> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cashmere.exchange/llms.txt
> Use this file to discover all available pages before exploring further.

# How positions are run

> Which pool, how wide a range, and what is measured on-chain before capital moves.

## Three decisions

Running a liquidity book is three questions repeated: which pool, how wide a range, and when to
move. Each is answered from chain data rather than from a target yield.

<Steps>
  <Step title="Pick the venue">
    The same pair often exists at several fee tiers, and the busiest pool is not always the best
    one. A tier with a fifth of the volume but five times the fee rate can pay more per dollar
    deployed, and a small pool pays well until your own capital dilutes it. Every candidate venue
    for a pair is measured the same way: what the pool's own fee rate over the last day would pay
    on the capital being deployed, **after** accounting for the dilution that deploying it causes.
  </Step>

  <Step title="Set the range">
    A position earns only while the price is inside its range, and the liquidity a fixed amount of
    capital buys falls as the range widens. Fees per dollar therefore keep rising as a range
    narrows, without limit — so the fee-maximising range is always the tightest one, and it is also
    the one that spends most of its life out of range holding nothing but whichever side just lost.

    The range is set instead to the narrowest band that would have contained the price for **95% of
    the last week**, rounded outward to the pool's tick spacing. Tight enough to earn, wide enough
    to survive a week untouched.
  </Step>

  <Step title="Decide when to move">
    Repositioning is not free: it realises whatever impermanent loss has accumulated and pays a swap
    fee to rebalance into the new range. A position that is holding its range is left alone. One
    that has drifted out, or whose pool has changed character, is moved.
  </Step>
</Steps>

## What gets measured

Every number behind those decisions is read from the chain — pool state, price history, and fee
growth — and published on the dashboard rather than summarised into a single yield.

| Measure             | What it answers                                                          |
| ------------------- | ------------------------------------------------------------------------ |
| **Fee share**       | What fraction of everything the pool collects reaches this position      |
| **Earned APR**      | What the position has actually made over its own life                    |
| **Run rate**        | What it would make at the pool's fee rate of the last day, at this range |
| **vs. hold**        | Whether providing liquidity beat simply holding the two assets           |
| **Would hold**      | What fraction of the last week this range would have contained the price |
| **Price travelled** | How far the pool's price actually moved, sampled over a week             |

<Callout type="info">
  **Earned APR and Run rate are not meant to agree, and neither is the APR a venue shows you.** A
  venue's APR column is a property of the *pool* — its fees divided by its whole TVL — so every
  position in that pool shows the same number regardless of where its range sits. Run rate applies
  that same pool rate to *this* position's range. Earned is what actually happened. Early in a
  position's life the gap between them is mostly the gap between a short window and a long one.
</Callout>

## Fees are not the return

A position can show a large fee number and still be a loss. Fees only ever rise; the capital
underneath them moves both ways, and a concentrated position converts steadily into whichever asset
is falling as the price crosses its range.

This is why **vs. hold** is reported next to every position and why the two are never added into a
single figure. A pool that pays spectacularly is usually paying for volatility, and volatility is
paid for out of the capital. The book is sized accordingly: the pairs it holds are chosen because
their price travel is contained enough that a workable range holds for a week, not because their
headline yield is the highest on the chain.

<Callout type="warning">
  Past measurement is evidence about the future, not a forecast. Price history is sampled, so the
  travel measured over any window is a floor on the real travel and never a ceiling — a range that
  only just held last week is a range that breaks on the first week that is worse.
</Callout>

## Everything is verifiable

The positions are held by a treasury address that is published, the pools are public, and every
figure on the dashboard is derived from contract state at load time. There is no indexer to trust,
no database that can drift, and no reported number that cannot be recomputed from the chain by
anyone who wants to check it.