Funny, because I just listened to the podcast you’ve posted in another thread, and that guy somewhat points out that many strategies may have only 40% win rate (while winning larger amounts than losing), because high accuracy strategies are hard to come by, or they don’t find opportunities often. So therefore you’d have to have 100 high-accuracy strategies that can together find something to trade each day and offset each other’s losses, vs one or two “good ones” that lose often but make some money. Also, I checked his website and it was showing less than 4% annual return for his best strategy with around 55% win rate, which isn’t much and does seem costly in terms of risks and not doing something better with time and money. And “better” could mean running 100 high-accuracy strategies. Though indeed this can be costly when trying to find and possibly pay for a bunch of strategies. So really there isn’t an easy solution, and maybe that’s why large capital goes to hedge funds that can pull this off.
I didn't mean that "good" strategies are high win rate ones. I mean they have good risk adjusted returns. The article sounds like - hey, don't sweat too much about finding strategies with an edge ... Just trade as many good enough strategies together. But what is good enough? Kind of dangerous message seems to me.
I think that if you trade 10 or more profitable strategies, then overall you’d smooth out your risk/drawdowns, while amplifying profit potential. So “good enough” would simply need to be profitable, at the least. I’m currently mixing hundreds of strategies of all kinds, and the only requirement is for each to be profitable each year, or consistently profitable in some way. The results of mixing anything that’s profitable, no matter how little or how often, are amazing. But a string of a few bad trades can wipe out the profits, so there may be additional criteria for those strategies. Generally the bigger problem is when a strategy stops working, not how good it works when it is profitable even at 0.5%.
While your approach is nice and something I agree with, the shortcomings are clear - you don't have unlimited leverage and the correlation of returns at certain times. The first problem forced you to skip trades in certain strategies and that is difficult to model when you have variable margin requirements..
These are literally the problems I’m working on solving every day. But I see the light, and the light still leads to mixing multiple strategies. Though I guess they just cannot be sh*t. I prefer high accuracy (win rate) strategies, though not because they are better for mixing, but because they seem to be more reliable and continue performing going forward. At least the ones I’m looking at. I don’t use margin, just available cash, and small position sizing (though I’m also looking at those that would use full cash).
A "gaggle of PhD quants with hundreds/thousands of indicators"? That's laughable, considering they/we all need only ONE. That is... "be in tune with price".... and you can do that with "Price TA".
Originally I didn't want to post their url here as I thought the consensus would be overwhelmingly negative. It wasn't, so I guess no harm. Few people requested, so here it is https://robotwealth.com/
Here's an idea...use a bunch of strategies to smooth the equity curve (hopefully uncorrelated). When one of them starts to falter, swap out for another. That way, you don't put all your eggs in one strategy. Downside being it takes more effort to manage the trades. Even heavily curve-fitted strategies tend to work for awhile, the hope is that they don't all fail at the same time.
Baiting newbees with quant lingo. Awesome. Every day another trash product. A quick glance at the site makes that apparently clear. What is not clear is what those guys are even selling. A lot of mumbo-jumbo but nothing of substance. Example in case, the article on deep learning. Big bait title and nothing that relates deep learning to trading, something I do every day and claim to have some expertise in.
This might be good or bad strategy. Nobody can tell if you dont try it yourself. Eventually, any proposed new trading strategy should be well tested. You can use any demo account for this purpose. Dont rely too much on historical performance