What does this continuous latency monitor measure?
Every 500 milliseconds we fire a tiny HEAD request to Cloudflare's speed endpoint and record the round-trip time. The chart plots each sample, and the readout aggregates mean, median, p99, and jitter (the standard deviation of inter-sample differences). Unlike a one-shot speed test, this keeps running — leave it open during a stream or a game and you'll see exactly when and how badly the link wobbles. Packet loss is inferred from requests that exceed a 2-second timeout.
What is a good ping for gaming, video calls, and general browsing?
For competitive online games, under 30 ms to the game server is excellent, 30–60 ms is fine, and above 100 ms is felt as input lag. Video calls tolerate up to about 150 ms one-way before conversation rhythm breaks. General browsing is barely affected below 200 ms. Jitter matters more than absolute ping for real-time apps — a steady 80 ms beats a chaotic 40 ms that swings between 20 and 120. Aim for jitter under 10 ms and zero packet loss for a connection that feels solid.
Why is p99 a more useful number than average ping?
Average smooths over the spikes that actually ruin calls and games. p99 is the latency that 99 percent of your packets stay under — in other words, the worst one in a hundred. If your average is 25 ms but your p99 is 400 ms, every few seconds something is hanging for nearly half a second, and that is exactly when your voice cuts out or your character teleports. Watching p99 over a few minutes is the fastest way to tell a calm link from a flaky one.
How do I troubleshoot a noisy or lossy result?
First, switch to wired Ethernet and re-run — if the chart flattens, the problem is Wi-Fi (interference, distance, or a saturated channel). If wired is also noisy, try a different DNS server and check whether a household device is uploading in the background, since upload saturation triggers bufferbloat. Run /network/bufferbloat to confirm. If idle latency itself is high, traceroute from a terminal will show which hop is adding the delay — often the ISP's first-mile equipment.
How does this compare to your other network tools?
/network/speed gives you a peak-throughput snapshot, this latency tool gives you a stability timeline, and /network/bufferbloat stress-tests how ping degrades when the pipe is full. Latency is the right tool when something feels off but speed tests look fine — flaky video calls, sporadic game lag, intermittent page-load stalls. Leave it running in a background tab and check back after the next bad call.
Does running this test use much data?
Almost none. Each HEAD request is a few hundred bytes including headers, and at two per second that adds up to roughly 200 KB per minute. You can leave the monitor open for hours on a metered connection without making a meaningful dent. The samples stay in your browser tab and are not uploaded anywhere — only the HEAD pings themselves leave your machine.