This guide is for users who have imported their own connection details into Shadowrocket and want to choose a server from their existing options for everyday use. Follow this sequence: standardize test conditions → check the actual egress → verify the apps you use → keep a backup. Distinguish latency from throughput and stability, and make sure Global Routing isn't sending test traffic around the selected server. We don't provide, sell, or recommend servers or subscriptions.
1. What latency does—and doesn't—tell you
A latency test measures the time taken by a specific probe; it isn't an overall server rating. Different test methods can follow different paths. A TCP probe typically measures how long it takes to establish a connection to the server port. An HTTP request sent through a proxy may also include the proxy handshake, the connection to the destination, and time waiting for a response. Before comparing results, check the test method and target. Don't rank numbers from different test types side by side.
Low latency doesn't mean high bandwidth. A connection may come up quickly but slow down during a large download because of congestion, throttling, or packet loss. Another connection may respond slightly later but transfer data reliably. A single Timeout doesn't prove a server is down: the test target may be unreachable, DNS resolution may be failing, or your local network may have changed.
| What you measure | What it tells you | What to test next |
|---|---|---|
| Port connection time | Whether your device can establish a connection to the specified address and port | Whether authentication, proxy forwarding, and the destination website work |
| HTTP probe time | How the specified test request responds over the current route | Real-world domains, persistent connections, and sustained transfers |
| Page load behavior | Whether page resources load completely without interruptions | Repeat results at different times and on different local networks |
| Sustained transfer rate | Available throughput to the target during the test period | Short-connection response and stability during peak hours |
2. Standardize conditions before comparing servers
Change only one variable at a time. If you test one server over Wi-Fi and another over cellular, the results also reflect a change in access network, so you can't attribute the difference to the server alone. Background downloads, system updates, and heavy traffic from other devices on the same network can also affect results.
-
Check your connection details
Open Home and confirm that your own connection entries are imported. For manually added entries, check Type, server address, and port under Add Server. Don't change parameters required by the server just to get a lower number.
-
Keep the network constant
Stay on the same Wi-Fi or cellular network and pause large transfers. Go to Settings → On Demand and check your automatic connection conditions. Note the current settings to avoid an automatic connection change during testing.
-
Use the same test
Run Connectivity Test from Home using the same method and target each time. Test each candidate three times and record the success rate and variation. Don't compare results from different test methods as if they were equivalent.
-
Verify the egress
Select a candidate on Home and connect, then check the status of Global Routing. Confirm that actual requests use the selected candidate. If Config has a separate policy selection, also check which policy the target rule uses.
-
Test the apps you use
Open your usual websites and apps one by one. Check initial response, whether resources load fully, and whether connections stay up. After switching servers, reconnect and, if necessary, close and reopen the test app to avoid reusing an old session.
For each round, record five things: local network, test method, three latency readings, whether the site opens, and whether use is interrupted. If two candidates perform similarly, keep the one with less variation and that reliably handles your tasks rather than switching back and forth to save a few milliseconds.
3. Region labels, connection addresses, and egress are different things
A region in a server name usually comes from the naming convention used in the connection details; it isn't a guarantee from the client about the actual egress location. The connection address may be an entry point, with additional forwarding along the route. The address seen by the destination website may differ from the address your device connects to first. So neither a nearby-sounding name nor geographic proximity on a map proves that access will be faster.
If a service requires a specific region, first make sure you're allowed to use it there, then check the actual egress and whether the service works. IP geolocation databases can disagree, and a website's region detection may also depend on account settings, location permissions, and an existing session. Don't draw conclusions from a single lookup.
Choose regions based on what you need to do
Everyday browsing and apps
- Compare success rates and latency variation across your existing candidates first.
- Test with the domains you actually use, not just a generic test address.
- Test each candidate separately, even when the region label is the same. Don't assume the routes perform alike.
Services with specific region requirements
- Check the service's supported regions and account requirements first.
- After connecting, check the egress and start a fresh session in the service.
- Once the region requirement is met, compare response times and sustained transfers.
Use region to filter for eligibility; use the actual route and its stability to choose what you use day to day. Neither can replace the other.
