beaver quant

for it's first weekend, @tradexyz's S&P500 did pretty well volume-wise compared to other markets' first weekends, becoming the biggest equity launch weekend ever on hip-3

This is one of the sickest thing I've read here. For the ones wondering what was the opportunity like, here is some simple backtest of @Cbb0fe's first described strategy: executed on a single core/evm pair for a single day. It consists of: - monitoring pool price at a block - if the spread between core price and pool price is > threshold (10bps here), execute on the pool - if at the end of the block we get executed, hedge on the other leg (disclaimer: no possibility to 100% say that we would have so here we take the whole opportunity and consider that we would have). Only thing to consider on the backtest is that hedge price is taken based on orderbook snapshot at block end and not at opportunity timestamp - finding the size of the opportunity, with a max size of 20k per trade. This means simulating on both ob snapshots and pool tick map the exec price This backtest version is obviously only a small portion of the way more complex strategy they used. We only take the opportunity of arbing HYPE perp vs HYPE / USD₮0 (0x337b56d87a6185cd46af3ac2cdf03cbc37070c30) on hyperswap. In reality, the arb strat should include every pool on every dex, and every pair available on core. The pool was chosen as the most liquid pool to date, and the backtested date as a high volume one. The profit chart, based on our "always executed on evm" assumption, highlights the total part of the pie that all competitors share: an annualised $5M opportunity on a single pool. In reality, no one would be able to access 100% of the opportunity. For these results, we took 5bps fees on the pool, 2bps taker fees on the perp (platinum tier4). The opportunity would be bigger with a lower threshold and lower fees. If there's some interest, I'll provide backtests for a multi pool arb (the whole opportunity), or the maker perp strat cbb talks about. Don't hesitate to dm if you want to discuss about it 🫡

It's now been 3 weeks since @HyperliquidX released HIP-3 on mainnet, and @tradexyz launched XYZ100. Three weekends have passed since, and except for a few funding flips, the oracle behaved perfectly across closed-market periods. With on-chain stocks about to expand, I wanted to explore how tradexyz's weekend oracle design behaves in extreme black swan events, and double check if there is a risk. These events are rare, but happen. Most equities have seen double-digit gaps between Friday close and Monday open at least once in the past years. I thus built a simulation engine that reproduces all HIP-3 and tradexyz design, and mimics the product behaviour during a stressed market. The full paper, with methodology and results, is here: github.com/ymsynn/Unit-Perps… but here is a breakdown of major results. The setting: We simulate a market for a product with: - 10x max leverage - 30% price drop - compressed 3h simulation (instead of full weekend) for computation ease All assumptions are listed transparently in the paper. Liquidation Heatmap: A key part of tradexyz's design is to bound mark price to +- 1/max_leverage% from the latest oracle price available. This bound prevents liquidation from temporary price manipulation or oracle shocks. In bounded mode, liquidations remain stable and contained within the safe zone: Without the bound, liquidation cascades appear instantly. The bounded version delays liquidation until the oracle syncs on market reopen, which is precisely its purpose. When the market reopens, ADL will likely kick to preserve protocol integrity. A key question rises at this stage: are the bounds only for mark price computation, or are trades also blocked outside of the bounds? Average leverage asymmetry when trades are permitted outside mark price bounds: If trades can occur outside the bounds, but liquidations and unrealised p&l use the bound mark price, the opposite side has a capped maximum leverage (see in the paper for the maths 🤓). This leads to an asymmetry on the average leverage per side: This asymmetry is not a bug or an issue, but could be worth managing. Bad Debt forming when trading out of bounds: An other risk that rises if trading out of bounds, without liquidations as mark price is bounded, is that underwater positions could close the position, realising the bad debt (which wouldn't be the case if trading is limited outside of bounds, and these positions' bad debt would be socialised by ADL when market reopens). Key Takeaways: tradexyz's internal mark price design is really well thought, and as the first weeks of trading has shown it's robust. The simulation highlighted that some real life events, outside market manipulation, could convey a risk for the protocol. Some possible design improvements, such as having a dynamic max leverage, or preventing from trading outside the bounds, could be a solution to add an additional safety net in my opinion. By preventing trading outside the bounds, the product would not check the "price discovery" use case, but I don't think that it's the objective. Instead, it would add more transparency as traders would be aware of their maximum losses and maximum p&l over the weekend (computation in the paper). Now that I’ve done the background check, it’s time to join the orderbook 🫡

From what I get, the ±5% weekend bound means the mark price just stops moving once the underlying moves too much. So if the market nukes / is manipulated while it’s closed, most traders won't get liquidated — right?