Various Snippets

News Driven Means Reversion
Let's say FTSE100 is 7350 and you fancy an upward movement. You'll trade, say, GBP10 a point long, and GBP7 a point short. This reflects the fact that you fancy upward movement but also see the (lower) chance of downward movement. You can collect a lot of cash with either direction e.g. 10 points up means you collect GBP100 in cash when you close the position but also have "paper" loses of GBP70 on the short position. You don't worry about this because:

1. you are GBP30 net up anyway (and have GBP100 in your cash account), and
2. you believe there's a good chance the market will, quite soon, head back down to 7350 points and below.

Note that the challenge with this approach is keeping the number and size of trades in proportion.

I have failed in this regard because I have a GBP7500 cash account (based on an initial capital of GBP700) but am sitting on paper losses of GBP5300. These loses are mainly from very out-of-the-money short positions.

So, on one level, I can close all my positions now and feel happy that I traded GBP700 to 7500-5300=GBP2200 in five weeks. A superb return, albeit on small money.

Or, better, I can be correct in predicting that my slight current bullish weighting will make me more cash. I then close my long positions and wait for the Sept/Oct crash to come (caused by North Korea, Trump, etc.) to negate all my loses and create profit besides. I then have traded GBP700 to GBP7500+ within around 3 months. A very strong performance. But note that, maybe, no crash comes, and strong profits/fundamentals continue…That could wipe me out.

But here's the thing. With GBP7500 in cash and GBP5300 in paper losses and a margin requirement of GBP1800, my "trading resources" are now only GBP400.

Not good. I am over-trading.

So you see, this "volatility trading" approach is not pure Taleb (he bets on the probability of cataclysm being underestimated; I bet in a near-market-neutral way), and it needs more discipline that I have applied…although I am still doing okay on it.

I could, of course, close many or all of my current positions, accept a reduction in my cash account, and return to more prudent strategies.

I'm summary, for sustained profits, low stress and a comfortable life, I recommend my first strategy i.e. bet on big news and - for even more safety - have a view on fair/typical value of the underlying asset too. That strategy has already been communicated to you and it's in the "boring but very good" classification. I commend it to you.

Big News
Test:

1. extract 10 years of new stories on one big asset e.g. FTSE100. spider (old) news where FTSE is mentioned and that have precise date time stamps
2. filter for strongly pos or neg. classify news articles on sentiment (pos & neg) and ignore low values and conflicting values (both pos & neg)
3. scan for max returns (e.g. % up if long, % down if short) within 2-200mins of time elapsed since news (or just take returns every 10 mins for 3 hours).
4. ignore all complexities e.g. both bullish and bearish stories within minutes of each other. devise some algo that ignores polarity changes within a short timeframe
5. sum all the %'s across all the trades.

If my experience is correct, you'll trade 2-3 times per week and make 1%-20% returns per week (if spread betting, and % returns here is calculated as returns on capital).
Abstract of the mean reversion + daily sentiment (from the BBC) system for daily trading.
There are four categories of data:

• above/below mean (in fact there is also "within 0.2% of mean too i.e. no meaningful diversion)
• positive/negative sentiment (in fact there is also "no interesting data" too)

1. below mean + positive sentiment is a LONG signal.
2. above mean + negative sentiment is a SHORT signal.
3. the other 2 quadrants we can leave to the computer to evaluate.

KISS
Second order derivative. Fragility is not measurable but the speed of volatility is. Averages hide fat tails

Nassim talk first ten minutes

Jenkins inequality

Not timing the market

Instead of picking up nickels lose nickels

Price of stock or an option cannot be based on probabilities but based on contexts. So probabilities are meaningless since contexts keep changing.

What is a new news article; i.e. Non stale article?
The third constraint that Tetlock (2011) imposes on the news data is that the news text
must have at least 5% single-word or 2% bigrams (i.e., adjacent word pairings) in common to
the previous 10 news items.
After classification (extraction) of SYMBOL

Share Turnover: Shares Traded / Shares Outstanding

The variables pos and neg for different texts. Colors indicate direction of stock movement.