RSI Calculator, Wilder RSI for Any Period, Step by Step

Almost always because of smoothing. Wilder’s original RSI averages gains and losses with a smoothed moving average that carries the whole history forward, while some platforms use a simple or exponential average over the last n bars instead. The first reading also depends on how much data you feed in: RSI needs a seed period, and a series that starts 30 bars ago will not match one that starts 300 bars ago until the smoothing converges. Feed at least 5× your period for a stable figure.

Wilder used 14, and that remains the default almost everywhere, which matters in itself, a level that many participants watch becomes self-reinforcing. Shorter periods (7, 9) react faster and produce far more overbought and oversold readings; longer periods (21, 25) produce fewer but leave you later. Change the period for a reason you can state, not to make a signal appear.

No. In a strong trend RSI can sit above 70 for weeks while price keeps rising, and selling each time it crosses 70 is one of the most reliable ways to exit a trend early. Overbought means the recent gains have been large relative to recent losses, a statement about momentum, not about value or about what happens next.

Not in any way that survives out of sample, and the search for one is how most people lose time with this indicator. Shorter periods reach the extremes more often and produce more signals of which a larger share are wrong; longer periods produce fewer and later. Fourteen has no statistical claim — Wilder offered none — but it is the default everywhere, which means a great many participants are acting on the same line. That coordination is a real effect and it is the strongest argument for leaving the setting alone.

Feed a continuous series and let the gap be an ordinary change. A weekend or a holiday is not a missing bar, the next close simply follows the previous one, and the difference between them enters the calculation as any other change would. What does need fixing is genuinely absent data: a symbol that did not trade for a week leaves a hole, and interpolating across it invents changes that never happened.

For the level, yes, and it is the property that makes this indicator unusual. Because the scale is bounded and the calculation contains nothing external, two instruments at 75 have had a comparably one-sided run of gains against losses. What you cannot infer is comparable magnitude: a stock that rose two per cent on twelve of fourteen sessions and one that rose twenty per cent on the same pattern produce similar readings. The measure describes consistency, not size.

Because the purpose is verification rather than a data dump. Eight rows is enough to check the recursive step by hand and see the trajectory, and a full table of several hundred lines would be scrolled past rather than read. The computation runs over everything you paste; only the display is trimmed.

It is the standard thirty-close series used to verify RSI implementations, and this page reproduces its published values exactly, 70.53, then 66.32, then 66.55. If you change nothing and press Calculate, matching those three figures confirms the arithmetic on this page against an external reference before you trust it with your own data.

More than most people feed it. Wilder smoothing is recursive: each average is built from the one before, so the influence of the seed value decays but never disappears. In practice the difference between a series started 50 bars ago and one started 500 bars ago becomes negligible after roughly five times the period, about 70 bars for RSI(14), and is still visible in the second decimal before that. If you are comparing two implementations, give both several hundred bars before concluding anything.

The first average gain and average loss are a plain arithmetic mean of the first n changes; every value after that is recursive. That single simple mean is the seed, and it is where two otherwise identical implementations most often diverge, some platforms seed from the first n bars of the data you supplied, others from the first n bars of their own history for that symbol. Neither is wrong, and the results differ until convergence.

The average loss reaches zero, relative strength is undefined, and RSI is defined as 100 by convention. Implementations handle it differently, some return 100, some return an error, some quietly produce infinity and then render nothing. It is rare on liquid instruments over 14 periods, and common on short periods: RSI(2) hits the boundary regularly, which is one of the reasons that variant behaves so differently.

Wilder, the recursive form, which is the original and the default nearly everywhere. Cutler's variant uses a simple moving average of the gains and losses instead, forgetting everything older than n bars entirely. Cutler's is reproducible from a short series and Wilder's is not, which is why spreadsheets and academic papers often use it without saying so. If your figures disagree with a published table, this is worth checking before anything else.

Yes; the arithmetic does not care what a bar represents. What changes is interpretation: RSI(14) on five-minute bars covers just over an hour, and the same thresholds that are conventional on a daily chart have no standing there. If the series spans a session boundary, note that the gap between the close and the next open enters as a single large change, which can distort the averages for the rest of the window.

Using an exponential moving average with the standard 2 ÷ (n + 1) multiplier in place of Wilder's smoothing, which is equivalent to an EMA of period 2n − 1. An RSI(14) built that way behaves like an RSI(27) and looks plausible (the line is smooth, bounded and shaped roughly right), so the error survives review. The check is to compute one value by hand from the columns this page prints; if the recursive step does not match, the smoothing constant is the culprit.

Yes, and the result is a different indicator with the same arithmetic: substitute volume for the closing price and you get Volume RSI, which measures whether up-volume or down-volume has dominated recently. It answers a question price-based RSI cannot: whether participation is confirming the move.