Labelling warnings

Inspired by the chap at value and opportunity I thought I should increase the frequency of blog posts. I’ve not been blogging on my site much because I’ve writing 2500 words a week for ShareScope / SharePad  on company results. As I’m being paid to write, I’ve been prioritising that.
As an aside SharePad is an excellent product. It appeals to people like me who want to look at company ratios like Returns on Invested Capital over decades. They offer a bespoke 1-hour one-to-one training session with a SharePad or ShareScope expert to help you get the most out of the system.* If you’re a new to the product, sign up through the link you get the second month free (or 13th month free if you’re an annual subscriber).

Ideas for a book

I originally started my blog, because I thought that I might like to publish a book. A blog is a good way of finding which topics interest people, and then you can focus on that. Otherwise a writer risks the experience of David Hume, who published A Treatise on Human Nature in 1739, hoping for vehement attacks – but instead the book was ignored.
As Bertrand Russell describes it in A History of Western Philosophy, Hume then switched to writing of essays (the equivalent of blog posts of their day) and made an unsuccessful application to become a professor at Edinburgh University.

“Having failed in this, he became first tutor to a lunatic and then secretary to a General. Fortified by these credentials, he ventured again into philosophy. He shortened the Treatise by leaving out the best parts and most of the reasons for his conclusions; the result was the Inquiry into Human Understanding, for a long time much better known than the Treatise.”**

Writing a book takes years, and it is very hard to know if the subject will capture readers’ imagination or “fall dead-born from the press” as was the case with the Treatise. Blogs on the other hand are a way of “failing fast.”

Natural Language Processing

I then became interested in the programming language “R”, and was scraping company announcements and examining financial language using “natural language processing” techniques. I was using the blog not just to gauge interest, but also as motivation to investigate and share interesting results.
I ran into a couple of problems.

  • I bought a craft beer bar in Berlin, which has required considerable more time and attention than I first thought it would require pre-pandemic.
  • It became harder to webscrape company announcement.
  • I needed to find a way to label price sensitive announcements like profit warnings

The last problem is an interesting one, because it is boring. Profit warnings are never labelled “profit warning” at the top. As an example look at this profit warning from Northern Rock in June 2007. It was titled “PRE CLOSE PERIOD STATEMENT AND BASLE II STRATEGIC UPDATE”. Within a month the company would seek help from the Bank of England, yet the RNS is full of sentences like

“Northern Rock will continue its core strategy of strong growth in high quality residential mortgage assets which will be retained on the balance sheet.”

“Credit quality continues to be robust across all loan categories. Arrears levels are slightly higher than at the 2006 year end reflecting rising interest rates, but the loan impairment charge is anticipated to be in line with market expectations.”

“The current consensus from 19 analysts for 2007 is for underlying attributable profits, before the preference share dividend, of £403m – £457m, with a mid-point of £430m, an increase of 17%. Northern Rock expects its underlying attributable profits to increase by around 15% when compared to 2006.”

This was not just a profit warning, the bank failed within a couple of months. The lender would only achieve it’s 15% growth in “underlying attributable profit” because management entered into a swap arrangement to artificially create mark to market profits in a rising interest rate environment (very weird – normally banks report mark to market profits when market conditions are favourable).  I don’t think there’s any way a computer could figure out.

The easiest way to know if something is a profit warning is see if the shares drop more than 10% on the morning of the announcement; then you can be pretty sure the company has announced something disappointing that investors don’t like the look of. So I would have to spend a lot of time looking for announcements where stocks fell, and then labelling the data.

Positive language

Labelling profit warnings are just one example. On the positive side, I’ve noticed that a company like Games Workshop writes communicates very clearly.  It’s possible to go back and take a sample of high performing companies 5-10 baggers and compare the language they use to help investors understand what management are thinking and doing. This too would involve significant amounts of labelling though because only a few announcements mark a sea change in performance.

Games Workshop shares drifted for over a decade before revenue took off after 2016.  Richard Beddard points out that revenue has grown revenue at a Compound Annual Growth Rate (CAGR) of 23% and profit at a CAGR of 46%. Before 2017 it was chugging along earning a return on capital in the mid-teens. Since 2017, return on capital has been about 50%. So significantly positive announcements that mark turning points are not flagged either.  They also would require labelling. 

Computers are supposed to be good at boring tasks

That’s a pretty boring task, and computers are supposed to be good at automating boring tasks.
On the other hand, reading the 2 page announcement is fairly simple and interesting task for a human being. You’ve just done it yourself.

But I was training the computer to do the interesting part of the task, while doing the boring part myself. That seemed like the wrong way round. So I stopped, to think about what to do next.

Photo by Tobias Keller on Unsplash


*I’m biased because I’m being paid by Sharescope, but there is also an independent trust pilot site where you can read the reviews.
** Russell’s A History of Western Philosophy is full of these quotes. Another favourite is:
IN every history of philosophy for students, the first thing mentioned is that philosophy began with Thales, who said that everything is made of water. This is discouraging to the beginner, who is struggling–perhaps not very hard–to feel that respect for philosophy which the curriculum seems to expect.