Dec 10, 2010: Short Term Prediction Experiments

home   faq   articles   premium   contact  

A few weeks ago I've turned bullish for the next several months. There are several reasons for this shift:

So my macro view for the next several months (at least) is:

I expect more of the same: excess fiat-money printed and lent to financial institutions, who still need to 'recapitalize' better at near zero rates. This excess money should keep flowing into the markets and keep propping most stocks up, up and away. Given such a backdrop, I would simply buy the dips, and be aggressive to the max in doing so. I would even consider a limited exposure to ATM calls on stocks (on dips) or ATM puts on volatility on spikes.

An experiment in short term prediction of the stock market

About a year ago I signed up to an interesting experiment on a site called "":

Among other things, runs an experiment in short term market prediction.

Each day after the close, players are asked to predict how the S&P500 index would close the next day: up, or down. In my first few weeks attempts I quickly rose through the ranks into the main "G100" group and managed to stay in the ~50-70% correctness range for as long as I kept playing.

Then, I lost interest left my account idle for most of the year. As a result and my rank drifted down for inactivity and I dropped out of the main group for nearly a year.

About 6 weeks ago, I decided to get back to the site and see if, given all that I learned in the past year, I can do better. Indeed, I was doing much better. A few weeks after returning to activity, on Nov 20, 2010, I was at the top of the "members in qualification stage" group with a clear lead, a 80% (ratio of 4 out of every 5 guesses correct, see screenshot), and a chance of returning to the top league:

The detailed record on Nov 21 was:

So, I decided to persist for a couple of weeks longer, which got me into spot #3 in the top "G100" group, with a longer record, and a slightly better correctness ratio of 81% (screenshot):

The detailed record at the close of Dec 10, 2010 was:

One of the techniques I noticed the leaders tend to use is to bet less often. The site allows you to skip betting from time to time, as long as you don't exceed too many "passes" which will get you penalized as a wrong bet. I have been too aggressive skipping only one day out of the last 22 sessions. Even that skip wasn't intentional, I was simply too busy and forgot to vote on Nov 18th. In the coming weeks I will try to see if I can improve my record. Wish me luck.

The algorithm I use for these short term predictions is mostly "manual": I look at high resolution short-term charts (1-minute and 5-minute ticks) and from them, I try to figure the answers to all of:
  • What is the prevailing trend/momentum?
  • Are there clear oversold/overbought conditions within the trend?
  • Are recent leaders weakening or staying the course?
  • Breadth: are mid and small-caps outperforming the S&P500?
  • Are volume-weighted trends: accumulation/distribution, VWAPs agreeing?
  • Volatility trends: is the volatility index (VIX) waxing or waning?
  • Are correlations between asset classes increasing or decreasing?
  • Are any divergences developing in any of the above indicators?
It is a pretty simple and sensible algorithm, and it seems to be pretty reliable as the ~80% correctness results demonstrate.

My plan is to code this algorithm so I don't have to work as hard and so I can cover many ETFs as opposed to only looking at the broad market.

home   faq   articles   premium   contact  

The above merely reflects my own thinking and actions at the time of writing. Every investor should make up his own decisions based on his risk tolerance, time-frames, comfort-zones, convictions, and understanding. Never investment advice.

Any feedback is welcome.

-- ariel