Saturday, April 07, 2018

Reviving Old Trading Models

I dug into my computer files and updated the trading models I last used 10 years ago. One of them which is fairly simple flashed a strong warning sign at the January high in the market. It does tend to have false positives where it is fooled by a very strong trend into thinking that that is a top in the market, but this time the market actually did fall of course after the warning. This is a very negative signal. There was a minor buy signal at the recent low about a week back but there are typical several buy signals on the way down in bear markets. I had been looking for signs of a recession before taking de-risking action in a big way on our portfolio - for example, an inversion of the yield curve. There hasn't been any sign of a recession. But Trump's trade war and the Fed's unwinding of its inflated balance sheet are having a negative effect on the market.

I have another much more complex model that attempts to forecast the day ahead direction of the market - despite what standard investment theory says, that the stock market is a random walk and can't be predicted this is actually possible to some degree with some insight from econometrics into how to turn it into a predictable problem. I updated the model using the last ten years of data and reoptimized the parameters - they hardly changed. That is a good sign. However, though I have all the past predictions and the trade directions I decided on based on them, I can't remember how I used the model to actually choose market direction. Unless I can find something I wrote about that, I'll have to reverse engineer that from scratch.

I found a folder of handwritten research notes on my trading model from 2006-8 in my home office. This should help a lot.

I predict the US will go into recession in 2019. In 2007 the stock market peaked in Summer-Fall but the recession didn't really get started till Bear-Stearns failed in March 2008. In 1999-2000 the stockmarket peaked in March 2000 but the recession didn't really get going till September 11, 2001.

No comments: