RSI — advanced mechanics, the 70/30 zones, the 80/20 modification and divergences

In the summer of 1978, working out of Greensboro, North Carolina, J. Welles Wilder Jr. published a book that quietly reshaped how traders look at a price chart. Among the indicators he unveiled — ATR, ADX, Parabolic SAR, DMI — the one that built the longest career was the RSI. Half a century later virtually every charting package, from TradingView to MetaTrader 5, ships it in the default palette, and on the trading-account statements reviewed in MyBank.pl's reader inbox it would be hard to find a retail trader who does not have it switched on. And yet only a fraction of those users truly understand how the indicator is calculated, why the textbook 70 and 30 levels fail inside strong trends, and what Constance Brown actually meant when she described the 80/20 modification. In this article we strip the RSI down to its building blocks and walk through every signal it produces — both the way Wilder originally intended and the way later analysts extended his work.

What the RSI is actually built from

The RSI (Relative Strength Index) is an oscillator bounded on a 0 to 100 scale. Its whole mechanic boils down to one question — to what extent have closing-price gains outweighed losses over the last fourteen periods. A reading near one hundred means almost every session in that window closed higher; a reading near zero means almost every session closed lower. In real markets the RSI almost never touches the extremes — it spends most of its time between twenty and eighty.

One subtlety quietly missed by most educational materials deserves underlining. Wilder does not use a plain arithmetic mean for his average gain and average loss — he uses his own recursive smoothing, described in detail in the second half of this article. That is why an RSI computed by hand on the last fourteen candles in a spreadsheet returns values one to three points different from what TradingView or MetaTrader display. Around the 30 and 70 thresholds that gap can be the difference between a signal arriving one candle earlier or one candle later.

Wilder smoothing, step by step

The very first value of the average gain and average loss is computed as a plain arithmetic mean of the first fourteen periods. Sum of positive changes divided by fourteen — that is all. From the fifteenth period onward the recursive machinery takes over. Each new average gain is yesterday’s average multiplied by thirteen, plus today’s reading, with the result divided by fourteen. The same logic applies to the average loss.

Mathematically that is an exponential moving average with a smoothing factor of one over the period, rather than two over the period plus one, which is the convention for a classic EMA. Wilder engineered the method specifically to give the indicator a longer memory and a slower reaction to one-off noise spikes. The consequence is that an RSI smoothed Wilder-style typically needs two or three consecutive candles in the same direction to move noticeably. Put differently — the indicator filters impulses rather than mirroring them.

The 70 overbought and 30 oversold zones — and where they fail

Wilder’s classic decision thresholds are the 70 overbought zone and the 30 oversold zone. The rule sounds simple: once the RSI clears 70 the market is "overheated" and statistically near a local top; once it drops below 30 it is "oversold" and near a local bottom. In a range-bound environment that reading works beautifully — a threshold breach precedes a reversal in well over seventy percent of cases.

The trouble starts the moment the market leaves the range and enters a trend. An overbought reading inside an uptrend stops being information about a top and becomes information about strength. The textbook case is the USD/JPY rally of spring 2024, when the daily RSI sat above 70 for more than six straight weeks while price climbed by more than ten big figures. Every trader who shorted on the basis of an overbought reading was stopped out, often several times in a row. The same dynamic played out on EUR/USD in the autumn of 2020, when the daily RSI lived above 70 for almost five weeks and retail traders who fought it recorded record drawdowns.

The statistics here are unsparing. In a trending market — defined as price above the 200-period moving average together with a series of higher highs on the next higher timeframe — contrarian RSI signals win on roughly forty percent of trades. That is below a coin flip, and once transaction costs are accounted for it becomes a clear losing strategy. The fix is either to ignore contrarian RSI signals in a trend or to apply the modification described in the next section.

The 80/20 modification in trends — Constance Brown’s insight

Constance Brown, in her book Technical Analysis for the Trading Professional, first published in 1999, was the first to formally describe how the RSI’s working range shifts with the market regime. The idea is the following: in a clear uptrend the RSI rarely sinks to the 30 oversold zone — its lower touches during pullbacks tend to stall in the low forties. Its upper touches, by contrast, regularly clear eighty and can reach into the high eighties. In a downtrend the relationship mirrors — upper touches during bounces stall in the low sixties, while lower touches dig down into the twenties.

The practical consequence is fundamental. Inside an uptrend an RSI touching forty during a correction is a genuine buy signal — it behaves analogously to a touch of thirty in a range-bound market. An RSI touching eighty in the same uptrend is not a sell signal but rather confirmation that the upward move is strong. A trader who learns to switch between the two threshold pairs as the regime changes statistically removes roughly sixty percent of the false signals that rigid classic settings would have produced.

