Best perp house vaults in 2026

Perp House Vaults evaluated across how much has this vault already lost when a trader won — and can i check that number myself, today, for free?, when one trader's win is bigger than the pool, who pulls the plug — a rule i can read, or a vote i'll never hear?, who eats the loss before i do — and if i try to leave first, how much do they take?, and is the number they advertise what i actually earned, or does it quietly leave out what the traders took from me?.

#1 of 9 · published ranking

GMX GM / GLV pools

81ChainChoice Score

Why it leads

3 points ahead of Avantis avUSDC: +2.0 pts verifiable loss of record, +2.0 pts fee versus pnl honesty

Cost

Not priced · No comparable price is published

Below the table

How this ranking works

Everything the table draws on continues here: how firm the #1 is, the per-criterion arithmetic behind each score, who pays ChainChoice, and the full guide to choosing.

Direct answer

What is the best perp house vaults in 2026?

GMX GM / GLV pools ranks #1 overall for perp house vaults on ChainChoice. Risk-isolated per-market house vaults where the LP is the traders' counterparty, protected by a named automatic on-chain deleveraging rule (MAX_PNL_FACTOR_FOR_ADL) rather than a human vote, with a full-life realised-PnL series retrievable keylessly but only by assembling it yourself. It holds that rank under an affiliate-blind methodology scored across 4 published, weighted criteria — the code that ranks providers physically cannot read affiliate payouts (CI-enforced), so a payout can't move a rank. The verdict re-computes on every fee change, incident, or regulatory action; full reasoning and the audit receipt are below.

Best picks

Best perp house vaults in 2026

Weighted on a verifiable record of losses already taken (32), what stops a loss bigger than the pool (28), who eats it before you do and what leaving costs (24), and whether the advertised yield separates fees from trader PnL (16). The heaviest axis contains no 9 and no 10, because not one of the nine publishes a dated, sized account of a loss it has absorbed — the ceiling is 8 and two products hold it. The finding that should change a deposit: every junior tranche that could be sized was at or below 1.21% of the senior capital it claims to protect, or already gone, and only three of nine document any capital beneath the depositor at all. Advertised APY is shown as a fact and never scored.

Best overall

GMX GM / GLV pools

Data checked Aug 2026

Risk-isolated per-market house vaults where the LP is the traders' counterparty, protected by a named automatic on-chain deleveraging rule (MAX_PNL_FACTOR_FOR_ADL) rather than a human vote, with a full-life realised-PnL series retrievable keylessly but only by assembling it yourself. Strongest on when one trader's win is bigger than the pool, who pulls the plug — a rule i can read, or a vote i'll never hear? (9/10): Read character-exact on 2026-08-17 at https://docs.gmx.io/docs/trading/liquidations.md: 'Auto-Deleveraging (ADL) protects pool solvency by automatically reducing profitable positions when the ratio of pending PnL to pool value exceeds the market's configured `MAX_PNL_FACTOR_FOR_ADL` threshold. Weakest on who eats the loss before i do — and if i try to leave first, how much do they take? (6/10): All read character-exact on 2026-08-17 at https://docs.gmx.io/docs/providing-liquidity.md. FIRST-LOSS IS STATED PLAINLY, under a heading literally called 'Risks': 'Counterparty risks: The GLV / GM pool is the counterparty to traders. Published price: liquidity providers earn the majority of the fees generated from trading, liquidations, borrow fees, and swaps (63% on Arbitrum and Avalanche) ... On Arbitrum and Avalanche, 63% of collected fees go to the pool and 37% go to the protocol.

Best for: Pool protection mechanism — 9/10

Why this score4 published criteria · leads 3 of 4

Published criterionWtScore, and the best hereGap/10Pts

How much has this vault already lost when a trader won — and can I check that number myself, today, for free?7.7·88.2

When one trader's win is bigger than the pool, who pulls the plug — a rule I can read, or a vote I'll never hear?6.7·98.1

Who eats the loss before I do — and if I try to leave first, how much do they take?5.8−464.6

Is the number they advertise what I actually earned, or does it quietly leave out what the traders took from me?3.8·94.6

Σ methodology points25.5/32

Each bar is the score on that criterion’s own 0–10 scale, never rescaled to the pool. The dark line is the best any product here reached on that axis. Wt is the most the criterion can add to the 86-point weighted total. Pts is weight × score × 32; the sum is the methodology score, and each weighted point behind the leader costs 2.6 on the displayed score. how these are weighted

Considered and not ranked

10 products we looked at and left out

A shortlist is only honest if it says who it turned away. Each of these was assessed against the same published criteria as the ranked table and excluded for a stated reason — not overlooked.

We assessed 14 products here and rank 4 — 29% of what we looked at. That share is of the products we assessed, not of the category: how many exist is not something we can count, so we do not claim a number for it.

