Comments for Systematic Edge

Comment on Max Decorrelation Portfolio by Portfolio Optimisation with MlFinLab: Mean-Variance Optimisation - Hudson & Thames

Tue, 01 Sep 2020 10:44:01 +0000

[…] You can read more on maximum decorrelation portfolio in the following blog post: Max Decorrelation Portfolio. […]

Comment on Queue Position Simulation by aman kumar

aman kumar — Wed, 12 Aug 2020 17:14:11 +0000

In reply to Author.

One way is to have a fixed component of latency and a variable component which will be decided by tick flow.

Comment on The Man Who Solved the Market – Notes by Quantocracy's Daily Wrap for 11/12/2019 | Quantocracy

Quantocracy's Daily Wrap for 11/12/2019 | Quantocracy — Mon, 13 Jan 2020 06:48:37 +0000

[…] The Man Who Solved the Market Notes [Systematic Edge] […]

Comment on The Man Who Solved the Market – Notes by Book Summary: The Man Who Solved The Market – AYAK Consultancy™

Book Summary: The Man Who Solved The Market – AYAK Consultancy™ — Mon, 30 Dec 2019 13:47:55 +0000

[…] Original Post: https://systematicedge.wordpress.com/2019/11/12/the-man-who-solved-the-market-notes/ […]

Comment on The Man Who Solved the Market – Notes by What I am Reading – Nov19 – Week 3 – Talking Macro

What I am Reading – Nov19 – Week 3 – Talking Macro — Thu, 21 Nov 2019 19:43:12 +0000

[…] Notes on Jim Simons book (Systematic Edge) […]

Comment on The Man Who Solved the Market – Notes by Author

Author — Tue, 12 Nov 2019 16:58:59 +0000

In reply to Tho Do (@thodoha).

Ya I agree with all your points. He was a great manager of the brightest minds in the industry. There’s an argument you can make that managing a team is harder then trading itself….

Comment on The Man Who Solved the Market – Notes by Tho Do (@thodoha)

Tho Do (@thodoha) — Tue, 12 Nov 2019 15:48:43 +0000

Cool summary! Thank you for sharing! To me, some other points off the top of my head are also quite important are:
1) The need for a super clean and efficient database to query all kinds of data (Straus’s team)
2) Pair programming + no hardcoding variables! (Margeman)
3) Other lessons on building and managing a superstar team.
4) … more that I have not recalled yet.

Comment on Queue Position Simulation by bwc150

bwc150 — Sat, 24 Nov 2018 05:03:17 +0000

In reply to Author.

The places that are best at it use the latency from their own orders at that time (order send to ack receive for example) as a proxy for matching engine latency.

Data replay engines usually just blast through the updates as fast as possible and all the components that care about time use the timestamps on the updates rather than “wall clock”.

Comment on Queue Position Simulation by Author

Author — Fri, 23 Nov 2018 16:28:36 +0000

In reply to bwc.

100% agree.

Simulating latency is hard. The way I do it is to apply a fixed delay (which is a function of my latency) when I send orders. I’d love to know how others are doing it as my way is not a smart one given volume bursts.

Building a proper market data replay engine is hard as it’s not easy to re-create the time between the events (ie you’d get a lot more packets around numbers than normal times).

Comment on Queue Position Simulation by bwc

bwc — Fri, 23 Nov 2018 15:36:34 +0000

How to handle queue position in backtests/simulations has been an important topic inside HFTs for years. First it depends on what kind of market data the exchange supplies and what you have available. If it’s order-by-order based, then you know exactly what happens to each price level’s queue as time passes. If it’s price-qty based then you need to estimate how each change to the qty at a particular price will impact the queue.

In either situation, you need to make some assumptions about your latency placing orders and getting in to the matching engine. If you’re placing new orders in to the queue right at the same time competitors are, then you’ll need to have a way to assume whether you get in to the queue before or after them.