A reader of my blog, Matt B., commented recently on an old post I’d written about momentum bursts. Like me, Matt was intrigued by the short 3 to 5-day momentum bursts he saw described time and again on Pradeep Bonde’s stockbee site. Those bursts look so pretty, so elegant, and more to the point: so profitable.
My previous research felt a little like banging my head against the wall. Trading on an end-of-day basis showed poor results in backtests. And on two separate multi-month periods over the past year, I paper-traded versions of the momentum-burst trade that relied on intraday data.
I kept coming up with the same result: those big momentum bursts just didn’t seem very predictable. The conditions might be right and you find a stock that shows a breakout…but will it follow through?
Matt B. was experiencing the same thing with paper-trading on an EOD basis. We started a dialog about this phenomenon. We wondered if these bursts were reliant on overall market conditions to achieve ‘follow through’.
My first look was to create a simple trading system and see how it did over time. This system found stocks that had a recent burst, followed by some consolidation. The system buys in anticipation of another boom, but there’s no look-ahead to ensure success. This system tests the quick follow through on a previous burst.
• liquidity requirements: Close > $5, average 20-day volume > 50,000 shares, and the stock was a member of the Russell 3000 index.
• The period between 01/01/2000 and 04/01/2016
• the stock was up at least 50% over the past 100 trading days.
• Find a day that closes more than 5% above the previous day, followed by three days of low volatility ( < 2% absolute change from close to close). Buy at the close of that third low-volatility day.
• Hold for five days, sell at the close.
This is not meant to be the ideal momentum-burst swing trade system. It’s only a simple method to see if these sorts of events are evenly distributed through time, or if they clump.
That’s what I call clumping!
All the major gains happen in three places: the 2000 dot-com blow off, the end of the bear market in 2003, and the end of the bear market in 2009. Everything else is flat or worse.
Matt B. described the system he’d been paper-trading. I’ll summarize without specifics, so as not to reveal any proprietary information. I coded up a backtest for his initial proposal, which is something like:
• price is up over the past five days as well as the past 20 days.
• the stock surged at least 5% today, but not some crazy blow-out percentage.
• the previous three days were not burst days.
• volume was greater than the previous day’s volume.
• Enter the next day on a buy-stop order, if the price went a couple of cents above the signal day’s closing price. The theory here is that a buy-stop shows the stock means business.
• hold for five days. Matt B. had a much more nuanced exit strategy, but for testing purposes I went with a 5-day hold. If the entry is good, a simple n-day hold should show results.
I won’t bore you with multiple charts, but this initial attempt was not a winning system. Matt B. had commented that the system often had trades where it bought near the high of the day, only to plummet to the close. My results showed this to be true: bursts often end up with mean-reversion as traders take profits, at the expense of those joining the party too late.
I switched the entry to a simple buy-at-open, without regard to stops or limits. The results improved, but were still not profitable.
Matt B. also had a profit target in his exit, so I added a simple close-greater-than-5% exit (on that close). The results were finally profitable, but not what you’d call inspiring.
Interesting that this system showed good results during the dot-com bust, as well as the usual market turn around points. But it languishes during the middle part of bull markets.
Up to this point, we’ve been thinking about entrances and exits and whatnot. Maybe we should just see if there are enough of these momentum bursts to go around. Maybe we’re not identifying them correctly, and that’s one thing. But if they’re not there at all, then we should consider something else.
So let’s just go look for every single 3-day momentum burst we can find. After all, if we’re going to enter after the first day, we’d better have another day or two of gains to make the trade worthwhile.
Here are my search criteria for the ‘unicorn’ burst (because it angles upward like a unicorn horn, and is just as elusive):
• liquidity requirements: close > $2, median volume of the last 19 days is > 50,000 shares and the stock was a member of the Russell 3000 at the time of the trade.
• The stock has 3 days in a row that are up at least 3% each day, from close to close.
You may quibble whether that constitutes a momentum burst, and that’s fine. But the frequency of these puppies will give us some indication as to where the good hunting is. Not that we’re hunting puppies, because that would be mean.
There certainly have been better years for 3-day momentum bursts, that’s for sure. Interesting that they were prevalent in bear markets, not just at the start of bull markets. However there is more activity now than a few years ago.
Maybe a meta-filter could be created. For example, only trade your momentum-burst system if the number of “burst” stocks n the past month is above a trailing average. The burst-count does seem to persist for longer periods of time.
Or it may come down to an inconvenient truth: momentum bursts may require intraday entries to be profitable. Or require some external input beyond price and volume to be worthwhile, for example fundamental analysis. But for now, I’m exhausted and will go back to thinking about my mean-reversion swing trades.