Payout anomaly on recent rounds

[lukmen74]

Hi everyone,

I’ve been analyzing my performance matrix because I noticed a drawdown on my account over the last 48 hours that didn’t make any sense based on the old calendar.

I just checked the leaderboard to cross-reference the data, and the evidence is staggering: 9 out of the top 10 models globally are simultaneously suffering massive, unprecedented crashes over the last 3 rounds.

Even though the official dashboard states that the new atomic multipliers (3xCORR60 + 15xMMC60) should only fully resolve weeks from now, it strongly looks like the new payout logic and the 60-day volatility index are already hitting recently resolved rounds that were technically bound to the old rules.

Has anyone else noticed this? The math on the live balance isn’t lining up with the official schedule.

[ark]

If by “crash” you’re referring to their stake values, these are decreasing in accordance with the Atomic Blockchain Staking migration. If you’re referring to scores, then please link an example of an account with an outsized negative score which does not align with your expectations. I’ve found no evidence of the 60-day scores being considered for payouts in recent rounds.

[lukmen74]

Hi Ark,

Thanks for the feedback. Let me share my specific case because the numbers simply don’t add up mathematically, which strongly points toward a sync or data-pipeline anomaly rather than a real market burn.

Look at my models (LUKMEN74 and LUKMEN74_LAB):

• The “Crash”: My live balance on the dashboard has been dropping by roughly 3% per day over the last 4 rounds, resulting in an overall ~12% drop.
• Checking 60 rounds back my performance is completely stable and positive. There is absolutely no outsized negative score or sequence of 4 disastrous rounds in my history capable of burning 12% of the stake under pure linear rules.
• Last resolved round tell me that I have about 15 nmr at risk.

Yet, my daily 1D indicators are green today (+0.60% and +0.99%), but the total live stake keeps dropping on the main UI.

Even looking at the latest officially resolved rounds (1321 and 1322), there is absolutely nothing pointing to a crash. Paradoxically, these rounds are both positive, and both their MMC60 and CORR60 metrics are in the green, for the two models.

Could this be a data migration lag or a node synchronization bug related to the Atomic Blockchain Staking rollover? It looks like the main dashboard UI is failing to fetch or inject the cumulative live payouts into the current node balance.

Are you or the team seeing any pipeline delays or cached state issues on the user profiles since the 5-rounds-a-week switchover?

I’ve also noticed that many top accounts are showing similar patterns, with unexplained drops of 10-15% in their recent rounds. It feels really strange that nobody is bringing this up on the forum. Perhaps it’s just my lack of experience making me worry, and for veterans who have been here for years, this kind of behavior is normal during system upgrades. I don’t want to alarm anyone; I just hope this post and your insights can be helpful to others who share the same concerns but have less experience with Numerai’s switchover dynamics.

[ark]

My live balance on the dashboard has been dropping by roughly 3% per day

Yes, the “total stake” values are decreasing in accordance with the Atomic Blockchain Staking migration + the transition from 20-day scoring to 60-day scoring. Specifically, each 20-day round that resolves, releases ~1/24th of your stake total into your new strategy contract. Then, each new 60-day round only stakes ~1/64th of your stake total to ensure you’ll be able to cover future 60-day rounds.

So you effectively get a 1/24 - 1/64 = ~2.6% reduction in total stake each day until the end of September - this NMR is in your strategy contract as idle balance. Then, as 60-day rounds continue opening (and while you wait for the first 60-day round to resolve), you’ll use that idle balance to continue staking and your total stake amount will continue to rise until you’ve fully staked on all 64 concurrently unresolved rounds, reaching your previous total stake value.

[lukmen74]

Hi ark,

Thank you very much for your detailed explanation. I sincerely apologize for missing the core details regarding the recent version upgrades and the transition to the new staking mechanisms.

Following your instructions, I checked the platform and successfully located the missing NMR under the idle balance section. Your breakdown makes perfect sense and completely clarifies what is happening with the dashboard balances.

Regarding this setup, I have a couple of follow-up questions to make sure I understand the new payout logic correctly.

First, to recap: until the end of September, nothing changes regarding the payouts as the old rules still apply. After that, payouts will be calculated based on MMC60 and CORR60 metrics (represented by the dots and dashed lines in the chart).

Take the highlighted round on Oct 28, 2026, as an example, assuming it resolves with these exact same figures (MMC60: 0.011, CORR60: 0.016). Based on my multiplier settings, I calculated a gross return of roughly 21% for that specific round. However, since we are no longer staking the full capital but only ~1/64th of the total stake per round, this means the net return on my total capital for that single round would be around 0.3%. Is my math and understanding of the net payout scaling correct?

Looking at my current dashboard, I can clearly see the transition starting around the late August rounds (like Aug 28), where the ‘At-Risk’ stake drops from around 7 NMR down to 0.11 NMR. Am I correct in understanding that this sharp decline is precisely due to the capital being divided by 64 across the simultaneous rounds?

Furthermore, since the Payout Factor (PF) shifts from around ~0.09 to a full 1.0, this implies that the payouts will become much more aggressive. Even with a smaller nominal stake per round, we will face a significantly higher burn risk paired with much higher potential returns if the model performs well. Is this interpretation correct?

Lastly, I’ve noticed that many top models on the leaderboard have been suffering heavy burns lately. Is this widespread drawdown directly related to the target shifting from predominantly Ender20 to Ender60, or do you think it’s just a temporary coincidence?

[ark]

Is my math and understanding of the net payout scaling correct?

Yes

Am I correct in understanding that this sharp decline is precisely due to the capital being divided by 64 across the simultaneous rounds?

Yes

Is this interpretation correct?

Not really. As you calculated, the per round burn risk across all deployed stake would be a maximum of 1/64 ~= 1.6% per round.

Is this widespread drawdown directly related to the target shifting from predominantly Ender20 to Ender60, or do you think it’s just a temporary coincidence?

If you’re referring to 1D returns, then it’s a combination of the v3 staking transition and the ender 60 switch - both of which are not fully completed yet, so the 1D returns (which are showing changes in unresolved rounds) are going to be somewhat unreliable indicators. If you’re referring to actual scores of resolving rounds, those are going to be unrelated since they’re still scoring and paying on ender 20.

[lukmen74]

Thank you so much Ark, this is incredibly helpful and clarifies almost everything. Really appreciate your time and kindness.