Research Note 09

KuCoin Earn: 486% APR and What the Rate Actually Buys

We announced a KuCoin Dual Investment lab after auditing Binance's AI trading skills, and asked a simple question: can a retail trader actually trade these offers profitably? This is the first answer. The headline rates are arithmetically honest, the payouts are a known class of options trade, and the menu contains no strategy at all. One product, Range Bound, priced worse than the historical frequency of the event it sells.

Status of this work: mechanics reconstructed, profitability not established. We read public data and priced it. We did not subscribe to anything, and we did not verify a single executed payout.

What we found

The highest rate in the catalogue, 486% APR, is a one day product paying 1.33% for the trade.

The BTC Buy Low ladders we examined are parameter variants of a familiar option-selling mechanism, not hundreds of new strategies.

Model implied volatility from those rates sits about 5 points below Deribit mark volatility on approximately matched BTC contracts.

Range Bound's own barriers, term and loss field imply break even odds that BTC's history did not meet in any of 156 live offers.

The number on the banner is a year, the trade is a day

Annual Percentage Rate normalises a payout to a year. Divide by the year and the headline shrinks fast.

The catalogue counts and APR distribution above use the 19 October 2026, 23:48 UTC snapshot. The Deribit, menu-anomaly and Range Bound checks use a separate snapshot around 20 October, 00:30 UTC. Interface screenshots were captured separately that morning; they illustrate specific offers rather than a synchronised catalogue.

The loudest figure in the snapshot, 486% on a one day ARB contract, pays 1.33% if it works. An earlier BTC quote of 181.24% APR for one day corresponds to 0.497% for the term. Nothing here is fabricated: multiply the rate by days and divide by 365 and the arithmetic is consistent with the annual scale. We did not verify an order preview for Dual Investment. On the Simple Earn promotion the daily figure is visible on the card itself, and 100 USDT at 100% APR shows 0.2739726 per day, which matches 100/365. But a rate is a scale, not a result, and the retail promotion sits on the scale rather than the result.

The same arithmetic applies to the promotional offers. A 100% APR campaign on 100 USDT for seven days pays about 1.92 USDT. Shark Fin at 1.99% to 19.15% APR over 92 days pays 0.50% to 4.83% for the full term, and the upper bound is conditional rather than fixed.

KuCoin's own support pages are less careful than its calculators. One help article ↗ computes interest as 1000 USDT at 20% for 10 days and prints 54.79 USDT. The correct figure is 5.48, ten times smaller. The Dual Investment FAQ ↗ links to Shark Fin in its order history section. The Russian product description says you convert when the market "reaches your target price", while the FAQ ↗ settles on an index averaged at the end of the term, so touching the strike mid term is not the condition at all. KuCoin's own playbook ↗ compares the product to a limit order, which executes on a different condition entirely. These are documentation defects. We found no evidence of a wrong payout.

We asked the API instead of the banner

KuCoin's documented API ↗ exposes these products at /api/v1/struct-earn/dual/products, and that endpoint requires a signed request. Its own web front end does not: the route it calls, /_pxapi/struct-main/v4/products/category-currency, returns the catalogue to an ordinary GET with no login and no keys. That is how the numbers in this article were produced, and anyone can repeat them.

A snapshot taken on 19 October 2026 at 23:48 UTC contained 3,741 products across 13 currencies: ETH, BTC, BNB, LTC, DOGE, XRP, SOL, ALGO, ATOM, NEAR, OP, ARB and ORDI, in 28 series, across 30 distinct terms from 1 to 80 days. Those counts are valid for that timestamp only. We re-requested the same route four times in twelve seconds and got byte identical counts, then saw the menu shift within minutes.

Two caveats we hold to. This route is an internal front end interface, not a documented API, so its path and schema can change without notice. And a product listed as available is not proof that it is executable for a given account, in a given region, at a given size.

Dual Investment is a covered option sale

Strip the naming and the payout is standard. Buy Low takes your USDT. If the settlement index finishes above your target, you get the USDT back plus the interest. If it does not, you get coins at the target price. Sell High takes your coins and mirrors it.

With C invested, a target K, a settlement index X and interest r for the term, the terminal value when coins are valued at X is

Buy Low = C(1+r) - [C(1+r)/K] × max(K - X, 0)

Sell High = Q(1+r) × X - Q(1+r) × max(X - K, 0)

That is cash secured put selling and covered call selling with a scaled notional. You are the seller of the option, and the interest is the premium.

This matters because it bounds what can happen. Take an offer captured in the interface: target 82,000, 73.04% APR and seven days. For an illustrative 1,000 USDT allocation, the interest component is equivalent to 14.01 USDT before conversion. Break even against your original 1,000 USDT sits at a settlement index of 80,867. At 80,000 the payout is worth 989. At 70,000 it is worth 866. A pretty coupon and a capital loss are fully compatible, and the coupon is the smaller number in any serious move.

