a new perp dex launches almost every week. most are forks of something older. some are genuinely new. a few are dangerous. the good news is that you do not need to be a contract auditor to filter out the worst ones. you just need to run a structured checklist before you fund. this is the exact list we use ourselves before we touch any new venue.
we built uponly. we route execution to a third party protocol (avantis on base), so we have very direct experience with how to evaluate a perp dex from the perspective of both a user and a product team that has to integrate with one. the list below is what we actually run.
1. what is the counterparty model
this is the first and most important question. when you open a position, who is on the other side of the trade. there are three common answers: a peer-to-pool model where liquidity providers take the risk, an orderbook where other traders are the counterparty, and a "we route to an underlying protocol" model where the venue is a wrapper on top of someone else. each is fine. each has different failure modes. but if a platform cannot give you a clear, one-sentence answer to this question, that is a hard stop.
2. is execution actually on-chain
some perp dexes execute trades on-chain. some run an off-chain orderbook and only settle on-chain. some are fully centralized with a non-custodial wallet sticker slapped on the marketing site. none of these are necessarily bad, but they have very different trust assumptions. the docs should tell you, in plain language, where execution lives. if the docs are evasive on this point, treat it as a yellow flag.
3. what chain is it on, and what are the actual contract addresses
every legitimate perp dex publishes the contract addresses of the core protocol. you should be able to click through to a block explorer and see the contracts deployed, ideally verified, ideally with a clear ownership and upgrade pattern. if the contracts are not published, or if they are published but unverified on the explorer, that is a meaningful negative signal.
4. how is the price oracle structured
every perp dex uses some form of price oracle to drive entries, fundings, and liquidations. you want to know which oracle. chainlink, pyth, an internal twap, a multi-oracle aggregator. each has different latency and manipulation characteristics. the worst possible answer is a single proprietary oracle controlled by the team with no public methodology. that is the kind of design that gets exploited or, worse, used adversarially against the user base in a stress event.
5. what are the fees, and when do they apply
open fees, close fees, borrow fees, funding rates, withdrawal fees. each of these is fine on its own. the question is whether they are documented clearly and whether they apply to losing trades. some platforms charge fees only on net winnings (uponly is one example). most charge fees on every trade regardless of outcome. neither is wrong. but you should know which one you are using before you fund.
6. what is the leverage band, and is it actually available
a lot of perp dexes advertise a leverage maximum that they never let you actually reach because of position size caps, asset caps, or risk overrides. before you fund, open the trade ui and try to dial leverage to the advertised maximum on a small position. if the slider stops at a meaningfully lower number, that tells you something about how the platform really works. for context on how to think about leverage at the top of the band, see max leverage trading 101.
7. how does the withdrawal flow work
fund a test amount, do a tiny trade, then withdraw. seriously, do this before you fund a real amount. the withdrawal flow is where bad platforms reveal themselves. you want to confirm that the funds leave the platform on-chain to your wallet without a kyc gate, a hold period, or a "review" stage that the docs did not mention. legitimate non-custodial perp dexes have instant withdrawals to your wallet.
8. is the team identifiable, and is there any track record
anonymous teams are normal in crypto and not automatically disqualifying. but you want some form of public record. a github with real commits over months. a twitter account that is more than three weeks old. a discord with users complaining about real product problems, which paradoxically is a healthier signal than a discord full of moonboy emojis. the absence of all of these is a flag.
9. how do they handle a stress event
search the protocol name plus terms like "liquidation cascade," "downtime," "oracle exploit," or "stuck position." every protocol that has been live long enough has had at least one event. what you are looking for is whether the team posted a public post-mortem, whether they made users whole if they should have, and how the community responded. silence in the aftermath of an incident is the worst possible signal.
10. what does the ux feel like for closing a trade quickly
this sounds trivial. it is not. the closing flow is where every perp dex shows you what it really thinks of its users. some venues hide the close button behind two confirmations and a slippage warning that you cannot dismiss. some sneak in a fee adjustment at the close stage. some have outright failed to close positions during volatility because their off-chain infra could not keep up. open a tiny position and try to close it. that fifteen second test is more informative than fifty pages of docs.
the bonus rule that most lists forget
read the front page carefully. specifically, read what the platform promises and what it does not promise. does it promise zero fees but bury "only on losing trades" in a footnote. does it promise the highest leverage but only on three obscure pairs. does it promise instant withdrawals but route through a centralized off-ramp. the front page is a contract with the user. trust the platforms whose front page is provably true after you have actually used them.
a working example, run end to end
as a worked example, here is how this checklist applies to uponly. counterparty: on-chain market on avantis, not us. execution: on-chain on base. contracts: avantis contracts are published and verified. oracle: pyth. fees: zero on open, zero on losing trades, small variable fee on net winnings. leverage: 75x to 500x defaults, the slider reaches the top. withdrawals: instant, non-custodial, no kyc gate. team: public team and public github. stress events: any incidents are posted publicly. close ux: one tap. you can verify each of those points yourself in under five minutes. that is the bar you should hold every perp dex to, including ours.
the final call
most new perp dexes will not pass this list. that is the point. the goal is not to be paranoid. the goal is to filter out the bottom 80 percent of new launches so that you can experiment safely with the top 20 percent. when you find a venue that passes, start tiny anyway. if you want to start with a venue that passes the checklist cleanly, try uponly and run the test withdrawal as the first thing you do. that is the right habit for any new perp dex.