Open source off-shoring

Following a post in Enypher blog, I came across imatix.com, the company behind OpenAMQ. OpenAMQ is an open source implementation of AMQ, a messaging middleware protocol designed for low latency, critical applications. JPMorgan contributed to the project from 2004 and migrated trading desks globally over the last 3 years. iMatix claims zero downtime since 2006.

The middleware open source offering is spreading its wings. We are now seeing a number of projects challenging commercial applications: I mentioned RedHat MRG in an earlier post, OpenMPI is the product of years of research in multi-core systems and parallelism, Hadoop is a data management and clustering solution sponsored by Google, Yahoo and HP. 0MQ is the last open source initiative for very low-latency middleware, backed by FastMQ and Intel.

This prompts for the question: How much money has been saved by financial corporations sponsoring open-source initiatives vs. structural off-shoring?

For the last 10 years (ever since I got involved in trading systems development) the refrain of every business manager was about IT development going to India. Entire IT departments were counting their days and Bangalore was going to become the new mecca of finance IT development. Starting with back-office systems and gradually moving to the middle-office and desk applications, the future of IT management was in off-shoring. Keyword: cost-cutting. Ah..., the commoditization of finance IT...

Today Bangalore is the Silicon Valley of India and --guess what-- production costs there have risen. At the same time we have seen a number of open source projects emerging as the de-facto standard in finance: in the 90s, X11 and Motif were the state-of-the-art for front-office GUIs; Linux is now used in critical server applications; Cygwin X Server offers a no-cost alternative to Exceed; SOAP, XML and HTML are open standards; Samba and SVN are the best tools I know for remote file access and source code management in terms of price/quality (hint: price = 0). Is MySQL next?

In my view, the fundamental error lies in seeing software development as an industrial process, rather than a creative intellectual effort. Not surprisingly, the IT and the Business units in most investment banks have been organized as two separate entities, with the latter framed mainly as a support center rather than core-business. Today the move that saw venerable technology institutions (HP, IBM, DEC, SUN...) supporting academic efforts in the past might spread to financial institutions supporting open-source projects. After all, it's all about cost-cutting.

Powered by ScribeFire.

Technorati Tags: open source

Bill Ackman Open Source Research

So now even Bill Ackman is embracing the open source model :) In this open letter to MBIA's and Ambac's state insurance regulators, the SEC and the Federal Reserve he states:

"Our primary goal is to initiate what we call “Open Source Research” where all market participants can have equal access to the primary source data and construct their own views of losses without reliance on the analytical judgment of rating agencies or the bond insurance industry. By focusing the discussion on a fundamental, data-driven approach, we expect that the dissemination of the Open Source Model will enable market participants and regulators to accurately estimate probable losses by relying on rigorous fundamental analysis of specific credit exposures, a departure from relying on the opaque, faith-based pronouncements that the bond insurance industry has promulgated to the marketplace."

Bill Ackman runs Pershing Square Capital Management, a hedge fund returning 22% in 2007 betting on the demise of MBIA and Ambac Financial.

Powered by ScribeFire.

Commodity model

Today I opened the first position in the new COM model, a ranking system based on commodities stocks that will replace HFCF in the Demo portfolio. I made this change based on the 6 months GSG momentum and the recent small cap underperformance. The 2008 Demo portfolio will thus be composed of NH, LKS, PCF, COM (5 positions each) and SHORT (10 positions). I might add a large-cap model going forward.

The idea for the COM model comes from the Motley Fool Mechanical Investing forums where the screen was originally designed as a commodity stocks universe sorted by low Short Interest ratio. I made a few changes to include my liquidity and cash flow requirements and replaced the regular SI as SI % of outstanding shares but for the rest the screen didn't change substantially. The goal is to overperform GSG in 2008.

Powered by ScribeFire.

Technorati Tags: investing, quantitative trading, commodities model

2008 Demo Portfolio

The 2008 Demo Portfolio lists all the real trades executed in one of my brokerage accounts net of commissions costs and YTD P/L. I generally update the spreadsheet on the same day orders were executed.

The first page of the spreadsheet shows the portfolio YTD P/L and the return for a benchmark I built based on a set of equity indexes. The portfolio includes 4 long and 1 short model rebalanced weekly, with each long model trading 5 stocks, for a total of 20 securities on the long side. The short model triggers a limited number of trades and is composed of around 8 stocks on average. Currently the funds allocation is 120% NAV long and 25% short, but I might change the allocation in the future according to market valuation.

All the models have been backtested on more than 8 years of data and include a post-discovery analysis period of 1/2 years. Each model universe is built based on a set of liquidity, fundamentals and price filters and the final candidates are selected applying a ranking system based on fundamental data. Tools that I use consist of a combination of freely resources available online, such as keelix.com and backtester.org, and commercial tools like portfolio123.com and MathLab. The data used by keelix.com is based on AAII weekly downloads, which consist of a revisited version of Reuters data (previously Mutex). Reuters is also the main data source for portfolio123.com. backtester.org use data published by ValueLine.

HFCF (High Free Cash Flow) and NH (New Highs)
Both HFCF and NH are long momentum screens looking for stocks with good fundamental ratios and close to their 52-weeks highs. They are in facts very similar: HFCF has more stringent value rules selecting stocks with higher free cash flow then their industry peers; NH requires a positive 3 months and 6 months uptrend.

