Stochastic oscillator

Jana looks at a bitcoin chart and wants to know whether the latest price is nearer a recent peak or bottom. First she selects a window, a number of candles including the one being evaluated. A candle summarizes a time interval: High is its highest price, Low its lowest, and Close its last. Fidelity calculates basic %K from Close and the highest High and lowest Low across the whole window. Using only the latest candle’s extremes is not enough. As the window moves, an old extreme can leave and a new one can change the scale. [1]

TradingView describes a bounded scale: for raw %K, 0 is the lower edge and 100 the upper. The calculation first measures Close’s distance from the lowest Low, then compares it with the highest High minus the lowest Low. Jana gets a relative position, not a percentage profit or a probability. If the extremes are equal, the denominator is zero and the mathematical ratio is undefined. Any substitute value an application supplies is an implementation convention, not new market information. [2]

For the fast variant, Fidelity distinguishes basic %K from its moving average, %D. Averaging combines neighboring values and softens abrupt changes. In TradingView’s slow variant, a 3-period SMA of the original series first produces slow %K, followed by another 3-period SMA for %D. SMA means simple moving average. Jana therefore checks whether she is seeing the original calculation or an already smoothed line. The lines are related, not two independent votes for the same trade. [1] [3]

Jana sees the label overbought and considers selling. Yet TradingView describes 20 and 80 as commonly used indicator thresholds, not a valuation of bitcoin. For the raw calculation, they identify positions near the lower or upper part of the range; smoothed lines also carry previous values. A strong trend can keep the indicator near an edge for a long time. A low reading alone therefore does not establish a bargain, and a high one does not announce the end of a rise. [2]

Fidelity discusses %K crossing %D and divergence, meaning price and the indicator developing differently. An example is a new price peak without a higher oscillator peak. Jana can watch that disagreement, but the market is not obliged to reverse. She needs to define in advance which peaks she compares and when they count as recognized. Selecting them only after seeing subsequent prices would introduce future information into the test. [1]

While searching, Jana also finds Stochastic RSI. TradingView explains that it applies the same idea of relative position to a series of RSI values, not directly to price. RSI is another indicator that compares average upward and downward changes. RSI is calculated first, then its value is located within its own recent range. This is a different calculation: similar-looking lines do not justify swapping settings or expecting identical signals. [4]

Until a candle ends, its High, Low and current Close can change, as TradingView explains. That can alter both the latest price and an extreme defining the window. Jana may see a crossover during a candle that no longer exists at the close. Reacting continuously and waiting for the final result are different rules. A historical test cannot pretend that a candle’s final values were known at its beginning. [5]

Jana now knows what %K and %D measure. To evaluate a trading rule she also needs entry and exit conditions, position size and execution assumptions. TradingView distinguishes signals from simulated fills and discusses fees and slippage, the difference between expected and actual price. Overfitting means settings fitted too closely to past results. Further testing should therefore use data that did not shape the rule. A crossover by itself does not establish profitability. [6]