October 10th: The Day ADL Broke. What the Crash Revealed About Perp DEX Risk Models

Transcript

Hey guys. Um I'm I'm PB. I work for Purple. Purple is a fully on-chain perp decks on Monad. Um it's a central limit order book uh very similar to what you see in Hyperliquid. Um the difference here is we don't run the underlying chain. So, Monad is L1. Uh we're purely a protocol uh EVM perp decks. My talk is about ADL and um what happened October 10th and uh some of the things that we're doing to basically protect traders from uh wrongfully being liquidated and ADL'd.

So, just a quick [clears throat] review. Liquidation versus ADL. So, in a liquidation, what's happening is um the loser is being removed from the system uh and replaced with a different trader. So, usually that other trader is also in the order book and they're matched against the winner. Um what happens here is other traders are not affected other than the loser being removed. In an ADL situation, uh there's no other resting orders on the book. Um so, you can't replace the loser. So, uh the winning side is also removed. Um so, both sides are removed from the system.

Uh this is a pretty drastic measure and the reason why you have to do this is if you don't ADL people out, the protocol becomes insolvent and at that point, you're basically unable to use the exchange. Um so, the invariants that the protocol is built on, i.e. like winners and losers get paid out, um that invariant doesn't get held. So, it's a very important reason why ADL exists. Um >> [clears throat] >> So, a big part of uh liquidation mechanisms like how perp decks is do margin.

Um the simplest one is isolated margin where every winner and trader brings their margin uh and then the most common one that you see is cross margin where you're able to share your profits and losses across different uh markets. So, for example, if you're long Bitcoin and short ETH and your long is up 50k and your short is down, let's say, 25k, the protocol nets the profits and losses out. So, you're not forced to close out the losing position and you can hold it longer.

Um traders do this for multiple reasons uh for collecting funding, for uh doing some sort of spread trade. Um a lot of reasons are there for these types of mechanisms. Um the thing with isolated margin is uh in that situation I described where you have a profit in Bitcoin and a loss in ETH, um you would be forced to close out some of the profits in Bitcoin to hold your ETH position. So, you would take some profits, move that over, move the PNL over to uh your ETH position to make sure both positions stay open.

>> [clears throat] >> Now, this all seems fine, especially cross margin. It's a nice way for people to keep positions open, but the issue comes up when there are these market crashes and October 10th is a great example. Um so, people are putting on spread trades or funding uh trades where they're long Bitcoin, short ETH. Um but what happens is when the short ETH position, which is basically in profit when the market's crashing, uh ADL's out because a counterparty needs to be uh replaced, but there's no counterparty to add.

Uh the short position gets closed out. Then the long position, which was being hedged with the short, uh is also liquidated out. So, you this this trader who technically was not risky to the protocol because they had a long and a short and they were hedging themselves, are wrongfully liquidated. And this was a complaint that a lot of traders had uh during October 10th where they had a spread trade, these funds had specific strategies out and because of ADL and the aggressive way they were ADL'd, uh even their profitable positions were liquidated. So, they lost on both sides even though their strategy was actually correct and not risky for the protocol.

This boils down to the fact that most DEXes just look at risk in individual markets. Um they don't look at the fact that in that spread trade example I gave you that the that trader is not actually risky to the protocol because they're actually solvent. They don't consider insurance funds. It's pretty alarming that the uh you know, most popular perp decks is don't even have the concept of insurance fund. There is a LP pool that is generally prioritized higher than insurance.

Um they don't think about mobility of liquidity around these different uh ex- different markets. And losses are basically uh given to the traders even though they were doing the right thing. So, I think unfortunately, there's just uh a lack of visibility of when these protocols uh misbehave because people are so focused on price and vanity metrics that they don't understand that the underlying mechanisms could be broken. So, what [clears throat] the solution that we're proposing is called hybrid margin.

Uh what we're trying to do is get best of both worlds. So, again, to recap isolated margin, you can't move your unrealized PNL across markets and in cross margin, unrealized PNL is automatically netted, but that also causes cross contagion risk. Hybrid margin is a system where you as a trader are able to move unrealized PNL to the different markets to protect your positions. So, in essence, you get uh protection of isolated margin when ADL happens and what I mean by that is it's not path dependent. You know exactly where you are in the ADL queue. Uh but you also uh benefit from the capital efficiency of

cross margin where you're able to move your margin to these different markets as needed. So, the trader has a little bit uh has a lot more control and obviously they need to be able to exercise the control. Um and if they don't, the worst case scenario is the protocol gracefully just falls back into isolated margin mode anyway. So, the counterparties are still safe. This [clears throat] diagram kind of just highlights uh what I was talking about. So, each market has its own separate insurance fund. There's not a global insurance fund. Again, that is a potential way where these mechanisms fail where one single insurance fund can be drained in, for example, like the

jelly jelly attack or when James Bond takes like a crazy position. Uh but each market has its own individual insurance fund and again, you're able to move capital between markets based on parameters that I will go into later. But uh the the important point to note here is the sweeping ability uh and the reason why we've done these tradeoffs. Again, because we are fully on-chain and gas is a measurable metric that we need to like understand, uh you have to take these things into account and risk can't be fully baked in and uh simple like a cross margin netting of risk. Um there's other parameters that I highlighted before that needs to be

accounted for before you allow traders to just simply move margin around. So, [clears throat] this example kind of highlights what I was talking about. So, in here, you have a trader that has a $80,000 uh in one market and $75,000 loss. In the isolated example, what the trader would have to do is realize some of the profits in the ETH market here and move it to the Bitcoin market. In the cross margin example, uh the protocol does this automatically and most of the times this is good, right? Until the extreme scenarios happen or you you expect not to be ADL'd, but you get ADL'd.

