Fully automated futures trading
- Thread starter globalarbtrader
- Start date
To roll, I use the past year's history of rolls as general guidelines. So, if an instrument rolled on March 15th 2017, on March 1st 2018, I will start looking at the volume and the ba spread of the logical next contract. Once the ba spread and volume of the next contract are fairly close to the one I'm holding, which would probably be close to March 15th 2018, I elect to roll. I have a separate python program that gathers prices each day from the quadl api. So, I can tell it to roll, and I specify the new contract I will be holding, as well as the next carry contract, with which I'll measure carry. The program then starts collecting the prices on the new contracts, each day. For the continuous price csv file, the program takes the prior day's price of the new contract minus the prior day's price of the old contract and adds the result to every historical price (before the date of the roll) in the continuous price series. I believe this is the correct implementation of the Panama method, which GAT recommends.
GAT,
I had a few comments on your posts about the (dis)advantages of being large and small when it comes to extracting alpha:
[retail]- a smaller player has a massive advantage when it comes to cross-sectional strategies due to smoother weights
[retail] - there is a garden variety of regulatory arbitrages that are not accesible to the institutional investor, like anything that has to do with Basel II/III
[fund] - a large player has an economy of scale when it comes to technology costs or data costs
I am especially looking for your thoughts on the first and the second.
I had a few comments on your posts about the (dis)advantages of being large and small when it comes to extracting alpha:
[retail]- a smaller player has a massive advantage when it comes to cross-sectional strategies due to smoother weights
[retail] - there is a garden variety of regulatory arbitrages that are not accesible to the institutional investor, like anything that has to do with Basel II/III
[fund] - a large player has an economy of scale when it comes to technology costs or data costs
I am especially looking for your thoughts on the first and the second.
I think he's saying he can trade microcaps in the same $size as largecaps.
First one is similar to what truetype said. Let’s say you find some cross-sectional alpha in a set of 500 mid-cap stocks. In general, you will see that equal weighted performance is much better than liquidity weighted performance. So it would be attractive to a guy trying to deploy a million dollars in gross notional value, but not to a PM like me that’s trying to deploy a yard. It’s especially visible when you trying to play in an asset space where liquidity decreases exponentially like corporate fixed income.
Second one is just as important. As a retail trader, you don’t care about arbitrary restrictions. Take a simple example of Russian sanctions on companies like Lukoil - as a retail trader one should have been running around building div carry portfolios against other oil firms. I can name a few more off the top of my head. Many such opportunities persist for years.
In some things, these firms are going to be providing the required risk premium. In some, they would (and definitely do) align themselves with other larger industry players. There is such an abundance of opportunities in that space and new ones are being “built” as we speak.
One of the ways your can roughly classify your alpha is by source of diversification. In a cross-sectional strategy you have a lot of dislocations across some universe. So you are not doing many trades per name, but do a lot because you got multiple names in your universe. In a longitudinal strategy, you look at a small universe but trade do a lot of trades along the time axis. Obviously, you can think of examples somewhere in between the two.
@globalarbtrader :
Do you clip forecasts before weighting them? Currently, my vstoxx forecasts are:
'ewmac8': -0.50
'ewmac16': -9.45
'ewmac32': -23.78
'ewmac64': -43.51
'carry': -12.88
I weight them, and then clip them, which gives me a solid -20 overall. Doing it this way improved performance on backtests iirc, but I'm now 60 contracts short here (450k, 25% vol), so not sure whether I should be OK with that or not.
Do you clip forecasts before weighting them? Currently, my vstoxx forecasts are:
'ewmac8': -0.50
'ewmac16': -9.45
'ewmac32': -23.78
'ewmac64': -43.51
'carry': -12.88
I weight them, and then clip them, which gives me a solid -20 overall. Doing it this way improved performance on backtests iirc, but I'm now 60 contracts short here (450k, 25% vol), so not sure whether I should be OK with that or not.