New leaderboards are wonky

The code which calculates the values has changed (mostly because the old code was a stopgap measure which had stopped working properly). I have more confidence in the new code.

Quarterly gains as they stand should represent the total net gains from making predictions. There may be some discrepencies compared to the old code where predictions were resettled - we let you keep the cash from a prediction that was wrongly called winning, but we don't want to let you keep the credit :) . I believe that's the only difference, but feel free to do some adding up and tell me I'm wrong.

Short term means predictions which were/are open for less than 48 hours, and long term means ones which were/are open for more than a week (which is just the kind of prediction you like to avoid infernalmachine!). These boards are mostly a proof of concept, so feel free to tell us how you feel about our choice of metric (though bear in mind that I'm very attached to the long-term one because I'm top :D )

The main reason for not clawing back the cash was that we were reluctant to put people into negative credits or inconvenience their investment strategies. Neither of these are reasons to wrongly give them credit for something they were wrong about.

We are currently trying to work out exactly how to treat these positions though, since as it stands if the market is reopened and left open you get to keep the cash but not the credit, and you cannot sell out of the prediction which is a little harsh. Real use of the site throws up so many amazing edge cases....