Systolic blood pressure (BP) variability at daytime and nighttime, also known as BP dip, has shown its clinical value in diagnosis and treatment of cardiovascular diseases (CVDs). A model agnostic meta learning (MAML)-based 24-hour personalized approach is proposed in this paper for BP estimation using wrist photoplethysmography (PPG) signals from smart watch in free-living context. Detection of cuffinflation is adopted to avoid reactive hyperemia effect after deflation and to accomplish synchronization between reference BP values measured by the ambulatory BP monitoring (ABPM) device and recorded 24-hour PPG signals in the processing flow during the experiment. The assessment of signal quality and activity occurrence is incorporated to indicate the applicability of BP estimation. The fast adaptability of personal model is enhanced by pre-training with the MAML technique and thus only few data for fine-tuning in the testing task are required. From the experimental results, our approach achieves diurnal and nocturnal SBP estimation error of 0.89 ± 7.78 mmHg from 14 recruited subjects with ages from 20 to 90 years old and shows the feasibility of tracking BP variability with smart watch in daily life.