The P wave is caused by the depolarization of the atria. It is an important feature of the electrocardiogram (ECG) because it can be used to diagnose atrial fibrillation and other arrhythmias. Many previous groups have developed algorithms to detect and analyze properties of the P wave. However, most of these algorithms are focused merely on detection time of the P wave, and occasionally the times of the P wave onset and offset. The accuracy of the P wave amplitude is rarely studied or reported. Software implementations are rarely available, the published descriptions are almost always incomplete, and many of the algorithms are very complicated, thereby requiring a large investment of time to understand, implement, and evaluate. We propose a simple algorithm for estimating the P wave amplitude that overcomes these limitations. The algorithm is designed to estimate the P wave amplitude, rather than merely the times of the onset, peak, and offset. Our algorithm is based on a simple parametric model that can be globally optimized, is computationally efficient, only requires a single lead of the ECG, is easy to understand, and has great performance. We used the QT Database for this study, but found that the P wave annotation times were inaccurate. This database also lacks annotations of the P wave amplitude. To overcome these problems, we annotated over 950 P waves in the QT Database. These annotations include both the time and amplitude of the P wave onset, peak, and offset. The Pearson correlation of the P wave amplitudes estimated by our algorithm with expert annotations was 0.94 (p<0.0001). An implementation of the algorithm and our expert annotations are available on GitHub.