Fortune‘s Dan Primack and Reuters’ Felix Salmon have two great post-mortems on this week’s Snapchat venture capital deal. Primack stresses the misalignment of VC (venture capital) and LP (limited partner) interests in how transactions are structured. Salmon points out that VC deals are framed for “growth and exit” rather than “entrepreneurs building for the long-term.” The exchanges between Primack and Salmon — and between their commenters — are an example of what can make blogging a faster learning experience than traditional sources. Time to brush up on your copy of Venture Deals, The Business of Venture Capital, or Term Sheets and Valuations.
For the past several months I’ve been learning about machine learning and algorithms to build a quantitative trading capability.
Jeff Dean and his team at Google Research have developed DistBelief: a deep network that uses unsupervised learning, thousands of computers, and billions of parameters to train new models:
Recent work in unsupervised feature learning and deep learning has shown that being able to train large models can dramatically improve performance. In this paper, we consider the problem of training a deep network with billions of parameters using tens of thousands of CPU cores. We have developed a software framework called DistBelief that can utilize computing clusters with thousands of machines to train large models. Within this framework, we have developed two algorithms for large-scale distributed training: (i) Downpour SGD, an asynchronous stochastic gradient descent procedure supporting a large number of model replicas, and (ii) Sandblaster, a framework that supports a variety of distributed batch optimization procedures, including a distributed implementation of L-BFGS. Downpour SGD and Sandblaster L-BFGS both increase the scale and speed of deep network training. We have successfully used our system to train a deep network 30x larger than previously reported in the literature, and achieves state-of-the-art performance on ImageNet, a visual object recognition task with 16 million images and 21k categories. We show that these same techniques dramatically accelerate the training of a more modestly- sized deep network for a commercial speech recognition service. Although we focus on and report performance of these methods as applied to training large neural networks, the underlying algorithms are applicable to any gradient-based machine learning algorithm.
DistBelief has several interesting implications. It might underpin a new wave of consumer-oriented apps which use a machine learning architecture. It will likely discover new combinatorics and permutations of model parameters. Applied to historical and real-time data from financial markets it might discover new model parameters, usable for quantitative trading. This would potentially change our mental models of financial markets in an evolutionary way just as high-frequency econometrics disrupted indicator-based technical analysis models over a decade ago.
Collaborator Roy Christopher has released his annual Summer Reading List for 2013, with contributions from danah boyd, Janet Murray, Richard Kadrey, Howard Rheingold, Mark Amerika, Matthew Kirschenbaum, Gareth Branwyn, Peter Lunenfeld, RoyC, myself, and others.
My contribution this year is a mix of books by allies, PhD research, and trading system development.
Portfolio Design focuses on return drivers and risk premia for the major asset classes used to allocate assets in diversified portfolios. The ‘modern approach’ includes a focus on data analytics and the available evidence base; a review of recent (Fama-French-influenced) quantitative finance studies; and insights on index benchmarks and portfolio construction.
I’m a couple of chapters in: Marston focuses on smallcap and value return drivers, in contrast to largecap and growth stocks. Although he doesn’t explore this Marston’s insights in his opening chapters are also relevant to two other areas:
- Penny stick promoters rely on ‘crowded trades’ and rational herding: this is often what the ‘free’ investment newsletters and watchlists are for, to create drawdown market dynamics in relatively illiquid stocks.
- Venture capital’s asymmetric returns can have large payoffs if opportunity evaluation screening works, and if stage one financing backs a break-out company that creates a new, large market or is a successful disruptive entrant into existing markets.
It’s interesting to see how an experienced, endowment investment adviser handles index benchmarks, portfolio construction, and specific return drivers. Since he is also on Yale’s investment committee Portfolio Design also makes an interesting counterpart to the influential Yale Swensen model detailed in Swensen’s Pioneering Portfolio Management: An Unconventional Approach to Institutional Investment (rev ed.) (New York: The Free Press, 2009).
