Notifications
Clear all
Topic starter
28/08/2017 3:20 pm
Hey guys (and gals),
A couple of house keeping things as I dive into the next round of upgrades, features, and additions to Build Alpha.
First, I changed the Sharpe calculation to be the (average(trades) / stdev(trades)) * sqrt(252). Originally, it was this calculation but on daily marked to market results as opposed to just the trade results. A few of you requested this change and now a few of you have since noticed! I'd like to put it to a vote as there appears to be some dissention on which is preferred.
Secondly, multi-timeframe analysis is not yet supported as there are some nuances we need to take care of. I also want to put that to a vote as I'm under the impression after talking to a few of you that this is done in different ways.
Multi-timeframe (MTF) style 1. Assume you have market A that trades 24 hours per day. Market B only trades 600am to 2400.
At 600, assume we have the signal close[0] > close[1] for market A. The question is... should market A's close[1] be the bar prior to 600 (where no market B data exists) or should close[1] be the last market A bar where BOTH market A and market B exist? Most likely around 2400 of previous day?
I can clarify this, but I assume sum of you will be able to understand this and can provide feedback.
The main question is.. Would you prefer me to calculate the signals on each market independently and then combine the signals - OR - combined the data for both markets based on matching timestamps and then calculate their respective signals?
It is a small nuance but it can make a difference.
Thanks (forum response or email is fine - I know some value a more private life) as always,
Dave
PS if you're receiving forum updates via email and wish not to then please alter your profile settings. thanks all.
02/09/2017 12:33 am
market A's close should be the bar prior to 600 where not market B data exists.....
24/09/2017 5:33 pm
1. Marked to market daily.
2. Why not make it configurable?
16/10/2017 1:57 am
1. Marked to market daily
2. Bar where do B data exists. I think you'd want to know what happened in the last bar of trading activity while markets were open
17/10/2017 9:23 pm
I think sharpe should be closed trades only, aggregated by day for intraday systems. For longer term systems that trade much less frequently or hold for long periods of time daily MTM makes more sense
For MTF I think the larger timeframe signal should always persist until it can be calculated again, even if not updated. So if market A is daily bars, and market B is 5 minute bars, and the larger timeframe signal is close[0] > close[1] = 1, during the trading day the market A (daily) timeframe cannot update until close, so the signal will effectively return the same value for all bars on market B throughout the day. That way for the market where data does exist (either shorter timeframe or active trading hours, etc) can always calculate a signal on its own data and will always have the latest available signal from the larger timeframe or last traded bar.
I believe this also mirrors how many people do discretionary MTF analysis where they may look at larger timeframe data that is somewhat stale relative to the short timeframe data they are using to make a trading decision.
I also think this approach would work for combining signals from one market with a time based bar type and another market with event based bar types
So to answer your question above Market A close[1] should be the bar prior to 600 where no Market B data exists.
Thx - kyran
WN reacted
