This book is free today only: Machine Learning for Asset Managers Marcos M. López de Prado Cornell University https://www.cambridge.org/core/elem.../6D9211305EA2E425D33A9F38D0AE3545/core-reader
scoop, https://www.cambridge.org/core/serv...R.pdf/machine_learning_for_asset_managers.pdf downloader
Neither SLE nor Kevin Schmit are fans of this guy. Just fair warning...their reviews of his work seem legit to me.
Yeah, all such books and research are just resources to borrow from. And not sure that anyone is a fan of anyone else in this business
Actually, I used to have this approach. No longer... With material like De Prado’s, for a layperson like me, you can be lead very very far astray. Without the requisite technical background, you aren’t equipped to judge the quality of what you’re learning. So you build the wrong foundation around nonsense because everything looks shiny and exciting AND plausible (again because you might not have the requisite background). When learning something new, I focus on on vetting what other practitioners say about an author. I’d rather learn 4-5 basic things from a simple book than learn many advanced and wrong concepts form a De Prado just for the chance of learning a couple sexy/complicated concepts. I feel like many of these books should be peer reviewed. My 2c
Just got it, seems like an interesting read by the reviews I've just read. Thanks! Anyone here read it already? What are your thoughts? Is it worth the time spent reading? (I know every book is valuable and worth reading lol, just curious if you'd read something else over this or?)
You can just quickly scan through the book to see what's in it. That's what I did. Generally I was impressed by his story of how his team handled a flash crash and how his system predicted it shortly before it happened. I also think that his main tool for dealing with overfitting is Monte Carlo, while evaluating multiple Sharpes. This and other type of stuff could be helpful or an eye opener to some, or could lead astray as @Straitjacket mentioned. I guess a lot depends on your own approach and/or already utilized methodology.