Arrival Prediction Accuracy
Each week, we measure how accurate our arrival predictions were for the bus and subway. We continuously collect data to make our predictions more accurate.
We consider a prediction to be accurate when a train or bus arrives within minutes of a prediction. The exact number of minutes depends on how far out into the future a prediction is.
How We Measure the Accuracy of Arrival Predictions
Generally, it’s harder to predict events further out into the future because there’s a higher chance of something unexpected happening. The closer a bus or subway car gets to your stop, the more accurate its predicted arrival tends to be. That’s why we use four different ranges of predicted arrival times to measure the accuracy of our predictions.
We consider predictions to be accurate if a bus or train arrives within the following time frames:
| Predicted arrival time | Actual arrival time for a prediction to be counted as accurate | ||
|---|---|---|---|
| Predicted arrival time | 0 - 2 minutes 59 seconds | Actual arrival time for a prediction to be counted as accurate | 1 minute early to 1 minute late |
| Predicted arrival time | 3 - 5 minutes 59 seconds | Actual arrival time for a prediction to be counted as accurate | 1.5 minutes early to 2 minutes late |
| Predicted arrival time | 6 - 11 minutes 59 seconds | Actual arrival time for a prediction to be counted as accurate | 2.5 minutes early to 3.5 minutes late |
| Predicted arrival time | 12 – 30 minutes | Actual arrival time for a prediction to be counted as accurate | 4 minutes early to 6 minutes late |
A single bus or train has multiple arrival predictions as it travels to a station or stop. We measure the accuracy of each of those predictions by comparing them to the time the bus or train actually arrived.
