RSI compares smoothed price gains with smoothed magnitudes of price losses. Smoothing reduces the influence of individual changes by maintaining an evolving average. It describes a single price series; despite its name, it does not compare bitcoin with another asset.
Petr opens a chart and sees another line below bitcoin’s price, labeled RSI. TradingView calculates it by separating upward changes from the magnitudes of downward changes. A rising period contributes zero to the loss series, and vice versa. What matters is not just how often prices rise, but by how much. Several small rises may carry less weight than a sharp fall; RSI is neither a count of green candles nor an average of the prices themselves. [1]
TradingView calls the ratio of smoothed gains to losses RS and converts it to the RSI scale. If Petr sees 75, that means neither a 75% return nor a 75% chance of further gains. The indicator describes a ratio of past movements. Small and large price changes can produce the same ratio, so the number alone cannot tell Petr how much money a purchase would put at risk. [1]
TradingView uses length 14, Close as the input and rma smoothing in its example. Close is a candle’s closing price: the candle summarizes a chosen time interval. An hourly chart therefore uses hourly changes, not automatically daily ones. The rma calculation carries part of the previous result forward, so its starting history can still affect today’s number. If two charts disagree, Petr first compares the market, interval, price input and calculation settings. [1]
Fidelity describes the traditional regions above 70 and below 30 as overbought and oversold. Those labels can sound like simple instructions: sell at the top, buy at the bottom. Yet a strong trend can remain in an extreme region. For Petr, this describes the balance of past movements, not a guaranteed reversal. Nor is it a valuation showing bitcoin to be objectively expensive or cheap. [2]
The RS ratio described by TradingView divides by the average magnitude of losses. If that value is zero, ordinary division is not enough; if both gains and losses are zero, the ratio itself has no unique result. Software needs an edge-case rule. When comparing platforms, check their treatment of zeros and the available history. A missing observation is not the same as a measured zero price change. [1]
TradingView notes that Close on an open candle follows the latest price. As the price moves, the calculated RSI changes too. Petr may see a threshold crossed during the candle, only for that crossing to disappear before it closes. Deciding on a finished candle and deciding partway through it are different rules. A historical test must not treat the final value as information that was available earlier. [3]
TradingView’s Smoothing option adds an average of already calculated RSI values. Petr may therefore see the original RSI alongside another, smoother line. This is not the same as smoothing gains and losses inside the underlying calculation. Any discussion of a crossover or threshold touch needs to specify which lines and settings it refers to. [4]
Fidelity classifies RSI as a momentum indicator, describing the strength and pace of past price movements. A trading rule still needs decisions about entry, exit and position size. TradingView also warns about overfitting: tailoring rules too closely to known history. A useful test includes separate data, fees and realistic execution—the prices and quantities actually traded. A carefully chosen attractive chart is no substitute. [2] [5]
What an RSI reading actually means
In a fictional example, the smoothed average gain is 30 USD/BTC and the smoothed average magnitude of losses is 10 USD/BTC. Their ratio is RS = 30 / 10 = 3. Conversion to the scale gives RSI = 100 − 100 / (1 + 3) = 75. Gains in this calculation are three times the losses, but the result means neither a 75% profit nor a 75% probability of a further rise. This is not a calculation from scratch using a candle list: the already smoothed averages are supplied as inputs. The model says nothing about position size, fees or the future price.
For the clearest picture, read this entry together with Technical analysis, Exponential moving average, Moving average, Backtesting, Overfitting, Technical indicator. The reverse links also lead from MACD, Stochastic oscillator, Average Directional Index, Ichimoku Cloud.
01Does oversold RSI mean I should buy?+
No. A low RSI describes losses dominating the selected past data. The price may keep falling and the extreme reading may persist. The threshold alone specifies neither the timing of a purchase, the amount to invest nor the risk of further losses.
02Why do two apps show different RSI readings?+
Compare the exchange and pair, candle interval, price input, length, smoothing and starting history. Also check whether you are viewing a completed or still-forming candle, and the basic RSI or an additional average of it. A difference need not mean an app is faulty.