on checking a broker.

In short

Technical analysis predicts prices from past prices. Indicators, Fibonacci levels and pivot points add no new information and only present the same data differently.
Early studies found profits from simple rules, but after correcting for fitting to history and for costs they mostly disappear. According to research, the best rule cannot be picked in advance.
Only slow trend following in diversified futures portfolios has a robust result. It is a tool for funds and has nothing in common with day trading.
Among real day traders, fewer than 1% make money consistently; in Brazil, 97% of those who traded more than 300 days lost money.
A high share of winning trades does not guarantee a profit: what matters is the ratio of profit to loss and the costs.

Figures in this article

Retail CFD accounts losing money, lower bound74%
ESMA data, 2018
Retail CFD accounts losing money, upper bound89%
ESMA data, 2018

Checked against official sources on October 30, 2026. If a figure changes, we update the article and the «Updated» date changes automatically.

Sources

This article is for information only and is not investment advice or a recommendation to trade. Trading with leverage and CFDs is high-risk: most retail traders lose money. The numerical examples are illustrative and show only the maths of trades.

Spotted a mistake or an outdated figure? Let us know and we will check it against the primary source and fix it.