Planemo Trading

2.9K posts

Planemo Trading banner
Planemo Trading

Planemo Trading

@PlanemoTrading

quantitative algorithmic trading platform for perpetual futures. alpha live on hyperliquid (hip3), extended, grvt & lighter. stat arb: https://t.co/XfLlM5UQAS

Katılım Ocak 2023
21 Takip Edilen5.3K Takipçiler
Planemo Trading
Planemo Trading@PlanemoTrading·
When you read about latency tests of perp dexes, those are usually done in a vacuum, not in a production system with up to hundreds of orders per second and across 100+ users. It took us almost 2 weeks to get to this level of latency for our upcoming @extendedapp <> @HyperliquidX stat arb algo. Right now around 30% of orders fill within <250ms of submit timestamp, with 95% being faster than HL (without priority fees), which is good enough for our purposes given the conservatively edge decay model. It's very frustrating to not make (visible) progress for many days in a row. Then one day you refresh your dashboard and see data like this after your latest update.
Planemo Trading tweet media
English
1
0
13
1.5K
Planemo Trading
Planemo Trading@PlanemoTrading·
HIP-4 internal basket arb update: Here you can see the last 1.5 hours for the BTC range market. 1 tick = 50ms "Historically" (as in the last few days), the biggest opportunity was in buying the "NO" in all 3 markets for below $2, for a guaranteed $2 outcome. This happens much more frequently than getting all 3 "YES" for <$1. As you can see, there were some opportunities, especially in the very "on the margins" range of 1.995 - 2.000. But since liquidity is so limited right now, and we are definitely not the only ones running this basic strategy, the risk of something going wrong (mainly no fills on one of the legs, or bad fills) already makes this not very attractive. For reference: On day 1, you could buy $2 for <$1.70 for several hours. Even <$1.50 in the first few hours after launch. We are looking forward to more markets launching and hopefully being able to construct a much more advanced strategy that can create a higher, sustained edge.
English
0
3
25
1.6K
Planemo Trading
Planemo Trading@PlanemoTrading·
HIP-4 markets continue being fun with steady profits. However, we are already clearly seeing edge decay compared to day 1, and this will only get (much) worse in the coming weeks. It's much easier to build a basic working algo for prediction markets which can run at least at break-even compared to regular markets. Not to mention the possibility of a fee switch activated at any time. The true opportunity will likely be in cross-exchange. But for now, it's a fun little stream of bonus bps which could still grow into something big once more markets are launched.
Planemo Trading tweet media
English
0
1
29
3.6K
Planemo Trading
Planemo Trading@PlanemoTrading·
Let's see what we can cook here in terms of new algos. EURC will be especially interesting for a basis trade given the funding of the last few weeks.
rf.extended@rf_extended

What's unique about Extended’s native money market. With multi-asset collateral now live, @extendedapp users can deposit ETH and wBTC (with USDT and EURC coming shortly) and use them as margin to trade perpetuals. When a user's USDC balance goes negative, borrowing is triggered automatically with the Extended Vault serving as the primary lender. The key difference is how borrowing rates work. Traditional crypto money markets typically operate through isolated lending pools, where interest rates depend only on utilisation of a specific pool. Collateral risk is managed separately through haircuts and borrowing limits. Extended's setup is fundamentally different. The Vault lends against multiple collateral assets simultaneously, while borrowing demand emerges dynamically from unrealised PnL. In this environment static global borrowing caps are not practical. As a result borrowing rates depend on two dimensions: - overall vault utilisation - utilisation against a specific collateral asset. This means borrowing USDC against ETH can be cheaper than borrowing against BTC if system-wide exposure to ETH is lower. The second layer is how borrowing is allocated. When a user has multiple collateral assets, the system automatically routes borrowing through the lowest-rate collateral first, minimising the effective cost of capital. Example: if a user has a negative USDC balance backed by both ETH and wBTC collateral, and ETH borrow rates are lower than BTC, the system will allocate borrowing against ETH first before routing the remainder against BTC, continuously reducing the effective cost of capital. The result is a system where: - users automatically receive the cheapest borrowing allocation across their collateral portfolio - Extended maintains granular risk control over exposure to different collateral assets backing borrowed USDC - vault depositors earn additional yield directly from trading activity.

