Stocks have now rallied nicely, and SPY appears to have broken free from the confines of its bottom Bollinger Band. Let’s model the action of the past four days and see what has typically happened going forward.
Four days ago, SPY made a new 50 day low. Since making that low, SPY has rallied, and the 4 day rate-of-change (ROC) is at 1.91%. The index ETF closed above its 200 day moving average and closed beneath its 20 day moving average.
The Rules:
- Buy SPY at the close when 4 days ago SPY made a new 50 day low and today the 4 day ROC is > 1.5%
- Sell X days later
- No commissions or slippage included
- All SPY history used
The Results:
Summary of Results:
First, let’s look at sample sizes:
- Baseline setup had 61 trades held for 1 day and 36 held for the full 50 days
- Adding C<MA20 to the baseline gave us 39 trades held for 1 day and 26 held for the full 50 days
- Adding C>MA200 to the baseline gave us 25 trades held for 1 day and 21 held for the full 50 days
- Adding both C<MA20 AND C>MA200 to the baseline narrowed things down quite a bit with 17 trades held for one day and 14 held for the full 50 days
- If you are confused by the sample sizes and why holding trades longer reduces the sample size, let me know in the comments section
Sample sizes should be a concern for the model that is most similar to recent action, which is the baseline with C<MA20 and C>MA200. That is too bad because that is the model with the best performance. However, forgetting about the MA20 and just requiring a close above the MA200 gets the samples up to 21, and it also yields decent performance.
Something that is confusing about this study to me (but I didn’t really investigate it thoroughly), is why just requiring the baseline and a close beneath the MA20 has such mediocre results (the blue line). Results are very similar to the baseline. Yet requiring a close above the MA200 improves things significantly. I’m guessing that when this setup occurs, it almost always occurs with SPY closing beneath the MA20, but it doesn’t always occur with it trading above the MA200.
Bottom Line: When SPY rallies shortly after a new 50 day low, some of that rally is usually retraced over the next 5-9 days. We’ll call that a re-test of the bottom or low. However, after this re-test or retracement, SPY has typically exhibited higher than average performance over the next 40 days. If we couple this study with my other study about bottoms and volume surges, it appears we just might have a tradeable bottom in the making.
This is great help navigating the markets and my essential read of the day besides the PPT-scores.
Your finding today fits nicely with the PPT-OB indicating a dip the next few days.
Thanks!
Wood,
Studies like this are quite useful and thought-provoking, though I do have one question regarding these that won’t go away – what does the SPY do on average compared to the study conditions? For example, the Sell 25 Days later in the chart above for C > MA200 and C < MA 20 looks good at ~3.5% or so in absolute terms, but if the SPY's typical (average) 25-day net over the backtest period was 3.0%, how significant is the backtest? In other words, are the backtest results due to the conditions of the setup or the market's general tendency to go up over time (or does this comparison even matter)?
KTB
KTB, you are exactly right, about needing to separate the setup from the study conditions. From time to time, I will put up average SPY performance separate from the study conditions, for comparison. As the charts tend to get cluttered, I don’t always put it up there.
Right now, over the average 50 day period of time, SPY does about 2% (going from memory). While I don’t always put this up for comparison, it always in my head, and I will typically hedge with comments like “under-performance is expected” or “consolidation expected.” On the next post, I will include SPY over an average of all 50 day periods so you can check it out.
Thanks!
Never comment but enjoy your analysis, good sir.
Thanks Dubz. I hear you are quite the pelican master in the PPT.
Wood – good stuff as always…
Re 20/200 – first q – how much overlap was there with the sets? Of the 39 20s, how many of those were 200 trades and vice versa? If there is a great deal of overlap, this could be signaling that the 200 trend is far stronger in terms of overall market direction.
Did you happen to try c+20sma instead of just lower? That might echo the 200 results and be signaling that a bounce which shows strength enough to close over major trends follows thru – but if the trade doesn’t have enough rally (or the drop leading up to this was so strong downward) that the rally fades and downtrend continues
My broker gave me the election year cycle charts back in May. The chart showed SP 1250 in June or July and then a rally starting in September lasting until Q4 2012. Imagine, he made this prediction when the SP was 1350.