What a rebalance costs

A range move looks like an adjustment. It is a trade. When a vault pulls its liquidity out of a band, it does so at the price prevailing right then, in the ratio the pool has left it with — which is to say, holding more of whatever has fallen and less of whatever has risen. The move crystallises that. Every rebalance is forced selling, and it is the largest line on the bill.

The four costs

• 01Divergence, realised. The difference between the vault's inventory and a fifty-fifty hold, locked in at the moment of the move.
• 02Slippage. If the new band needs a different ratio, the difference has to be swapped, and the swap moves the price against itself.
• 03Gas. Small on an L2, never zero, and paid whether or not the move turns out to be right.
• 04Foregone fees. The seconds between removing and adding are seconds the capital is not working.

Against that sits one number: the fees the new band is expected to earn before the next move. Estimating it means estimating volume through the band and the share of the pool's depth the vault will own inside it. Neither is knowable. Both are estimable from the last few hours, which is what we do.

The rule we actually use

We rebalance when the expected extra fees clear the full cost with room to spare — not when the price crosses a line. This has an unglamorous consequence: in thin markets, or when gas spikes, the correct action is very often to do nothing and let the position sit slightly off-centre, earning less than it could, because getting to the better position costs more than the improvement is worth.

The most expensive rebalance is the one that was not worth making.

Every rebalance is written to the vault's log with the cost it paid. Over enough moves that log, and not a headline rate, is the only honest test of whether the rule is any good.