Screens · Momentum
Overbought Stocks and RSI Above 70
In a rising market a reading above 70 is the normal state of the strongest instruments, not a warning about them. The screen returns leaders, which is useful, and close to the opposite of what the word implies.
What the screen is doing
An overbought screen ranks a universe by an oscillator and returns the members at the top of its scale. With RSI, which is what nearly every published version uses, that means instruments whose average gain has substantially exceeded their average loss over the lookback period.
So the output is a list of instruments that have risen persistently and without much interruption. The word "overbought" adds a claim on top of that (that the buying has gone too far), and the arithmetic contains no such claim. Nothing in the calculation knows what a fair price is.
That is the whole argument of the page, and it is arithmetic rather than opinion. RSI is bounded, so it cannot keep rising, but it can sit near its ceiling for as long as the gains keep outweighing the losses, and in a trend that is a long time. The bound is on the indicator, not on the price.
What the list is actually good for
| Reading | Defensible? | Why |
|---|---|---|
| "These have been the strongest instruments recently" | Yes, this is what the screen measures | A direct restatement of the arithmetic. Useful for seeing what is leading and in which sectors. |
| "These are stretched and due a pause" | Not from the reading alone | The reading persists through trends. A statement about what happens next needs evidence the level does not carry. |
| "These are candidates to sell" | No | It inverts what the screen found. In a rising market this is a list of the instruments that have been rising, and selling strength on a bounded oscillator is the classic way to be early for months. |
The middle row is worth being careful about rather than dismissive. Extreme readings do mean-revert, because a bounded measure must; the difficulty is that the reversion happens in the indicator whether or not it happens in the price. RSI can fall from 78 to 60 while price makes a new high, because the average loss has crept up, which is the mechanism behind most published divergences, and it is arithmetic rather than a change in behaviour.
What to check before using the list
The period. Fourteen sessions is Wilder’s default and the reason two screens disagree is usually that one of them is not using it. A shorter period crosses 70 far more often and produces a much longer list.
How much history was supplied. Wilder smoothing carries its seed forward with a decaying weight, so a screen computed from three months of data and one computed from three years can report different readings for the same session. The calculator prints the intermediate columns for exactly this reason.
The universe. If the screen was not restricted to common stock, part of the list is funds and preferred shares whose readings moved together on a rate move. And if it had no price floor, part of it is low-priced instruments whose oscillator readings are tick-size artefacts.
Whether the move was one event. A reading built over six weeks of steady gains is a different situation from one produced by a single announcement, and the level cannot distinguish them. This is the check that takes ten seconds and is skipped most often.
The market-wide version of the same screen
Counting how many instruments in a universe are above 70 turns this screen into a breadth measure, and that is a more defensible use of the threshold than any individual reading.
The count answers a question no index level can: whether a rise is being made by many instruments or a few. A market where a large share of the list is above 70 is one where participation is broad, which is the opposite of the fragile condition the breadth section describes; a market making new highs with very few instruments above 70 is being carried by a handful of names. Both are descriptions of the present, both are countable, and neither needs the word "overbought" at all.
The same cautions apply as to every breadth count. The proportion has to be computed on a stated universe (common stock only, with a price floor), or funds and preferred shares move the number on a rate day. And the level of the count means nothing borrowed from another market or another decade: express it as a percentile of its own history, which is the instruction every breadth page here ends with.
Pairing it with volume
The reading says a rise happened. Volume says how much stock it took, which is the part that makes the rise easier or harder to interpret, and the two together are the closest this screen gets to being informative.
A strong advance on expanding volume describes broad participation: a lot of stock changed hands to produce the move. The same advance on contracting volume describes a narrowing one, with fewer participants each session. Neither is a forecast, both are checkable, and the distinction is real in a way that the 70 level is not.
That is the honest summary of what to do with an overbought list: read it as a description of what has been strong, look at the volume beside it, check whether one event produced it, and do not invert it. The oversold screen needs the opposite warning: its list is full of businesses that are cheap for reasons the oscillator cannot see.
Frequently asked questions
What does "overbought" mean?
That an oscillator — almost always RSI — is at the high end of its scale, which means recent gains have outweighed recent losses by a wide margin. That is all it means. The word implies that buying has gone too far and something is owed in return, and the arithmetic contains no such implication: it is a description of the last fourteen sessions, not a statement about the next one.
Does an overbought reading mean a stock will fall?
No, and this is the specific error the screen invites. In an uptrend a reading above 70 is the normal condition of a strong instrument, not an exception to it, the figure on this page shows a synthetic trending series holding the reading for many consecutive sessions while the price continues rising. A reading that persists for weeks cannot be a signal about the next few days.
So what does the screen select?
The strongest instruments in the market over the last few weeks. That is genuinely useful information if you want to know what has been leading, and it is close to the opposite of what the label suggests. An overbought list in a rising market is a momentum list; treating it as a list of things about to reverse is reading it backwards.
Why is the threshold 70?
Because Wilder proposed 70 and 30 in the 1970s as reasonable boundaries, and the convention stuck. There is nothing special about the number: it is not derived from anything, and on any given universe the proportion of instruments above it varies enormously with the state of the market. The defensible use is to compute where 70 sits in the distribution of readings for the instruments you actually screen, which usually turns out not to be where the literature assumes.
Is a divergence more informative than the level?
It is a more interesting observation and it carries the same burden of proof. Price making a new high while RSI does not is a real feature of the data, it has preceded many reversals, and it has also appeared repeatedly in advances that continued for months. Nobody publishes the count of divergences that led nowhere, which is exactly the number needed to know what the observation is worth.
Does the period matter?
Considerably, and it is the first thing to check when two screens disagree. A 14-period RSI and a 7-period RSI on the same instrument cross 70 at different times and different numbers of times, and a shorter period crosses far more often. Wilder’s smoothing also means the reading depends on how much history preceded it, so a screen run on two years of data and one run on three months can report different figures for the same day.
What should be paired with the reading?
Volume, and the reason to look. A strong move on expanding volume describes broad participation; the same move on contracting volume describes a narrowing one. Neither is a forecast, and the pair is more informative than either alone. Checking whether the instrument had news is the other half, a reading driven by one announcement is a different situation from one built over six weeks.
Is the mirror screen symmetric?
Not in what it selects. An oversold screen fills with businesses in genuine difficulty, because sustained selling and deteriorating fundamentals frequently coincide, the oversold page covers that. An overbought screen fills with leaders. The measure is symmetric; the populations it finds at the two ends of its scale are not, which is worth remembering before applying the same interpretation to both.