Topic: Do we have a list of best practices and/or other protocol standards?
The deeper that I go into EA Studio, the more impressed I get. Although I'm familiar with some of these tools from past data experience, the fact that they're all packaged together in seamless fashion cuts out much of the grunt work and Excel spreadsheets.
That said, it also opens up a new world that hand-coded data can't easily move through. For example, I can re-optimize and re-randomize as much as I want, diminishing (or negative) returns can drift in at any time, and the real work becomes keeping track of seemingly random variables. I can dilute OOS or not, I can randomize X via Monte Carlo or keep it intact while randomizing something else, etc., which means there's an infinite variety of ways to make money with EA Studio, with or without following best practices.
But it's been years since the software has been released- do we have a list of best practices yet that aren't so controversial?
For example, these settings make sense to me in the Reactor:
1) Monte Carlo- randomize most settings, especially random bar start, but are the defaults too restrictive? For example, I can imagine 5-10% randomization working for many settings since you will not likely get too-steep deviations (depending on overall TF) with 30+ years of data (which I recognize is itself controversial).
2) OOS- there are arguments for any % setting offered, with the trade-off being fewer validated results at 50% (although the safest choice). I am thinking of sticking to 30-40% to be as conservative, but, again, I'm new to the software, and don't know if results are typically better with more aggression here given the huge # of diversified strategies we can run to further ameliorate risk elsewhere.
3) Is over-diversification a concern? Some only trade one pair with lots of strategies, others do a dozen or more. My first batch of tests will be on EURUSD + USDJPY since this not only captures 3 highly liquid markets, but also sets USD against itself, in a sense. Not that these two pairings are a hedge, but they will likely capture good upside for both over the next few years and allow you to limit downside.
4) Not sure how EA Studio's multi-market validation works. Different pairs do exhibit some different behaviors, but this is sometimes overstated. As a robustness tool, is MMV simply a way to test for generic strategies- meaning, for getting strategies that have a verifiable yet more generic edge across all markets (meaning, the chances of diluting get less and less)? For example, simply buying every daily pivot in a modified Martingale style will work for all pairs if you're comfortable with quick 1-2R wins. I can't imagine such an edge going dull since the concept of expensive/cheap will continue to exist along pivot lines.
5) Anything you've learned about these EA Portfolios more broadly?
6) Finally, on an unrelated note- can we have multiple instances of EA Studio running across multiple browser tabs? What's the best way to simultaneously test multiple time frames on multiple pairs, adding them to the total portfolio as I go along? I typically do 6-8 hours for strategies for each time frame on a pair.