Every January, we conduct a comprehensive review of our forecasting and signal performance during the previous year. Accountability is very important to us and this performance review is another way for us to demonstrate our commitment to providing you with a highly reliable service. The audit process itself is admittedly tedious and time-consuming, but we believe it ultimately provides great benefit to both you and us.
In every short-term, intermediate-term and long-term forecast, we provide outlooks based on technical and cycle analysis of the markets that we monitor. Our forecasting methodology identifies two scenarios, one bullish and one bearish, along with the approximate probability of the more likely scenario, providing there is one. These outlooks fall into one of four broad categories.
- If both scenarios are equally likely, they each have about a 50% chance of occurring.
- If one scenario is slightly more likely, it has about a 60% chance of occurring.
- If one scenario is more likely, it has about a 70% chance of occurring.
- Finally, a highly likely scenario is at least 80% likely to occur.
The likelihoods associated with these outlooks are intentionally imprecise. We do not believe it is possible to calculate probabilities with a high degree of precision when it comes to financial market forecasting, so each category covers a relatively large range and offers only a rough estimate. However, rough estimates are all you need as an investor or trader to be successful over the long run. A disciplined strategy of aligning yourself with the likely scenarios and protecting yourself from the unlikely ones will invariably produce gains over time. With the odds consistently on your side, you will be successful.
The key, of course, is to have reliable probabilities on which to base your decisions. We may claim that a given scenario is ~70% likely, but what assurance do you have that this assessment is reasonably accurate? The only way to know for sure that our forecasting process is reliable is to review the results of past predictions. We performed this auditing process for every forecast during 2013 and the results are summarized in the following table.
|Click here to download the forecast audit data spreadsheet|
Each column displays the actual performance of a given forecast category during 2013. The top row, labeled “All Markets,” compiles the results for every forecast in every market that we monitor, and then subsequent rows display the results for individual markets. As usual, there was a great deal of variance from market to market, especially in the equally likely category, reflecting underlying long-term trends. If you would like to check the results of the audit yourself, click on the link below the table to download the spreadsheet containing the data.
Our computer models produced 1,817 forecasts during 2013 and our methodology performed very well during the year. The equally likely and ‚Č•80% likely categories were very close to their theoretical values at¬†45.3%¬†and¬†84.5%, respectively. The ~60% and ~70% likely categories were well above their theoretical values at¬†69.5%¬†and 81.8%, respectively. Our computer models are intentionally conservative with their projections, especially when it comes to scenarios that are at least 70% likely, but these results suggest that our models remain a little¬†too¬†conservative for the ~60% and ~70% categories. We will continue to use these audit data to fine-tune our forecasting process moving forward, but an important takeaway from these results is that our computer models continue to perform exceptionally well when it comes to identifying likely scenarios. In other words, when we forecast that a given scenario is ~70% or ‚Č•80% probable, you can be confident that it is at¬†least¬†that likely. We have now evaluated more than 10,000 forecasts and the actual probabilities for all categories since we began auditing our forecasts are displayed in the following table.
Inflection Point Identification Audit
Our cycle analysis (CA) methodology identifies temporal inflection points across all time frames for the markets that we monitor. Once a confirmed cycle low is in place, we use CA to determine the window during which the next low is likely to occur. We have tested our methodology using several decades of historical data and the results of this validation process indicate that approximately 70% of all cycle lows should occur within the predicted window. Of the lows that develop outside of the window, about 20% typically occur before it begins and 10% typically occur after it ends. The following table summarizes the performance of our CA during 2013.
There were 69 cycle lows across all markets and 44 (63.8%) of them occurred in the predicted window, which was slightly lower than theoretical expectations. Of the remaining lows that formed outside of the window, 18 (26.1%) occurred before the window and 7 (10.1%) occurred after it, which was slightly higher than expected for the occurrences before the window. While it is important for actual forecasting results to align closely with theoretical expectations, the ultimate goal of CA is to reliably identify temporal inflection points as they are developing. In this regard, we had another very successful year, as 66 (95.7%) of the 69 lows were correctly identified in real-time as they formed. Even when a given low does not develop within the predicted window, having an accurate time frame for its arrival still enables it to be reliably identified, regardless of whether it is “early” or “late.”
With regard to all inflection points (including alpha highs, alpha lows, beta highs and beta lows), our methodology correctly identified 92.6% of all turning points, while issuing false signals only 4.3% of the time.
The following series of charts displays the identified turning points for each market, along with the accompanying false signals.
S&P 500 Index Short-term Inflection Points
10-year Treasury Note Yield Short-term Inflection Points
US Dollar Index Short-term Inflection Points
Gold Market Short-term Inflection Points
Oil Market Short-term Inflection Points
S&P 500 Index Intermediate-term Inflection Points
10-year Treasury Note Yield Intermediate-term Inflection Points
US Dollar Index Intermediate-term Inflection Points
Gold Market Intermediate-term Inflection Points
Oil Market Intermediate-term Inflection Points
Overall, our inflection point identification process performed exceptionally well once again in 2013 and we do not foresee making any meaningful changes to the methodology.
Our forecasting based upon technical and cycle analysis provided highly reliable outlooks throughout the year. Our inflection point identification process had a very successful year as well, correctly identifying developing lows in real-time 95.7% of the time and identifying 92.6% of all inflection points while issuing false signals only 4.3% of the time.
Overall, it was another violent, volatile year in the markets that we monitor, which is precisely what we anticipated during this stage of the secular bear market in stocks that began in 2000. The current cyclical bull market in stocks that began in early 2009 is displaying typical characteristics of a late-stage bubble and it will almost certainly terminate sometime during the first half of 2014, so it will be important to monitor market behavior closely for the development of the overdue cyclical top. The secular bear is still several years away from its terminal phase, so we expect the current environment of volatility and violent price behavior to persist during the coming year.