ManticoreExit liquidity as a mandate

Every large holder of an illiquid token faces the same problem: the price on the screen is the price of the next dollar out, not the price of theirs. Selling size into an automated market maker is a mechanical transfer of value from the seller to everyone standing on the other side, and the size of that transfer is knowable in advance, exactly, from the pool invariant.

Manticore is a liquidity desk built around a single observation: a seller with patience is not a taker. They are a maker who has not been paid yet. This paper describes three mandates that convert a position which has to be sold into liquidity that gets paid for standing there, the arithmetic behind each, and — at equal length — the conditions under which each one is the wrong thing to do.

Take a constant-product pool holding x units of a token against y units of quote, with invariant k = x · y and spot price p = y / x. A seller pushing Δx tokens through the pool receives the quote the invariant gives up:

Define τ as the size of the sale measured against the depth standing against it — that is, the position's value at spot divided by the quote-side depth of the pool. For a balanced pool, x · p = y, so τ = Δx / x. Substituting collapses everything into a function of that one number:

The consequences are unintuitive until they are written down. Selling a position equal to a quarter of the quote-side depth costs a fifth of it to slippage and leaves the price 36% lower. Selling a position equal to the depth itself costs half, and leaves the price down 75%.

Splitting the sale into pieces does not escape this. The invariant does not care about the number of transactions; a TWAP over a day is the same arithmetic run in slices, plus the risk that the market notices the pattern and front-runs the remainder. What actually changes the answer is not slicing the trade — it is changing which side of it you are on.

A limit order and a single-sided liquidity position do the same thing: they offer to sell an asset above the current price. They differ in what happens while nobody is buying.

• A limit order that is not filled earns nothing. It is a free option written to the market, and the market exercises it precisely when it is worst for the writer.
• A liquidity position that is not filled earns fees on every trade that transits its range — including trades that come up, cross, and go straight back down without ever consuming it.
• Both fill at the maker's price rather than the taker's. Neither pays slippage; both are paid it.

Discretised liquidity — DLMM bins, or v3 ticks — makes the comparison exact rather than approximate. Liquidity deposited in a bin above spot is entirely in the token being sold; it needs no quote capital to establish. When price rises through that bin, the pool swaps that token for quote at that bin's price. Functionally it is a rung of a limit ladder that has been collecting rent while it waits.

The cost of the position is symmetrical and should be stated plainly. A maker is compensated for taking the other side of flow, and flow is on average informed. In a trending market, a maker sells the asset all the way up or buys it all the way down — which is the entire content of divergence loss, and the reason mandate II is gated on regime rather than run continuously.

The position is deposited as one-sided token liquidity across a geometric ladder of bins, from a floor a few percent above spot to a ceiling set by the mandate. Nothing is placed at or below market, so nothing can be filled by a sell-off.

If price reaches R, every bin below R fills and the realised average is the weight-weighted mean of those bins' prices. Bins above R do not fill; that portion of the position is still held, and still in the token.

Distribution shape

Weight can be leaned toward the near bins or the far ones, and this is the only real parameter choice in the mandate. Front-loading fills sooner at a lower average. Back-loading raises the average and lowers the probability of filling at all. There is no setting that improves both; the mandate's job is to pick the one that matches why the client is selling.

Sweep discipline

Filled rungs are swept to quote and not redeployed into the same range. Without this rule a retrace runs the ladder backwards, buying the token back with the proceeds and undoing the exit — which is the single most common way a well-designed ladder turns into an accidental range position.

Failure modes

1. 01Price never reaches the floor. Nothing fills. The client remains long, having collected fees on whatever volume transited below — the same outcome as an unfilled limit ladder, plus the fees.
2. 02Price gaps through the whole ladder. Everything fills at the mandate's average and the market trades higher without the client.
3. 03The client needs proceeds on a date. A ladder is a price commitment, not a time commitment. Forcing it early means market-selling the remainder and paying the arithmetic in Section 02.

A two-sided position in a deliberately wide band around spot, held only while the token is going nowhere on real volume. Fees accumulate roughly linearly with volume; divergence loss is a function of how far price ends up from where it started.

Those two expressions are the entire mandate. One grows with activity, the other with displacement, and a token can have a great deal of the first with almost none of the second. That combination is what the desk is looking for, and it is not rare — it is simply boring, which is why manual liquidity providers do not sit in it and do not leave it in time.

Width

Concentration multiplies fee income and divergence loss by approximately the same factor, 1 / (1 − (lower/upper)^¼). A tight band is not a better trade — it is the same trade with the volume turned up, and it exits the range faster. Width is therefore set against realised volatility and rebalance cost, never against a target yield.

