Relative strength vs RSI: compare crypto pairs correctly

Relative strength compares one crypto asset with another asset or benchmark, commonly through a price ratio. RSI measures the balance of upward and downward changes within one chosen price series. A token can have a high RSI in USDT while losing value against Bitcoin. To compare them correctly, name the denominator, quote currency, observation times and calculation window before interpreting the number.

The word “relative” hides two different comparisons. In a ratio chart, it means A compared with B. Inside the Relative Strength Index formula, it means average gains compared with average losses. Treating these as the same measurement can turn a correct indicator reading into an incorrect account of performance.

Choose the comparison before the indicator

A price-ratio version of relative strength divides the price of asset A by the price of asset B. A rising ratio means A has gained against B over that interval. Fidelity describes this ratio construction and its direction of relative performance in its relative strength comparison guide.

For crypto prices in a common quote currency, define Q at time t as A's price divided by B's price. If A costs 100 USDT and B costs 50 USDT, Q is 2 B per A: one unit of A has the same quoted value as two units of B. A ratio above one says nothing about which asset is a better investment. Token denomination alone can change its level.

RSI starts with changes in one input series. Positive changes contribute to average gain; negative changes, expressed as positive magnitudes, contribute to average loss. RSI equals 100 minus 100 divided by one plus average gain divided by average loss. TradingView's RSI calculation documentation shows this construction with smoothed gain and loss averages.

You can also apply RSI to the A/B ratio itself. That oscillator describes changes in the ratio series. It is a different input from A/USDT and needs its own explicit label. Neither an RSI level nor the direction of a ratio supplies a probability of a profitable next trade.

Work through six matched crypto closes

This is an original hypothetical example, with fictional assets A and B rather than observed market prices. Both are spot prices quoted in USDT. Each row represents the same completed daily interval on an identical UTC calendar, with both observations available when the row is evaluated. There are no fees, distributions or changes to token units.

Rebase the ratio to 100 at Day 0 by calculating 100 times Q at the current day divided by Q at Day 0. Keep full precision in the calculation; the table rounds only for display.

A's USDT price increased by 6%: 106 divided by 100, minus one. B increased by 20%: 60 divided by 50, minus one. Rebased separately, A ends at 106 and B at 120. A ended higher in USDT while underperforming B over the same window.

The normalized ratio ends at 100 times 106 divided by 120, or 88.33. Its change is therefore −11.67%. One A bought the quoted equivalent of two B initially, but only 1.766667 B at the end. This is a price comparison, without an assumption that either conversion could be executed at those closes.

Now calculate a five-period RSI seed for A using exactly those five close-to-close changes: +3, −1, +3, −1 and +2. Total gains are 8 and total losses are 2. Divide each by five to get an average gain of 1.6 and average loss of 0.4. Their quotient is 4, so RSI is 100 − 100 ÷ 5 = 80.

The RSI of 80 and the relative decline of 11.67% are consistent. A's own gains outweighed its losses under this calculation, while B appreciated faster. The five-period setting and simple-average initialization are teaching choices. A platform with a longer warm-up or different smoothing history may display another value. The RSI context worksheet examines those implementation differences and unfinished candles in detail.

Keep the return gap separate from the ratio return

Subtracting 20% from 6% produces a gap of −14 percentage points. It does not produce the −11.67% return on A measured in B. These are two legitimate descriptions with different denominators.

If rA and rB are the two simple price returns in decimal form, the ratio return is (1 + rA) divided by (1 + rB), minus one. Equivalently, divide the return difference rA − rB by 1 + rB. Here that is −0.14 divided by 1.20, or −0.116667.

The distinction also matters in a falling market. In a second hypothetical endpoint comparison, A falls from 100 to 90 and B from 50 to 40. A loses 10% in USDT, while B loses 20%. The A/B ratio rises from 2 to 2.25, a 12.5% relative gain. The return gap is +10 percentage points. “Outperformed B” remains compatible with losing money in USDT.

For a report, save all three fields: A's quote-currency return, B's quote-currency return and the ratio return. Include the percentage-point gap only under that name. The simple and log returns guide extends this unit discipline to compounding across multiple periods.

Check when the common quote really cancels

Dividing A/USDT by B/USDT cancels the common USDT unit at each matched observation. This cancellation also holds after converting both prices into another common currency with the same contemporaneous conversion factor. It is an algebraic property of those synchronized converted values.

It fails as a shortcut when the inputs differ. Dividing an A/USD price by a B/USDT price leaves a currency conversion embedded in the result. Labeling both columns “dollars” does not remove it. Convert to a common quote at the corresponding times, retain that conversion series and inspect any missing observations. The prices' USDT returns also remain distinct from their USD returns when the quote currency moves against USD.

A ratio constructed from two closing prices is not automatically the executable A/B market price. Each close can be the last transaction at a different moment within its interval. A direct A/B order book, or two separate conversion legs, can involve spreads, fees, limited depth and price movement between fills. Use the order-book imbalance guide when the next question concerns displayed liquidity rather than historical relative price performance.

Choose one price definition for the comparison. Mixing A's trade-weighted average with B's last close changes what the ratio measures. The trade VWAP versus typical price explanation helps identify those inputs before they enter a comparison.

Align the observations before calculating strength

A common interval label is necessary for this example, but a date string alone is insufficient. Binance Spot's kline endpoint identifies bars by open time, returns open and close times, and allows interval interpretation in a specified timezone. Its documentation says request start and end timestamps remain in UTC regardless of that timezone setting. These are endpoint-specific details from the Binance candlestick documentation.

For a reproducible comparison, retain both instruments' venue, market type, price field, interval boundaries and completion state. Compare spot with spot unless a deliberate research question calls for mixing instruments. A perpetual-contract mark price and a spot close introduce different price definitions.

If B's newest close is missing, carrying its previous value forward while updating A creates a mixed-time ratio. Dropping the missing row instead changes the sequence of price changes used by RSI. A five-row window can then span a different elapsed period. Declare the missing-data policy before computing either metric, and retain the original timestamps. The multi-asset alignment guide provides selection and availability checks for this step.

Use a short comparison receipt

Before describing an asset as strong, fill in this record:

1. Write the question: performance against B, performance in the account's currency, or momentum within one series.
2. Name both instruments, venues, market types and quote currencies. Specify any currency conversion and whether a ratio is synthetic or directly observed.
3. Save the common start and end, completed interval boundaries, source prices and missing-row policy.
4. Calculate both price returns, the normalized ratio return and any percentage-point gap. Keep their units visible.
5. For RSI, record the exact input series, period, smoothing, initialization and sufficient preceding observations to reproduce the state.

Then run two quick checks. Multiplying every A price by a fixed positive unit-conversion factor should change the raw A/B level but preserve its normalized return. Reversing the ratio should reverse its economic direction, while its percentage return generally has a different magnitude: the inverse of this example's 88.33 ratio index implies a 13.21% gain for B measured in A, rather than 11.67%.

Finish the note with a sentence that preserves the distinction: “A gained 6% in USDT over five matched daily changes, lagged B by 14 percentage points, declined 11.67% in B terms, and had an RSI-5 seed of 80 under the stated method.” That sentence answers four precise questions without turning any observation into a trade instruction.