Beyond Leverage: What Else a Derivative Is For
In Read the Spec Sheet I spent twelve minutes on one reason people reach for derivatives, and it was the obvious one. Leverage. Put down $50,000, hold $150,000 of exposure, and watch the result move three times as fast in both directions (more than three times, if the contract happens to be inverse and you happen to be long, which was rather the point of that post).
Leverage is the reason derivatives get advertised. It's the number on the banner, the 100x and the 125x, the thing every exchange competes on. It's also, I'd argue, the least interesting thing a derivative does, because all it does is make a bet you could already place bigger.
A derivative is a contract whose value is taken from the price of something else. You hold the contract and you never hold the something else, and that separation is the whole trick. Once a price has been lifted off the thing it belongs to, it becomes a number, and numbers can be operated on. Leverage is multiplication. This post is about the rest of the arithmetic: turning the price upside down, changing the units it's paid in, and dividing one price by another.
Most of the post goes on that last operation, because it's where derivatives do something spot markets can only do clumsily, and because it ties back to something I wrote over a year ago. In The Long Shadow of Markets I described the secondary market effect, the habit the same underlying value has of disagreeing with itself when it's priced in two places at once. A price separated from its asset gets a market of its own, and a market of its own has a life of its own. We'll watch that happen inside a single exchange, between three prices that are arithmetically bound to agree and never quite do.
As before, there are no trade setups here; those are coming in their own series. This is about the mechanics, and about what the mechanics allow. The numbers carry over from last time where they can: BTC at $50,000, $50,000 of capital, and no fees, funding or slippage unless I say otherwise. ETH joins at $2,500, which puts the ETH/BTC ratio at a tidy 0.05.
The Price Without the Thing
Holding a bitcoin is a chore. There's the wallet, the seed phrase written on something that won't burn, and the small private terror of sending it to an address you've already checked three times. A perpetual contract on bitcoin is a row in an exchange's database. It moves with every dollar BTC moves, and you never touch a satoshi.
For crypto that's a convenience. The real reach of it shows when the underlying is something you couldn't hold even if you wanted to. In February 2026 Binance listed USDT-settled perpetuals on US equities, which trade around the clock while the stocks themselves only trade when Wall Street is open. In September it added foreign exchange perpetuals, starting with the dollar against the Brazilian real, at up to 100x. No brokerage account, no currency account, and no need for the underlying market to be open. You need a price feed and someone willing to take the other side.
That raises a question The Long Shadow would enjoy. What is a stock worth on a Sunday, when the only market trading it is a contract on a crypto exchange? Whatever the contract says, until Monday morning disagrees.
Keep the last clause of the paragraph before that in mind as well, the one about someone willing to take the other side. It comes back at the end.
Turning the Price Upside Down
To short something on spot you have to borrow it first. Someone has to own it, be willing to lend it, and charge you for the privilege, and they can want it back at an inconvenient moment. When BitMEX compared the two routes for shorting Ether in 2015, it noted that margin shorts on Poloniex and Kraken depended on other users lending out their Ether: no supply meant no short, and the interest rate swung about wildly.
On a derivative, long and short are the same button. Every contract has a buyer and a seller by construction, so nothing needs borrowing, and the short side exists the moment the contract is listed.
The cleanest demonstration I know of is also the oldest. Ethereum went live at the end of July 2015, and for the first weeks exchanges were slow to credit Ether deposits, so a great many holders physically couldn't sell what they owned. BitMEX listed a weekly futures contract, ETH7D, worth one Ether per contract, cash-settled in bitcoin against Kraken's ETH/XBT rate, at up to 5x. Because the margin was bitcoin, an early holder could short ETH7D and lock in their profit without depositing a single Ether.
ETH7D traded at a substantial discount to spot from day one. As each wave of deposits was credited and holders could finally sell, spot gapped down towards the contract. For those first weeks the most honest price of Ether was the one on a contract that never touched any. The secondary market effect usually has the derivative chasing spot; here spot did the chasing.