The Australian Transaction Reports and Analysis Centre (AUSTRAC) has an important role in financial intelligence and anti-money laundering initiatives in Australia.
ANAO’s audit opinion has some interesting reflections on AUSTRAC’s financial intelligence function:
- Partner agency access to AUSTRAC data should be regularly reviewed by AUSTRAC personnel.
- Processing time targets and processing backlogs for financial intelligence assessment should be monitored by AUSTRAC management.
- Key Performance Indicators and structured feedback from partner agencies should be developed for financial intelligence.
These are a mix of high-level strategic planning; operational review; and partner corporate governance. The ANAO audit opinion is an interesting read on AUSTRAC’s organisational evolution — its financial intelligence methodologies remain confidential.
For some possible answers, we can look to the current literature in intelligence studies on structured analytic techniques and methodologies. The US-based Central Intelligence Agency released A Tradecraft Primer in 2009 (PDF). Ricahrd J. Heuer Jr and Randolph H. Pherson’s Structured Analytic Techniques for Intelligence Analysis (Washington DC: CQ Press College, 2010), and Sarah Miller Beebe and Randolph H. Pherson’s Cases in Intelligence Analysis: Structured Analytic Techniques in Action (Washington DC: CQ Press College, 2011) are also excellent tradecraft primers.
EquaMetrics’ app is simply designed and since its software firepower comes from the cloud, it doesn’t require anything more than the typical PC. You can drag and drop colored tiles to assemble your own algorithm. Day traders can choose between 30 variables to build their formulas. The options are built on so-called technical indicators, metrics that reflect trading patterns as opposed to stock fundamentals such as the price-earnings ratio. After you’re done, you run the program to buy and sell stocks and currencies.
The web application is relatively inexpensive: it costs $99 a month or $250 a month, depending on how many algorithms you want to run. That’s a steal compared to the alternative of hiring a quantitative programmer for $200,000 a year. EquaMetrics gives you the stuff a programmer could produce. Then it’s up to you to assemble your own strategy.
I have been expecting apps like this for several months, and have been monitoring other initiatives like the Quantopian community. The popular literature on algorithmic trading strategies evolved from Technical Analysis mechanical systems (Tushar S. Chande’s Beyond Technical Analysis) to back-testing (Robert Pardo’s The Evaluation and Optimization of Trading Strategies) and then to algo trading using Matlab software (Ernie Chan’s Quantitative Trading and his new Algorithmic Trading; and Barry Johnson’s Algorithmic Trading & DMA). This period spans the post-dotcom collapse; the 2003-08 speculative bubble in real estate and asset-backed securitisation; and institutional experimentation with high-frequency trading platforms, and transaction and execution costs.
EquaMetrics’ strategy reflects this decade-long evolution:
- Its initial offering is Technical Analysis strategies: at a time when: (a) high-frequency trading has ‘broken’ many trend-following and momentum indicators; and (b) hedge funds and proprietary trading desks use predatory trading to clean out TA-oriented retail traders.
- The model is subscription-based software as a service — which could eventually disrupt or change the economics of agile software programming if this offering scales up in a significant way. Will the $US99-250 per month price point remain? Or will another platform develop a lower-priced offering and trigger a ‘race to the bottom’ competitive dynamic?
- It opens the way for the licensing of specific TA indicators and proprietary methods as ancillary revenue streams, and as a way to build a market around the core product offering (which NinjaTrader, MetaStation, and ESignal have all done with their respective platforms).
- The quality and scope of the back-tested data is important: quantitative hedge funds like Jim Simons’ Renaissance and David Shaw’s D.E. Shaw & Co each clean their own data.
- EquaMetrics’ move into fundamental indicators reflects some recently published work on the quantitative analysis of these strategies (notably, Richard Tortoriello’s Quantitative Strategies for Achieving Alpha, and Wesley Gray and Tobias Carlisle’s Quantitative Value).
