Learn · The Twelve Measurements — Lesson 5 of 12

RSI and momentum, explained

RSI is probably the most-quoted number in retail technical analysis, which is exactly why it deserves a careful lesson: what it measures is narrower than what it is commonly asked to do — and this product exists partly because we measured that gap.

What it measures

The Relative Strength Index compares the size of recent gains to recent losses on a 0–100 scale. It measures how far a move has travelled, not whether it is right.

The Relative Strength Index compares the average size of recent up-moves to recent down-moves, over a rolling window, on a 0–100 scale. At 70 the convention says “overbought”; at 30, “oversold”. Those words overpromise — the honest reading is stretched: gains (or losses) have dominated recently, and the move has travelled a long way in a short time.

Why it moves prices

Momentum is one of the most-documented effects in market history — assets that have recently risen tend, on average and over certain horizons, to keep rising, while short-term stretches tend to snap back. Those two findings pull in opposite directions, which is precisely why a single momentum gauge makes a poor strategy and a useful ingredient: it locates where in a move you are, without saying which of the two forces wins next.

Farlens has a specific, published reason for treating RSI this way. When the engine's strategies were re-tested with walk-forward validation — train on one period, test honestly on the next, repeat — a single 70/30 split had called an RSI-only approach robust. Three folds showed it was overfitted: it lost money out-of-sample in one fold. The composite retained its edge across every fold. That finding is published in full on the walk-forward page, and it is the reason this product scores twelve measurements rather than one.

How to read it

Readings near the top of the range mean gains have dominated and the move is extended; near the bottom, losses have dominated. Extended is not the same as wrong — strong trends can stay extended for a long time.

The most common misreading is treating 70 as a ceiling. In strong trends RSI can sit above 70 for weeks — “stretched” describes distance travelled, not a limit. The bands are also asymmetric in practice: readings spend more time high in bull markets and low in bear markets, so the same number means something different depending on the regime around it.

How Farlens uses it

Momentum (RSI) carries 8.6% of the Farlens composite (weight 1.0 of 11.6 across all twelve measurements). Coverage: Available for anything with enough price history.

Two conventions matter here. First, a measurement that cannot be observed for an instrument is skipped, never counted as zero — a zero would read as “neutral” and quietly dilute the composite. Second, on any given day a contribution is the measurement's score times its weight, divided by the total weight of the measurements present that day — so the published breakdown always sums to the score beside it. The full stack, with every weight published, is on the signals page.

A worked example

Farlens publishes one frozen example reading — AAPL on 22 July 2026, deliberately weeks old — so the shape of a reading can be studied without opening the live layer. That day the composite printed +14.7, “Mixed”, and this measurement read −75.7.

Measurement present that day	Reading
Trend	+100.0
Advance/decline pressure	+100.0
On-Balance Volume	+97.4
Momentum (RSI)	−75.7
Money Flow Index	−63.3
Sector rotation	−45.3
Volume surge	−29.7
Short volume share	−23.9

Momentum read −75.7 that day — deeply stretched to the downside — while the structural measurements stayed positive. Read together: a sharp pullback inside an intact uptrend, the exact situation where a single-gauge reading would have told only half the story.

A real reading, kept deliberately stale. Descriptive of one historical day — not a prediction, and not representative of current conditions.

What it cannot tell you

RSI knows nothing about volume (a stretch on heavy participation and one on air look identical), nothing about the trend it sits inside, and nothing about why the move happened. It is also the easiest measurement to over-optimise — tune the window and thresholds long enough and any backtest improves, which is the overfitting trap the walk-forward test exists to catch. Farlens deliberately runs it at standard settings.

Frequently asked

Does RSI above 70 mean the move is over?

The data doesn't support that folklore: in trending markets, readings above 70 have often been followed by further gains. It describes a stretched recent run — nothing more specific than that, and Farlens treats it as description, never as an instruction.

Why does Farlens still include RSI if RSI-only failed validation?

The finding was that RSI alone was overfitted, not that it carries no information. As one voice among twelve, at a moderate weight, its short-horizon stretch reading complements the slower structural measurements. The failure is published rather than hidden because that is the product's standard for itself.

What window does it use?

The standard 14-period convention. Optimising the window per asset is exactly the overfitting pattern the walk-forward finding warns against.