I was reading a blog post today about the ‘best’ way to catch a momentum breakout, and it got me thinking. Most systems that trade a breakout reference volume. “Only trade a breakout that has a sharp increase in volume” and that sort of thing.
Whenever I look at these systems however, I don’t seem to find any breakouts that don’t have higher volume. From my non-scientific survey, high volume seemed to be a prerequisite for a price breakout. The idea of a breakout without higher volume seems, well, weird and unlikely.
So is volume important for breakouts? Let’s do the numbers.
I took a look at the period from Jan 1 2010 through Oct 13th 2015. I looked at the current list of Russell 3000 stocks (I don’t see a need to look at delisted stocks, or those that were part of the list at the time of trade, because this isn’t a backtest).
I defined my breakout as:
• Where the close was at least 5% higher than the close of the previous day.
• The previous day was not a breakout day.
• The median 21-day volume was greater than 100,000 and the closing price was greater than $15. This is to ensure liquidity.
This is a very simple definition of breakout, with no massaging for long term momentum or anything else.
I calculated the return starting at the open of the following day, through the close 5 days later.
I then divided my data into two parts: those breakouts that traded on less volume than the day before the breakout, and those that traded with more.
Surprisingly, 23% of these breakouts were accompanied by a drop in volume from the day before. And here I didn’t think these even existed! 7124 instances out of 31,010 total.
So what’s the average gain or loss when comparing reduced-volume breakouts vs increased-volume breakouts?
Vol decrease from day before:
win rate: 48.9%
Vol increase from day before:
win rate: 50.3%
So it looks like it matters a tiny bit.
Then I wondered, does more volume improve the results? I examined the returns from only breakouts that had a 50% increase in volume (i.e. 1.5X), 100% and 200%. Here are the gain/loss averages and the win rates:
There does not appear to be a coherent pattern here. Increasingly higher volume does not yield increasingly higher returns. And the gain/loss average is so close to zero that I’m not sure filtering by volume increase is worth the trouble.
Volume spikes may have some value, but I’m not seeing it here. I think the next step is to construct a slightly more elaborate set of parameters. I’m thinking a simple trend filter, and comparing the breakout volume to a moving average of the volume. Stay tuned…
P.S. I’ve been quiet lately because I haven’t had any new ideas. I’ve been trading some mean reversion ideas, but mostly keeping my head down.