What it measures
Over a lookback window, usually fourteen periods, the RSI compares the average size of upward
closes with the average size of downward ones and expresses the result on a scale from zero to
a hundred. High means recent gains dominated. Low means recent losses did.
That is all it says. It is a description of the recent past, compressed into one number, and
every interpretation beyond that is something a reader adds. Keeping the measurement and the
interpretation separate is what makes the rest of this page possible.
- Scale
- 0 to 100
- Default period
- 14
- Measures
- Recent gain vs loss
- Predicts
- Nothing by itself
Misreading one: treating a level as a signal
Above 70 is conventionally called overbought and below 30 oversold, and both terms smuggle in
a conclusion. The reading states a condition, not an instruction. An asset can be above 70 for
weeks, and during that period the condition is describing a strong uptrend rather than an
imminent reversal.
Acting on the threshold alone means systematically selling strength and buying weakness, which
is a coherent strategy in a range-bound market and an expensive one in a trending market. The
indicator does not tell you which market you are in, which is precisely the information the
threshold reading assumes.
Misreading two: using it in a trend
The RSI was built with mean reversion in mind, the idea that a stretched move snaps back. In a
market that trends persistently, and crypto trends persistently, the indicator stays pinned at
one end and produces a continuous stream of signals that are all wrong in the same direction.
The practical adjustment is to establish the regime first and only then read the indicator. In
a range, threshold readings carry some information. In a trend, they mostly measure how strong
the trend is, and divergence is the only reading with much left in it.
Why it is worth knowing anyway
Because enough participants watch it that the levels become weakly self-fulfilling, and
because it appears inside other people’s arguments constantly. Understanding what it measures
lets you evaluate a forecast that cites it, which is more often the useful application than
trading it yourself.
How the number is actually built
Take the closing changes over the lookback window and separate them into gains and losses.
Average each. Divide the average gain by the average loss to get a ratio, then map that ratio
onto a scale from zero to a hundred. A period with no losses reads 100; one with no gains reads
zero.
Seeing the arithmetic makes the behaviour obvious. The reading is high when recent gains have
been large relative to recent losses, which in a strong uptrend is simply true and stays true.
Nothing in the formula contains a notion of “too far”, which is the meaning people attach to a
high reading.
The period setting changes what you are looking at
Nearly every chart defaults to fourteen periods, and nearly every discussion of the indicator
treats that number as though it were part of the definition. It is a parameter. Shortening it
makes the line reach the extremes far more often; lengthening it produces a smoother line that
rarely does.
That matters because the thresholds people quote were popularised alongside the default. Keep
the thresholds and change the period and the same reading now means something different, since
the frequency of reaching it has moved. A short period touching an extreme is close to
ordinary. A long one doing so is genuinely unusual.
The same applies to the timeframe underneath. An extreme reading on an hourly chart and the
same reading on a weekly one are not comparable observations, and screenshots circulate without
either setting being visible. Before treating any reading as information, check both numbers.
If a source does not state them, the reading cannot be interpreted, which is a more common
situation than it sounds.
Divergence, and why it is the defensible reading
Price makes a higher high while the indicator makes a lower one. That compares two
measurements rather than reading a threshold, and it says something specific: the second move
was made with less momentum behind it than the first.
It is also frequently early, sometimes by a long way, and early is indistinguishable from wrong
while a position is open. Which is why divergence is better used as a reason to pay attention
than as a reason to act, and why an indicator is at its most useful when it changes what you
look at rather than what you do.
Be the first to comment