The Relative Strength Index compares how much a stock has gained on its up days against how much it has lost on its down days, over a rolling window.
About the Numbers in This Tutorial
The figures below were measured on SPY in July 2026, across the full price history the tool loads. Numbers like these move slowly, since each new session is one more observation in a very large sample, so expect small differences rather than different conclusions.
The calculation
This tool uses the standard 14-day RSI. Over the trailing fourteen sessions it keeps a smoothed average of the daily gains and a smoothed average of the daily losses, then turns their ratio into a number between 0 and 100:
RSI = 100 − 100 ÷ (1 + average gain ÷ average loss)
Both averages are smoothed forward each day rather than recalculated from scratch, which is the conventional method.
A reading of 50 means the average gain and the average loss have been about equal. Above 50 the up days have been carrying more weight; below 50 the down days have.
RSI cannot leave the 0 to 100 band, which is what makes 70 and 30 usable as fixed reference points.
What it is not
Despite the name, RSI does not compare the stock to anything else. There is no index, no sector, no peer group. It is a measure of a stock against its own recent history and nothing more.
It is also bounded. Price can run forever but RSI cannot leave the 0 to 100 range, which is what makes it useful as a level and also what makes extreme readings behave differently from extreme prices.
The Detail That Matters Most
Today's close is one of the fourteen days in today's RSI. A large move today shifts today's reading immediately, and by roughly a fourteenth of the calculation. Hold on to that: Chapter 4 is entirely about what it does to the numbers on this page.
Unlike a zero line, RSI has no natural boundary. You pick one, and that choice changes everything the page reports.
The controls
The RSI Level stepper sets any threshold you like. Quick Select offers the three conventional ones: 30 for oversold, 50 for the midpoint, and 70 for overbought.
Open the tool on a ticker and it does not start at 50. It starts at that ticker's current RSI, rounded, so the first thing you see describes where the stock actually is. On SPY that opened at 52. Reach for Quick Select when you want a conventional level instead of the live one.
The three levels describe three different worlds
The same ticker, three thresholds. Measured on SPY, 21 July 2026.
SPY sits above RSI 30 on 98.2% of sessions. Dropping below it is rare, and when it happens it lasts an average of under two days. Above RSI 70 is the opposite kind of rare: 7.9% of sessions, in bursts averaging 3.3 days.
So "oversold" and "overbought" are not symmetrical conditions on this ticker, and neither is a place the stock spends much time. Picking 30 or 70 means studying a handful of unusual sessions; picking 50 means splitting the history roughly two to one.
Reading Tip
Before reading any percentage on the page, look at what share of sessions the level captures. A statistic drawn from 1.8% of history is describing something the ticker almost never does.
Chapter 3
The Period View
The first of the two views. It splits every session in the ticker's history by which side of your level the RSI was on, then describes each side.
What each column reports
The top pair covers time: average days per year on that side, share of all sessions, average consecutive days, and the longest stretch. The bottom pair covers conditions: average daily return, daily green/red odds, average VIX, and average RSI.
SPY across the three preset levels
Level
Sessions above
Green odds above
Green odds below
Avg VIX above
Avg VIX below
RSI 30
98.2%
56%
7%
19.4
32.9
RSI 50
68.1%
63%
38%
17.1
25.1
RSI 70
7.9%
82%
53%
14.2
20.1
Measured on SPY, 21 July 2026.
The volatility gradient is clean
Read the two VIX columns from top to bottom. Below RSI 30 the average VIX is 32.9; above RSI 70 it is 14.2. The relationship runs smoothly through every level in between, which is about as tidy as market data gets.
Low RSI and high volatility are the same market seen through two instruments. Stocks fall faster than they rise, so the sessions that drag a 14-day RSI down are the same sessions that push option pricing up.
Then there are the odds
The green/red odds look astonishing. Below RSI 30, SPY closed green on just 7% of those sessions. Above RSI 70, it closed green on 82%. Those are not the kind of numbers market data usually produces.
They are also not what they appear to be, and the next chapter is about why.
Chapter 4
The Circularity Problem
This is the most important chapter on this tool, and the reason its most dramatic numbers should be read with the most care.
The measurement contains its own answer
The Period view classifies each session by that session's RSI, then reports that same session's return. But today's close is one of the fourteen inputs to today's RSI.
So a day with a heavy loss pushes the RSI down and supplies the negative return being averaged. The two are not independent observations. They are two descriptions of the same day, and one of them helped decide which bucket the other landed in.
today's return → today's RSI → which column today lands in
The arrow only runs one way, and it runs from the thing being measured into the thing doing the measuring.
How much this explains
Look at the magnitudes rather than the odds:
Condition
Average daily return
Green odds
Below RSI 30
−1.94%
7%
Below RSI 50
−0.38%
38%
Above RSI 50
+0.24%
63%
Above RSI 70
+0.36%
82%
Measured on SPY, 21 July 2026.
An average daily loss of 1.94% is enormous for an index ETF. It is not a discovery about what happens when RSI is low. It is close to a description of what it takes to get a 14-day RSI below 30 in the first place: a run of heavy declines, of which today is one.
The 7% green figure says something similar. On a day SPY closed green, that gain lifted the RSI, which made it less likely to be sitting under 30 when the session was classified. The measurement quietly filters out most green days from that bucket before counting.
What This Does and Does Not Invalidate
The time rows are unaffected: how often a ticker sits above a level, and how long stretches last, are clean measurements. The VIX and RSI averages are descriptive and fine. It is specifically the same-session return and odds rows that carry this circularity, and the more extreme the level, the more of the result it explains.
Where to look instead
The Crosses view measures returns after the event rather than on the same day, which removes the loop. That is the half of this tool to reach for when you want to know what follows a condition rather than what defines it.
Switch views and the page counts transitions instead of sessions, then measures forward from each one at 5, 10 and 20 days.
The counts alternate
At RSI 70, SPY shows 120 crosses above and 121 crosses below. At RSI 30, 46 and 46. The near-equality is arithmetic, not balance: a cross above has to be followed by a cross below before another can happen, so the two counts can never differ by more than one.
What the count does tell you is how often the level gets touched. Six crosses a year at RSI 70 against 2.3 at RSI 30 says SPY visits overbought far more readily than oversold, which fits everything Chapter 2 showed.
At RSI 70, the two sides are indistinguishable
Crossing RSI 70
Occurrences
5 days green
10 days green
20 days green
Upward, into overbought
120
62%
68%
68%
Downward, out of overbought
121
63%
69%
70%
Measured on SPY, 21 July 2026.
Entering overbought and leaving it were followed by almost exactly the same outcomes at every horizon. Within a point of each other, three times running, on more than 120 events each. Whatever information the 70 line carries, the direction of the crossing is not it.
At RSI 30, the conventional reading is inverted
Crossing RSI 30
Occurrences
5 days green
10 days green
20 days green
Upward, out of oversold
46
54%
61%
67%
Downward, into oversold
46
61%
70%
76%
Measured on SPY, 21 July 2026.
Crossing down into oversold was followed by more green than crossing up out of it, at all three horizons, by a widening margin. The traditional trigger is the upward cross; the measured history prefers the other one.
Two cautions before anyone acts on that. Forty-six events is a modest sample. And both columns are green-leaning at every horizon, which on an instrument that drifts upward is what you would expect regardless. The gap between the two columns is more informative than either number alone.
Common Mistake
Reading a cross column as a signal because of what the level is called. Overbought and oversold are names borrowed from a convention, not conclusions the data reached. On this ticker one pair was indistinguishable and the other ran opposite to the convention.
The tool has its own guided walkthrough, and it adapts to whichever view is open: