Accuracy
How accurate is the Azuki model?
Every estimate is benchmarked against real sales the model never trained on (a seeded 80/20 hold-out).
Median accuracy
89.9%
Mean accuracy
85.2%
Within ±5%
28%
Within ±10%
50%
Within ±20%
74%
Bias
+1.0%
Weekly median accuracy
Hold-out accuracy by week
Error distribution
Signed % error on recent sales (0 = perfect)
Recent sales vs model
The 60 most recent sales, each valued as of its sale date
| Item | Type | Date | Sale | Model | Error | Accuracy |
|---|---|---|---|---|---|---|
| #5793 | Human | Jun 14, 2026 | 0.842Ξ | 0.844Ξ | -0.2% | 99.8% |
| #6835 | Human | Jun 14, 2026 | 0.805Ξ | 0.801Ξ | +0.5% | 99.5% |
| #3723 | Red | Jun 13, 2026 | 1.500Ξ | 0.896Ξ | +67.5% | 32.5% |
| #8306 | Human | Jun 13, 2026 | 0.795Ξ | 0.878Ξ | -9.5% | 90.5% |
| #8306 | Human | Jun 13, 2026 | 0.786Ξ | 0.878Ξ | -10.5% | 89.5% |
| #2019 | Human | Jun 13, 2026 | 1.100Ξ | 0.820Ξ | +34.2% | 65.9% |
| #665 | Human | Jun 13, 2026 | 0.786Ξ | 0.895Ξ | -12.1% | 87.9% |
| #411 | Human | Jun 13, 2026 | 1.499Ξ | 0.991Ξ | +51.3% | 48.8% |
| #6835 | Human | Jun 13, 2026 | 0.809Ξ | 0.802Ξ | +1.0% | 99.1% |
| #9286 | Human | Jun 13, 2026 | 0.810Ξ | 0.779Ξ | +3.9% | 96.1% |
| #1854 | Human | Jun 12, 2026 | 0.860Ξ | 0.857Ξ | +0.4% | 99.6% |
| #9595 | Human | Jun 12, 2026 | 0.810Ξ | 0.744Ξ | +8.8% | 91.2% |
| #8292 | Red | Jun 12, 2026 | 0.879Ξ | 0.890Ξ | -1.3% | 98.8% |
| #2660 | Human | Jun 12, 2026 | 0.888Ξ | 0.765Ξ | +16.2% | 83.8% |
| #2660 | Human | Jun 12, 2026 | 0.809Ξ | 0.765Ξ | +5.8% | 94.2% |
| #6832 | Human | Jun 12, 2026 | 0.800Ξ | 0.777Ξ | +3.0% | 97.0% |
| #6832 | Human | Jun 12, 2026 | 0.781Ξ | 0.777Ξ | +0.5% | 99.5% |
| #6504 | Human | Jun 12, 2026 | 0.816Ξ | 0.764Ξ | +6.8% | 93.2% |
| #7669 | Human | Jun 11, 2026 | 0.800Ξ | 0.838Ξ | -4.6% | 95.4% |
| #5793 | Human | Jun 11, 2026 | 0.880Ξ | 0.846Ξ | +4.0% | 96.0% |
| #3146 | Human | Jun 11, 2026 | 0.860Ξ | 0.865Ξ | -0.6% | 99.4% |
| #2356 | Human | Jun 11, 2026 | 0.815Ξ | 0.788Ξ | +3.4% | 96.6% |
| #5188 | Human | Jun 11, 2026 | 0.850Ξ | 0.759Ξ | +12.1% | 88.0% |
| #9703 | Human | Jun 11, 2026 | 0.780Ξ | 0.796Ξ | -2.1% | 97.9% |
| #6832 | Human | Jun 11, 2026 | 0.810Ξ | 0.778Ξ | +4.0% | 96.0% |
| #5379 | Human | Jun 11, 2026 | 0.840Ξ | 0.766Ξ | +9.7% | 90.3% |
| #5667 | Human | Jun 11, 2026 | 0.880Ξ | 0.742Ξ | +18.6% | 81.4% |
| #1967 | Human | Jun 10, 2026 | 0.810Ξ | 0.787Ξ | +3.0% | 97.0% |
| #1119 | Human | Jun 10, 2026 | 1.000Ξ | 0.920Ξ | +8.7% | 91.4% |
| #6354 | Human | Jun 10, 2026 | 0.840Ξ | 0.820Ξ | +2.4% | 97.6% |
| #4100 | Human | Jun 10, 2026 | 0.810Ξ | 0.817Ξ | -0.9% | 99.1% |
| #5172 | Spirit | Jun 10, 2026 | 25.00Ξ | 3.227Ξ | +674.7% | -574.7% |
| #1967 | Human | Jun 10, 2026 | 0.756Ξ | 0.787Ξ | -4.0% | 96.0% |
| #6835 | Human | Jun 10, 2026 | 0.800Ξ | 0.806Ξ | -0.7% | 99.3% |
| #4149 | Human | Jun 9, 2026 | 0.815Ξ | 0.822Ξ | -0.9% | 99.1% |
| #8796 | Human | Jun 9, 2026 | 1.200Ξ | 0.994Ξ | +20.7% | 79.3% |
| #2746 | Human | Jun 9, 2026 | 0.820Ξ | 0.798Ξ | +2.7% | 97.3% |
| #4886 | Human | Jun 9, 2026 | 1.000Ξ | 0.822Ξ | +21.6% | 78.4% |
| #1128 | Human | Jun 9, 2026 | 0.920Ξ | 0.815Ξ | +13.0% | 87.1% |
| #4149 | Human | Jun 9, 2026 | 0.818Ξ | 0.823Ξ | -0.7% | 99.3% |
Method details on the methodology page.