GMX GLP (V1)Vertex VLP· No longer operatingElixir· No longer operatingAark· No longer operatingParadex VTFsLighterExtended (ex-X10) Community VaultHyperliquid user vaults (as distinct from HLP)Drift Insurance Fund / DLP under the Drift nameSynthetix sUSD debt pool· Ranked elsewhere

Why it ranks first

Why GMX GM / GLV pools leads this category right now

Best for

Pool protection mechanism — 9/10

Main tradeoff

The record you need exists and is free, but every convenient route to it is closed and the one number GMX advertises can flatter you. GM tokens — the thing you actually hold — have no ticker on the public price API (2 of 127 tokens are GLV, none are GM), the keyless GLV candle series reaches back only 267 days when the same endpoint serves 1,758 days for ETH, and GMX's own fee-versus-PnL breakdown endpoints (marketsAprByPeriod, marketsPnlAprByPeriod, marketsFeeUsdByPeriod) return 'not allowed' to an outside caller. So checking what your pool has lost to winning traders means writing GraphQL against a 135-market indexer in 10^30 fixed point with an undocumented sign convention. Meanwhile the advertised 'Performance APY' is excess return over a geometric-mean benchmark, not a plain return: a pool can advertise positive annualised performance in a quarter when your position lost USD value, because the 50/50 benchmark fell further. And when the protection finally fires it fires on the trader, not for you — ADL closes someone's winning position without warning, which is what keeps you solvent, but there is no page anywhere in 126 that tells you the last time it had to.

Verify before signup

liquidity providers earn the majority of the fees generated from trading, liquidations, borrow fees, and swaps (63% on Arbitrum and Avalanche) ... On Arbitrum and Avalanche, 63% of collected fees go to the pool and 37% go to the protocol.

Recommendation summary

What should decide this category

What has this vault already lost when a trader won, and can you verify it yourself?

When one win is bigger than the pool, does a rule fire or does a person decide?

Is there real capital beneath you, and what does leaving cost when everyone leaves at once?

Quick picks

Strong options in this category

Start with the lead choice first, then use the shortlist only if you still need a challenger or stronger fit for a specific setup.

Best overall

GMX GM / GLV pools

Best for: Pool protection mechanism — 9/10

How much has this vault already lost when a trader won — and can I check that number myself, today, for free? · 32%

8/10

When one trader's win is bigger than the pool, who pulls the plug — a rule I can read, or a vote I'll never hear? · 28%

9/10

Who eats the loss before I do — and if I try to leave first, how much do they take? · 24%

6/10

Is the number they advertise what I actually earned, or does it quietly leave out what the traders took from me? · 16%

9/10

Quick pick

Avantis avUSDC

Single ERC-4626 USDC vault on Base that is the counterparty to every Avantis perp trade, protected by a published per-trade profit cap of 2500% of collateral and a solvency-indexed withdrawal-fee schedule. Strongest on who eats the loss before i do — and if i try to leave first, how much do they take? (10/10): All three elements the top band requires are present verbatim, and unlike the Ostium precedent the junior layer is SIZEABLE. Weakest on is the number they advertise what i actually earned, or does it quietly leave out what the traders took from me? (7/10): The headline is a REALISED number, not a fee slice, and I verified that rather than taking it on trust. The docs advertise 'Realized return inception-to-date is 35%.' and a spec-table row 'Return, Inception to date | 35%'. Published price: Deposits are free and there is no documented management or performance fee; the only price the depositor pays is a solvency-indexed withdrawal fee.

Best for: Loss waterfall and exit price — 10/10

How much has this vault already lost when a trader won — and can I check that number myself, today, for free? · 32%

7/10

When one trader's win is bigger than the pool, who pulls the plug — a rule I can read, or a vote I'll never hear? · 28%

9/10

Who eats the loss before I do — and if I try to leave first, how much do they take? · 24%

10/10

Is the number they advertise what I actually earned, or does it quietly leave out what the traders took from me? · 16%

7/10

Quick pick

Ostium OLP

The pool's only documented junior tranche — and, post-upgrade, barely a house vault: OLP now earns opening fees only, with the real counterparty an offchain Jump Trading mirror hedge. Its loss record is fully re-derivable from a keyless Ostium subgraph the docs never point depositors at, while the operator's own pages stay silent on the July 2026 vault drain. Strongest on how much has this vault already lost when a trader won — and can i check that number myself, today, for free? (8/10): CORRECTION TO THE BRIEF — the 3-4 band placement is wrong, and I can show why. The brief scored Ostium at 3-4 on the strength of the overview page routing APR to Dune. Weakest on who eats the loss before i do — and if i try to leave first, how much do they take? (7/10): TRANCHE HALF: best in the pool, unambiguously. On https://docs.ostium.com/vault/overview.md, character-exact: 'In structured-credit terms, OLP is the senior tranche and the buffer is the equity tranche beneath it. Published price: Opening fees split between two destinations: 30% flows to OLP holders at each daily settlement, and the remainder funds protocol operations and development. The OLP allocation is a tunable protocol parameter.

