Simple Forecast Averaging

Simple Forecast Averaging is the process of combining two or more independent predictions into one by calculating their mathematical average. The correct approach in the opening scenario is to average the forecasts to 125 units, as this new number is often more accurate than either individual guess. This works because individual forecasts tend to have unique errors—one might be too high, another too low—and averaging them helps cancel out this noise, leaving a more reliable signal.

Try this: Next time you hear two different predictions for the same event, like a game's score or a project deadline, calculate the simple average as a third, often better, alternative.

Simple Forecast Averaging is a straightforward but powerful technique for improving predictions. The process involves three key steps: gathering inputs, processing them, and interpreting the output.

Input: Start by collecting two or more independent forecasts for the same unknown quantity. The key here is independence; the forecasts should be generated using different information, methods, or assumptions. If all forecasters share the same flawed data, the average will also be flawed.

Process: The mechanism is just simple math. You add all the forecast values together and then divide by the number of forecasts you have. For example, if three models predict a company's quarterly revenue will be $1.2M, $1.4M, and $1.3M, you add them up (1.2 + 1.4 + 1.3 = 3.9) and divide by three.

Output: The result is a single, combined forecast ($1.3M in the example). This average forecast tends to be more robust because it smooths out the idiosyncratic errors of the individual models. One model's over-optimism might be balanced by another's pessimism.

A project manager can use this when planning. If one developer estimates a feature will take 8 days and another estimates 12, their average of 10 days becomes a more reliable estimate for the timeline, accounting for unseen complexities without being overly cautious.

Even in daily life, this is useful. If you are trying to decide when to leave for the airport, you might check two different navigation apps. If one says 40 minutes and the other says 50, planning for a 45-minute journey is a practical application of forecast averaging.

The surprising effectiveness of combining forecasts, often referred to as the 'wisdom of the crowd' effect, has been a subject of study for decades. Research consistently shows that a simple average of multiple decent, independent forecasts is exceptionally difficult to beat, even with more complex, weighted models. The core finding is that aggregation reduces variance, effectively canceling out random errors inherent in any single predictive model.

• Bates and Granger (1969) provided some of the earliest formal evidence that combining forecasts from different models could significantly improve accuracy over selecting a single 'best' model. [1]
• Armstrong (2001) conducted extensive reviews and found that simple, unweighted averaging is a robust strategy that performs well across various domains, often matching or outperforming more sophisticated combination techniques. [2]
• Clemen (1989), in a comprehensive review of over 200 studies, concluded that combining forecasts is one of the most valuable and practical strategies available to forecasters, with the simple average serving as a powerful and simple-to-implement benchmark. [3]

Simple averaging is not a silver bullet. Its effectiveness diminishes if:

• Forecasts are not independent. If all forecasters are influenced by the same information or groupthink, their average will simply reflect that shared bias.
• Expertise levels vary drastically. Averaging a true expert's forecast with several novices' guesses will dilute the expert's valuable insight.
• The system is non-linear or prone to extremes. For predicting events like market crashes ('black swans'), where historical patterns are poor guides, simple averaging may provide a false sense of security.