Does "overbought" mean the price is about to fall?

No. And the word itself is doing the damage.

What RSI above 70 literally says

From the previous lesson: RSI is the share of recent movement that was upward. A reading of 78 says that across the last fourteen sessions, roughly 78% of the average daily movement was to the upside. That is a description of the composition of the past two weeks.

There is nothing in the formula about supply being exhausted, about buyers running out, about fair value, or about tomorrow. Those ideas were smuggled in by the label. "Overbought" sounds like "bought too much, must come back". The arithmetic says "went up fast lately".

The empirical embarrassment

In strong uptrends RSI can sit above 70 for weeks — and the strongest trends are exactly the ones that do it. Consider a stock that rises around 1% on most days with the occasional small down day. The average gain stays large, the average loss stays small, RS stays high, and RSI stays pinned in the 70s and 80s for a month or more. Someone selling on each cross above 70 would have sold repeatedly into the best part of the move, then watched the reading go higher.

Drawn, it looks like this — an authored reading that crosses 70 and is still above it twenty-three sessions later:

The mirror case is uglier. RSI can stay below 30 through a long decline, and "oversold" invites people to buy something that is still falling — the classic way a small loss becomes a large one.

The reading describes strength, and strength persists more often than beginners expect. Wilder himself framed 70 and 30 as levels worth examining, not as triggers.

What experienced traders actually watch

Framed as description, not instruction:

• Whether extremes stick. Readings that stay extreme for many sessions look like trend; readings that flip in and out look like a range. That's a regime observation, available only in hindsight.
• The instrument's own history. Some tickers rarely exceed 65; others live above 80 for months. A universal 70 line ignores the distribution of the thing in front of you.
• Divergences — price makes a higher high, RSI doesn't. Worth noting; it resolves the "wrong" way often enough that it is a prompt to look at price, never a conclusion.
• The regime. The same reading means different things in a trend and in a range. This is the honest reason no single level "works": the level is conditional on something the level cannot see.

The language fix

Ban the conclusion from the label. Instead of "AAPL is overbought", write: "RSI(14) is at 78 — the last fourteen sessions were strongly one-sided to the upside." The second sentence is checkable, carries the same information and, notably, does not tell you to do anything. One is an observation; the other is a prediction nobody stated.

Try it now

1. Count the sessions above the line on the schematic above, then write the sentence a seller of signals would have written on the first of them. Twenty-three sessions is the answer the picture gives; the sentence is the thing being sold.
2. Now count it where you did not choose the data. On the five years below, with the 14-day RSI in the pane under the bars, mark every stretch that held above 70, and record the length of each. For each, record what price then did over the following ten sessions — continuations and reversals both, with no editing.

3. Go through your own notes and replace every "overbought" and "oversold" with the literal description. Notice how the urge to act quietly drains away once the word is gone.