Anatomy of a PunditScore Box

The PunditScore Box is a concise representation of a person's performance at punditude.com.

Total Positions: Beginning at the lower left, this is the total number of different Predictions this person has taken a Position on. In the example above, this person has taken 10 Positions.

Raw Accuracy: At the lower right is a fraction. The denominator(bottom) is the number of resolved Prediction this person has taken a Position on. The numerator(top) is the number of correct Positions this person has taken. In other words, this is the simplest measurement of accuracy. In the example, this person has made 1 correct Prediction out of 3 possible resolved Predictions.

Raw Average: At the bottom is a numerical representation of the raw accuracy score. Think of it like a batting average in baseball, but without the decimal point. This can range from 0 to 1000. In the example, 1 out of 3 is 0.333, which is 333 on the 0-1000 scale.

PunditScore: Finally, the PunditScore itself appears in the upper area. The PunditScore is the result of our proprietary algorithm which incorporates several factors, most importantly accuracy and the length of time between establishing a Position on a Prediction, and when that Prediction is resolved. In the example, the person has 1 Hit and 2 Misses, so it's reasonable (but not absolutely certain - see below) to expect that the PunditScore will be below 500.

The reasoning behind rewarding early predicting is simple. The difficulty of making any prediction is associated with the information available about the circumstances of that Prediction. As time advances, more information is inevitably made available. This necessarily makes it easier, on balance, to make accurate predictions as time goes on. Therefore, the PunditScore algorithm punishes those who would wait until the last minute, and rewards those who can make accurate predictions as far in advance as possible.

By the same reasoning, an incorrect prediction is "punished" more harshly the closer to resolution it is made. Consider the prediction that George W. Bush would carry Michigan in the 2004 presidential election (ultimately false). Made 6 months before the election, this error shouldn't count for all that much. But the same prediction made on election night (when all possible information is available) is a much more significant blunder, and is therefore "punished" accordingly. To summarize:

• Long-range correct predictions: Very good for PunditScore
• Short-range correct predictions: Somewhat good for PunditScore
• Long-range incorrect predictions: Somewhat bad for PunditScore
• Short-range incorrect predictions: Very bad for PunditScore

It's important to note that each resolved Prediction you've made will be measured against a neutral score of 500. Each correct Position ("Hit") will push your score above 500, and each incorrect Prediction ("Miss") will drive it below the neutral point. Therefore, even if all your Positions have been accurate, your PunditScore will almost certainly not be 1000. Depending largely on how short-ranged your Predictions have been, it may well be below 600 (but it will certainly be above 500).