Ninja Quant

[Ninja Quant @ninjaquant_]

Sharpe ratio without accounting for transaction costs is worthless. Most academic studies overstate Sharpe and returns by ignoring impact

[Ninja Quant @ninjaquant_]

Here’s some alpha:

Relax your hedge when you’re confident the counterparty is retail flow

[Ninja Quant @ninjaquant_]

AI progress during our lifetime:

Electrons -> Photons -> Wetware

[Ninja Quant @ninjaquant_]

One of the easiest ways to make some money in trading is to compete in markets where there’s not a lot of competition. Avoid swimming in a shark tank!!

[Ninja Quant @ninjaquant_]

For HFT you want to avoid async, GC and interpreted languages, the goal is to have the hot path latency as deterministic as it can be

[Ninja Quant @ninjaquant_]

If I was a junior looking for a job I would look into building things end to end, launching micro-startups, touching many techs…

Coding is basically dead, engineering isn’t (yet)

[Ninja Quant @ninjaquant_]

There’s no need to hire juniors anymore. Opus delivers the output of 10 junior engineers at roughly 10% of the cost

[Ninja Quant @ninjaquant_]

There’s no need to hire juniors anymore. Opus delivers the output of 10 junior engineers at roughly 10% of the cost

[Ninja Quant @ninjaquant_]

In trading, skip testnet and go straight into testing in prod. It’s cheaper and more efficient over the long run

[Ninja Quant @ninjaquant_]

Great interview

[Ti Morse @ti_morse]

My first interview w @sulaimanghori, Member of Technical Staff @xAI.

0:41 WTF is happening at xAI
1:46 Predicting future bottlenecks
3:05 Shredding conventional timelines
4:23 Experience joining xAI
9:23 Bootstrapping off the Tesla network
11:59 What is Macrohard
13:14 How Elon deals w fires
16:30 What it’s like working at xAI
20:33 Cybertruck bet with Elon
21:12 Using 80 mobile generators + battery packs to balance load at their data centers
22:45 How they built Colossus in 122 days
23:35 Work backwards & figure out the highest leverage thing you can be doing
25:51 How xAI hires
30:27 Challenging requirements
32:46 Experimentation
34:55 How Elon recalibrated his timeline estimates
39:15 AI engineers vs AI researchers
40:36 No one tells me ‘no’
42:09 Everyone’s an engineer
44:06 Why fuzziness between teams is an advantage
48:25 Testing human emulators as employees
50:00 Biggest blunders
53:23 What a meeting w Elon is like
54:22 How Elon gives feedback
56:44 Figuring out ‘what is truth’ for Grokipedia
59:21 What happens when Elon sees wrong Grok outputs on X
1:00:08 What a surge feels like & operating in xAI’s war room
1:02:53 Making fidget spinners & 3D printers in his bedroom
1:08:48 Creating a liquid fuel rocket engine

[Ninja Quant @ninjaquant_]

You are only crazy until you succeed, then, you are genius.

[Ninja Quant @ninjaquant_]

If you know how much time it takes to compute a signal (N secs, 99.999 percentile), you can add a new field in your aggs for the Last Price at time T[0] - N.

This way you can trade at T[0]-ish. With data as recent as it can be without lookahead.

[Ninja Quant @ninjaquant_]

Here's a tip

If you trade based on aggregates - i.e LastPrice every X mins - using an EMA (or any other feature) at time T[0], and the EMA is calculated from T[-20] to T[0], that's lookahead bias.

Technically you cant assume fills at the instant time data is received.

BUT‼️...

[Ninja Quant @ninjaquant_]

Guys I’m getting runtime errors after trying this new halal formatter, how to fix?

[Ninja Quant @ninjaquant_]

When I see a linked list implementation for an order book 🚩🚩

[Ninja Quant @ninjaquant_]

Headlines on its own are not a good idea to use as sentiment features.

Some headlines may intentionally leave out some key information (clickbait). And of course not all sources are reliable.

[Ninja Quant @ninjaquant_]

A thought on LLMs for trading

Let’s say you want to use the sentiment of $NVDA headlines from 2020 to 2025 as a signal. And the model was trained with data up to 2025.

What could go (very) wrong:

The results of your backtests are useless because the LLM could leak information from 2024-2025 to classify a headline from 2021.

The sentiment could be positively biased, falsely indicating that you have found alpha.

[Ninja Quant @ninjaquant_]

Quant dev interview question:

- What’s the difference between static and global variables?

[Ninja Quant @ninjaquant_]

To use LLMs in trading you must ensure the model wasn’t trained on data from the future. Treat it as just another feature where look ahead is forbidden!!

[Ninja Quant @ninjaquant_]

A thought on LLMs for trading

Let’s say you want to use the sentiment of $NVDA headlines from 2020 to 2025 as a signal. And the model was trained with data up to 2025.

What could go (very) wrong:

The results of your backtests are useless because the LLM could leak information from 2024-2025 to classify a headline from 2021.

The sentiment could be positively biased, falsely indicating that you have found alpha.

[Ninja Quant @ninjaquant_]

Here's a tip: If you're using leverage or short selling, don't do it under your name.

Trade through an LLC or similar to protect your ass(ets) if it all goes south

[Ninja Quant @ninjaquant_]

To make it in trading or anything in life you've got to deserve it