Pre-mortems help you consider why you might be wrong
As much as we may hate it, sometimes we are wrong. Our predictions are no different, and we are constantly trying to think of ways to make them better. One of the counterintuitive ways is to imagine that, once we know the answer, we are wrong. In other words, if your prediction is a 75% chance something will happen, you’re also saying there’s a 25% chance it won’t. What’s the “won’t” case?
INFER recently introduced a pre-mortem – an attempt to think about why your forecast could be wrong, to challenge your intuition and help avoid extreme forecasting when it’s not warranted. This method has been backed up empirically – pre-mortems are associated with a greater ability to identify future causes.
While the pre-mortem will always be optional, we strongly recommend taking advantage of it.
See your accuracy over time and how you compare to others
Another feature we have recently added is the ability to compare your performance to other forecasters. When viewing the Performance section of your (or anyone’s) profile, you can now compare yourself to various percentiles. Let’s look at @ben’s stats:
While the Relative Brier Scores are often an excellent way to compare yourself to other forecasters, it’s challenging to know the distribution of Brier scores. This issue is where the new feature comes in, where you can compare yourself to other forecasters. Let’s see how @ben stacks up against the median forecaster:
Much better! (Remember that the lower your Brier score, the better). We also can compare him to the 90th percentile:
Even the top forecaster of this season:
It looks like @ben may have a bit of work to do to catch up to our leader, but he’s still in the top 15% of all forecasters, as he’s just below that purple line.
INFER’s development and engineering team constantly releases new features - you can check out the latest release notes here.