Two things that are often said about this product and are both wrong. The loss is not unlimited: your downside is the capital you committed, because the position is collateralised rather than margined. And on the Sell High side what is unlimited is the forgone upside against simply holding the coin, which is not a cash loss and should not be reported as one.

Many contracts, a small number of mechanisms

Here is the part that speaks to the strategy question. We inverted each advertised BTC rate into the Black Scholes volatility that reproduces it, which is a descriptive transform rather than a reconstruction of KuCoin's own pricing engine.

Monotone, with the familiar upward skew as the strike moves away. The 14 day ladder behaves the same way, the 30 day ladder is flatter, and the BTC Buy Low offers we inverted sit between roughly 24% and 37% volatility. A smooth surface does not by itself prove a single pricing formula, and distinct contracts can also price smoothly. The contracts vary by strike and date within the same payoff family. This is a descriptive surface, not evidence that two fitted parameters explain every offer or that we recovered KuCoin's pricing algorithm. Other assets, terms and the Sell High side were not inverted here.

So the sophistication is real and it is in the wrong place. Snowball observes a barrier daily against an initial fixing set after you subscribe. Range Bound terminates on any breach at any moment. Discount Buy splits into three branches at a knock out price, and from the published branch definitions a higher BTC price can pay you less than a slightly lower one: we compute 1,032.68 just below the barrier against 1,002.47 at it. The data also carries a leverage field of 2 alongside the half capital branch, though the field alone does not establish the mechanism. These are genuinely non trivial payouts. None of them tells you when to take the trade.

That is the distinction worth keeping. There are payout formulas, in quantity. There is no method. The product converts a view you already hold into a contract and pays you for the risk, and the view stays entirely your problem. A catalogue of 3,741 offers is a parameter grid of known mechanisms, not a set of strategies. That is a claim about this shelf, not about everything KuCoin offers.

What the risk is worth elsewhere

If you are selling an option, the fair question is what similar risk costs on a market that quotes it. We compared the volatility implied by KuCoin's rates with Deribit's mark volatility on approximately matched BTC contracts, pairing within 600 USDT of strike and 1.6 days of term, using KuCoin's integer term label:

Median gap 5.5 volatility points in this snapshot, widening with term. Three limits on reading it. Mark IV is a model value, not a bid at which a client would sell the risk. The pairing is approximate, and a 1.6 day tolerance is loose on a one day contract. And the two settle differently: Deribit on its own index, KuCoin on an average over the 07:30 to 08:00 UTC window. So this says KuCoin's model implied volatility sits below the mark volatility of roughly comparable listed options, which is a direction, not a measured premium gap, and it was measured on BTC only.

A second, weaker check points the same way. Pricing each offer against Black Scholes on realised volatility, the median at the money ratio of actual to model premium is 0.65, with BTC at 0.92 and ETH at 1.00 but ARB at 0.44 and NEAR at 0.31. We treat this as a scenario benchmark rather than a fair value: it depends on the volatility window chosen, it ignores the averaged fixing, costs and the smile, and an earlier version of this calculation was wrong because it included an unclosed daily candle. The cross market comparison above does not depend on that choice, which is why we lead with it.

The honest summary: against both benchmarks the premium sits below them almost everywhere, with the smallest gap on BTC and ETH and the largest on the high volatility alts where the banner numbers are loudest. We are not claiming a proven underpayment, and the alt figures rest on the model benchmark rather than on the cross market check, which we ran on BTC only.

We could not construct a retail arbitrage from this. Selling expensive and buying back cheaper would need the opposite sign, and you cannot take the other side: the product only lets you sell the risk, and it cannot be closed before expiry. That is a statement about what we tried, not a proof that no arbitrage exists anywhere in these products.

Range Bound: the odds did not line up

Range Bound pays interest if the price stays inside a corridor and returns less principal if the corridor breaks at any moment. The size of that haircut decides everything, and the public data carries a field taking values 0.01 and 0.02. That it scales the risk is visible in the menu: doubling it doubles the rate exactly, from 318.18% to 636.36% and from 46.05% to 92.11%, while the break even odds stay put. We read it as the loss fraction, and that reading is the load bearing assumption of this section.

With a loss fraction L and a term return r, the corridor has to hold with probability at least L/(L+r) to break even before costs. In the separate 00:30 UTC snapshot, we took the actual barriers of each listed BTC offer, expressed them relative to the spot at the time of the snapshot, and applied those relative bounds to every historical entry point across 11,200 hourly candles, about 467 days.

