Fully automated futures trading

Discussion in 'Journals' started by globalarbtrader, Feb 11, 2015.

  1. newwurldmn

    newwurldmn

    You need some method to determine if there is any risk premia to be harvested. How else would you do it?
     
    #1121     Nov 13, 2017
  2. Well backtesting tells me if a particular risk premium has paid off in the past. And we'd normally build our trading rules so that forecasts are proportional to expected return.

    If I take something like value in the form of buying low PE stocks then yes, you could construct a trading rule like this:

    E(sharpe ratio) = b * (EP - rf) / vol

    Where EP is the earnings yield (inverse of PE ratio) and rf is the risk free rate, and vol is the standard deviation. b is then the 'price of risk'. But how do we determine the price of 'value' risk? The only way to do is to backtest and take something like a long run average. But in my trading system b is the amount of capital allocated to this trading rule (multiplied by a scalar which isn't important). So in an implicit way I don't need to work out the price of risk.

    So I think it's pointless, although if you like this kind of thing then "Efficiently Inefficient" by Pedersen is full of it (and a great book as well).

    GAT
     
    #1122     Nov 13, 2017
  3. newwurldmn

    newwurldmn

    Thanks for your time so far in helping me understand how this strategy works.

    How do you know what to buy or sell? Are you looking for moving average relationships to determine if you should buy oil for example? Or is it some persistent phenominen (like convertible arbitrage - where embedded options are typically cheap to their listed counterparts). It seems like all the CTA's make and lose money together so whatever they are doing, it seems they are all basically doing it the same (like convertible arb funds).
     
    #1123     Nov 13, 2017
  4. To pick up trends then yes, you'd use something like a moving average. Most CTAs use pretty similar methods for picking up trends so they have correlated returns.

    GAT
     
    #1124     Nov 13, 2017
  5. srinir

    srinir

    GAT allow me to post one of your link.

    Here is the link from his website which combines trading rules like break out, MA and carry which generates buys and sells.

    https://qoppac.blogspot.co.uk/2016/05/a-simple-breakout-trading-rule.html
     
    #1125     Nov 13, 2017
  6. Elder

    Elder

    Hi GAT, minor thing. Just doing a comparison between your legacy stitched prices and mine and everything looks to be fairly consistent. However I noticed what looks like a rogue datapoint in your EUR file. On 11/06/2015 it shows a ~ 20 pts fall from previous day (1.329 -> 1.139). The carrydata file does not show this jump so its probably just a bad stitch.

    Probably does not make a blind bit of difference to your system but I though I would mention it, just in case.
     
    #1126     Nov 16, 2017
  7. Thanks for the heads up

    GAT
     
    #1127     Nov 16, 2017
  8. ETA

    ETA

    Hi GAT,

    I am concerned about the data in your trade algo pysystemtrade. The whole system is largely self contained with the data in the legacycsv folder. However, how do I update it with more recent data. I used Quandl and use Vix_price as an example, I am unable to replicate the same price history from Quandl and the file data file in the legacycsv folder.

    Is that because you made a back adjustment everytime you do the update. Wonder if you could share on how do you update your data files for your system.

    Wanted to figure out myself how to sync data downloaded from Quandl to the existing price data but I got more confused.

    Thanks

    ETA
     
    #1128     Nov 16, 2017
  9. Personally, I have found this post of GAT's very helpful: https://qoppac.blogspot.co.uk/2015/05/systems-building-futures-rolling.html?m=1
     
    #1129     Nov 16, 2017
  10. ETA

    ETA

    #1130     Nov 17, 2017