When stocks are moving gently from one day to the next, there is often no discernible pattern. However when they start rockin’ and rollin’ one direction or the other, they show certain similarities.
I’m always curious how stocks behave when they show a significant drop, or when they pop upward unexpectedly. I ran some simple statistics and noticed a couple of things.
First, I decided to look at what happens when stocks drop significantly. Rather than look at a fixed percentage, I instead used Average True Range of the stock. This shows the average price movement over the previous days, and is a measure of volatility. I took the ATR(20) before a drop, and corralled all the stocks that fell at least 3x the previous day’s ATR(20) value. I also looked at stocks that dropped at least 5x the previous ATR. Here’s a visual:
I then recorded the percent gain or loss for each day’s close following the drop, for five days after. This resulted in thousands of rows of data, but you know I gladly suffer through spreadsheet hell so that you can have pretty graphs.
It turns out there’s a clear pattern:
For each drop that was at least 3x bigger than the ATR(20), the above graph shows the average gain/loss for each day, expressed as a percentage of the original ATR value.
For a 5x ATR drop, the results flatten out. Day 2 still shows a loss on average.
Can we make a trading system out of this? Turns out we can.
- Stock must be a member of the Russell 3000 index at the time of the trade.
- Stock closing price is greater than $15 and the 20-day average volume is > 100000 shares.
- Look for a stock that drops at least 3x its previous day’s ATR(20) value.
- If the second day after the drop day is below the original big drop, buy at the close.
- Sell at the first profitable close, or at the close of the third day.
- If you have more choices than you want to trade, pick the ones with the biggest drop (or the smallest gain if none are down) on the day you’re going to enter.
As you can see, there are a couple of specifications there that can’t be derived from looking ata bar graph of statistics. I optimized over the period of 2000-2009, then took a look at the results to the present day. Here’s the equity chart (with no commissions):
Looks pretty good, even in the OOS period. Mind you there are a lot of short-term trades. You’d really need a large position/commission ratio to make this practical. TheCAR/MaxDD and Sharpe are merely so-so, and I’m not proposing this as an actual trading system without further refinement.
Like me, you’re probably now wondering what happens in the other direction. If a stock pops up significantly, are there any general patterns that emerge?
That, of course, is the subject of the next blog post!