The prediction. The actual sale.

See how ApePredictor’s last pre-sale valuation compared with the price paid.

Sales Predicted—verified sales in this analysis
Mean Accuracy—average closeness to sale price
Median Accuracy—typical closeness, less affected by outliers
Model Bias ⓘ—tendency to over- or underpredict

Loading the verified performance record…

Prediction-before-sale records only. Every chart uses the same selected evidence.

Weekly Accuracy

Average accuracy and number of sales per week

Weekly results are also available in the sales table below.

Accuracy Over Time

Cumulative median accuracy as more sales are verified

Cumulative accuracy is calculated from the sales table below.

Error Distribution

Negative = underestimate · positive = overestimate

The signed error for every sale is available in the table below.

How we measure Prices in ETH

Ape
Methodology, coverage, and limitations

Mean Accuracy = 100% − Mean Absolute Percentage Error. Median Accuracy = 100% − Median Absolute Percentage Error. Accuracy can be negative when the estimate is more than twice the sale price; errors are never clipped.

Model bias is the average signed error: (estimate − sale price) ÷ sale price. Negative values indicate underprediction. Weekly accuracy uses the mean; accuracy over time uses the cumulative median. Only recorded weeks and sales occupy space on the horizontal axes. Spacing represents successive observations, not elapsed time; labels and tooltips show the original UTC dates.

Only verified sales with a valid prediction timestamp earlier than the sale are included. Wash trades and unresolved transactions are excluded. This selection can affect the result; any known coverage gap is disclosed above. Predictions are estimates, not guarantees of an executable price or future return.

Search and date filters apply to the metrics, charts, and table together. Prices are shown in ETH to keep historical comparisons consistent.