Estimate local input-to-display latency and network-added timing from polling, FPS, refresh and processing.
Model uses average half-interval waiting for polling, render frame and refresh scanout. Actual latency varies with engine queues, V-Sync/VRR, GPU load, display processing and netcode.