English
1
0
7
1.5K
Planemo Trading
Planemo Trading@PlanemoTrading·
First day on HIP-4. Accounted for ~20% of total volume in yesterday's new range market. This will be huge. Wen leverage?
Planemo Trading tweet media
English
2
0
15
1.7K
Planemo Trading
Planemo Trading@PlanemoTrading·
New HIP-4 outcome market released ~3 hours ago. It took us less than an hour to launch new arb algo (internal only so far). That's locked in, guaranteed profit upon resolution. Proper release once more HIP-4 markets are live.
Planemo Trading tweet media
English
3
2
28
1.8K
Planemo Trading
Planemo Trading@PlanemoTrading·
Our @HyperliquidX algo is the only publicly available one that consistently prints a positive PnL. 0 directional risk (look at that drawdown). Coming to more exchanges very soon.
Planemo Trading tweet media
English
0
4
12
848
Planemo Trading
Planemo Trading@PlanemoTrading·
@allaboutvol i like it in extra tab tbh could always have in options to put in under regulas positions i guess
English
0
0
1
128
vol
vol@allaboutvol·
ngl unecessary to have outcomes as its own tab, should've just been under positions
English
5
1
23
2.8K
Planemo Trading
Planemo Trading@PlanemoTrading·
3,161 trades 0.42% max drawdown (good for nerves) Worst fee tier on HL + builder fees enabled. Now would be a good time to enable HIP3 support for vaults so we can make this public. (If you didn't know, you can't trade HIP3 via vaults).
Planemo Trading tweet media
English
0
3
6
611
Planemo Trading
Planemo Trading@PlanemoTrading·
@web3_pastel @ArrakisFinance Hey, okay that's weird. Builder is correct. For that timeframe it should actually be 50% ALO 50% IOC. That's when algo still executed first (maker) and second (taker) leg back then. Mid-April we changed to 100% IOC.
English
1
0
2
65
Arihant
Arihant@web3_pastel·
@PlanemoTrading @ArrakisFinance Interesting, could you confirm if this is your builder code address? 0x12c50cb693dfb1ee8c2a71bbbac883ae79dac9de We analysed fills tagging this builder address between 10th and 31st March and 99%+ of the orders were ALO
English
3
0
0
93
Planemo Trading retweetledi
Planemo Trading
Planemo Trading@PlanemoTrading·
We at @PlanemoTrading were among the first 15 users to implement priority fees on @HyperliquidX for internal testing and are a top10 wallet by fees spent so far. From our own tests across 1,000+ transactions, we can see that this update serves effectively as a speedbump for IOC orders on HIP3. Our latency increased by roughly the 8x45ms that is outlined in the docs as maximum possible speed increase. It would be interesting to analyze the current user base of priority fees to find out what strategies are being run that justify implementation. The docs mention the reasoning behind priority fees is to prevent the escalation of an arms race of ultra-fast market makers. We are not ultra-fast nor a market maker. We so far don't see the value of paying 8bps extra just to get the speed we got "for free" a week ago. Our strategy does not necessarily benefit from being faster than other takers, but it certainly hurts performance if we are slower while order cancel/refresh stays the same as pre-update. So looks like some strategies could end up as collateral damage here as an unintended (?) side effect.
Yaugourt.hl@Yaugourt

Priority fees update. Live on mainnet for 4 days now. Here's what the data shows. Last 24h: Total priority gas burned: 50.41 HYPE Fills with priority: 267,323 Annualized burn (linear): 18.40K HYPE Avg priority gas: 0.000189 HYPE Max single priority gas: 0.318423 HYPE Who's paying: 0x3999...3336 · 29.54 HYPE · 83,333 fills · ~59% of all priority gas 0xefd3...28df · 12.78 HYPE · 11,090 fills 0x10fb...632c · 3.60 HYPE · 346 fills 0x57dd...494b · 3.10 HYPE · 2,871 fills 0xce97...7e78 · 0.66 HYPE · 106,929 fills What's interesting in the activity feed: cash:TSLA, cash:SILVER, xyz:BIRD, xyz:MSTR, xyz:EWY, hyna:SUI. Priority gas is being used across HIP-3 equities, commodities, crypto, and spot. Not concentrated on one pair, spread across the board. Compared to day 2 data (first tweet): Total priority gas: 14.56 to 50.41 HYPE (+246%) Fills with priority: 102,424 to 267,323 (+161%) Max priority gas: 0.141 to 0.318 (+126%) Number of active payers growing, top payer share dropping from 86% to 59% Usage is spreading. More wallets paying, across more instruments, with higher max bids. Still early but the trend is clear. For context: My earlier projections had the LOW scenario at ~810 HYPE/day. We're at 50.41 HYPE/day after 4 days. Below LOW but growing fast. If the current growth rate holds for a few weeks, LOW scenario becomes realistic quickly. Track it live: liquidterminal.xyz/explorer/prior… Powered by @hypedexer Hyperliquid.

