I dont do anything fancy. I get a data set and create a dependent variable and independent variables. I then try to use simple (mostly linear) models to help me predict the dependent variable. The edge comes from good independent variables and good transformations. Once I get my coefficients I dont change them. Usually because I'm using like 10 years of data
How do you value vol?
- Thread starter .sigma
- Start date
Linear regression like OLS is a linear model. You can do better by using ridge,lasso, robust etc... one thing that I have learned from a member on this board that has helped me out is, try and get your dependent variable IID(independent and identically distributed) and make sure your independent variables are stationary (demeaning or scaling).
With this in mind dont get caught up in all the fancy BS Marcos Lopez de Prado talks about.
"Everything should be made as simple as possible, but no simpler"
Last edited:
Genuine question TBS - does any of this mathematical modelling have a measureable effect on increasing your profitability?
There are algos like Historic Volatility and then there are simple algos like Highest High minus Lowest Low over a period of time. Finally, there is VIX.
Which does everyone use to help set their stop loss points ? With recent volatility, fixed stops have been getting decimated IMHO.
Which does everyone use to help set their stop loss points ? With recent volatility, fixed stops have been getting decimated IMHO.
Last edited:
Firstly, nothing I do is complex. Simple things like linear regression is 90% of what I do. Things like PCA, scaling etc..are used to transform variables.
To answer the question. It helps me A LOT. Let's say for example I am trying to figure out what the variance risk premium is likely to be in the future for SPX. How would I do it without some type of mathematics or data science?
Here is one of Blair Hull's employees on how they try to forecast the equity risk premium.
What I am curious about is how do you trade without doing any modeling?
I model using ONE software - this may be laughable to someone who likes to wrestle with differential calculus, but it's pragmatic.
I can drive a BMW car, and all I need to know is how it handles under various conditions of rain, wind, slippery roads, bends, hills etc. I do not need to understand the thermodynamics of the engine, nor the heat dissipation functionality behind it - I don't even need to know what the letters BMW stand for.
I'm not questioning the advanced quant theorams - I just want to know how useful the mathematical analysis is in terms of $$$. And where. I await the day when someone on ET gives a real-life trade example where quant analysis improved the profit figure.
Second answer - quite successfully. As you know, one of my portfolios is trading the strategies learnt from a subscription service - for Jan and Feb, I had remarkable success (34% and 31% return on portfolio) trading earnings calendars. And for those trades I didn't even need ONE.
But as they say, different strokes for different folks.
Happy trading.
IIRC, this was steadyOptions software you were using. I thought he buys options into earnings yet I see you have mentioned trading post-earnings quite a few times (double calendars).
LOL I am not very smart when you compare me to others in this field (and on this board). Its actually super basic stuff.
Give me a few hours, I will do an example. Hopefully I will be able to get some criticsm from others. Spoiler alert* there will be no quant theorems. Just simple statistics while using financial knowledge.
You are correct. SO mainly trades pre-earnings. I run numerous portfolios (SO strategies, short selling, commodities, futures, currencies etc), one of which is based on the SO calendars. The post-earnings double cals were my own ideas in a different portfolio.
Who is SO?