Hi everyone, am I the only one feeling confused about how this competition actually works? 😅

Let me break down my confusion:

We're currently in week 47 (Nov 17-22) and we're supposed to predict weeks 48 and 49. But here's the thing - those weeks haven't even happened yet! The actual prices for week 48 will only be known on Dec 2nd, and for week 49 on Dec 9th.

So my big question is: what is Zindi using RIGHT NOW to calculate our leaderboard scores? Like, if nobody has the real prices for weeks 48-49 yet, how are we getting MAE and RMSE scores?

And about the submissions - the template shows weeks 48 through 52 plus weeks 1-2. But the rules say we only predict 2 weeks at a time. So should I be:

- Submitting predictions for ALL those weeks every time?

- Or just the 2 weeks we're actually forecasting and put zeros for the rest?

Right now I'm only predicting weeks 48-49 and leaving the others as zero, but I'm not sure if that's correct.

Also, I'm wondering how some geniuses are already scoring 0.9's on the leaderboard! ✨🎉🎊 how is that even possible when we're predicting future prices that don't exist yet? make me wonder is there something am missing?

And this is what worries me - sometimes when the LB score looks bad we dismiss an approach, but then later find out it was actually the best one. By then it's impossible to retrace our steps because we've moved on to other methods.

My training MAE is like 0.11 but my leaderboard MAE is 1.65 - that's a huge difference! Makes me wonder what data is actually being used for scoring and if I'm completely misunderstanding something.

Can anyone shed some light on this?