Notice the unit ETH7D was quoted in: bitcoin. The first derivative ever listed on Ether was a contract on the ETH/BTC ratio, two sections before we get to ratios properly.
Changing the Units
The last post ended its bear case on a position that does almost nothing, and that was its appeal. A 1x inverse short on BTC, opened with 1 BTC at $50,000, holds 2 BTC worth $50,000 if the price halves and 0.5 BTC worth $50,000 if it doubles. The coin count moves so the dollar value doesn't. It's a synthetic dollar built out of bitcoin.
Generalise that and you get the third operation. Every contract has a settlement currency, and the settlement currency decides what your profit is made of. A linear contract pays you in stablecoins, an inverse one pays you in the coin, and a holder can move between the two without moving the coin itself. A BTC holder who wants a dollar-stable fortnight doesn't have to sell and buy back later; they can short against what they hold and leave the coins where they are.
It sounds like a footnote to hedging. In the next section it decides whether you're trading a ratio at all.
Dividing One Price by Another
For years I've called these correlated products. The textbook name is cross pairs, or cross rates if you come from FX, and the textbooks have a point, since ETH/BTC is defined by the prices of ETH and BTC, which is a good deal stronger than being correlated with them. I'll keep the old name in my head and use the right one here.
Write the dollar price of ETH as
At $2,500 and $50,000,
Three Prices that Should Agree
Any large exchange lists ETHUSDT, BTCUSDT and ETHBTC side by side. Divide the first by the second and you get the cross the dollar markets imply:
The third market quotes its own number, and the two rarely match. Call the gap the drift:
Suppose ETHUSDT prints $2,500.00 and BTCUSDT $50,000.00, so the implied cross is 0.050000, while the ETHBTC book is quoting 0.050030. The drift is +0.06%. (Those numbers are illustrative. The real ones move every second and I'd rather you looked at your own screen.)
Why doesn't arbitrage close it? Because closing it costs money. The triangular loop runs: start in USDT, buy BTC, use the BTC to buy ETH, sell the ETH back for USDT. Each leg pays a taker fee
which rearranges to
Running the loop the other way covers negative drift. With a 0.1% fee on each leg the band is roughly ±0.3% before spreads, and inside that band nobody is paid to fix anything. The cross is free to wander.
The band isn't the same for everyone, either. A market maker on the lowest fee tier sees a narrow one, while a retail account sees a wide one, so the cross is held in place most tightly by whoever trades it most cheaply. Then there's the tick. Binance's ETHBTC perpetual launched with a tick of 0.00001, which at a ratio of 0.05 is 0.02%; the quote physically can't sit closer to the implied value than the nearest tick. And on perpetuals, each of the three contracts is tethered to its own index by its own funding rate, so three separate sets of crowd pressure decide how far each one leans.
This is the secondary market effect without the second market. Same venue, same moment, same two underlying coins, and three prices that are arithmetically bound to agree, each living its own life inside a band the fees have drawn around it.
What the Ratio Tells You
The drift is the ratio disagreeing with itself. The ratio's actual movement is a different thing, and the more useful one.
The ratio trends on its own, independently of either coin. Both can rally for months while the ratio bleeds, which tells you the rally belongs to BTC and ETH is being carried by it. Read alongside the dollar charts, it shows strength: if ETHUSD prints a new high and the ratio doesn't, the high came from bitcoin's tide lifting everything, and ETH did little of the lifting itself. It also shows positioning, meaning which way money is leaning between the two, and rotation from one into the other often shows in the ratio before it's obvious in either dollar chart.
A recent example puts numbers on it. CME reported that in Q3 2026 Ether returned roughly 70% and Bitcoin about 43%, and the ratio rose around 19% over the same quarter:
Subtract the two returns and you get 27 points. Divide the two growth factors and you get 19. Hold on to the gap between those two numbers, because the last part of this section is built on it.
The Ratio as a Consistency Check
This is something I used to do for my own trades, and I used to call it confirmation bias, which is close to the opposite of what it does. A confirmation bias only ever says yes. This test can say no.
