We know from previous tests that when a market is making new highs, it continues to make new highs. On April 2nd, SPY made a significant new high and then pulled back, with Friday’s close down -3.3% from its April 2nd high. Is this just a pullback, or the beginning of a significant market top?
Every market top is different. When modeling a top, we want to be sure not to curve-fit the model, mainly because if it is, there will be very few usable samples from the past.
One thing all market tops have in common is that the market recently made a new high. The 2nd thing they have in common is that after the new high, the market begins to move lower. These two elements are very simple ways to describe a market top.
On April 2nd, SPY closed at a new 50, 200, and 500 day high. To make sure we are not curve-fitting our model, we will use the new 50 day high. Friday, April 13th marks 8 trading days since a new high was made. For this test, we will look for occurrences when it has been fewer than 10 days since SPY made a new high. Finally, SPY has fallen -3.3% over the 8 days since making the new high. Therefore, we will look for instances when SPY’s 8 day rate-of-change is <-2.99%.
So there is the model of our current market top. Let’s see what has happened in the past after this setup.
The Rules:
Buy SPY at the Close If:
- A new 50 day high was made fewer than 10 days ago
- Since making a new high, the 8 day rate-of-change is <-2.99%
Sell SPY at the Close X Days Later.
No commissions or slippage included. All SPY history used.
The Results:
Using the new 50 day high parameter, there were 30 trades held for the full 50 days. I have found that to be a decent sized sample for this type of test. For fun, I added the new 200 day high and new 500 day high (200DH and 500DH) to the test/graph. This lowers the number of samples but makes the model more specific to our current environment.
To put it plainly, the results show that it very bullish for SPY to have recently made a significant new high. Even a sizable pullback similar to what we have seen over the past week has not negated the bullishness. After a month or so of consolidation, SPY has tended to begin climbing again.
These results suggest we will likely see an uptick in volatility and some consolidation and range trading over the next few weeks. This makes sense for a market that has not really taken a breather in over 4 months.
As we go forward, I’ll adjust the parameters of the model and report back on any new developments.
Another awesome study.. Senor Fly and Woodshedder are the reasons I come to this site.
Thanks!
Wood,
Question for you: do you ever compare your results to out of sample data? When I think about what you’ve presented, half of me tends to understand what you did and agree with the methodology. The other half starts to wonder how we can study historical gains or losses in the stock market as presented; to what extent is the effect of the the market top that you’ve modeled really responsible for gains that occured a month after a certain trigger? How do you isolate the uptrend that you’re seeing taking place after the event you’ve coded for, from the markets historical tendency to the upside?
I would ‘think’ that a more accurate measure would be to compare risk adjusted returns to random data. Just a thought. Curious to hear your take.
Steak, with these kinds of studies, it is not easy to compare them with out of sample data as the events they model do not happen very often. For example, the setup described in this study has occurred 58 times in all SPY history. I guess I could create a database of all different studies and peruse it nightly to see if any current market events would be picked up by a previous study.
The market top I’ve modeled is not really responsible for the gains that occurred a month later after a certain trigger. What is the most likely to be responsible for the gains is momentum. A market making new highs tends to keep making new highs. There is very good evidence of that.
I added the pullback to see what effect, if any, it has on a market that recently made a new high. See the link in the first paragraph to compare a market making new highs without a recent pullback to a market making new highs with a recent pullback.
The point about the market’s natural tendency to trend up is a valid point. It is very hard to generate a study that results in bearish implications for the next 50-100 days. Again, the link in the first paragraph shows the SPY average plotted with the averages after making new highs.
As far as risk-adjusted returns go, trading this setup like a strategy, buying at the close and holding for 50 days, results in an annualized return of 3.73% with the risk-adjusted return at 12.24%. However, almost all of the gains were generated over the past 5 years. This setup has been very strong over the past 5 years and didn’t do much before that.
I will look at generating random segments of 50 day periods of SPY performance to compare for future tests.
Thanks!
looks like you were right on
I would expect more back and forth and volatility, but I will take the rocket ship instead!
BRAVO!
Thks, I try to write a AFL as your system, but can’t not. Can you help me how to write this AFL
How does this compare to the tops in April 2010 and 2011?
can you upload AFL code?
Hi TG, what part of it are you having problems with?