Discussion in 'Journals' started by globalarbtrader, Feb 11, 2015.
page 155 print edition
What you do is backtest your system, which you can do for different signals. This gives you r_a r_b... r_z where each r_ is the return of a signal over some period. You then add these up and get r_sim. You then measure your actual total performance r_actual, and calculate the total difference r_diff = r_sim - r_actual. You then attribute r_diff amongst your forecasting rules using the forecast weights, and subtract that from each of r_a .... r_z.
You now have a series of backtested signal returns, which if you add up will give the live trading performance.
This will do particularly poorly if your rules have very different speeds, and also (obviously) your backtest is rubbish.
sorry. I should have kept reading
Yes. Please finish it before asking any more questions. I don't want to spoil the ending.
The final twist was excellent. I simply didn't see it coming. Who'd have thought the one-legged transsexual dwarf was the key to everything!?
It was the bi-gnome-ial all along!
Is there a reason why you went down the swigibpy route vs using FIX? Would you do the same again?
My initial, cursory, research revealed a couple of different ways to use python with the API (there are a few more now). I chose swigiby purely because I couldn't get IBpy to work. Even now after googling I can't see any python FIX -> interactive brokers packages out there so I'd have to write something myself. And I have zero experience with FIX.
Here https://github.com/robcarver17/pysystemtrade/blob/master/sysdata/legacycsv/VIX_carrydata.csv and here https://github.com/robcarver17/pysystemtrade/blob/master/sysdata/legacycsv/WHEAT_carrydata.csv the maturity date of CARRY_CONTRACT is less than the maturity date of PRICE_CONTRACT. However, here https://github.com/robcarver17/pysystemtrade/blob/master/sysdata/legacycsv/SP500_carrydata.csv it is the other way around. My understanding is PRICE_CONTRACT should have earlier maturity than CARRY_CONTRACT, isn`t this correct?
No it depends on the contract. Ideally CARRY would be earlier than PRICE but that's not aways possible if you're already trading the front contract. I suggest you read this http://qoppac.blogspot.co.uk/2015/05/systems-building-futures-rolling.html ("The need to measure contango") or appendix B of my book on the carry rule p.285 print edition
Separate names with a comma.