Nexfin RSI Divergence Indicator

This indicator plots the RSI in its own window and locates fractal pivots on price, reading the RSI at those pivots. A bearish divergence is a higher price high against a lower RSI high with the earlier RSI peak in overbought territory. A bullish divergence is the mirror.

Most divergences never resolve, so the signal is withheld until the RSI crosses back through the midline, optionally with MACD agreement. That is the difference between marking every divergence on the chart and marking the ones that turned.

The divergence itself is drawn, not only the outcome: a line connecting the two price pivots, and a matching line connecting the two RSI pivots. You can see the disagreement that produced the signal instead of taking the arrow on trust.

A green arrow marks a buy, a red arrow marks a sell. Signals are evaluated on completed bars, so an arrow that has printed does not move or disappear afterwards. Signals are drawn back through the loaded history; by default the most recent 500 are kept on the chart, and that limit is an input.

Higher timeframe RSI

Plotted alongside the main RSI by default, and can additionally be required to agree before a signal is drawn. That filter is disabled by default.

Levels and panel

Suggested stop and target levels for the latest signal are drawn as rays that start at the bar the signal printed on. The panel reports the current RSI, the higher timeframe RSI, whether price is overbought or oversold, and the running count of buy and sell signals.

Signal statistics

The panel counts the divergences this configuration produced on the chart in front of you, and reports how they resolved against the stop and target the indicator itself suggests. Change an input, switch symbol or timeframe, and the numbers are recomputed for what you are now looking at.

It reports how often the setting fires as signals per year, the win and loss counts, and the expectancy in R implied by the reward-to-risk you configured. Below thirty resolved signals no headline percentage is shown at all, only the count, because a rate measured on a handful of signals cannot separate a good setting from a poor one. Next to the win rate it prints the break-even rate for that reward-to-risk, and colours the row by the expectancy rather than by the win rate: at 1:2.5, a setting that wins 33% of the time is ahead before costs, and the panel shows it in green. A last line adds up the result of every resolved signal, after the spread, into one total in R. Results differ by market: the panel shows each chart's own record, so read it on the symbol and timeframe you trade.

This is not a backtest. There is no position sizing, no compounding and no slippage. It answers one question: of the signals on this chart, how many reached the suggested target before the suggested stop.

Spread charged, not assumed

A second line charges your broker's current spread against every signal, scaled to price and applied at that signal's own risk, and prints the expectancy again net of it. A figure quoted before costs is not an answer, and a disclaimer under one does not turn it into an answer.

Same-bar outcomes

When one bar contains both the suggested stop and the suggested target, bar data cannot say which came first. The minute bars inside are read to settle it where that history exists. What stays ambiguous is excluded and counted separately, and the panel also prints a worst-case rate that scores every excluded signal as a loss, so you see the range rather than one number that hides a choice.

Stability check

Switch it on and the same rule is measured again with the RSI period and the pivot width each moved twenty percent in both directions, and the worst of those neighbours is reported next to your own result. A setting that only works at exactly its own numbers is reported as a spike rather than a plateau. The check declines to give a verdict when too few neighbours produced enough signals to judge, instead of quietly passing.

It is off by default, because each perturbation is a full re-scan of the history.

Alerts

Popup, push and email alerts, all disabled by default.

Try it first

Download the free demo and run it in the Strategy Tester's visual mode. Load enough history, watch the chart and the panel update bar by bar, and change one input at a time to see its effect. The demo runs in the tester, not on a live chart.

Technical

Any symbol, any timeframe. No DLL calls, no web requests, no external files.

Support: please use the comments section of this product, or send a message through the site.