Say I have a thesis that BTC reaches some target by a certain date, and a separate thesis for ETH over the same window. Each target, on its own, can look perfectly reasonable on its own chart. Divide them:
Then plot
Take BTC to $60,000 and ETH to $6,000 inside a month, starting from $50,000 and $2,500. Twenty per cent on bitcoin is a busy month. A bit more than doubling on Ether is a heroic one, but people have predicted heroic months before. Now divide:
The ratio would have to double in a month. Put that on the ratio chart next to anything it has done over comparable windows and the pair of theses falls apart, even though the BTC half on its own was unremarkable. Change the ETH target to $3,300 and the ratio only needs to go from 0.05 to 0.055, a 10% move over the month, which may well have precedent. Whether it does is the judgement, and "reasonable" keeps its quote marks for exactly that reason.
It has two limits. It tells you that something is wrong without telling you which leg, and it's only as good as your sense of how the ratio usually moves. It can also be run backwards: if you trust the BTC thesis more, the range the ratio could plausibly reach puts a ceiling and a floor on the ETH target that goes with it.
Five Ways to Own the Same Ratio
Now the practical question. You've decided ETH will beat BTC (or the reverse), and you want $50,000 of exposure to the race and nothing else. There are five ways to get it, and they don't pay the same.
Two scenarios, from the starting prices of $2,500 and $50,000:
- A: ETH rises 50% to $3,750 and BTC rises 20% to $60,000. The ratio goes from 0.05 to 0.0625, up 25%.
- B: ETH falls 50% to $1,250 and BTC falls 20% to $40,000. The ratio goes from 0.05 to 0.03125, down 37.5%.
$50,000 of ETH at the start is 20 ETH, which is also 1 BTC.
1. Spot ETHBTC. Swap 1 BTC for 20 ETH. If you keep score in bitcoin, scenario A leaves you with 20 ETH worth 1.25 BTC, a gain of 0.25 BTC, and B leaves you with 0.625 BTC, a loss of 0.375 BTC. If you keep score in dollars, you sold your bitcoin and bought Ether, so you're simply long ETH: +$25,000 or −$25,000. Spot is a ratio trade only for someone who already measures their wealth in BTC.
2. A BTC-settled ratio contract. This is ETH7D's descendant: BitMEX's later ETH/XBT quarterlies, settled in BTC and quoted in XBT, and Binance's ETHBTC perpetual, launched in April 2023 with BTC as its settlement asset and up to 75x. Binance files that one under its USDⓈ-M tab, and Multi-Assets Mode lets you margin it with USDT, so from the order screen it looks like a dollar product. The profit and loss arrive in bitcoin regardless. On 20 ETH of size, scenario A pays 0.25 BTC, which is $15,000 at $60,000, and B costs 0.375 BTC, which is −$15,000 at $40,000.
3. A dollar-matched pair. Long $50,000 of an ETHUSDT perpetual and short $50,000 of a BTCUSDT perpetual. In A the ETH leg makes $25,000 and the BTC leg loses $10,000, for +$15,000. In B the ETH leg loses $25,000 and the BTC leg makes $10,000, for −$15,000. If those numbers look familiar, they should.
4. Bybit's ETHBTCUSDT. Bybit listed it in June 2024, at up to 50x. It quotes the ETH/BTC ratio as if it were a USDT price, so a ratio of 0.05464 is quoted as 0.05464 USDT, and it settles in USDT. The ratio is measured in bitcoin and paid out in dollars, a design the industry calls a quanto. Sized to $50,000 of notional at a ratio of 0.05, A pays $12,500 and B costs $18,750.