The filtered API rows specify investCurrency: BTC. The table normalises to 1,000 units of deposited capital, not 1,000 USDT; it is a coin-denominated coupon/haircut scenario and does not include changes in the dollar value of BTC. The meaning of the loss field and the applicable valuation rules still need confirmation.

All 156 BTC offers in the menu came out negative on this conditional historical test, by 4 to 48 percentage points of probability. On the widest corridors the shortfall is around 10 points. On the narrow ones it is not close: a corridor needing roughly 50% held 1.6% of the time. A 1% loss can erase a coupon many times smaller than the headline annual rate suggests; it is an outcome for that term, not an annual loss forecast.

An earlier version of this calculation was wrong, and the correction matters enough to state. We first tested a symmetric corridor around each entry price, which throws away where the spot actually sits between the barriers. The numbers above use the real relative bounds. The direction survived the fix; several individual frequencies moved by up to 8 points, generally against the product.

What this is not. It is a frequency estimate on overlapping windows, not an independent test, and not a probability forecast. It is one asset over one period that included high volatility regimes. Hourly highs and lows approximate a continuous barrier; shifting the window length by an hour or stepping entries every six hours moves the frequencies by about a point, far less than the shortfall. We did not model early settlement, re entry, or the published formula tying order value to spot. And if the loss field means something other than the loss fraction, this section needs redoing rather than editing.

One real inefficiency, and it is not a trade

Checking every Buy Low ladder for monotonicity, since a strike further from spot carries less risk and must pay less, we found 7 violations among 1,622 offers. One NEAR contract quoted 72.06% APR against an 8.77% average of its adjacent rates, at a safer strike than its neighbour. The identifiers suggest an explanation rather than prove one: the anomalies carry older batch prefixes than the current offers, which is consistent with stale quotes the new curve has overtaken. We did not verify expiration and accrual times, availability, or simultaneous executability across the neighbours, and the comparison curve here is the average of the two adjacent rates, not a fitted fair curve.

This is a candidate relative improvement, not a demonstrated profit opportunity. Dominance within a menu does not establish positive expectation, and the available size is not returned by this route, so we cannot say any of it is executable. We also did not track how long such quotes persist. The practical value is narrow: when two offers differ only in strike and rate, do not take the strictly worse one.

The rest of the shelf, briefly

Dual Investment and Range Bound are where our numbers are. The other shelves follow the same pattern, a known mechanism plus terms, and they deserve their own article rather than a paragraph each. Our working map:

None of these product labels supplies a complete, independently tested rule for selecting profitable trades. All of them are terms attached to a mechanism that existed before the product page.

So who is this for

If you already hold the coin and would genuinely sell it at a given price, Sell High writes that intention into a contract and pays you for it. If you hold stablecoins and would genuinely buy at a lower price, Buy Low does the same. That is a legitimate use, and it is the same use a covered call or a cash secured put has always had.

What you give up is specific. You cannot exit the position. You can only choose strikes from the menu. You take the quoted price without negotiating it. Settlement runs on an averaged index rather than a price you can watch. The BTC comparison put KuCoin's model implied volatility below approximately matched Deribit marks. That is a benchmark observation, not an executable price for an identical contract or a measured fee for the wrapper. Our altcoin comparisons are model scenarios, not observed alternative sale prices.

Before any of these offers, the questions that decide the outcome are not about the rate. How much comes back in each branch, in which currency? When is the money actually available? Who is the counterparty and what are the fees? And what is the alternative you are giving up for the same capital over the same period?

Is KuCoin deceiving anyone? We found no evidence of incorrect payouts. The arithmetic checks out, the capital risk is disclosed in the fine print, and the BTC and ETH premiums were closest to our historical-volatility model benchmark among the examples shown. Only BTC was compared with Deribit marks. What we did find is marketing that leans on an annual scale for day long trades, documentation with real errors, and a slogan describing an obligation you are selling rather than a trade you are winning. Range Bound is where our own numbers came out worst, with the caveats above attached.

What we did not check

This article covers the Earn and Wealth shelves, and only what public data supports. We did not verify any executed payout, any order preview, or whether any listed offer is executable at size. We did not audit the quant funds: their published NAV history, fee schedule and the source of their multi year track record are still unexamined, and a reported APR that disagrees with a visible NAV change is a question, not a finding. We did not revisit grid bots, which we covered previously. And we did not study KuCoin's Skill Hub, which is a separate product from the Earn shelf, so nothing here should be read as a verdict on it, nor as a comparison with the Binance and CoinMarketCap catalogues we audited earlier.

Next in this lab: order previews without subscribing. One on a Dual Investment offer, for both branches, to check how the coupon would be credited and whether the card's rate survives to confirmation. A separate one on Range Bound, because only that product's own preview or contract can confirm what the loss field means, and the first preview cannot answer it.