Imagine Jana opening a bitcoin chart to understand its movement. She looks at where prices stalled, how quickly they rose and how much was traded. These market observations are the starting point of technical analysis. It helps to separate describing what happened from making a forecast and from placing a trade that could actually execute. Each step needs its own evidence.
Instead of a simple line, Jana sees colored candles. Each summarizes a chosen interval: the first and last price, the high and low; volume records the amount traded. To read it properly she needs the product, exchange, time zone and interval length. Different charts may therefore describe different trades. Coinbase also warns that its historical data may be incomplete and that an interval with no trades may have no candle. A gap in the data does not mean the price fell to zero. [1] [2]
Jana turns on a Moving average. It calculates a smoother curve from selected prices, making the broader movement easier to see. The chosen period and input data affect the result; the calculation adds no new information about events in the world. Other indicators, meaning calculated measures, may describe the speed or range of changes. Several curves can agree simply because they all use the same price data. Their number is therefore neither a count of independent evidence nor a measure of certainty. [1]
Support names an area where declines previously stalled; Resistance names an area where rises met resistance. These are observed zones, not promises made by the market. The same applies to a trend line joining selected points. To test a rule, Jana must state in advance which points count and what invalidates the pattern. Redrawing the line after every failure to fit the past does not test the original idea. It creates another one with knowledge of the outcome. [3] [4]
A backtest simulates how a rule would have performed on historical data. It can easily use information that was unavailable at the time. A candle’s closing price is final only once the interval ends. If Jana waits for that close, she cannot simply assume she can then buy at the same price from a moment already past. The signal, meaning the condition being met, must come before an executable trade. Otherwise the test gives the trader the advantage of knowing the future. [2] [6]
Jana tries different averaging periods and entry thresholds until one version looks excellent. The more attempts she makes on the same data, the easier it becomes to find a winner by chance. This is overfitting: a rule adapts to a particular past but may fail elsewhere. An honest record must include unsuccessful variants too. A fixed rule needs testing on data not used to select it. Repeatedly adjusting it to those data, however, makes them another tuning set; splitting the data alone does not solve the problem. [5]
Lo, Mamaysky and Wang studied automated pattern recognition in US stocks during 1962–1996. Some patterns provided additional information about the distribution of returns, meaning how different outcomes appeared in the data. That is narrower than claiming that charts can make money. It does not show that the same method earns a profit in bitcoin today. A different market, period or execution process requires its own evaluation. A statistical relationship may be interesting, but it does not pay the fees by itself. [4]
Even if a rule captures a price move, Jana needs to specify when to buy, when to sell and the size of the position. Results must include fees and realistic execution prices. The spread is the gap between the best buying and selling offers; slippage is the difference between the actual and expected execution price. Both can affect results, for example for a large order. If actual fills are already used, the same effects must not be deducted twice. Comparison with a reasonable alternative and the risks taken also matters. [6]
Look beyond a selected winning chart for the data source, exact rule, test period and list of attempted variants. An analysis should show costs and losing periods. If the rule changes later, it needs a new version, not a claim that it predicted correctly all along. This makes the idea open to scrutiny and comparison. It provides no guarantee of future earnings, but helps distinguish a clear question about data from an impressive story told after the event. [5] [6]
Jana waits for the candle to close
In this fictional example Jana buys only after a candle closes above 100 USD. The interval ends at 12:00 UTC at 101 USD; only now does she know the signal. She subsequently buys at an actual price of 102 USD and sells at 103 USD. The difference per unit is therefore 103 − 102 = 1 USD before costs, not 103 − 101 = 2 USD. Treating the closing price as the purchase price would unjustifiably improve the result. These figures explain timing; they are neither a current price nor a recommendation to trade.
For the clearest picture, read this entry together with Price action, Backtesting, Overfitting, OHLCV, Moving average, Fundamental analysis. The reverse links also lead from Fundamental analysis, OHLCV, Timeframe, Candlestick chart.
01Do I have more certainty if all indicators point upward?+
Not automatically. They may just recalculate the same price history in different ways. Whether the combination adds useful information must be tested with rules set in advance. Counting green curves cannot replace that test.
02Does a losing trade mean technical analysis does not work?+
One trade neither confirms nor disproves a general method. A specific rule defined in advance can be assessed on sufficient data, including losses, costs and uncertainty. Even a good historical record is no promise of another profit.