Can you beat Ellie?

Most daily episodes end with a prediction: a number, a threshold and a date, scored automatically from archived on-chain data. Here you can make your own call on each one: how likely is it to come true?

You can change your mind until forecasting closes: 48 hours after the prediction is made, or halfway to its resolution date if that comes first. That's the same moment the YES/NO comment votes close. The forecasters' average and the crowd's vote count stay hidden here until then, so nobody copies them.

These predictions track live numbers you can watch move. If forecasting stayed open to the end, anyone could wait until the answer was nearly obvious, say 95%, and top the leaderboard without forecasting anything. Ellie commits the moment she makes the claim, with the whole wait ahead of her. Closing after two days keeps everyone close to that footing, while still giving you time to join in after the episode airs.

The Brier score: the square of the gap between your forecast and what happened. Saying 90% on something that comes true scores 0.01; saying 90% on something that doesn't scores 0.81. Lower is better, so confident and right beats cautious, and confident and wrong costs the most.

The headline is right calls: a forecast over 50% counts as a yes, under 50% as a no, and Ellie's call is always yes, because each claim is her prediction. Ellie also has a number for each prediction: the chance her claim-picking model gave it when the claim was made, logged before the outcome. It shows here only after forecasting closes, so nobody can copy it, and it is capped at 5–95% and scored with Brier exactly like yours.

The crowd is everyone who replies YES or NO to the prediction comment under each YouTube episode. Its call is the majority, and a tie is no call. One vote per account, counted until forecasting closes here. Votes have no percentage, so the crowd has no Brier score, only right calls. You need no account to join the crowd, and you can join both.