🏁 May 2026 Competition & Season 1: Closing Updates and What Comes Next

Greetings, fellow AlphaNovans! 👋

First of all, thank you to everyone who has participated in our competitions. We’ve seen some amazing work and a wonderful range of ideas. The creativity, persistence and thought you’ve put into your submissions have made this a very rewarding experience for us.

October 31 was a bit of a “double witching hour” for AlphaNova: it marked both the close of Season 1 submissions for our biweekly competitions and the end of the three-month live evaluation period for the May 2026 competition.

We’ll announce the May competition winners and the latest completed biweekly results shortly, once our final checks are complete. The remaining Season 1 live evaluation windows will continue as scheduled.

Before those announcements, we have a few updates to share. The first two concern mistakes on our side, and we want to explain them clearly.

📅 1. Season 2 begins on November 15

Our biweekly rules say that cycles begin on the 1st and 15th of each month. Our overview also describes a season as five cycles followed by a one-week break. Those two schedules don’t line up.

To preserve the established start dates, we’ll take a two-week break between seasons. This also gives us the time we need to prepare the next season properly.

Season 2 will therefore begin on November 15, 2026.

Apologies for the inconsistent wording, and thank you for bearing with us.

🏆 2. A correction to the Cycle 2 prizes—and a larger Cycle 3 pot

In Cycle 2, we paid the top three prizes from a full $2,500 pot. However, that cycle produced 11 quality signals, below the 13-signal threshold needed to reach the full pot.

This was my mistake.

Under the published formula, the correct Cycle 2 pot was:

100 + 2,400 × (11/13)^0.75 = $2,217.38

The difference of $282.62 should have rolled into Cycle 3. We will honour that rollover, bringing Cycle 3’s prize pot to $2,782.62.

By coincidence, the first- and second-place finishers in the provisional Cycle 3 results are the same people, in the same positions, as in Cycle 2. We can therefore reconcile their previous overpayments against their new awards.

Their announced Cycle 3 prizes will reflect the larger pot. Their actual payments will account for the amounts already paid in excess during Cycle 2:

Place	Cycle 3 prize	Previous overpayment deducted	New payment
1st	$1,669.57	$169.57	$1,500.00
2nd	$695.66	$70.66	$625.00
3rd	$417.39	—	$417.39

The Cycle 3 third-place winner will therefore receive $417.39, rather than the $375 payable from a $2,500 pot.

The Cycle 2 third-place winner’s $42.39 overpayment will be recorded and deducted from a future prize if they win again. We’ll explain the adjustments individually to the affected winners.

I’m sorry for the confusion. We want both the published awards and the payments to reflect the prize rules correctly.

🔍 3. Additional checks for models trained offline

We’ve noticed a number of submissions using models trained offline, with little or no training performed by the submitted code. Offline training is allowed. We are developing additional overfitting checks suited to these submissions.

Our first step will be to compare performance on the data we provide to participants with performance on our additional unseen data. That unseen history extends well beyond the training, validation and test datasets available in the competition package.

A sharp, material deterioration that lines up with the boundary between supplied and unseen data will be flagged for an overfitting review. We’ll take ordinary statistical variation and changes in market conditions into account when assessing those flags.

Longer term, we plan to develop a way to programmatically reconstruct a training method for these submissions, where feasible, so that they can also undergo our existing retraining-based overfitting tests.

The aim is to reward signals that hold up on unseen data and keep the competition fair for everyone.

---

Thank you again for the energy and ideas you’ve brought to AlphaNova. We’re looking forward to sharing the results—and to seeing what you bring to Season 2 on November 15! 🚀

NONIUS
On behalf of the AlphaNova team