Best for: Verifiable loss of record — 8/10

How much has this vault already lost when a trader won — and can I check that number myself, today, for free? · 32%

8/10

When one trader's win is bigger than the pool, who pulls the plug — a rule I can read, or a vote I'll never hear? · 28%

8/10

Who eats the loss before I do — and if I try to leave first, how much do they take? · 24%

7/10

Is the number they advertise what I actually earned, or does it quietly leave out what the traders took from me? · 16%

8/10

Quick pick

gTrade gToken Vaults (gDAI / gUSDC)

The longest-running pure house vault: ERC-4626 gToken shares whose price is literally 1 + cumulative fees - cumulative trader PnL, both accumulators readable free onchain, with a withdraw lock that lengthens to 3 epochs exactly as the collateralization ratio falls - and gUSDC sits at 99.23% today. Strongest on when one trader's win is bigger than the pool, who pulls the plug — a rule i can read, or a vote i'll never hear? (7/10): THE HEADLINE FINDING: gTrade USED to have the exact protection Avantis is benchmarked on, and has published that it removed it. Weakest on who eats the loss before i do — and if i try to leave first, how much do they take? (6/10): PLACE IN THE LOSS ORDER: the depositor is senior to a named, sized, onchain-checkable buffer - but that buffer is currently gone. Published price: Vault (Liquidity Providers): 15% for underwriting trades

Best for: Pool protection mechanism — 7/10

How much has this vault already lost when a trader won — and can I check that number myself, today, for free? · 32%

6/10

When one trader's win is bigger than the pool, who pulls the plug — a rule I can read, or a vote I'll never hear? · 28%

7/10

Who eats the loss before I do — and if I try to leave first, how much do they take? · 24%

6/10

Is the number they advertise what I actually earned, or does it quietly leave out what the traders took from me? · 16%

7/10

Frequently asked

Questions people ask before choosing house vaults

Is this not just a yield vault?

No, and the distinction is the reason this page exists. A yield vault earns from lending, staking or market making. A house vault takes the OTHER SIDE of leveraged traders — when they win, you pay. That produces two questions no yield product has to answer: how much has this vault already lost to a single trader's win, and can anyone force-settle a market to protect it. Both are N/A for seven of the eight products in yield_vaults, and that pool's own methodology marks market making down rather than scoring it. The one house vault ranked there sits as 1 of 9 passive-yield products; here it is ranked as what it actually is, and lands mid-table.

Everyone quotes an APY. Why is that not the ranking?

Because it is the number that moves fastest and explains least, and in this category it is frequently not what you earned. Seven of the nine contradict their own live data somewhere. The deeper problem is blending: a headline that mixes fees collected with trader PnL absorbed tells you nothing about which part is a durable spread and which part is you being on the losing side of a directional bet. Only one of the nine publishes a numeric per-trade profit cap. APY is shown here as a fact with its source and date, and it is scored nowhere.

Why weight the loss record so heavily if nobody publishes one?

Because the absence is itself the finding, and it is measurable. The 32% axis has a ceiling of 8 and no 9 or 10 anywhere in the pool: across seven documentation corpora totalling more than a thousand enumerated pages, there is not one incident page. What separates the products is whether the loss can be RE-DERIVED by you — some expose keyless endpoints where the drawdown is reconstructable, and others do not. Three of the nine never state that principal loss is possible at all. A category where nobody discloses is exactly where a buyer needs the scoring to be strictest.

Why is the biggest vault in the category not ranked first?

Because size is not seniority. The largest house vault in existence holds over $768m and scores 3 out of 10 on what stops a loss bigger than the pool — below four smaller products. Scale tells you how much capital trusts a venue; it says nothing about what sits beneath you when one trader wins big, what the operator may do to a live market to protect the pool, or what it costs to leave while everyone else is leaving. Deposits are shown here as a dated fact with their source, and scored nowhere.

Free advisorNo signup needed

Still unsure? Get your best house vaults pick

Answer a few quick questions and get one clear recommendation based on how you actually plan to use crypto — then review the evidence before deciding.

No signupFree first pass · PrivateNo spam · No account

Not financial advice · Independent · Always do your own research

Browse this network

How this ranking is built

Reviewed on how much has this vault already lost when a trader won — and can i check that number myself, today, for free?, when one trader's win is bigger than the pool, who pulls the plug — a rule i can read, or a vote i'll never hear?, who eats the loss before i do — and if i try to leave first, how much do they take?, and is the number they advertise what i actually earned, or does it quietly leave out what the traders took from me?.

Data checked Aug 2026 · Independent rankings · We show our work

Not financial advice · For informational purposes only · Always do your own research