I'm making neither of those assumptions. Also, HFT is pretty much as far from making markets "efficient" prediction mechanisms as you can. HFT is a part of the plumbing of financial markets, they arb out the tiny but obvious things. For example they're the people that make ETFs possible - Vanguard or what have you doesn't actually go and buy all the thousands of securities that they have in their products, they outsource it to so called Authorized Participant (read: some org with HFT skills).

I wouldn't say that HFT arbs out only the "obvious" things. We try to arb the non obvious things as well, with more or less success.

HFT just means you trade a lot and you have low tick to trade latency. It doesn't say much about why you are trading.

For example you could have a HFT stat arb strategy in which you are arbing higher dimensional factors that are pretty far from obvious.

In general, the speed wars are over and everyone is reaching deep to become as smart as possible. Or in the other words, just being fast is no longer an edge, it's a commodity.

> In general, the speed wars are over and everyone is reaching deep to become as smart as possible. Or in the other words, just being fast is no longer an edge, it's a commodity.

I'm curious what the best techniques for being "smart" are when you have to have such quick turnaround. Are these rules-based techniques, statistical but interpretable, or black box?

I would say all of the above, with a general transition from rules -> statistical -> black box as time goes on.

But you're right that being fast and smart is pretty hard. You almost always have to sacrifice one for the other. Different firms operate on different spectrums of this speed vs smarts divide: some are faster and dumber (relatively), some are slower and smarter. But everyone is trying to up skill (analogous to moving up on the value chain), because the markets have gotten fiercely competitive at the nanosecond to millisecond scale, especially on american venues (CME for example is a straight shark tank).

I am so glad for all this work that HFTs are doing, let's me buy ETFs and all kinds of things cheaply. But maybe not now, I need the valuations, rates, risk premia to get a little more attractive. Maybe even next year, who knows when's the next March/April 2020 coming. I am hoping for a much better selloff next time!

> For example you could have a HFT stat arb strategy in which you are arbing higher dimensional factors that are pretty far from obvious.

What does this mean (HFT stat arb strategy)? What higher dimensional factors ? Ty!

> HFT is pretty much as far from making markets "efficient" prediction mechanisms as you can

? HFT facilitates liquidity, liquidity facilitates efficiency.

Yes of course, but i meant as far as you can [while still being a market participant].

Nobody really cares (except for traders) if the US 10 year note is at 1.54% or 1.52%, but rather if it's at 1.5%, or 2.0%, or 3.0%. Or even 10%.

>Both of these assumptions are demonstrably false.

Can you demonstrate?

Not the OP, but this book does a pretty solid job of arguing against it:

https://www.worldscientific.com/worldscibooks/10.1142/11484#...

Yes solve any market equilibrium model and compare to the real deal.