This is a review of Nerds on Wall Street: Math, Machines and Wired Markets
UPDATE: A website has been put up with more info on the book, nerdsonwallstreet.com
I got a pre-release copy from Dave Leinweber, the author, who I know through Berkeley's Center for Innovative Financial Technology (CIFT).
The three parts:
1) the history of financial technology
2) alpha generation ideas
3) what quants can do about the financial crisis
The first part is hilarious. One of the many pictures is of a past president of the CME banging on a gong with a giant salami.
It's a really irreverant look at Wall Street from the start- "In 1792, the New York Stock Exchange was a bunch of guys standing around a buttonwood tree at 68 Wall Street shouting at each other on days when it didn't rain or snow. We like our markets to be liquid, efficient, resilient, and robust. But this is hard when all the participants have to crowd around a tree and hope for good weather. So in 1794, we see the first big technological solution: the roof. ..."
Dave was one of the fathers of AI-based trading so the second part "Alpha as Life" really stoked my imagination. I was reading along with a pencil and paper to look for typos in the pre-release copy but I ended up writing down a new multi-factor trading strategy for small cap stocks along the way. The book is especially great for text-based strategies. It's amazing how many details he shares about old trading software companies he founded which were vendors of billion-dollar funds.
The book was fascinating because of the author's insider status over an already long and prolific career. He's come into contact with so many famous personalities in the trading world and he has a great way of characterizing people, as, well, people. It's almost 400 pages but I had no trouble reading it in a day.
Check out its reviews too- the first is by David Shaw (founder of D.E. Shaw). Here's the book jacket, with the table of contents and other reviews.
Please leave a comment or question about anything. I'm a little worried I scared everyone away with the past three notes on the Dirichlet Process.
7 comments:
Yeah,Your worry is inevitable.I guess Couple of Chaps fainted immediately while reading Dirichlet Process.
Chintan,
I hope not- there will be at least one more note on the DP, with code, which could be even more dangerous.
In the book, Dave mentions that each equation lowers a book's sales by 50%. Hopefully this doesn't apply to websites.
Regards,
Max
Max - I am looking forward to the book. It sounds good. I read a ton of varied stuff & get a lot of ideas from them.
Regarding your worry about the previous posts, I suspect that the topic was just way over most people's heads. I know it was for me.
I think one area of discussion that would be very useful to many people would be how to combine multiple trading ideas/systems into one consolidated system. Often that improves ones results better than coming up with a single 'best' strategy.
Eric
Great idea Eric. I've been thinking a lot about that too. Classification and regression trees plus bootstrapping seems especially promising to me. It naturally handles categorical and numerical data, it's inherently nonlinear, overfitting/complexity can be partially controlled a priori, and bootstrapping gives us a number of extra benefits, most importantly confidence values for each prediction. Support vector machines have none of these specific features. I'll try to accelerate work on that and implement it in code to share.
Regards,
Max
I am really interested in estimating liquidity. As an example maximum position one can build/exit in xxxx without affecting price too much.
Is there any formula for this?
Pratik,
There is no formula liquidity. It's actually a huge problem for institutional investors. I haven't looked into it much but there is lots of literature out there.
Regards,
Max
Pratik:
I have a very simple idea on this, I think it could be a good point to start from.
Simply take the average volume of the last 25 trading days and substract with the avarage (close) price of this period, then divide the result with 10 or 20.
When you have this, you will have intraday volume data too. Analyze "volume distribution", then write an algorythm which randomly distributes packets in size and by time.
And there is an even easier way. Spread bad news about the company, and buy in the downtrend. But don't cry if your ass is being kicked :D
Post a Comment