Trending pairs
A ratio with a drift term pushes concentrated liquidity to one side and keeps it there. Screened on cointegration, not correlation.
KO/PEP · V/MA · JPM/BAC · XOM/CVX · four pairs, no prices
Trade the ratio, not the price.
A KO/PEP pool is a pair trade with the plumbing removed. Sell KO into PEP and your book moves from one to the other in a single transaction — long the cheap side, out of the rich side, priced entirely by the ratio between them.
Nothing is borrowed, so nothing can be called. No funding leg to bleed, no margin to post, no liquidation price to defend. It clears like any other spot swap, because that is all it is.
KO/PEP0.6641+2.14σV/MA0.5975−0.37σJPM/BAC5.5681+0.92σXOM/CVX0.7248−1.61σKO/PEP0.6641+2.14σV/MA0.5975−0.37σJPM/BAC5.5681+0.92σXOM/CVX0.7248−1.61σSIMULATED FEED · OU PROCESS · NOT MARKET DATAKO/PEP0.6641+2.14σV/MA0.5975−0.37σJPM/BAC5.5681+0.92σXOM/CVX0.7248−1.61σKO/PEP0.6641+2.14σV/MA0.5975−0.37σJPM/BAC5.5681+0.92σXOM/CVX0.7248−1.61σSIMULATED FEED · OU PROCESS · NOT MARKET DATA
F1 · Reading a spread
The chart is the ratio.
There is no price on this screen. A pair trade has no price — it has a ratio, a mean, and a distance from that mean. So the y-axis is KO divided by PEP, the flat line through the middle is where that number has spent the last thirty days, and the dashed lines are one and two standard deviations either side of it.
A z-score replaces the APR because an APR would be the wrong question. You are not asking what this pays. You are asking how far from ordinary it is, and how long it has historically taken to get back.
Grab the trace and pull it — the screen degrades with the z-score
- Ratio
- 0.6641
- Mean 30d
- 0.6380
- Std dev
- 0.0122
- Z-score
- +2.14
- Half-life
- 6.2 d
- Correlation
- 0.86
The reference pair. Same shelf, same consumer, same input costs.
SIGNAL
KO is 2.14 sigma rich against PEP. Swap KO into PEP and you are short the spread. Expected reversion half-life 6.2 days.
Series are simulated with a mean-reverting process at each pair’s stated half-life and volatility. Nothing on this page is a live quote.
This chart is read-onlyTrade the same pairs in the terminalQuiet, quiet, quiet — two sigma — reversion. That is the entire trade, and it is why the chart is worth watching instead of the price.
F2 · The mechanic
Both legs, one transaction.
A pair trade normally takes four moving parts: a long, a borrow, a short, and a margin account that can go wrong at three in the morning. Here it takes one swap. Sell the rich side into the cheap side and the position exists. Sell it back and it is closed.
Your profit and loss is the ratio, and only the ratio. If both names fall twenty percent together, you are flat. That is the point of a pair trade, and it is the one thing perpetuals make expensive to hold.
- No funding. Nothing is being financed, so nothing accrues against you overnight.
- No liquidation. Nothing is borrowed, so there is no price at which someone else closes your position.
- No borrow to source. The short leg is not a short — it is simply the asset you no longer hold.
> SWAP 1,000.00 KO -> PEP
ROUTEUNISWAP V4 · KO/PEP · 0.05%
RATIO IN0.6412 (+2.14σ)
RECEIVED641.20 PEP
BOOK BEFOREKO 1,000.00
BOOK AFTERPEP 641.20
NET POSITIONSHORT THE KO/PEP RATIO
LEVERAGENONE
FUNDINGNONE
BORROWNONE
LIQUIDATION PRICEDOES NOT EXIST
SETTLEMENTSPOT, ONE TRANSACTION
> CLOSE: SWAP PEP -> KO AT ANY RATIO
P&L(RATIO OUT / RATIO IN) − 1, IN KO
F3 · Listings
The pair list is the product.
Anyone can deploy a pool. The work is deciding which two tickers belong in one, and the answer is almost never the two everybody names first. A pair earns a listing by reverting, repeatedly, with a half-life short enough to trade and a spread wide enough to pay the fee.