English
3
9
48
16.8K
Planemo Trading
Planemo Trading@PlanemoTrading·
@0x_snatch you should run same test on extended and hyperliquid (with and without priority fees)
English
1
0
4
327
Insilico Terminal
Insilico Terminal@InsilicoTrading·
New episode with @crypto_hades A conversation with someone who has seen almost every era of crypto market structure firsthand. We talk through the shift from early inefficiencies and manual trading to stat arb, HFT, execution, and competing in a much harder market. 00:00 Getting into Bitcoin in 2010, mining, Mt. Gox, and coming back in 2017 13:37 ICO mania, BitMEX, and the Bitcoin fork trade 17:14 Why 2018 pushed him toward quant trading 24:52 Building a profitable stat arb business on FTX 40:21 Stat arb vs HFT, and why crypto got more competitive 49:09 What his trading operation looks like today 55:13 The future of crypto market structure and where edge is going 01:05:04 Advice for newer traders and why quant thinking differs from retail
English
7
3
99
46.2K
Planemo Trading
Planemo Trading@PlanemoTrading·
@hftgod Not a maker here, and yea "makes it even harder to take" is spot on x.com/PlanemoTrading…
Planemo Trading@PlanemoTrading

We at @PlanemoTrading were among the first 15 users to implement priority fees on @HyperliquidX for internal testing and are a top10 wallet by fees spent so far. From our own tests across 1,000+ transactions, we can see that this update serves effectively as a speedbump for IOC orders on HIP3. Our latency increased by roughly the 8x45ms that is outlined in the docs as maximum possible speed increase. It would be interesting to analyze the current user base of priority fees to find out what strategies are being run that justify implementation. The docs mention the reasoning behind priority fees is to prevent the escalation of an arms race of ultra-fast market makers. We are not ultra-fast nor a market maker. We so far don't see the value of paying 8bps extra just to get the speed we got "for free" a week ago. Our strategy does not necessarily benefit from being faster than other takers, but it certainly hurts performance if we are slower while order cancel/refresh stays the same as pre-update. So looks like some strategies could end up as collateral damage here as an unintended (?) side effect.

English
0
0
4
2.9K
yang
yang@hftgod·
I want to share my thoughts on hyperliquid’s latest priority fees update. I think it’s going to significantly change the market structure there going forwards longer term. A lot of the best market makers on HL - Alber Blanc, Pinely etc. are latency edge guys. They’ve already invested significant resources into trying to simulate what the next TOB is going to look like from pre-validated transactions, like by emitting raw mempool transactions from the node or by reverse engineering the binary to listen to the gossip data. Infra like this is all useless now, because you can literally just pay for it. There's a Dutch auction for the slot, you send your IP with your bid, and if you win, the IP gets whitelisted and prioritised, so the gossip data comes 10ms before everyone else. The change to order priority fees is also interesting - you pay up to 8 bps to reduce the latency of your order. They say it’s empirically 45ms of latency reduced per bp paid, so max of 360ms faster - I haven’t tested this. Under the new system, if a maker cancels their order, that still gets processed first, but for the remaining executable orders, whoever paid more is matched first. Currently it looks like there aren’t that many people bidding for the market data, and even fewer (if any) bidding for the order priority. I think this is because Hyperliquid already implements a speedbump on taker orders which is roughly 250ms but can spike up to half a second so it’s already harder to take - the block time is ~70ms so this is ~4+ blocks as it is, and makers have plenty of time to cancel their liquidity. These fees are paid in HYPE and burned. This isn’t really new - when protocols see trading firms making money they often want to extract more of a cut. Arbitrum did a similar thing with the timeboost auctions, and I’d say this was pretty successful, in that roughly 20-30% of the latency sensitive dex flow moved onto timeboost, and volumes didn’t meaningfully change. We’re sitting at around a year since they shipped this, and timeboost brought in around $7mn in revenue in this time. This is on (roughly) $300bn in volume, so with Hyperliquid at approx. 10x this, I can see the bull case for them to move to this kind of model. Objectively, I think it’s a smarter version of Lighter’s design - Jeff basically just waited to see what Lighter’s business model would look like, improved it, shipped it to testnet on April 10th, and rolled it out to mainnet within a week. I don’t think any firm has had time to think about the game theory of how to bid yet and optimise this, but it should basically completely collapse the gap between the top MMs. Disclaimers - I’m long some HYPE and haven’t sold LIT airdrop (yet).
English
26
42
598
110.9K
Lookonchain
Lookonchain@lookonchain·
The largest on-chain $ASTER long, 0x1527, is now long $CHIP. He also holds longs in $MON and $TAO. Positions: • 24.93M $ASTER($16.72M) • 36.18M $CHIP($3.43M) • 81.55M $MON($2.52M) • 9,026 $TAO($2.22M) #perps" target="_blank" rel="nofollow noopener">hypurrscan.io/address/0x1527…
Lookonchain tweet media
English
24
62
378
63.7K

Keşfet

@extendedapp @HyperliquidX @0xArcyl @BrianInCrypto @allaboutvol @web3_pastel @ArrakisFinance @elonmusk