Pull, do not re-centre

When the regime breaks, the position is pulled to spot rather than re-centred around the new price. Re-centring realises the loss and immediately re-establishes the same exposure one step further into the trend, which is how range positions become unintentional directional ones. This is the mandate's most important rule and the one that is hardest to follow while it is happening.

A constant-product position has a closed-form delta. Its value is V(p) = 2·√(k·p), so the exposure to price is:

Shorting that delta on a perpetual removes most of the direction and leaves the fee stream. The mathematics is settled and has been for years. The reasons this desk has not deployed it are operational, and they are the reasons that matter.

1. 01Funding is a tax set by whoever is most levered that week. It is not constant, it is not always in the hedge's favour, and over a holding period long enough for the fees to matter it will change sign.
2. 02Most tokens where mandates I and II are useful have no perpetual with meaningful depth. Hedging into a thin perp reintroduces the exact price impact the desk exists to avoid.
3. 03Delta drifts as price moves, so the hedge must be re-struck continuously — and the re-hedge loop lags hardest precisely during the fast move it exists to protect against.
4. 04The hedge leg can be liquidated; the LP leg cannot. Combining them creates a failure mode that neither position has alone.

Placing liquidity is trivial. Deciding whether to be placed at all is the product. The engine monitors four things and produces one output: in, or flat.

• Regime — is realised movement bounded relative to volume, or is a trend developing? Mandate II is only ever live in the first case.
• Depth — what share of the liquidity standing in the traded range belongs to this position? A fee tier is meaningless without it, and it can collapse without the price moving at all.
• Reach — how far has strength been carrying? This sets mandate I's ceiling, which is the difference between an average that leaves money on the table and a ladder that never fills.
• Flat — the default. Most tokens do not deserve liquidity most of the time.

None of this is a prediction engine and it is not presented as one. It is a set of conditions for being in the market, applied consistently, including on the days when applying them is annoying. The alpha is not in knowing where price is going. It is in leaving before the answer becomes obvious to everyone else.

Divergence loss

Quantified in Section 05 and unavoidable in any two-sided position. It is not a fee and cannot be hedged away without importing the problems in Section 06. It is the cost of the fee income, and in a sustained trend it is unbounded.

Adverse selection

A maker is systematically on the wrong side of informed flow. Fee income is compensation for this, and whether the compensation is adequate is a question about the specific token and week, not a general property of liquidity provision.

Non-execution

Mandate I can end with the position unsold. This is a real outcome, not an edge case, and any client whose exit has a deadline should treat it as the base case rather than the tail.

Smart contract and venue

Liquidity sits in third-party pool contracts on a third-party chain. Contract failure, venue failure and chain halts are outside the desk's control and are not insured against.

Execution and operations

Bins must be placed, swept and pulled by software that can fail, mis-sign, or lag. Key management, transaction landing under congestion, and the sweep loop are all operational risks that sit with the desk and are not eliminated by the mandate design.

The token launches on pump.fun with no allocation to the desk, no team unlock and no private round. The entire supply enters as public float on the bonding curve.

1. 01The desk buys 20% of supply on the open market, at the same prices and with the same slippage as anyone else. This is an acquisition, not an allocation, and it is the only mechanism by which the treasury ever obtains supply.
2. 02That 20% is locked for thirty days. During the lock it cannot be sold, moved, deployed as liquidity, or voted with.
3. 03At unlock, the entire position is airdropped to holders. It does not return to the treasury and it is not sold into the book.
4. 04On an ongoing basis, creator rewards accrued through the pump.fun rewards programme are used to buy supply back off the market. Bought supply lands in the treasury, where its disposition is decided by vote.

Governance here is deliberately narrow. Holders decide what happens to treasury supply. They do not set strategy parameters, approve client mandates, or direct desk operations — those are underwritten by the desk and cannot be voted into or out of existence by people who do not carry the risk.

The desk drafts the proposals and holders choose between them. This is not a hedge against holder judgement; it is a constraint on the ballot. Every option that reaches a vote is one the desk has confirmed it can actually execute, which is the difference between a governance process and a suggestion box.

This document describes mechanisms. It is not an offer, a solicitation, investment advice, or a projection of returns, and it does not constitute a track record.

• Every numerical figure on this site is computed from published automated-market-maker formulas — the constant-product invariant, the standard divergence-loss identity, and the concentrated-liquidity amplification approximation — and is shown to illustrate a mechanism.
• No performance history, assets under management, or realised return is claimed anywhere on this site, because none exists to claim.
• Digital assets are volatile. Providing liquidity can lose money, including through divergence loss, adverse selection, contract failure, and total loss of the underlying asset.
• Mandate III is documented and not deployed. No client capital operates under it.