Start with the formula
Before calculating, state price type, interval, parameters and initialization. Two applications may show slightly different numbers because they use different history lengths or smoothing variants. Comparison is meaningless until these conditions are known. A missing new trade is also not necessarily new price information if a platform carries forward the last observation.
What MACD contains
MACD is the faster exponential average minus the slower exponential average. Common parameters are 12 and 26 periods. The signal line is usually a 9-period exponential average of MACD. In this teaching notation, the histogram is MACD minus the signal line. A period refers to the selected observation frequency, not necessarily a trading day.[1]
At a fictional checkpoint, the fast average is 108, the slow average 105 and the signal line 2. MACD is 3 and the histogram is 1. These are supplied values for checking subtraction, not a calculation from a complete market series. A positive histogram says only that the average difference is above its smoothed version; it does not logically imply a positive next return.
Multiplying all prices by one factor also scales the difference of averages. Absolute MACD values of expensive and cheap shares cannot be compared as the same “strength” without a separately defined normalization. After a stock split, first check input-series comparability so a technical event is not mistaken for an economic decline.
RSI and positive versus negative changes
For a teaching version, split closing changes into positive amounts and absolute negative amounts, then compute average positive and average negative change under the chosen rule. If their ratio is RS, RSI is `100 − 100 / (1 + RS)`. In this editorial model, after the initial average, a new value receives a weight reciprocal to the number of periods. That is a model assumption, not a verified description of every platform; a simple rolling average produces another version.
In the author’s example, average positive change is 2 and average negative change is 1. RS is 2 and RSI is about 66.67. This is not a 66.67% probability of a rise or a share of winning trades. It describes the ratio of past directional-change magnitudes. If there are no negative changes, the usual formula requires division-by-zero handling; with positive growth, the limiting value is 100. A completely unchanged series requires a separately disclosed software convention.
An extreme value is not a reversal date
“Overbought” and “oversold” can create a false impression of a fair economic price. In fact, an indicator threshold belongs to its own construction. A prolonged one-sided series can keep RSI near the edge. Changing the window changes sensitivity and the value even without new fundamental information.
MACD, its histogram and RSI partly share inputs. Agreement is not three independent confirmations. Divergence from price describes a relationship between transformed series that can be formalized and tested, but it is not a reliable reversal from one image. The [moving-average guide](/guides/moving-averages/) shows the source of this dependence.[2]
Test implementation separately from the hypothesis
First run a constant series, steady growth and one jump. Check initial values, rounding precision and missing-data handling. That is an implementation test. If a future-result relationship is claimed, a separate research protocol, holdout sample and cost model are required; the [technical-analysis evidence guide](/guides/technical-analysis-evidence/) addresses that task.
Separate definitions from interpretations in the sources. “Signal line” names a component; it is not an independent study of usefulness. Financial literacy here means explaining why a number changed and what the formula cannot establish, rather than treating a convenient transformation of history as a promise of future movement.
Special-case check and sources
On a constant series, MACD after consistent initialization should reflect no difference between fast and slow estimates. For RSI, constancy requires an explicit convention because positive and negative changes are both absent. Save the test result with the program version. If the provider hides zero handling, exact comparison is limited. Do not choose the implementation that best explained one past turn.
[1] Fidelity MACD: construction of MACD and RSI; page trading recommendations are not adopted. [2] Fidelity Exponential Moving Average: average definition and weight differences; no signal validation. Additional primary materials: Fidelity Simple Moving Average and Fidelity RSI, definitions only. The original CME materials were not obtained in full; only the named concepts and teaching arithmetic were checked, and trading interpretations and predictive promises were not validated.