🪶Dulumu²🪶 retweeted
We have built DeepStack, the first AI to beat humans in no-limit poker.
We are now building the next generation AI company for algorithmic trading!
We are growing our research and engineering team.
If you want to work with a world class AI team led by ex-DeepMind researchers and engineers.
If you want to build something that has never been done before - join us.
equilibretechnologies.com
🪶Dulumu²🪶 retweeted
In market making, average latency only tells part of the story. You don’t get picked off on your average order. You get picked off when you’re in the tail.
A system doing 1ms most of the time but occasionally taking 10ms can be worse than one consistently doing 2ms.
That’s why reducing jitter matters as much as reducing latency itself. And why deterministic infrastructure all the way from market data ingestion to order entry matters.
Whether you’re a solo trader, prop shop, market maker, or hedge fund, I strongly recommend giving this series a read.
Shoutout to @Tristan0x. It’s hard to explain complex subjects in such a clear, and practical way.
How I made $6K PnL in 30 days on $10K principal market making perps ($158m+ volume).
The strat was entirely built by me and my quant, Claude (he won a national math competition in China).
Last night I asked if you'd want a guide. Hundreds of you said yes, so here it is: HFT for Dummies.
The honest timeline when I did it:
- Days for basic harness and venue integrations.
- Weeks to get it semi-competitive.
- Months to really tune and get it properly profitable.
This series can hopefully shortcut the learning curve for you.
I've been writing trading bots since 2017. It's how I got into crypto (free market data!), and how I met my co-founder.
Back then the barrier was high. You had to know how to code, to study data to find alpha, and to grind devops to have systems not blow up in production. Thankfully I taught myself Python and later Rust, but not everyone has a CS degree.
Nowadays with LLMs, anyone can have a skilled quant in their pocket. I use it for data analysis, strategy tuning, research and alpha generation and infra optimisation. It's a ridiculous time saver.
I'm extremely bullish on one-person quant shops. I think this will be a huge trend this next decade as finance comes on chain and market data and connectivity are accessible to anyone. I've already met many hobbyists who've made a living trading on e.g. Hyperliquid, enough so to quit their day jobs.
What we'll be building
A delta-neutral, reference-price market maker.
Binance is where most price discovery happens. We stream its market data, use it as a reference price, and quote a bid and an ask around it on another venue. I'm using @Bulletxyz because it's the one I know best.
Market making is the business of picking up pennies. Mine are under half a basis point per dollar. Enough pennies add up at the scale of millions of volume.
The catch is speed. If the market moves and you're slow to update, someone lifts your stale quote and you're the exit liquidity. It's a fun game of algorithmic cat and mouse.
On Bullet, takers get a 100ms speed bump. Why? It gives soon-to-be market makers like you a safety buffer so you don't get steamrolled by billion dollar firms with microwave towers.
The goal
The goal: tick to order sent in under 10ms. Stretch goal: under 1ms, which is very achievable if you're determined. We'll also focus on minimising jitter (randomness) because your slowest orders are the ones that will lose you the most pnl.
Agenda:
I'll aim to post one topic a day, going from easy wins, and getting progressively more deep in the weeds. If you work at any HFT firm this will all be elementary stuff, but hopefully it's helpful for newcomers.
Saving seconds
1. Basic intro to algo-trading (this post)
2. Running a simple market maker from your laptop
3. Measure once, cut twice: telemetry and Amdahl's law
4. AWS: Infra and colocation
Saving millis
5. Stop polling: websockets are your friend
6. Optimising order execution
7. How to not get rekt by API rate limits
8. Linux tuning: CPU pinning, IRQs and other nerd stuff
Saving micros (diminishing returns, but fun)
9. Optimising the hot path with CPU profiling in Rust
10. AWS: the last mile
11. Bonus: nonces (and other DeFi quirks)
I'll try share some open source code along with it. Don't expect to be beating Jump in the orderbooks just yet. It should be a good baseline that you can extend and take it as far as you want (you can take it pretty far!).
Tomorrow: setting up a basic MM in Rust, the architecture of which will roughly follow the diagram attached below.
🪶Dulumu²🪶 retweeted
i decompiled the contract that drained binance wallet 14
gist.wavey.info/Ta3FgAxastgj…
🪶Dulumu²🪶 retweeted
My cousin asked me for book recommendations to learn about trading.
I dug this post up which I published a year ago. It was also one of my most popular articles.
As always I hope you find it useful or at least entertaining.
moontower.substack.com/p/so-…
🪶Dulumu²🪶 retweeted
let's ride
This is the big week.
@sinjoh 2.0 is imminent.
A lot of hard work has gone into this & I can’t wait to finally release it to the public 💫
🪶Dulumu²🪶 retweeted
despite the fact that almost any memecoin etf can properly obtain sec approval, traditional issuers and sponsors are incapable of properly running memecoin etfs, as evidenced by some abandoning s-1s or even full etf liquidations after just one year
while you will see many memecoins such as useless break to unimaginable valuations, the more realistic approach would be sponsoring a bucket memecoin approach to rebalance coins as attention will inevitably keep changing
🪶Dulumu²🪶 retweeted
if your girl is impressed by money you're bound to lose her to literally any other dork with a little more loot than you
🪶Dulumu²🪶 retweeted
Good old times.
The first strategy we ran did something like 20x in one month. It worked like this:
Bitmex had pegged orders and contingent orders. A pegged order would be placed one tick below the opposing price by the matching engine on arrival. For example, a pegged buy order one tick below the best ask. A contingent order would trigger the placement of the pegged order upon a fill.