PCF (Price/Cash Flow)
This model can be described as a value mean reversion strategy looking for stocks with high operating cash flow and low debt and disregarding recent price action, thus in general favoring oversold stocks. I introduced the strategy in real trading in October 2007 after I had around 10 years of simulation available. So far it has been a disappointment but the entry time was bad given the recent market performances. I'll have to trust simulation data on this.

LKS (Less Known Stocks)
A large winner in 2007, this model seeks to anticipate large institutional moves in low-liquidity stocks, providing good factors diversification to the portfolio. I got the original idea by looking at the Reuters PowerInvestor screens (it still carries the same name) and then added some other insider trading and price/volume data rules.

SHORT
This is by far the most complex model in terms of number of factors employed in the final ranking system. It was backtested for a period including 2003, a difficult time for short strategies. The model requires stocks with poor fundamentals and avoids short-term oversolds, thus limiting the selection to only a few names every month.

I am currently working on a number of new models to improve portfolio diversification. In particular I would like to add a Commodity and a Healtcare model to the set.

Powered by ScribeFire.

Technorati Tags: investing, trading, quantitative model

6 Months Asset Class Monitor Updated

I added Global Bonds, Emerging Market Bonds and a Managed Futures fund (RYMFX) to the 6 months momentum monitor. I will keep the monitor updated as I intend to change my portfolio models accordingly.

Markets Momentum: Momentum

		ETF	18-Jun-2007	18-Jan-2008	6-Months		Alt. ETF

Commodities GSG $43.03 $51.10 18.75% DBC, GSP Global Bonds BND $71.87 $78.32 8.97% AGG Managed Futures RYMFX $25.56 $27.05 5.83% Emerging Mkt Bonds ESD $17.05 $18.04 5.81% Emerging Markets EEM $130.86 $132.33 1.12% GMM, VWO Large-Cap Growth JKE $70.10 $64.02 -8.67% World Markets EFA $79.22 $71.30 -10.00% Mid-Cap Growth IWP $115.66 $99.92 -13.61% Small-Cap Growth IJT $143.23 $121.86 -14.92% Large-Cap Value JKF $88.93 $73.62 -17.22% Mid-Cap Value IWS $159.84 $126.37 -20.94% Micro-Cap IWC $61.49 $47.60 -22.59% Small-Cap Value IJS $80.95 $62.08 -23.31% Real Estate IYR $80.06 $59.48 -25.71% RWX

Powered by ScribeFire.

6 Months Asset Class Returns

The following table shows returns from Mid-June (6 months) for some of the main asset classes and US value/growth segments. Commodities returned an impressive 21%, emerging markets are still positive after the last downturn:

	ETF 18-Jun-2007 17-Jan-2008 6-Months Commodities GSG $43.03 $51.80 20.38% Emerging Markets EEM $130.86 $135.28 3.38% Large-Cap Growth JKE $70.10 $65.01 -7.26% World Markets EFA $79.22 $72.46 -8.53% Mid-Cap Growth IWP $115.66 $101.82 -11.97% Small-Cap Growth IJT $143.23 $124.14 -13.33% Large-Cap Value JKF $88.93 $74.97 -15.70% Mid-Cap Value IWS $159.84 $128.82 -19.41% Micro-Cap IWC $61.49 $48.16 -21.68% Small-Cap Value IJS $80.95 $63.17 -21.96% Real Estate IYR $80.06 $59.41 -25.79%

Real-time updates here.

For 2008, I am going to introduce new Commodity and large-cap models. Also I am looking at ADR-based screens, given the relative strength of emerging and international markets.

Powered by ScribeFire.

2007/2008 January Effect

This year the January trade returned a negative 0.34% before transaction costs for the period from 12/20/2007 to 1/9/2008: chart. Microcaps continued the downtrend that started in Q2: 2007 IWC:SPY chart.

Emerging markets on the other hand, are off to a good start in 2008 continuing a long-term uptrend vs. the US markets that started in 2000: here's a chart from 2005.

2007 Demo Portfolio Results (Revisited)

The Demo portfolio finished 2007 with a return of 17.62% net of transaction costs, for the 9 months from 4/1/2007 to 12/31/2007, well above the 9.63% benchmark results. I will continue trading the same model in 2008 and will keep the short model allocation to around 20-25% of the NAV to reduce portfolio risk.

The short model started trading at the end of August and so far has met my expectations despite a 145% loss in TMTA in October. Given the low capitalization of the stocks selected by the model, the risk of another TMTA going forward is high, but the simulation includes these jumps in the backtested results*.

*I checked the simulation log from 2/24/2003 to 1/7/2008 and found a total of 251 transactions for this 4.8 years period, averaging 47 calendar days holding period p/position or around 7 stocks per week. Contrary to what I thought initially, the highest simulated loss is only 35% for OPWV on 1/12/2004. Interestingly, OPWV is currently my best position carrying a 60% positive return from 8/2007.

This is a complete log of the highest simulated losses:

Symbol	Open	Close	Days	Pct
OPWV	12/29/2003	1/12/2004	14	35.40%
MDTL	9/17/2007	10/1/2007	14	33.76%
PRSF	12/22/2003	1/26/2004	35	27.44%
SSRI	1/18/2005	2/14/2005	27	25.23%
COSI	6/6/2005	6/13/2005	7	23.77%
MONE	7/26/2004	8/2/2004	7	22.28%
PRSF	12/22/2003	2/2/2004	42	22.26%
SYNM	11/7/2005	11/28/2005	21	21.44%
INPC	3/20/2006	4/10/2006	21	20.79%
MENT	12/29/2003	1/12/2004	14	20.24%