Category Archives: Bollinger Bands
We are entering historic territory with the S&P 500 closing for the 12th consecutive day above its upper Bollinger Band (50,2). This has only happened 25 times since 1928. When the market reaches historic extremes, it is ordinary to wonder what it all means. Let’s take a look.
- Buy $SPX at the close after 12 consecutive closes above its upper Bollinger Band (50,2).
- Sell $SPX at the close X days later.
- No commissions or slippage included.
- $SPX history starts in 1928.
With only 25 samples, we should be careful about drawing conclusions from these results.
If we ignore the amount of time this rally has persisted and only consider the number of consecutive closes above the upper Bollinger Band, we see that in the past the rally has been able to continue.
The $SPX buy-n-hold results were created by cutting the history into 50 day segments and then averaging those segments together.
For those who might be interested in digging a little deeper into this setup, I’ve posted below all the previous instances.
Today, despite closing slightly lower, the S&P 500 closed for the 5th day above its upper Bollinger Band (50,2). I have written previously how closes above the the upper band are typically bullish and may signal an abnormal market. This evening we’ll examine what happens after 5 consecutive closes above upper Bollinger Band.
- Buy $SPX at the close after it has made X consecutive closes above its upper Bollinger Band
- Sell Y days later
- $SPX history starts in 1928
- No commissions or slippage included
The results are more bullish than I expected, and they improve as the number of closes above the upper band increases.
- This setup was not nearly as rare as I had imagined.
- There were 126 occurrences of 5 closes above the upper band with 110 trades held the full 50 days.
- After 5 closes above the upper band, $SPX has closed higher 50 days later 67.27% of the time.
- Buy-n-Hold results are calculated by cutting $SPX history into 50 day segments and then averaging the segments.
The number of samples should be sufficient to lend reliability to the results, but I am honestly a little disappointed because I suspected the results would be either extremely bullish or extremely bearish. I was looking for some drama and didn’t find it. Instead, this study just presents another reason to lean bullish.
Yesterday evening, Frank Zorrilla @ZorTrades and I shared a few tweets where we discussed the use of Bollinger Bands (50,2) to time various market constituents. The conversation turned to Gold, and since I’ve been considering establishing a position, I thought it would be a good time to run a few tests. Frank agreed.
Unfortunately, the $GLD ETF does not have a great deal of history with which to test, so I’m going to use 2 proxies: $GOX (CBOE Gold Index) and $DJPR (Dow Jones UBS Precious Metals Subindex). Clicking on that last link will take you to a chart that allows a comparison between $GLD and $DJPR. They have tracked each other fairly close over the last year.
Although there are many different ways to assess whether Gold is a buy or a sell here, Frank and I were talking about using Bollinger Bands (50,2). Setting Bollinger Bands to (50,2) is a favorite of mine, especially as an abnormal market indicator. $GLD has closed beneath its lower Bollinger Band for the last 5 closes. Today, both $GOX and $DJPR closed for the 4th consecutive day beneath their lower Bollinger Bands. This is very significant. More about that in a bit. Let’s get to testing.
- Buy $GOX or $DJPR at the Close when it has Closed beneath its lower Bollinger Band (50,2)
- Sell X days later
- No Commissions or Slippage Included
- $GOX history starts on 4.22.1996
- $DJPR history starts on 1.2.1991
The Results using $GOX:
- There were 128 occurrences of this setup with 21 trades held the full 100 days. (Blue line)
- There were only 10 occurrences of $GOX making 4 consecutive closes beneath its lower Bollinger Band, with 6 held the full 100 days. (Red line)
I don’t see much to get excited about. It is a rare event for the index to make 4 consecutive closes beneath its lower band. When the index closes one day beneath the lower band and then quickly reverses to close above it, results are mildly bullish. Multiple closes beneath the lower band are telling us that something abnormal is happening. A bounce will eventually occur, but the results above show that on average, the index slid for over 3 more weeks before the bounce.
Let’s take a look at the results using $DJPR:
- There were 163 occurrences of this setup with 30 trades held for the full 100 days. (Blue line)
- There were 19 occurrences of 4 consecutive closes beneath the lower Bollinger Band, with 14 trades held for the full 100 days. (Red line)
The results leave some things to be desired, like profits. Again, one close beneath the lower Bollinger Band seems to be a fairly normal occurrence, and we can see some reversion to the mean takes hold, with average gains of 3% after 50 days. However, four consecutive closes beneath the lower band is not normal, is rare, and seems to indicate that a new trend is emerging or has emerged.
Based solely on the above analysis, I’m not comfortable establishing a position in Gold. Bollinger Bands (50,2) work well to show that a strong trend is developing when $SPY closes above them. It is not difficult to understand that consecutive closes beneath the lower Bollinger Band seem to indicate that at best, Gold will languish and trade sideways over the intermediate term. At worst, this setup may signal a new downtrend is emerging.
As noted above, this is only one study. Perhaps over the weekend we can examine the impending death cross in Gold. If there are other tests you would like to see run on Gold, let me know in the comments section. I can’t promise anything, but if you come up with something interesting and I can code it, I might give it a shot.
Finally, we get some market action that leads to less-than-bullish projections over an intermediate time frame.
First, we’ll start with the hard-to-quantify Honey Hole Setup. See below….