- EquaMetrics’ choice of FXCM and Interactive Brokers as prime brokers to process client trades is significant: brokerage transaction and execution costs can mean a potential, new trading strategy is actually unprofitable to execute, or that its profit-taking ability declines over time, especially in correlated and ‘crowded trade’ markets.
- The focus on TA and fundamental indicators does not address some of the quantitative, statistical or machine learning strategies that quantitative hedge funds use to develop algorithms; how correlation testing of model variables might occur; and what might happen to retail investors once several different competing firms have back-tested and issued dueling algorithms (a factor in high-frequency markets where scalping and order front-running occurs).
Still, the EquaMetrics offering has me interested: I’ve been waiting for algorithmic trading to ‘value migrate’ (Adrian Slywotzky) to retail traders, for awhile. It’s a first step towards post-human trading (Charles Stross’s novel Accelerando).
There are two kinds of Wall Street films: the techno-science thriller involving financial risk management and VaR (The Bank; Margin Call; Arbitrage); and the elite trader’s excessive lifestyle (Wall Street; Boiler Room; Wall Street: Money Never Sleeps). Martin Scorsese’s forthcoming film adaptation of Jordan Belfort’s ‘pump and dump’ memoir The Wolf of Wall Street (New York: Bantam, 2007) goes for the latter, because it appeals to a far broader audience. The trailer benefits from Kanye West’s pulsating new song ‘Black Skinhead’ (from Yeezus).
Belfort’s friends got him the film deal with Scorsese and actor Leonardo DiCaprio whilst Belfort was resurrecting his sales career on the Australian seminar circuit (PDF). In just a few years Belfort transformed his Straight Line Sales technique from free MP3 interviews into slick website and seminar circuit presentations on sales psychology. Josh Brown outlines in his book Backstage Wall Street: An Insider’s Guide to Knowing Who to Trust, Who to Run From, and How to Maximize Your Investments (New York: McGraw-Hill, 2012) how pitch books like Befort’s Straight Line works. Maybe Belfort’s seminar clients in Australia need to read Brown’s book and blog. It all reminds me of telemarketing stints I had in the mid-late 1990s, whilst reading Tom Hopkins and Zig Ziglar, and watching friends get recruited into Amway’s multi-level marketing schemes. Maybe Belfort will be able to pay back his clients, after all, if his deal includes a share of Scorsese’s film residuals.
The elite trader image of ‘conspicuous consumption’ has its roots in Thorstein Veblen‘s leisure class economics, and Reagan era money managers like Henry Kravis, Peter Lynch, Victor Sperandeo, and Martin Zweig. The traders I know enjoy the Hollywood media imagery. But they also know it is a myth that artificially inflates the subjective expectations of how trading actually works and what it is really like. The ‘excessive lifestyle’ myth actually serves as an entry barrier. Most novice and retail traders will more likely ‘blow up’ their trading accounts.
I’m a late-comer to Vince Gilligan’s television series Breaking Bad. Most people are waiting for Season 5’s second half in August. I’m at the end of Season 1. Walter White’s (Bryan Cranston) transformation from “Mr. Chips into Scarface” (Gilligan) might find its way into a PhD chapter.
A pivotal scene from Season 1 occurs during the finale of episode 6 ‘Crazy Handful of Nothin’‘. White has just shaved his hair due to chemotherapy. He confronts crystal methamphetamine dealer Tuco Salamanca (Raymond Cruz) who beat up White’s partner Jesse Pinkman (Aaron Paul). White’s bargaining leverage is the fulminated mercury he has bought with him as a small incendiary explosive, and which the episode foreshadowed in a high school chemistry class. The scene is a great negotiation clip that illustrates the madman theory in grand strategy and nuclear deterrence.
More significantly, it is the moment that White takes Action, and first transforms from a mild-mannered chemistry teacher (and former graduate student researcher) into his alter ego, Heisenberg. It’s a moment of Metic intelligence (craft, cunning, skill, wisdom). I’m looking forward to how White evolves, and what the consequences of his decisions are, in the remainder of Breaking Bad.