Before you start
Define the target address and protocol, representative networks and times. Record date, timezone, provider, connection type and test settings. Separate the website, administration port and game service. Use authorised, repeatable tests with low traffic volume; diagnostics should not become additional load.
Separate RTT, variation and processing time
RTT measures a round trip. It does not directly give the delay in each direction, because routes and processing can differ. One-way measurements need an appropriate method, including clock handling. Half a ping should not be presented as an exact measurement of the outbound path.
Compare several samples instead of the best result. Keep the minimum, a central value and high values alongside the sample count. Variation and interruptions can matter more than a small average difference. Separate network delay from the game’s time to simulate and respond.
Test the right service and protocol
An HTTPS TCP test describes that entry point, rather than necessarily the UDP gameplay path. Websites, proxies and game servers can use different destinations or processing. ICMP ping supplies another observation with its own rules and limits.
Pair measurements with a normal service connection. Keep in-game network statistics and context when available. Intermediate equipment may limit diagnostic replies while forwarding traffic; a silent traceroute hop alone does not prove loss reaching the player.
Build a baseline before an incident
Measure at several times from networks people actually use. Keep destination, protocol and method consistent. Datacentre measurements do not automatically represent home, mobile or Wi-Fi connections. Samples from a country do not describe every provider in that country.
Retain raw results and failures instead of selecting favourable values. Include attempt count and observation period. A short baseline shows behaviour at that time; it is not a permanent latency or availability guarantee.
Compare during mitigation
Correlate failures and changes with timestamped network events. Mitigation can alter routing or traffic handling, but distinguish that from route failure, false positives or application load. Ask which events and addresses are affected.
For tunnels, also examine MTU and symptoms involving packet size. For proxies, check timeouts and downstream connections. Do not change several components at once. Confirm legitimate traffic before removing a rule that contributes to protection.
Check the game workload at the same time
A server may answer ping quickly while processing players slowly. Compare CPU, memory, disk, logs and activity at the time of the issue. World generation, blocking scripts or slow databases can create delays without a routing fault.
Where a compatible tool exists, profile during the slowdown. Paper’s spark reports can identify expensive work. Bound measurements and copy configuration before making changes. Buy more resources only when an observed constraint justifies it.
Present comparable evidence
Prepare a table with target, protocol, provider, time, sample count and results. Add game messages, server measurements and recent changes. Separate observations from hypotheses and request comparison with matching routes or events.
After a fix, repeat the same method and check a normal session. One improved sample does not establish that all paths are fixed. Describe measured markets and connections without promising latency for an entire country or confusing geographic proximity with routing quality.
For your situation
| Situation | Next check |
|---|---|
| Stable ping, slow gameplay | Examine simulation time, scripts and resources. Ping alone does not measure application work. |
| One provider is affected | Compare from other networks, keep timestamps and request a check of the affected path. |
| Degradation during mitigation | Correlate samples with events and protected addresses. Check protocol, rules and routing without globally disabling security. |
Official sources
These documents describe the relevant protocols and software. Your service terms and configuration still need to be checked.