Issuepedia:Prediction Registry/About

From Issuepedia

Introduction

A lot of people make predictions about the future, but some people seem to predict more accurately than others. There are many cases in which knowledge of whose vision of the future has been clearer would be very valuable, especially in discussions of policy; the opinion of someone with a good predictive record should obviously hold substantially more persuasive weight than that of someone with an average record, or no record at all, while that person's advice should in turn probably be heeded more than that of someone with a comparatively poor record.

Until now, however, there has apparently been no serious attempt made to keep track of various predictions – who made them, and how they came out – and thus build up a record regarding which views of the future seem to be clearer than average. This area of Issuepedia is a first take at building up such a record.

Guidelines

In the spirit of Wiki and Web 2.0, there are no hard-and-fast rules; the registry will be largely depending on CITOKATE for maintaining accuracy and relevance. It would seem a good idea, however, to observe the following guidelines:

  1. Posts of predictions should indicate:
    • The date the prediction was originally made (if the date is uncertain, then whatever information is available)
    • The date the prediction was registered in the Prediction Registry
    • The original author of the prediction (the one who should get the credit if the prediction turns out to be accurate)
    • A link to the original source, if the author is not the poster (if the original source is not available online, then it may be quoted -- if it is long, then the poster might set up a separate page titled "Prediction nnnn" and link to it)
  2. Each prediction should be patrolled to see if there have been any relevant events which may indicate progress towards either satisfying the prediction or contradicting it
  3. Any given prediction should have one of the following status indicators:
    • open: conditions have not changed substantially, since the prediction was made, towards either verifying or falsifying it
    • updated: some conditions have changed which increase the likelihood of verification or falsification, but it is not yet clear how the cards are going to fall
    • [un]likely: conditions have changed to the point where the prediction is likely or unlikely to come true
    • TRUE: the prediction came true (should include specific details or a link to a relevant news story)
    • FALSE: the prediction has been falsified (should include specific details or a link to a relevant news story)
  4. Past predictions do not have to be accompanied by proof; if proof or evidence of a prediction's date is scanty or unavailable, however, this should be indicated. Lack of evidence diminishes a prediction's credibility, but far less so if the lack of evidence is stated at the outset; such predictions may still be useful data.

Plans

Eventually, if the quantity of predictions entered makes it worthwhile, this data will be entered into database whose contents and design/code will be available online under the GNU FDL.

Related Links

  • Long Bets: "Accountable Predictions"
  • Marketocracy uses statistical analyses to find individuals who best predict the stock market, and then uses the predictions of those individuals to obtain better-than-average returns on investment
  • Technovelgy: these folks seem to have a lot of the right spirit, but their site is a traditional one-way "walled garden" design. They do, however, accept user contributions, at least in the form of corrections. Some of the phraseology in the FAQ implies that the site is run by a single person; perhaps this person could be persuaded to wikify the site?
  • probably on the wrong track:
    • Prediction Company: seems to be attempting to apply algorithmic approaches to market prediction. I don't see any discussion of their model nor their track record, so this doesn't seem to be terribly useful. They also seem to be of a proprietary nature, which would limit the usefulness of their model even if it were a good one.

[_pw5_]

nv