Hyperliquid has partially addressed the concerns I raised two days ago. @pejisdev has mempool data from a node and ran some great analyses on how the the chain has behaved in the past few days, with a little input from me. TLDR: cancels are fixed, modifies still broken.
The team at @HyperliquidX took rapid action to address the prioritization of cancels. They were seemingly able to do so without a network upgrade which is really interesting - there must be a mechanism to control certain parameters/behaviours of the chain without a redeploy of the binaries.
Unfortunately ALO modifies still seem to lag when the chain is congested. This is problematic because it incentivizes market makers to cancel + post instead of modifying under load, which increases the load further.
Overall, very pleased to see how responsive the team at HL has been to this issue and others that have been flagged on discord as they continue to scale the chain, and am confident they can continue to deliver and improve.
Under load, Hyperliquid cancels were losing to orders signed after them. Thanks to @FrothyBeverage the issue went public, and the goated Hyperliquid team was fast to fix it.
We then worked together to go deeper and see whether anything else had broken.
Modifies were hit too, to the point where it is now faster to cancel and place a new order than to modify one. Here is the data.
Paper, code and data, so anyone can check it from Hyperliquid's public archive, or straight from the histograms we collected on the @femtosxyz node:
Paper link : github.com/pejisdev/hyperliq…
Decentralization and open nodes let anyone check data like this independently. We see that as a strength, not a weakness, next to other perp DEXs. Thanks to Hyperliquid for making it all available.
Hyperliquid.
It is important to clarify that this is a fixable technical issue on Hyperliquid. The exchange has on the whole been wildly successful at pioneering a fair and efficient market design that benefits smaller makers and retail compared to the more traditional style.
Replying to @HyperliquidX
@HyperliquidX's IOC speed-bump mechanism has a flaw. Normally designed to protect makers in the interests of a more liquid market, it breaks down in periods of high activity. When blocks become full, pending modifies and cancels queue up for later blocks, so their latency increases. However, IOCs behave as though they have a separate blockspace dedicated for them, and their latency barely budges. You can see this in the lower left chart, showing latency of different order types during 19 activity bursts.
The result is that under relatively modest load (25-30k tps), IOC + 8 bps (priority fee) orders can pick off resting liquidity before it has a chance to get out the way. Meanwhile, under higher load (35-40k tps), even IOCs without any priority fee can pick off resting liquidity.
This, in my opinion, is antithetical to the ethos and design of the ALO and cancel prioritization mechanism and needs investigation by the team. It exists to reduce toxicity from HFT takers, and is breaking down precisely when most needed.
It may also be exploitable. If IOCs really are insulated from the congestion affecting cancels and modifies, a malicious user could spam orders, delaying makers’ cancels while their own IOCs go through, and taking the liquidity left stale.
One single example that occurred on AVAX today caused 1.3M$ of volume to be adversely selected for a total loss of ~13k$ over the subsequent 10 seconds. This affected us materially. Faced with such episodes of adverse selection, spreads and liquidity are undoubtedly worse than they could be otherwise.
Would appreciate if anyone with a contact at HL could share this with the team.
Regularly clocking a tick (time the process knows it has data to process) to order (time we call the kernel to send) latency of under 1 microsecond now, including updating pricing on and checking every instrument in our tradable universe
2
28
2,388
Asked models about the possibility of oil stabilizing at 100, here is how they rated each other's responses. Opus came out best, but curiously every model except Gemini rated its own response as the best, which indicates consistency between the model preferences and how they write
4
940
And let me guess, any person who doesn't contribute to bringing about this regime should be punished?
I honestly think that America needs 20-50 years of a right wing authoritarian regime, Latin America style, with no real elections and suspension of habeas corpus to fix this.
100 million people need to be deported and entire networks of woke activists need to be put in jail permanently. Leftist media needs to be permanently dismantled. A non-woke Church or something like it needs to regain control over morality, shut down onlyf🐱ns, shut down the leftist universities, etc.
Full year zero transformation of society into something that's actually functional, healthy, white and growing again.
Hyperliquid OI in $ holding steady or even slightly climbing post 11 Oct actually means that a lot of new positions are opening up if you take into account the decline of many assets' prices by about a third since then
3
656
Latency spikes on hyperliquid getting kinda ridiculous, I now routinely get filled on orders i've attempted to cancel 2-3 secs ago whenever volume picks up. I hope the team is working on optimizing the chain for throughput..
















