Issue in SMARD day-ahead challenges: Affected rounds cancelled

We discovered a serious issue affecting our 24-hour and 72-hour SMARD day-ahead market price challenges, and we sincerely apologize.

Between 28 April and 10 September, these challenges asked participants to forecast prices that had already been published by the original source. A participant noticed that some models were reproducing the published prices almost exactly and reported it to us. They were right, and we are grateful that they brought it to our attention.

The root cause was a TimescaleDB configuration error on our side. We did not change the default behaviour of its continuous aggregates to use real-time aggregation. As a result, the aggregates excluded values with timestamps later than the current time, even when those values were already in our database.

This matters specifically for day-ahead markets: prices for the next delivery day are published the afternoon before. By excluding those future-dated but already known values, we provided less historical context than we held. The system then placed the forecast window too early, overlapping with prices that were already public.

We have corrected the configuration, and the aggregates now include everything we hold. New forecast windows therefore begin after the latest published values.

We hold ourselves to the forecast pre-registration protocol, so we cancelled all 262 potentially affected rounds, 131 in each challenge. This covers every round between 28 April and 10 September, not only those where we found evidence that published prices were used in their forecast submissions.

The challenge leaderboards and all related rankings are being recalculated. This process may take some time. For the next few days, some models may not appear on the leaderboards if they only competed in the affected rounds. They will return once new evaluations are available for the two challenges.

No action is required from participants. We understand that cancelling this many rounds is significant, but we believe it is the fairest solution. We apologize again and hope for the community's understanding. A special thank you to the participant who identified and reported the issue.