5. CME's Ether/Bitcoin Ratio futures. Listed in July 2023, each contract is $1,000,000 times the ratio, cash-settled in dollars, so at 0.05 one contract is exactly $50,000 of notional (a coincidence I'll gladly take). A pays $12,500 and B costs $18,750. One quirk: it settles to CME's own Ether futures price divided by CME's own Bitcoin futures price for the same month, so the ratio it tracks is a quotient of two derivatives, one step further from spot than anything else on this list.
| Route | Paid in | Scenario A | Scenario B |
|---|---|---|---|
| Spot ETHBTC, scored in BTC | ETH held | +0.25 BTC | −0.375 BTC |
| Spot ETHBTC, scored in dollars | ETH held | +$25,000 | −$25,000 |
| BTC-settled contract (Binance, BitMEX) | BTC | +0.25 BTC (+$15,000) | −0.375 BTC (−$15,000) |
| Dollar-matched perp pair | USDT | +$15,000 | −$15,000 |
| Bybit ETHBTCUSDT | USDT | +$12,500 | −$18,750 |
| CME Ether/Bitcoin Ratio | USD | +$12,500 | −$18,750 |
| The ratio itself | +25% | −37.5% |
Only the last two rows track the ratio. Routes 2 and 3 agree with each other to the dollar, in both scenarios, and that's no coincidence.
Let
The dollar-matched pair holds
Identical. A contract that pays in bitcoin, once you value the payout in dollars, is the same trade as two dollar legs. Compare both with a contract that pays the ratio itself in dollars,
Check it against the table:
So the BTC-settled ratio trade is the pure ratio trade with its result multiplied by bitcoin's growth factor. Whatever you make or lose is a quantity of bitcoin, and its dollar value rides on BTC whichever sign it has: in A the profit grew because BTC rose, and in B the loss shrank because BTC fell. It's the inverse-collateral effect from the last post, turning up again on your PnL. And it's the gap from the Q3 figures. The pair, and the BTC-settled contract, would have made roughly the 27 points you get by subtraction,
You can make two dollar legs approximate the pure ratio, but only by resizing them continually as prices move, paying fees and spread each time. The settlement currency from the last section turns out to decide which trade you actually own. Only routes 4 and 5 hand you the ratio and nothing else.
What the Arithmetic Costs
Every operation in this post has a price, and some of them hide it well.
Funding is the first. A pair means two perpetuals and therefore two funding streams, and nothing stops both of them charging you at once: crowded longs on the ETH leg, crowded shorts on the BTC leg. A ratio perpetual has its own funding on top of whatever drift its index carries.
Leverage is the second, and on ratios it's sneakier than usual. Two coins that mostly move together produce a quotient that mostly moves less than either, so a 2% day on the ratio feels small. Bybit offers 50x on its ratio contract and Binance launched its own at 75x, and at 50x a 2% move against you is the whole of your margin. The quietness is exactly what makes the leverage look safe.
The third is correlation, which is the only reason the ratio is quiet. When the two coins decouple, the ratio moves like an outright position, and a dollar-matched pair can lose on both legs at once if ETH falls while BTC rises. Liquidity is the fourth. Cross books are thinner than either dollar book, and the tidy fee band from earlier ignores the slippage you'd pay getting a size of any substance through them.
The last cost belongs to the first operation, holding the price without the thing. A price on a contract exists only for as long as the ledger that records it. BitMEX, the exchange that listed ETH7D and taught the market the perpetual swap, announced in July 2026 that it was closing. New positions were blocked from 26 August, and any position still open on 23 September was force-closed, whatever its owner's thesis. That someone willing to take the other side from the first section was, in the end, the venue itself. A bitcoin in your own wallet doesn't care whether an exchange decides to close.
Which brings everything back to the spec sheet: the settlement currency, the contract size, the index, the funding, the venue. It's still the cheapest page in trading to read, and each operation here adds another line to it worth reading.
A bitcoin, held, does one thing. It sits there and is worth whatever it's worth. Its price, lifted off and written into a contract, can be multiplied, turned over, paid out in something else, or divided by the price of something else entirely. Each operation makes a new number with a market of its own, a little out of step with the thing it was taken from and a little out of step with its neighbours. Three prices on one screen, and the triangle never quite closes.