Our sponsor has made the sensible request that we discount forecasts after the answers are known — even if it took us awhile to pause the question. We have now instrumented SciCast to record and show ‘Date Known.’ Forecasts after ‘Date Known’ will still be part of the historical record, but they will neither gain nor lose points. Remember to check this during the mandatory 48-hour comment period following a proposed resolution. (In the future it may be possible to soften the cutoff, or create a fixed bonus for first correct report of a resolution.)
The Official SciCast Blog
A crowdsourced science & technology forecasting project from George Mason University.
nice. that makes it safer to trade automatically, as i don’t have to worry about my program “fighting” with people who know the outcome (even if the outcome i thought most likely was the one that occurred.)
Users who have LIKED this comment:
What is “known”? Known to whom?
@dvasya: Publicly known, for example published, announced, reported, usually in accordance with the resolution criteria. Credible leaks (e.g. in the New York Times) may count, if they remove most of the uncertainty / information. Hypothetically, if Murphy and Akroyd “forecast” based on the stolen orange crop report, their forecasts will count up until the point the actual report is released. But if they prominently published the report ahead of time, their forecasts might not count, depending on how prominently.
Note: SciCast does not advise stealing orange reports, and no oranges were harmed in writing this blog. However, a large number of blueberries were ingested.