When choosing a server in Shadowrocket, first confirm it connects, then check its region and protocol, and finally test it with the apps you use. Assuming you already have a provider and subscription, this guide explains how to keep test conditions consistent, interpret latency readings, and retest with the same websites and videos. This site does not provide, sell, or recommend servers or subscriptions.
1. Understand what the latency reading measures
Latency is usually measured in milliseconds (ms), but whether readings are comparable depends on the test method, destination, and current network. A TCP connection test mainly measures how quickly a connection is established. A test that sends a request to a URL through a proxy may also include the protocol handshake, connection to the destination, and response time. Check which test method is in use; don’t rank results from different methods together.
Results in Home aren’t a bandwidth report for the entire route. A short request returning only shows that the path responded at that moment. It doesn’t measure sustained throughput or prove that every website uses the same route. Different results are to be expected when the test destination and the website you’re using are on different networks.
2. Keep conditions consistent when testing a group
Start with a small set of candidates from your existing subscription—for example, two servers in one region and one in another. Comparing too many at once can mix up test duration, network changes, and server load. The steps below use common iPhone and iPad menus; on other platforms, look for the equivalent options in the current interface.
-
Check your setup
Open Home and confirm your server list has loaded. To import your own subscription, go to Home → “+” in the top-right corner → Type → Subscribe, enter your existing subscription details, and save. After updating, check the list first so you don’t compare against outdated entries.
-
Keep the network unchanged
Use the same Wi-Fi or cellular network throughout, and pause large file transfers on other devices. Go to Settings → On Demand and note the current settings. Temporarily disable automatic connection conditions while testing to avoid changes in connection state; restore them when you’re done.
-
Record the results
In Home → Connectivity Test, use the same test method for each candidate and record the results. Test each one three times, leaving a short pause between tests. Note how often Timeout occurs instead of recording only the lowest reading.
-
Check real-world traffic
In Home → Global Routing, temporarily select Proxy, choose a candidate server, and connect. Then test the same website or video. If your configuration uses a policy group, make sure its automatic selection hasn’t changed the actual route.
-
Restore your settings
When testing is complete, restore Global Routing and Settings → On Demand to their original settings. When using Config as usual, check that requests still match the intended policy rather than assuming the temporary Proxy setting remains appropriate.
Three tests aren’t a statistical guarantee; they simply help reveal obvious fluctuations. If the results are close, compare again during your usual hours. A low reading during the day doesn’t predict evening performance. Don’t change servers, DNS, and Wi-Fi all at once, or you won’t know which change affected the result.
3. Use region labels to filter, not to infer the entire route
Region names in server labels are usually set by the provider. They may describe the exit location or simply name a group; on their own, they don’t prove the server’s physical location, entry point, or full network route. Proximity can help narrow down candidates, but carrier peering, congestion at the entry point, and cross-network routing can all affect the actual result.
- For everyday browsing: Compare connection success rates and time to first content, then check whether pages load reliably across repeated visits.
- For services that require a specific region: Check which region the service actually detects for your exit IP, and follow its account and usage rules. Identical server labels don’t guarantee identical detection results.
- For video: Use the same video, resolution, and playback duration. Record startup delay and mid-stream buffering, not just whether the home page loads.
- For mobile networks: Test Wi-Fi and cellular separately and keep separate records. The best option on one Wi-Fi network may not be best on every network.
Even a single website may load different resources from different CDNs. DNS results, proxy exit location, and server-side routing can all affect where content comes from. After filtering by region, retest against the actual destination. Don’t keep switching away from a server that reliably meets your needs just to get a lower number in the list.