NVDA/AMD is the instructive rejection. The correlation is real and it is high, right up until one of them wins the cycle. There is no stable mean underneath it, so a ratio chart of that pair is not a spread — it is a scoreboard, and supplying a range against a scoreboard is how liquidity providers get run over.
| Pair | Thesis | ρ 90d | Half-life | Ratio | Z | Status |
|---|---|---|---|---|---|---|
| KO/PEP | The reference pair. Same shelf, same consumer, same input costs. | 0.86 | 6.2 d | 0.6641 | +2.14 | Trade |
| V/MA | The tightest ratio on the board. Two tolls on the same road. | 0.91 | 4.8 d | 0.5975 | −0.37 | Trade |
| JPM/BAC | Reverts on rates, drifts on credit. Widest bands we list. | 0.88 | 7.9 d | 5.5681 | +0.92 | Trade |
| XOM/CVX | One barrel, two balance sheets. The ratio is a refining spread. | 0.89 | 5.4 d | 0.7248 | −1.61 | Trade |
| NVDA/AMD | Correlated, not cointegrated. There is no mean to revert to. | 0.74 | — | — | +3.05 | Rejected |
Listing rule — a pair is admitted when its log ratio is stationary with a half-life under ten days and a spread vol the fee tier can pay for. Correlation alone is not enough and never was.
F4 · Liquidity
A tight range is a short vol position.
Concentrate liquidity in a narrow band around the mean and you have written straddles on the spread. Every oscillation through your range pays you a fee, and every one of them leaves you holding a little more of whichever side just got cheap. On a mean-reverting pair, that is the trade — the inventory you accumulate is the inventory you want.
v3 already proved the shape on stETH/ETH: two assets pinned to each other, ranges measured in basis points, impermanent loss thin enough to disappear under the fee income. Correlated equities are the same geometry with a wider band and a slower clock.
- Mean-reversion market making, in one deposit. No bot, no rebalancer, no off-chain loop.
- Impermanent loss is bounded by the pair, not the market. Both legs move together; only the ratio can hurt you.
- Fees compound on chop. Sideways is the good outcome, which is unusual and worth saying out loud.
- Capital efficiency
- 176×
- Time in range
- 45.1%
- Fully one-sided at
- ±1.15%
Efficiency is the v3 multiple against a full-range position on the same capital. Time in range assumes the ratio stays normal around its mean — which is the whole bet, and the only thing that can break it.
F5 · What breaks this
Structural decorrelation.
One risk matters more than the rest and it is not smart contract risk. It is a pair that stops being a pair. A management change, a lost patent cliff, a regulator that only touches one of the two names — and the ratio starts trending instead of oscillating.
When that happens the mean you were trading against is retroactively fictional. Traders lose on a spread that never comes back; liquidity providers get pushed out of range and hold nothing but the losing side. There is no hedge inside the pool for this. The defence is upstream, in the listing decision, which is why the pair list is the product and not an afterthought.
Half-life drift
Reversion speed is not constant. A pair that took six days last quarter can take thirty this one, and the fee has to survive the wait.
Underlying tokenisation
Everything here inherits the risk of the tokenised equity itself: issuer, redemption, corporate actions, and whatever the venue does on a halt.
Thin books at the edge
The moment a spread is most worth trading is the moment providers are furthest out of range. Depth at two sigma is the honest metric.
F6 · Specification
Why it can only be built here.
A ratio pool needs two tokenised equities on the same chain, settling in the same block, with the same custody assumptions. That is not a design choice — it is a precondition, and it does not exist on a chain without listed equities on it.
The mechanism is ordinary. Uniswap v4, concentrated liquidity, spot settlement. What is not ordinary is the pair of assets you can put in the pool.
- Chain
- Robinhood Chain
- Protocol
- Uniswap v4, concentrated liquidity
- Instrument
- Spot swap between two tokenised equities
- Quote convention
- Ratio A/B, never a fiat price
- Signal
- 30-day rolling z-score of the ratio
- Fee tiers
- 0.05% staples and payments · 0.30% financials and energy
- Leverage
- None. Positions are fully collateralised by construction
- Liquidation
- Not applicable
- Custody
- Non-custodial. The pool holds the tokens, not us
- Token
- $SPREAD