Looking at the Subnet Metagraph, it appears that the incentive is nearly identical to consensus (corr ~ 0.999) and strongly tied to Trust (corr ~ 0.85). Which demonstrates that the intended distribution is achieved. Although we also notice:
- Distribution is very flat right now:
- Gini (Incentive among miners) ~ 0.03 (very equal)
- Top-10 miners’ share ~ 4.1% of total Incentive
This suggests either
- Most active miners perform similarly on your rolling windows, Although wandb reports do not align with this.
- Our performance → weight mapping is gentle, so even clear performance gaps don’t translate into sharp payout differences, which is the likely case.
We should review and ensure better performances result in significantly higher rewards, consequently increasing the motivation to improve miner models.
Looking at the Subnet Metagraph, it appears that the incentive is nearly identical to consensus (corr ~ 0.999) and strongly tied to Trust (corr ~ 0.85). Which demonstrates that the intended distribution is achieved. Although we also notice:
This suggests either
We should review and ensure better performances result in significantly higher rewards, consequently increasing the motivation to improve miner models.