> ## Documentation Index
> Fetch the complete documentation index at: https://continuum-ec12e897.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Peg maintenance

> Why v2 venues self-peg, what bounds any mispricing, and the residual arbitrage the keeper still runs.

## Why v2 self-pegs

Every venue prices off the same NAV, so there is no external pool to drift:

* **Instant oracle swap** quotes NAV ± spread directly - it cannot diverge.
* **Committed orders** fill at the *next* oracle print - the price is set after the commitment, so there is nothing stale to trade against.
* **The books** are the only venue with a free-floating price between repositions, and three things bound it: the keeper re-pins the base to NAV whenever drift exceeds the per-market gate (20-40bps); any external arber lifting a stale bin pays the bin ladder's real slippage; and closing a one-sided position risklessly requires the *paired* round trip through mint/redeem, whose profit is only the convexity gap of `f(P) = P + K/P` net of the fee hurdle - near the anchor that gap is second-order and almost never clears the fee.

The result: mispricing on any venue is bounded by `max(reposition gate, half-spread) + fee hurdle`, enforced by structure rather than by a bot winning races.

## What bounds a stale book (the depth cap)

If price gaps inside one keeper cycle, an arber can lift whichever bins went in-the-money - and that is the *worst case by design*: the loss is capped at the deployed depth of those bins, the books are deliberately lean (see [Book management](/keeper/book-management)), and the next cycle re-pins and refills. This depth-capped extraction is the property that makes the books safe to run on high-volatility assets where an uncapped at-NAV venue could be drained.

## Residual keeper arbitrage

The keeper retains a small scanner for genuine residuals - e.g. when its own book drifts past the paired-arb hurdle during a fast move, the keeper itself is the natural counterparty to collect the gap rather than leaving it to outsiders (the profit lands in protocol custody either way). This is housekeeping, not a profit engine: in steady state the gates re-pin books before the hurdle clears, and the scanner mostly logs that there is nothing to do.