What we're suggesting is the sweeping of unrealized PNL. So, here, the trader can move unrealized PNL from ETH to Bitcoin and not have to close out their position. Again, there are parameters here that we will talk about as to why they're able to do this that depend on how Bitcoin and ETH move together, insurance fund and other aspects. Yep. And that's exactly what this is about. So, here we are talk- we show how like between Bitcoin and ETH, for example, and these numbers are still theoretical and they will be dynamic, but in this example in Bitcoin and ETH, you're able to move 90% of profits over.

The way we come up with that number is how Bitcoin and ETH move together, uh how stable those asset prices are, how big is the individual Bitcoin ETH insurance funds. Um so, when we do move that margin around, the counterparty is still able to be paid out. And then again, if these markets diverge or there's insurance funds lacking or some other reason, then we would just haircut this down. So, it would be, for example, in this here, it's Bitcoin SOL would be 50% or if there's like Bitcoin Doge or some sort of meme coin, then we just won't even allow you to do that because that would just cause cross contagion risk where the entire protocol could potentially have like a bankruptcy

situation. So, another recap. Um each market has its own insurance fund. There is a global overflow fund where that is helps you determine how much profits can be moved across. And then there's also ways to haircut winners before we finally resort to ADL. We believe that resorting immediately to ADL is a bad experience for traders. That should be the last resort. But unfortunately, it's kind of become standardized where we just automatically go to ADL first case and just expect these traders to like not cry in the casino as meme goes. Um

But we think that we have a better model of doing this and uh you know, that is a trade-off that we made, right? So, we don't run our own blockchain because we want to be built on a fully decentralized L1. We care about resiliency. We don't want to be running on like a centralized L2 or AWS server for that matter. And again, with the margin and liquidation mechanisms, we want to have sane mechanisms in there. So, when traders put on trades, they're going to be paid out and they know exactly where they are in the ADL queue.

There's transparency and sanity to these mechanisms. And of course, eventually when the protocol is open source, all these all this information is visible. And the the numbers that I posted here, these are things that the best parallel for this would be like a risk oracle. Uh that would post at a certain frequency to the chain saying that these are the parameters for Bitcoin ETH, these are the parameters for ETH SOL, etc. So, all this information should be on chain and visible to traders so they can then make the best decisions for themselves.

>> [clears throat] >> Right. Uh so, the three constraints that we're optimizing for, um adversarial ordering. Uh this is basically is ADL path dependent or is it a predictable? Um unfortunately, the current mechanisms are all path dependent. Uh depends on which market gets uh closed out first for the winners to get paid out. Uh the solution that we're suggesting is predictable ADL uh where you know exactly if you when and if you're going to get ADL'd. Uh bounded compute. Again, a lot of the things that I've mentioned here are extremely complicated and to try to do that on a smart contract every block is nearly impossible. It doesn't matter how

uh fast your chain is. Unless you run your own blockchain, then yeah, you can do whatever you want. But we're not doing that. So, we have bounded compute. That's why That's why the risk oracle has to post uh these parameters at a certain frequency. And then um partial execution. So, in the worst case scenario, this should always fall back to isolated margin. Uh but in the best case scenario, you are getting ideally 70 to 75% of the benefits of cross margin where you're able to put on these advanced trades and have the safety of, you know, knowing that you won't be wrongfully ADL'd potentially.

Um some of the things that we're excited that this will unlock. Um I think one of the most interesting things that we would like to see is multiple collateral types. Uh right now it's stablecoin margin. Uh eventually we would like to have like Bitcoin and ETH and SOL as used as margin. Uh obviously that is something that That's where like the world is heading towards. Um and then advanced trades. We want to make sure that people are able to be comfortable putting on these advanced trades. Uh like the spread trades, delta neutral trades, and whatnot. And um again, the point to highlight here is this mechanism allows us to continue to be gas efficient. Um so, market makers are able to post orders and their solvency checks are done per market uh versus

cross markets, which would cost us a ton more gas to post these orders. Uh and you could always do solvency checks probably at settlement, but then the trade-off you're making there is that market makers could be placing a lot of orders that are just not going to get filled. You're clogging the order book with bad orders and traders are getting not filled at the price that they expected. Yep, that's uh pretty much the gist of it. Uh we believe that, you know, perps are obviously like a very exciting product, but there's a lot of improvement to be had uh with the mechanisms that underlie it, especially how we think about margin and liquidation.

Thank you. Questions? Sorry. I I came late. What's the protocol? What's the name of the protocol? What's the name of the protocol? >> It's It's Perp. Oh, Perp. P E R P L. Okay, thank you. Cuz we we can't read from here, so Oh, got it. Uh Would this work for a spot perps trading strategy?

Do you mean like equities, commodities type of thing? Uh no, with spot Bitcoin. If you're long spot and short the perp, would this uh system work with that? Um Well you Unless you can use that long spot as a underlying collateral, no. Um but um potentially if we have like a spot DEX as well, um then the liquidation engine could be aware of that. That's one way to do that, but um if the spot is held outside of Perp, the protocol it's there's no chance for us to be use that.

So, a similar analogy is like when people post collateral into a lending protocol and uh they borrow margin and use it on a Perp DEX. Uh unfortunately, liquidation engines don't compose. So, we're not aware of what the other protocol is doing and what your assets there are doing. And uh we can't use it um to account for margin on our DEX. How do you protect against uh MEV on Monad?

Sorry, who? Yeah, uh how do you protect against uh MEV on Monad? Oh, I think that's a better question for stakers, but uh since it's a club, there's a different angle for MEV. Um that's not like as simple as like AMMs um because we follow a price-time priority um and there's like different price levels, etc. So, it's not necessarily my area of expertise, but I'm sure there's a lot of staking and other protocols thinking about that. Cool.

Thank you.