So, all we did was place trigger orders on both sides at the bbo, along with the untriggered pegged orders. When the price changed, our trigger orders would fill, and the matching engine would immediately place our pegged orders at the new price level... with zero latency.
This meant that we (and, I think, one or two other teams) would always be first in the priority queue. Back in 2018, this was incredibly valuable, as the tick size was very large relative to volatility.
It worked for three months. Then Bitmex disabled both order types because we were putting too much load on the matching engine.
GG
Yesterday we closed the book on XBTUSD — the contract that started it all.
Launched on 13 May 2016, it was the longest-running perpetual swap contract in crypto. And until we are told otherwise, the longest running derivatives contract of any kind, given TradFi still trades on contracts that expire.
Let's take a trip down memory lane one last time 💔
🪶Dulumu²🪶 retweeted
One thing AI has done in the trading world is dramatically reduce the need for the traditional desk quant.
A desk quant’s role is very different from that of a quant researcher and or quant trader.
A quant researcher spends significant time building the infrastructure behind sustainable strategies: sourcing the right data, cleaning and storing it, identifying market anomalies, and determining whether those anomalies represent persistent, repeatable alpha. A quant trader is someone that mainly does all those things (varies across shops) but takes direct responsibility for the risk in a book.
A desk quant is traditionally closer to the day to day trading process. A core part of the role is helping traders quickly surface quantitatively supported information that can improve decision making.
An example might be: “The last four times the market dropped 1%, three-month fixed-strike vol significantly outperformed the rest of the surface.”
That observation probably isn’t a sustainable edge or something you would build an entire strategy around. But in that specific environment, it can provide valuable context for how a trader allocates risk.
AI has streamlined a tremendous amount of this work. With the right data infrastructure and prompts, agents can quickly identify these shorter-term relationships, run conditional analysis, and surface relevant information that previously required significant manual work from a desk quant.
The interesting shift is that the value is moving away from simply being able to retrieve and analyze information and toward knowing what questions to ask and how to translate the answers into risk. AI can increasingly provide the analysis. The trader still needs to understand what actually matters.
🪶Dulumu²🪶 retweeted
I've been working very hard on this Whiteboard series over the bear market so that it would be done by the time the bull run started and new people coming into the space would have somewhere to turn to for a full comprehensive free trading series.
So far I've made 21 episodes, there will be roughly 30, and as we get into the last few we are dealing with more advanced concepts and tying it all together.
It's everything that I wish I had when I fist started learning to trade, all in one spot. Currently almost 14hrs of content.
It comes out every Monday at 5pm EST. Check out the full series out here.
youtube.com/watch?v=OB5kMepC…
Mayne (@Tradermayne) dropped a full trading bootcamp on YouTube.
Free. Studio quality. Actual price action, not another “I 100x’d with this one trick” video.
Haven’t seen educational work like this in crypto since @CryptoCred .
If you just got here, start there instead of copying PnL cards.
🪶Dulumu²🪶 retweeted
Institutional Market makers are hired exclusively from the top art schools around the world in order to paint pictures on the chart in order to HOOK filthy retail traders into terrible positions. Never forget this and stay vigilant
🪶Dulumu²🪶 retweeted
Traders: The reason MBO is useless:
That "show your size" game died in 2010 - the ONLY reason you'd want to display size on the DOMin MBO is:
1) you are trading a non FIFO instrument, that allocates position based on order size
2) you are dicking people in a FIFO Queue
But literally - "follow the size" died in equities in 201x
🪶Dulumu²🪶 retweeted
If we are at circle 1 for $BTC and you are stressed about "missing out", think of all the opportunity that is about to come next.
A multi-year bull run where BTC and other coins go up and new things get created.
Hyperliquid didn't exist until 2023.
The biggest memecoins of the last cycle had their entire boom and bust cycles.
If the bottom is in we are in the first inning.
🪶Dulumu²🪶 retweeted
I dont think I know anyone hit by coldcard but...
If you were hit, do this. Most of it won't get coins back. Skipping it guarantees nothing ever will.
1. Off the device. Any Coldcard seed from March 2021 on is compromised. New firmware doesn't repair an existing seed. Fresh seed on different hardware, migrate. Check anything else set up in that window.
2. Create a packet. Your addresses, theft tx hashes, UTC timestamps, destination addresses, explorer screenshots showing the balance, short narrative, link to a current article.
3. File four ways. Each is a timestamp on the record.
- IC3 (ic3.gov), save the number
- Local police, as theft, get a report number
-FBI field office cyber squad, call directly
-Large amount: Secret Service, HSI, or IRS-CI
4. Preservation letters. No lawyer needed. Email exchange compliance: preserve KYC, deposit logs, IPs for this address. Ie the packet above
5. Leverage. It's Bitcoin. No issuer can freeze it the way Tether freezes USDT. The only chokepoint is cash-out at a custodian. The exchange paperwork is the whole game.
6. Attorney, if the sum justifies it. Sue persons unknown, TRO over the addresses, serve exchanges, subpoena to unmask.
7. Get named as a victim. Reports alone don't put you in line. Email the case agent your documented loss, get written acknowledgment, check forfeiture.gov monthly. Once notice issues you have about 30 to 35 days. Miss it and steps 1 through 6 were wasted.
Remission pays cash at theft-date price, not likely coins.
Anyone DMing recovery services for a fee is trying to steal from you.
🪶Dulumu²🪶 retweeted
After dealing with what is a complete shitshow in the @Collector_Crypt discord i cant say that i can ever recommend them again. I have had multiple problems, the community is awful and have now had multiple ppl get cracked slabs in the mail with no resolve. im sorry i ever gave them a platform.