The Honey Hole setup will be familiar to iBC old timers. It is a simple but seldom seen setup, except in bear markets, where it can be almost common. It is formed when price rises from beneath the 50 day moving average and is denied, usually several times, before reversing downward. I think I have tried to quantify its edge in the past, but I do not recall the outcome. For me personally, I automatically take a more cautious approach when i start seeing this setup appear across the major indices.
The quantifiable Bearshitter stuff is this: Today’s close just above the lower Bollinger Band (50,2) has bearish implications over the next 50 days. See results below…
The next 7 days show the propensity for a bounce, but after that, it is downhill. And don’t let this somewhat benign looking chart fool you! It is very difficult to find a quantifiable setup that leads to bearish results for the $SPY looking out over 50 days. Bearish results are uncommon as the markets have had a historical bias to the upside.
The 200 day moving average, the traditional bear market demarcation, sits roughly 1% beneath today’s close. This could get interesting…
Below are some other posts I have written on this or a similar topic:
Introductory Post: Dip Buyer’s Dilemma: Sometimes You Buy the Waterfall
I’ll assume the first two posts have been read and cut straight to the chase.
I’ve added a lower Bollinger Band (50,2) to the system and changed the rules so that the system will not buy $SPY even when the decliners indicator is greater than 80.
The test is going to run from 1.1.2007 to 10.4.2012. The first graph is of baseline system performance where $SPY is bought at the close when the decliners indicator is greater than 80 and sold at the close 10 days later. The baseline performance includes the waterfall trades as identified here. The second graph shows performance after the Bollinger Band filter has been added.
As predicted, the Bollinger Band filter reduces performance. However, it does lower the max trade % drawdown and the max system % drawdown. It also reduces the total number of trades by 25 which lowers exposure significantly.
But does the filter keep the system out of the waterfall trades? Let’s look. In the interest of transparency, I’ve included all the trades made. The list is long. The trades highlighted in yellow were on the original waterfall list and were NOT filtered out.
Below is the original waterfall list:
Using the lower Bollinger Band filter eliminated 7 of the 12 waterfall trades. However, a quick scan of the trade list shows that the system added other waterfall trades. Just scan the % Change column and look for the large losing trades.
While adding the filter did eliminate more than 50% of the waterfall trades, it hurt performance overall (as predicted). To see how much the filter hurt the system, just compare the two equity curves.
With Bollinger Band Filter
Without Bollinger Band Filter
In the next post I’ll apply another popular filter.
To be continued…
I have written many posts about abnormal markets. On Friday, $SPY met a basic criteria for being abnormal with a bias to the upside.
There are times to short and there are times sit on our hands. How do we determine when to short and when to sit? One method I like is to use the upper Bollinger Band (50,2) as an abnormal market filter. I have found through backtesting that a close above the upper BB signals a market that may be entering or continuing a sustained uptrend. In other words, it is a market that is less-likely to revert to the mean and more likely to trend. This may be counter-intuitive as we typically view a market that has been sharply rising as one that is due for a correction, or at the very least, a pullback. What I have found after a close above the upper Bollinger Band is that instead of a sharp pullback or correction, we see some consolidation and then a resume of the uptrend.
Let me demonstrate.
Buy $SPY at the close if the close yesterday was NOT above the upper Bollinger Band AND the close today IS above the upper Bollinger Band (50,2).
Sell at the close X days later.
These results compare buying $SPY according to the rules above vs. buy and hold. To generate the buy and hold results and have them comparable, I split all of the $SPY history into 50 day segments and then averaged the segments.
The buy and hold results show the clear bias to the upside that has been prevalent over the past 20 years. The setup results show a market that has closed more than 2 standard deviations above the 50 day mean and still manages another surge almost 2 months later to close higher on average than buying and holding.
And that is why this triggers the abnormal market filter. It is not a market that we want to aggressively short. We may also not want to be aggressively long. The results show we may have a few weeks to consider our positioning and bias.
The bottom line is that I am looking for the market to trend more than swing. Since the June 4th low, we’ve seen fairly predictable swings. These results suggest that the swings may be less predictable and the market may shrug off bad news as it climbs the wall of worry.
Here is the setup: Buy SPY at the close if it closes less than 0.5% above the lower Bollinger Band (50,2). As you will see below, unlike almost all studies looking 50 days forward, this one is bearish.
If one applies Bollinger Bands to SPY using a 50 day period and 2 standard deviations it is evident that SPY rarely closes near the lower band. Just look back over 5 years and you’ll see what I’m talking about. Usually when it does get near the lower band, it bounces. But sometimes it doesn’t bounce…
Since SPY closed on Friday less than 0.5% above the bottom Bollinger Band, I was curious what had happened when that occurred in the past.
The graph above shows the average trade spends most of its time in negative territory after this setup.
However, it is important to note that these results are heavily influenced by a few large losers, with the largest being a -29.84% loss on November 25, 2008.
In fact, with a 62.5% win rate (generated by selling the trade after 50 days), the setup has a decent edge in terms of predicting a higher close over the intermediate term. Unfortunately, in terms of the average winner compared to the average loser, the results are not nearly as positive.
The average winner was 4.77% The average loser was -8.99%.
So while over the intermediate term there is a better than average chance that 50 days later SPY will be higher than Friday’s close, if the setup does not work, the damages could be severe.
I might be a tad melodramatic when I write this, but we are nearing a bounce or die situation.