No doubt good data sources exist. They add too much complexity and some extra sources of deviations to my taste. Maybe for someone who is specifically looking for the "events" related edge that would be more useful. Btw, In a past I used a few and liked AAII fundamentals which is reasonably priced and I think was going back 5-10 years. Even wrote AmiBroker integration for their data. Can't find a screenshot, maybe will show it off later. I basically recreated MarketSmith screens with relative strength, industry info, earnings / fundamentals graphs alongside with prices. It looked very informative and cool. But I think it probably would have made me way more money if I would to sell it to other traders Absolutely nothing in terms of execution / positions management. But for backtests all data is dividends / splits adjusted to account for the impact. RealTest also accounts for dividends as my data provide includes that info. So backtest will +/- them depending on long/short position. EDIT: Found an old screenshot of my techno-fundamental screen for AmiBroker. Sorry, resolution is pretty low. Val
September was a good month. +8.83% over 179 trades. After a crazy winning streak early this month, 18 up days in a row, this week everything feels completely out of sync with the market. There is nothing to do about it as all is within expected boundaries. Val
"18 up days in a row" Congrats! "everything feels completely out of sync with the market." Maybe a side effect of Announcement City today and according to overnight, all week?
I haven't followed news or announcements for a long time so it is hard to say what is truly unusual, except the obvious unemployment stats. And even those seems like not the kind of things market cares about these days Val
"like" is probably a wrong word. I do specialize on stocks / daily / <5d days hold. Average hold is probably just around 2-3 days. All scans + executions are automated and I don't normally mess with them. There are 5 systems running at the moment. They have little correlation with each other and are designed to take advantage of different market conditions and not to loose too much while waiting for those conditions. Some systems have a bit of industry specific knowledge but it is pretty minimal. For instance certain industries will tend to have more fat tails than others. Part of it is cyclical, but sometimes it can be a fairly consistent tendency across most years. Val
Hi Val Really enjoying your journal, I have been playing around with RealTest over the past week or so and I like the multisystem testing capabilities. I was wondering if you could help me with the following dilemma. I am trying to test a long short mean reversion system that exits on the close of the entry bar. The maximum number of positions of I want filled is 10 across the portfolio. I am not concerned about the ratio between the two systems. The snippet below is allowing 20 positions across the portfolio, was wondering if you could point me in the right direction. i have tried a number of things just can quite get it. Great year you are having with the mean reversion trades too!
It's been a long day.. My quick guess will be: MaxPositions: NumPos MaxEntries: NumPos - Combined(S.Positions) Where "Combined" accesses all strategies' data and S.Positions stores the number of currently open positions calculated at the previous close, so if used together that should be give positions across all strategies. After you run the test look at results, Setups / Positions / Entries tabs. That will help you to see if it works the way you wanted. Though, I would suggest to use "MaxEntries: NumPos - S.Positions", which means you will have separate max per strategy and just see how often both strategies combined hit your total max. Could be never. The advantage is - you will remove a degree of freedom in your backtest related to which strategy will use up the slots first. Because those days when any of them actually need lots of slots might be the biggest sources of your model vs live discrepancy and potentially unexpected DD. So better look first at what happens in those cases and ideally eliminate this possibility. Might so happen that it is not even a problem, but them you will know for sure and there is nothing to worry about then. Val