Forums  > Trading  > Prediction horizon for high frequency trading  
     
Page 1 of 1
Display using:  

richardfatman


Total Posts: 1
Joined: Jun 2021
 
Posted: 2021-06-28 02:06
I wonder what is the typical prediction horizon of price change required for a high frequency liquidity taking strategy to make profits. Also, does using a fixed time horizon make sense in hft? What are your opinion on this paper "The Volume Clock: Insights into the High Frequency Paradigm" link
I am a total newbie to HFT. just curious

rickyvic


Total Posts: 249
Joined: Jul 2013
 
Posted: 2021-10-14 16:28
This is a good question actually and there is no answer.
I think sampling regularly makes sense especially if you have latency of millisecs.
I just do it very often, as often as possible.

Prediction horizon is also a function of the lag length you use to a certain extent at least.

I think the longer the horizon the better, so you are less trying to pick 1 tick which is difficult to get and you are maybe looking to catch the bigger swings.

"amicus Plato sed magis amica Veritas"

ronin


Total Posts: 690
Joined: May 2006
 
Posted: 2021-10-18 10:36
Partly showing my age here, but:

In my days, "HFT" meant that you are strictly interested in the current BBO levels, and that you are chasing per shares well below one tick. If you are looking at several ticks, that was called medium frequency.

I haven't been doing that for a while, and I think the terminology has changed a bit in that time. People now talk about HFT across several ticks, even tens of ticks.

Fixed time horizon is usually meaningless. Things have intraday seasonality. One minute after the open is very different from one minute during lunchtime.

The paper? Volatility as time change was more-or-less popular in the options world around ten years before that paper. See various papers by various combinations of Carr, Geman, Madan and Yor.

But, it was a dead end. Yes, it was an interesting parametrization. No, it didn't generate any new or interesting insights.

The whole thing was pretty old hat in 2012. I wasn't even aware of this paper, and I was deeply into that in 2012.

"There is a SIX am?" -- Arthur

rickyvic


Total Posts: 249
Joined: Jul 2013
 
Posted: 2021-10-20 07:28
Adding to what ronin rightly said... works for equities primarily though

It is a matter of cost, the higher your cost the longer the horizon.
Also capacity, if you want more capacity you need to take more risk and so longer horizons, at some point you are not calling it even hft.

It depends on the market also, otc markets you need to be careful not to hit your counterparties badly and to keep a fair game, sometimes they put speed bumps so you dont even have a chance.

Probably the most interesting topic is how to combine forecasts or how to allocate to each horizon, which is a complex problem when you also look at a system that ingests a lot of data at high speed (tick by tick sometimes is nearly impossible to keep up during bursts so you might well lag behind).

"amicus Plato sed magis amica Veritas"

rickyvic


Total Posts: 249
Joined: Jul 2013
 
Posted: 2021-10-27 09:57
There is no answer these are choices and need to be researched.

"amicus Plato sed magis amica Veritas"
Previous Thread :: Next Thread 
Page 1 of 1