(T2108 measures the percentage of stocks trading above their respective 40-day moving averages [DMAs]. It helps to identify extremes in market sentiment that are highly likely to reverse. To learn more about it, see my T2108 Resource Page. You can follow real-time T2108 commentary on twitter using the #T2108 hashtag. T2108-related trades and other trades are posted on twitter using the #120trade hashtag)
T2108 Status: 45.9%
VIX Status: 14.2
General (Short-term) Trading Call: Hold (1-day trade is to short $SPY based on the T2108 2-Day Decline Model. See below for more details)
Reference Charts (click for view of last 6 months from Stockcharts.com):
S&P 500 or SPY
SDS (ProShares UltraShort S&P500)
U.S. Dollar Index (volatility index)
VIX (volatility index)
VXX (iPath S&P 500 VIX Short-Term Futures ETN)
EWG (iShares MSCI Germany Index Fund)
I am FINALLY ready to announce a new beginning for T2108. On Tuesday night, I presented my final project in my data mining and machine learning course (using R as the programming platform), a project based on an analysis of T2108. I am now ready to start rolling out my results and methods. This remains a work in progress as I have already written up nearly a page of things I would like to do, try, and enhance. Thus, I will not yet publish my project until I feel I have addressed several additional modeling questions and done even more thorough data investigations. However, if you would like to see the presentation anyway, I will email it to you if you would kindly agree to send me feedback. (Go to my contact page to send me a message.)
I had one major objective: use data mining techniques to develop a more rigorous definition of oversold and overbought conditions. I produced some encouraging results but not yet as conclusive as I would like. 20% still works OK as a threshold for oversold and 70% is adequate as a threshold for overbought, but other threshold levels hold very interesting properties as well (like 40% on the downside and 75% to the upside) that produce relatively reliable predictions for short-term trades on the S&P 500. Also, the switching behavior I have noted for overbought periods – where the odds switch from the bears to the bulls once the duration hits around 20 days – is a common characteristic for many thresholds. After seeing these characteristics, I decided that the concept of oversold and overbought may be too narrow for how to interpret the crossing of threshold levels. Stay tuned on this.
A fortuitous objective appeared after the first few weeks of the class. Recall in mid-February I wondered whether a steep 2-day decline in T2108 could be considered a “quasi-oversold” condition. I predicted a bounce, and, voila, it happened. That event launched me into an early investigation using data mining to see whether 2-day declines in T2108 could meaningfully predict whether or not the S&P 500 would trade up or down the next day. My preliminary results were encouraging, but incomplete (and not quite correct given a very small flaw I found in my data!). I have now developed the model enough to make predictions.
T2108 declined 17.8% on Tuesday and has declined -18.4% over two days. Using tuned decision trees (classification), my model concludes that Thursday will be a down day for the S&P 500 with a 74% likelihood. Yes, this took my aback as well since I was hoping for more consistent behavior of bouncing after steep declines. But many factors are at work shaping this prediction, and it seems the VIX did not increase enough to generate an up prediction. (The error rate for this model is also better for down predictions than up predictions). Note well that I am not predicting the magnitude of the down day. The different machine learning models I tried were terrible at predicting the magnitude of the decline, and I expected this behavior. The predictors I used in the model do not suggest that they can have anything to say about magnitude.
The decision tree for this prediction is extremely deep (too deep!), and this is an area where I need to play around with modeling parameters to find the right balance between simplicity of the tree and prediction accuracy (or from a trading perspective, prediction utility, that is, the usefulness of the prediction). For example, the decision tree used to make the prediction for Thursday (April 4th’s) trade is as follows (all quantities are based on April 3rd’s closing trade):
The VIX changes by more than -.68% and
the VIX changes by less than 18.5% and
T2108 closes at 34.4 or higher (traverses three branches deep) and
T2108’s 1-day change is greater than -19.5% (traverses two branches deep) and
the S&P 500 is above its 50DMA and
FINALLY, the VIX changes by more than 5.5% (which was already established at the top of the tree)
Note that unlike my preliminary conclusions, the VIX CAN matter in these predictions. It is just not a consistent predictor. In this case, a VIX change ABOVE 18.5% would have led immediately to the terminal conclusion that the S&P 500 is likely to close up on Thursday.
These kinds of deep trees are not amenable to easy trading rules. I have an important goal to generate the simplest T2108 trading rules possible without sacrificing too much effectiveness. In the meantime, I will run the model to produce the predictions and report the results here. One to-do is to get more familiar with the model’s sensitivity to the data used to construct it (that was an early lesson from February!). For example, the trees I am currently building will change depending on the magnitude of the 2-day decline in T2108.
If none of that made any sense to you, then you are definitely someone who needs to ask clarifying questions below. If this made perfect sense to you, and you are shaking your head in horror at my methods, then you especially should be writing your critique below! I am looking forward to all feedback on this (living) project.
Now, back to the usual programming…
T2108’s sharp decline took it to a new low for the year. We last saw these levels right after Thanksgiving as the S&P 500 was starting a post-election recovery. The S&P 500 fell from all-time highs for a 1.1% loss. It also broke the primary uptrend at the 20DMA for the first time since mid-February (when it hit a quasi-oversold level).
The VIX popped but STILL remains under critical resistance around 15.4. It did not even touch a new high for the area of churn that has dominated the VIX for the past several weeks.
On twitter, I have been tweeting several warning signs. I have had some false positives earlier, but the bearish technical evidence just keeps mounting, including:
- Apple (AAPL) unable to sustain its breakout above its 50DMA and trading significantly lower just as the S&P 500 hit the all-time high last Thursday.
- Caterpillar (CAT) continues to sink ever lower. Never a good sign when cyclicals are in retreat. CAT is now DOWN for the year and off its 2013 peak by over 17%.
- A bearish divergence developed on Tuesday after the S&P 500 traded right to the all-time high, but T2108 surprisingly closed down slightly.
- Finally, today’s mini-breakdown on the S&P 500 was confirmed early by weakness in the Australian dollar against the U.S. dollar (FXA).
While none of these bearish indicators are necessary to raise the red flag, they are sufficient to do so. As I type, the Bank of Japan (BoJ) has added intrigue to tomorrow’s session that will also follow decisions by the central banks in Europe and the United Kingdom. On Friday, the U.S. will roll out another employment report. In other words, there will be PLENTY of fundamental news to give bulls and bears excuses for pulling quick triggers. The T2108 models will not be listening to any of it. 🙂
Daily T2108 vs the S&P 500
Black line: T2108 (measured on the right); Green line: S&P 500 (for comparative purposes)
Red line: T2108 Overbought (70%); Blue line: T2108 Oversold (20%)
*All charts created using freestockcharts.com unless otherwise stated
Be careful out there!
Full disclosure: long VXX shares and puts; long CAT shares and put spreads; long AAPL shares, calls, and puts