Picture: THINKSTOCK
Picture: THINKSTOCK

A NEW discovery (for me at least) is that JSE data are freely available from www.google.com/finance, though not from finance.yahoo.com.

My bias for Yahoo’s financial data service comes from a couple of free online university courses — University of Washington’s Introduction to Computational Finance and Financial Econometrics and Georgia Tech’s Computational Investing – which used Yahoo’s stock market data feeds in their lectures and tutorials.

The JSE pointed out its data are available from Google’s financial service after I griped it wasn’t offered via Yahoo’s in a previous column, so I started exploring Google’s stock market data offering and found it superior to Yahoo’s.

Hopefully, some free online university courses will appear soon on how to use Google’s financial data service because it’s even more complex than Yahoo’s. One thing I had to figure out with Google was that to get local share prices, you need to preface your search with "JSE:". For instance, to get Lonmin’s share data you would enter "JSE:LON" in Google finance’s search box. Getting it to work properly also seems to require using Google’s web browser, Chrome. Firefox, my preferred browser, could not load Google’s interactive graphs.

Among what I learnt from Washington University’s investment course is the importance of using "adjusted closing prices", which seems to be standard with US stock price feeds rather than historical closing prices traditionally used in SA and the UK.

A good illustration of why this is important is Lonmin’s recent share price history, which got badly distorted by a 100-for-one share consolidation on December 18. Simply graphing closing prices shows Lonmin rocketing from 16c to R14.50, whereas investors actually saw the value of their shares shrink about 10% after the consolidation.

So instead of showing a jump from 16c to R14.50, the data need to be backdated by multiplying all the prices before the consolidation by 100. The adjusted closing prices in US stock data feeds also reflect dividend payments, unbundlings and other corporate actions so working back from the most recent closing price accurately reflects how much an investment would have grown or shrunk from a given historical date.

Comparing Google’s Lonmin share price graph with Yahoo’s shows Google’s has several advantages.

Besides having JSE prices in rand while Yahoo’s are in pounds sterling from Lonmin’s London listing, the Google graph has a nifty feature of a timeline on the horizontal axis which is flagged with news events. Clicking on the flags highlights a link to corresponding report in a list to the right of the graph. Furthermore, in the case of Lonmin, the adjusted closing prices supplied by the JSE are correct while the London Stock Exchange has fed Yahoo unadjusted closing prices despite the label.

Both Yahoo and Google offer a "Historical prices" menu option from which you can download closing prices as a spreadsheet that can then be used in the computational finance techniques taught in university courses like the ones I’ve linked to above.

I’ve done nearly all the available free online university courses falling under the banners "big data", "machine learning", "data mining" and so on because I think these new technologies offer exciting prospects for journalism, but I’m dubious about their value at predicting future stock prices. A Random Walk Down Wall Street author Burton G Malkiel summed up how these techniques are used by hedge fund marketers as "beating the data set in every conceivable way until it finally confesses".

University of Washington lecturer Eric Zivot included the caveat that what he was teaching was statistically dodgy since not enough stock exchange data have been gathered to provide the "confidence interval" scientists in fields like physics or medicine would demand.

Georgia Tech’s Tucker Balch glosses over these problems, leading to the former F15 fighter pilot getting shot down by Reuters columnist Felix Salmon for making some highly improbable claims.

Computational investing is a pseudoscience, leading to fads like high frequency trading that have enabled stock exchanges to compete more effectively against casinos in the business of quickly separating fools from their money. This has prompted many modern exchanges to encourage this form of gambling by making their live prices freely available via the web. The JSE, on the other hand, insists on sticking to the old ways.

"Like all major exchanges, the JSE charges for market data which contributes toward the exchange’s revenue, regardless of the platform. New exchanges may not initially charge for data in order to attract new business, but this invariably changes over time," our local bourse informed me in an e-mail.