Articles on investing and capital management, with a quantitative focus.

#bullsignals - Algorithm-driven market timing

Timing the market is probably a bad idea


I'm aware that it's generally a bad idea to try "market timing". Trading in and out of the market usually gives worse performance than just holding an index, and I had serious hesitations about trying to use my BullSignals method with real money.

Even though it's often a bad idea, I'm trying it out using some software I wrote. I came up with the algorithm in 2006 while experimenting with stock data mining and technical analysis. My method is proprietary, but it uses relatively common trend detection approaches. The software wants to be long during an uptrend, and wants to be out (in cash) during a downtrend.

If timing the market is a bad idea, why am I still doing it? The big reason is that, unlike back-testing approaches, my algorithm has actually been running for 13 years with live market data. Over that time it successfully got out of the market before the 2008 crash and went long again in 2009. Back then, I didn't trade by this strategy, but I wish I had!

The long track record, with real (not back-tested) market data has given me enough confidence to deploy real money into the strategy. Some additional reasons I'm not terribly concerned are:

That being said, this is an experiment. Market timing usually doesn't work, so you're probably better off with index investing.

The result I expect

I think the most likely outcome over several years is lower performance than buying and holding the index, but also less volatility (milder drawdowns). I also suspect that the algorithm will be more successful in turbulent markets.