Why did my race prediction get slower after I ran a Personal Best?

It can be surprising (and even a bit discouraging) to smash a Personal Best only to see your predicted finish times for future events get slower. However, this is actually a sign of the AI's sophistication — it is reacting to your real-world physiological state rather than just your "paper" potential.

**1. The "Race Tomorrow" Principle**

Most running apps provide a "best-case scenario" prediction based on your fastest recent segments. TrainAsONE is different: it predicts what you could achieve if you had to race tomorrow. If you just ran an all-out 10k yesterday, your body is currently in a state of high fatigue. The AI knows that if you tried to race that same distance again 24 hours later, you would be slower. As you recover over the coming days, your prediction will "climb" back up as the fatigue clears, eventually reflecting your new, higher fitness level.

**2. Predictions don't "drive" your training**

A common misconception is that a slower prediction leads to "slower" or "worse" training. This is not the case. The prediction and the training plan are separate outputs. The AI doesn't give you a recovery run because the prediction dropped; instead, both the recovery run and the lower prediction are simultaneous reactions to the high workload detected from your recent effort.

**3. "Clearing the Decks" for Recovery**

When you see your training sessions get shorter or slower after a big effort, the system is "clearing the decks." It is prioritising the reduction of your Acute Workload to protect you from injury. Once the AI sees your heart rate and pace stabilise during these easier sessions, it regains "confidence" in your recovery. At that point, it will begin ramp your predictions back up to reflect your new "floor" of fitness.

**4. Accuracy and the Margin of Error**

Standard sports science formulas (like the Riegel or Vickers models) often have an error margin of 8% to 10% (or worse as race distance increases). TrainAsONE uses your second-by-second physiological data to be far more precise. If your actual race time is within a five percentage points of the AI's prediction, the model is performing with high accuracy. While there can be outliers — perhaps you had a "superhero" day or the data was noisy — the system will recalibrate, and it does not mean that next time it will be the same story.