4. Choose protocols for compatibility, not speed rankings
Shadowsocks, VMess, VLESS, Trojan, Hysteria2, and WireGuard use different connection and transport mechanisms; they aren’t six speed tiers. Performance also depends on the transport layer, encryption settings, server resources, and network policies. Keep the Type, port, and authentication fields from your existing subscription as provided. Changing Type alone can’t convert a server to another protocol.
| Protocol | What to check | Limitations to keep in mind |
|---|---|---|
| Shadowsocks | Encryption method, password, port, and any plugin parameters specified in the supplied details. | UDP availability depends on the configuration at both ends and network conditions. |
| VMess / VLESS | UUID, transport method, and the corresponding TLS, server name, or path parameters. | The same protocol can use different transport layers and produce different test results. |
| Trojan | Details such as password, port, TLS, and server name. | If the handshake fails, check the configuration and certificate requirements. Don’t disable verification just to get a connection. |
| Hysteria2 | Authentication, TLS, and whether the current network supports UDP. | Based on QUIC and relies on UDP; it may not connect on networks that restrict UDP. |
| WireGuard | Key, peer address, tunnel address, and routing scope. | Uses UDP. After the tunnel is established, check routing and DNS as well. |
5. If latency is low but pages load slowly, check these two sets of details
“Low reading in the list, slow to load in practice” can mean two different things: the test results may be unstable, or real traffic may not be using the tested server. The two cards below are for recording details, not importable configurations. Record the facts first, then decide whether to switch servers or adjust a rule.
Record tests under consistent conditions
- Local network
- Record the Wi-Fi network name or cellular network type; don’t switch during testing.
- Test method
- Keep the Connectivity Test method and destination the same.
- Repeat results
- Record all three latency readings and the number of Timeouts, not just the lowest value.
- Real-world task
- Use the same website, or the same video and resolution, and observe load time and buffering.
Results from two candidate servers are more useful to compare when these conditions match.
Check the actual route
- Proxy
- Use this to compare proxy paths, but also confirm which server or policy group is actually selected.
- Direct
- Direct connection; traffic in this state doesn’t test the selected proxy server.
- Config
- Traffic is routed by the current configuration rules; the destination may match DIRECT or another policy.
- Scene
- Behavior depends on the scene settings; check which scene conditions are currently active.
Check rule order in Config and active conditions in Scene first.
In Config, start by checking domain rules. For example, DOMAIN-SUFFIX,example.com,DIRECT sends requests matching that suffix through DIRECT. Even if a low-latency server is selected in Home, those requests won’t use it. GEOIP matches the destination IP’s region, IP-CIDR matches an IP range, and FINAL is the fallback when no earlier rule matches. Check rules in order while troubleshooting; don’t start by deleting the entire configuration.
- Only the first visit is slow: Check DNS, connection setup, and the TLS handshake separately. Browser caching may make the second visit noticeably faster.
- Starts quickly, but slows during transfer: Check available throughput, congestion, and server load. A short-request latency test can’t replace a sustained transfer test.
- Only one website has issues: Test another destination to distinguish a problem with the site from an exit restriction or a general connection issue.
- Only certain apps have issues: Check the app’s traffic and UDP requirements. A website working normally doesn’t mean every app will work normally.
6. Common questions and how to retest
Should I always choose the server with the lowest latency?
Not necessarily. First rule out candidates that repeatedly time out, then compare how well they handle the same task. If the lowest-latency server often drops the connection while another responds in all three tests and plays video without interruption, keep the latter as your preferred option and test again during your usual hours.
Why do two servers in the same region perform so differently?
The same region label doesn’t mean the same entry point, exit location, or server load. Keep Global Routing, the test destination, and your local network unchanged, then switch servers one at a time. If the difference persists, note the server names and time of day and contact your existing provider. Don’t assume there’s a configuration problem based on the region label alone.
What if my selected server disappears after a subscription update?
Go back to Home and check the updated groups and entries to see whether the server was removed or renamed. Don’t choose a replacement based on its position in the old list. Check the region, Type, and port again, then run Connectivity Test. If a policy group refers to the old name, check the current selection in your configuration too.
Can On Demand affect comparisons?
On Demand mainly controls when to connect; it doesn’t directly make a route faster. However, switching between Wi-Fi and cellular during a test can trigger a reconnection. Go to Settings → On Demand, check the conditions, and retest with the connection state fixed. Restore the original settings when you’re done.
Keep one everyday candidate and one backup, and record how each performs on Wi-Fi and cellular. Retesting when your needs, network conditions, or subscription change makes it easier to identify issues than constantly chasing the lowest reading in the list. The client’s purchase price is separate from the cost of any service you already use; buying the client does not include a service plan.