Quant Insider.io retweeted
Knight Capital Lost $440M in 45 Minutes. What Does That Say About Your Test Suite?
On Aug 1, 2012, Knight Capital deployed new trading software across eight servers.
Seven servers received the update. The 8th retained old code tied to a config flag that had been repurposed. 1/n
Quant Insider.io retweeted
New Approaches to (Safe) High and Mid Frequency Trading
(High Speed) Formally Verified Order Matching Engines
Provably correct code is no longer limited to just mission-critical systems. Now, formal proofs are at the heart of the revolution in mathematics and are making inroads in finance.
Sign up now:
topmate.io/quant_insider/225…
Quant Insider.io retweeted
(High Speed) Formally Verified Order Matching Engines
Provably correct code is no longer limited to just mission-critical systems. Now, formal proofs are at the heart of the revolution in mathematics and are making inroads in finance.
Sign up now:
topmate.io/quant_insider/225…
What does it mean when a quant says 'BOUNCE ALL THE STRATS'?
Find out here:
lnkd.in/djnxnA7Y
Sunday 9:30 eastern/ 2:30 london
From Market Making to Modern HFT: An Inside Look at How Markets Really Work
Date - 28th June 2026
Time 7:00 PM IST/1:30 PM GMT/8:30 AM EST
Register Now -topmate.io/quant_insider/216…
Use the coupon code "MM20" to get 20% off the early bird offer.
Join @annanay Kapila, former Quant Trader at Tower Research and Floor Traders and Founder of @QFEX (YC P25), for a hands-on session where you'll write the logic behind an algorithmic market-making strategy, line by line.
Drawing on years of building Trading Strategies at one of the world's leading high-frequency trading firms, Annanay goes beyond theory to show you how a market maker actually thinks in code. You'll start from the core question every market maker faces: where do I quote, and how much do I risk? and build up a working strategy in Python, piece by piece.
The core of this session: building it yourself
We'll walk through the real components of an algorithmic market making strategy, with Python examples you can follow, run, and extend:
Quoting around a fair value: setting bid and ask prices and choosing your spread
Managing inventory risk: skewing quotes as your position grows so you don't get run over
Reacting to the order book: adjusting to market depth, volatility, and order flow
Handling fills and PnL: tracking what you've traded and what it's worth
Tuning the knobs: how small parameter changes reshape the strategy's behavior
By the end, you won't just understand market making conceptually. You'll have seen the actual code that turns the idea into a strategy
We'll also set the stage and look ahead:
The role of financial markets in modern economies: why market makers exist at all
A brief history of market making: from open-outcry pits to electronic order books
The road to modern HFT: how speed and technology transformed everything we just built
Whether you're trading live, making markets, studying the field, or chasing your first quant role, this session meets you where you are. Reserve your spot today.
4
1,362
Why "Skew at 1st Percentile" Doesn't Mean Buy It🧵
One of the most common traps in volatility and derivatives markets is also one of the most seductive: you pull up skew, you see it's sitting at the 1st percentile of its historical range, and you think you've found a trade. Cheap is cheap, right? Buy it low, sell it high
How Dislocations Actually Get Found
None of this means skew is never tradable. It means you can't find the trade by staring at a percentile.
A genuine dislocation generally requires context usually an understanding of the flow that's driving a parameter to a relative extreme. It's generally hard, without any context, to just look at a volatility surface and be highly confident you're seeing a dislocation as opposed to the market's expectations, which are usually pretty reasonable. You need more information than the surface itself gives you.
Take a concrete example. Suppose there are very large call overwriting programs from ETFs in a name like Nvidia, the yield-type ETFs selling big size in the one-month upside calls. You know that flow exists. Now you can look at the volatility surface, see that it's relatively steep in the front where that part is getting depressed, compare the steepness in the front versus the back, understand where it's coming from, and then think about what kind of trade you could construct around it.
That's the difference. The flow gives you a reason to believe the price is dislocated. The percentile gives you nothing but a number that's far from its average, which, in an efficient market, is exactly where it's supposed to be most of the time.
1
1
354
First-order thinking is the biggest source of misconceptions in this space. "Skew is at the 90th percentile, sell it." "Skew is at the 10th percentile, buy it." People who haven't done the work, done the research, and put in the experience genuinely can't believe it's not that simple. But markets are not that inefficient.
If you want to trade skew, the question isn't "where is it in its historical range." The questions are: what flow is driving it here, am I getting paid enough to take the other side, and what's the carry while I wait. Answer those, and you might have a trade. Skip them and lean on the percentile, and you've just fooled yourself into a position the market was happy to give you.
302
How Auto-Callable Structured Products Distort the Long-Dated Vol Surface 🧵
Most vol traders know structured products matter. Few can explain exactly how their hedging lifecycle warps the surface and why the distortion flips violently when markets fall far enough.
4. The non-linearity is the whole point. Linear models of dealer hedging miss this completely. The shift from "long Vega, selling vol" to "covering hedges, buying vol" is a regime change tied to barrier proximity, not to spot level alone.
1
236