The fifty midline as a trend filter

The 50 level on the RSI carries a deeper meaning than most educational material suggests. That value corresponds mathematically to the moment when the sum of gains and the sum of losses over the last fourteen periods are equal — the market sits in a perfect balance of forces. Every reading above fifty means buyers are in control; every reading below means sellers are. That distinction is independent of the overbought and oversold thresholds.

The most practical use of the level is as a trend filter inside directional strategies. The rule reads: when the RSI on the next higher timeframe (the daily, say, when trading on H4) is above fifty, take only long setups. When it is below fifty, only shorts. Backtests on 2018 to 2024 data for EUR/USD, GBP/USD and USD/JPY show that simply layering this filter onto any directional strategy lifted its statistical win rate by an average of eight to twelve percentage points. With no extra indicator — purely by removing trades that fight the dominant regime.

Regular and hidden divergence — two different worlds

The divergence classification widely popularised by Constance Brown after her mentor Andrew Cardwell is one of the most important extensions ever built on top of Wilder’s original indicator. It splits divergences into four types, and the split is not academic — each type carries a different piece of market information and a different statistical edge.

• Regular bullish divergence. Price prints a lower low, the RSI prints a higher low. A signal that the downtrend is losing strength and may reverse upward. Sellers are running out of breath even though price is nominally still falling.
• Regular bearish divergence. Price prints a higher high, the RSI prints a lower high. A warning that the uptrend is fading — buyers are running out of fuel even though price is nominally still climbing.
• Hidden bullish divergence. Price prints a higher low (typically during a pullback inside an uptrend), the RSI prints a lower low. A signal that the uptrend will continue once the local correction is done — the buyers have not yet had their last word.
• Hidden bearish divergence. Price prints a lower high (typically during a bounce inside a downtrend), the RSI prints a higher high. A signal that the downtrend will continue.

Daily-timeframe statistics on the majors from 2018 to 2024 show that regular divergences hit on roughly fifty-five to sixty-five percent of trades once paired with a candle-pattern confirmation or a support or resistance level. Hidden divergences, which trade with the prevailing trend, return between sixty-five and seventy-five percent under the same conditions. The gap is in line with intuition — trading with the trend statistically beats trading against it.

"The Relative Strength Index can be considered one of the most useful tools in technical analysis. It is essential to remember, however, that no single indicator, taken in isolation from price itself, will ever replace a trader’s own reading of the market — the RSI is an extension of that reading, not a substitute for it." — J. Welles Wilder Jr., New Concepts in Technical Trading Systems, Trend Research, Greensboro, North Carolina, 1978.

The most common mistakes in reading the RSI

Years of watching retail brokerage statements through outlets such as MyBank.pl reveal three recurring mistakes that reliably turn an otherwise valuable indicator into a generator of losses. Each one stems from ignoring the broader market context in which the reading is being taken.

• Treating the 70 and 30 thresholds as entry signals. A simple threshold breach is not an order to open a position — it is information that the market sits near a statistical extreme for the current regime. In a range-bound environment that signal is strong; in a trending environment it can run directly against the trend and produce a string of losing trades.
• Ignoring the higher timeframe. RSI readings on M5 or M15 are largely noise — the indicator reacts to every micro-shift in momentum in a way that has little to do with a real change in the market’s balance of forces. Strategic RSI-based decisions are best made on H4 and above; lower timeframes serve only for refining the exact moment of entry.
• Watching the number without watching the structure. The bare fact that the RSI currently reads 54 tells you less than the shape of the RSI line over the past few weeks. Is the oscillator rising or falling? Is it above or below its own moving average? Is it forming divergences against price? The numerical value is only one of several components of a complete read.

Summary and where to go next

The original RSI, as Wilder built it in 1978, is a far more subtle tool than most popular write-ups suggest. Three elements are essential to reading it correctly. First — Wilder smoothing is recursive, which makes the indicator react more slowly than a classic exponential moving average and gives slightly different readings from a hand-calculated arithmetic version in a spreadsheet. Second — the classic 70 overbought and 30 oversold thresholds work only in a range-bound market; inside a clear trend they should be modified to 80 and 20 (or 80 and 40 in an uptrend, 60 and 20 in a downtrend), in line with Constance Brown’s insight. Third — divergences, both regular and hidden, are among the most valuable extensions of the original indicator; regular divergence warns of a trend reversal, hidden divergence warns of trend continuation.

A trader who internalises those three layers stops using the RSI as an automatic buy-and-sell generator and starts using it as a map of market context. The shift is fundamental — and it is precisely why seasoned traders, in private conversation, will quietly admit that the RSI remains one of the indicators they look at most often, even though the tool itself is now half a century old.

Related reading: RSI — how to read and when it fails — an introduction to the oscillator and its most common traps, a sensible starting point before this deeper article; MACD — mechanics, the 12-26-9 parameters and signals, step by step — the same kind of teardown for the second most popular momentum oscillator; RSI and MACD divergence — a trading system — a detailed walk-through of the four divergence types together with a concrete EUR/USD case study.