It seems that Excel breaks down for models this large. Clearly, computation doesn't scale well. And, even scrolling through all the tabs is a PITA. Do you think the model builders actually used Excel, or did they use something else and then convert it over so that they might "Open Source" it?
I'm interested as somebody who would like to build a better tool for working with such large models.
I would not be surprised if the model builders used Excel. I work with a few investment banks and hedge funds, and at least half use Excel -- though most of them are quantitative funds, so they have a few traders using Excel and many more quants implementing the actual models in C/C++. If Ackman were more of a PhD than an MBA, I would suspect him of building it in something else and releasing the kludgy model to outsiders.
Do you think there's a market for something in between Excel and raw code? I'm thinking of some sort of Excel/Rules/Code hybrid designed for large, complicated stuff. Good verification tools, multiple what-if analyses that could be run simultaneously, etc. I'm just trying to think through all the problems I've had when trying to use Excel for models even smaller (much!) and simpler than this one.
I'd like to think so, and there probably is, but the issue with Excel is not technical superiority but market dominance. If there were some tool that could be gracefully upgraded to harder and harder problems, but that was designed for this kind of thing, it would be ideal. Excel is a teetering mountain of compromises and (bad) hacks.
I have messed around with using Arc for something like this (my initial plan: a way to keep track of stock positions, but to treat arbitrage positions as a unit -- instead of "100 shares of XYZ, and -100 shares of ABC, which are merging next week in a 1:1 share swap," it would be "1 unit of XYZ/ABC merger.")
I'd be a bit shocked if people (particularly the more academic types) didn't use matlab or R. Anyone know if these two are prevalent in financial shops of various types?
I have seen specs calling for Matlab, R, S+, and Mathematica. But my impression is that since the finished product is going to be in a high-performance, low-level language, they don't care what you do your rough-drafting in as long as the finished product is C++. Exception: Jane St. Capital uses Ocaml; one of their top execs is a Lisp fan who has been credited in pg essays.
byrenesreview basically answered below where he said at least half of hedge funds and traders use Excel. Thing is, hedge funds, quants in general will be way more likely to use real modelling software than everybody else, so byrne's experience of them using anything but Excel is skewed.
In I-banking, Private Equity, management consulting, the financial industry generally, Excel is king. If you look at BankersBall they have a tab for excel tricks. Dealmaven got bought for $14m recently, and their product (apart from the bootcamp, which obviously doesn't scale) was a financial modelling package, or rather, training in modelling, along with macros, shortcuts and a manual for Excel, as it's used in finance (really very good training in modelling, as you'd expect from a firm with a Wharton prof on board)
Mssr.Byrne Hobart, thank you for an excellent link. Your blog is also quite good. I'm going to have to dig around more.
Having played with the actual Excel spreadsheet, this seems like something that probably could have been done with Mathematica using less horsepower...or am I missing something?
Great! Even the greatest trading strategy, once made available to the very general public (FT readers), is worthless, because you can't make abnormal returns by doing what everyone is doing.
It's a good learning tool, though. Strategies like this have been "opensourced" by academia for decades.
It may be personal. I remember reading his analysis of MBIA on valueinvestorsclub.com in 2002, and he was just as down on it then (though at the time he used kilobytes of prose instead of megabytes of numbers).
I'm interested as somebody who would like to build a better tool for working with such large models.