4. Protocol type determines compatibility, not a fixed ranking
Shadowsocks, VMess, VLESS, Trojan, Hysteria2, and WireGuard use different connection methods and require different parameters, but the protocol name alone doesn't predict real-world speed. Server setup, network path, encryption overhead, and current load all affect performance. Start with your existing connection details and use a configuration with complete parameters that authenticates successfully.
Don't change Type to another protocol to try to “speed things up.” For example, if the server provides Trojan, changing the client to VLESS won't convert the protocol automatically. The password, credentials, transport, and TLS settings must all match the server. The server also determines the port; don't change every entry to 443 just because it's common in examples.
Trojan parameter checklist
- Type
- Trojan; must match the server protocol
- Address and port
- Use the values from your original connection details. 443 is a common example, not a default requirement.
- Password
- Copy the full value, taking care not to add or lose spaces or line breaks.
- SNI / TLS
- Configure using the domain and certificate requirements provided by the server.
If the certificate or identity details don't match, correct the connection details first. Don't disable certificate validation as a substitute for troubleshooting.
Hysteria2 and network conditions
- Transport
- Uses QUIC and requires a working UDP path
- Port
- Keep the server-specified value; don't copy the port from another entry.
- Authentication
- Use the authentication parameters in your original connection details.
- When it fails
- Retest the same configuration on both Wi-Fi and cellular.
If the network restricts UDP, the connection may fail. That alone doesn't prove the configuration is wrong just because another protocol works.
- Shadowsocks: Check the encryption method and password; a resolvable address alone isn't enough.
- VMess / VLESS: Check the credentials, transport, and relevant TLS parameters. Complete fields don't necessarily mean they match the server.
- WireGuard: Check the keys, peer, and allowed IP ranges. Unlike some other configurations, it requires more than an address and password.
5. Check whether Global Routing is affecting the results
Selecting a server on Home doesn't mean every app request will automatically use it. Global Routing determines the routing mode, while rules and policies in Config may send different destinations through different egresses. If a website matches DIRECT, seeing no change after switching servers doesn't prove that the two servers perform equally.
Config
Rule-based routing (recommended)Use rules to route traffic after checking the configuration. Check which rule matches the destination and which policy that rule points to. Don't assume that every request to a destination outside your local region will use the server selected on Home.
Best for: users who understand their existing rules and are checking performance during everyday use.
Proxy
Routes requests handled by the client in proxy mode. This can help temporarily rule out the effect of standard routing rules, but it doesn't guarantee that the operating system routes all traffic through the client and doesn't replace egress verification.
Best for: short comparison tests. Restore your usual routing mode when you're done.
Direct
Routes requests directly. This isn't suitable for evaluating the forwarding performance of the selected proxy server. If the connection switch is on but your egress is still your local network, check this setting first.
Best for: comparing against a direct connection to your local network to see whether the issue occurs without the proxy.
If you use Scene, check the routing and policies associated with the active scene before comparing. A Scene isn't a fixed egress. On Demand controls whether a connection starts automatically when its conditions are met; it won't automatically replace a high-latency server with a lower-latency one.
6. Common issues and what to check next
Latency is low, but why does the page keep loading?
First check which test method produced the low latency, then go to Home → Global Routing and check the current routing mode. Reconnect and open the actual destination. If only one domain fails, check its rule, DNS resolution, and the destination site's response. Don't start by deleting all your servers.
Works on Wi-Fi but times out on cellular?
Keep the same configuration and retest after switching networks. Go to Settings → On Demand and check whether cellular conditions affect the connection. If the protocol relies on UDP, also check whether UDP is reachable on that network. Record each result separately rather than assuming the difference is due to region.
Switched servers, but the download hasn't changed?
An existing persistent connection may continue using its original session. Pause the download, reconnect, and test a new request. Also check whether a separate policy in Config pins the egress. Selecting another entry on Home alone doesn't prove that an active download has switched routes.
A server that used to work disappeared after a subscription update?
First check whether your own subscription updated successfully and whether the entry was renamed or removed. Return to Home and check the entries under the relevant subscription, then select a candidate again. Don't rely on the old name or test results. Subscription links contain access credentials; don't include them in public screenshots or troubleshooting reports.
7. Keep one primary candidate and one backup
After comparing candidates, keep one that meets your actual requirements as the primary, then choose a tested, working backup. A backup lets you switch quickly if something fails; a longer list isn't inherently better. If possible, choose entries from your existing connection details with different access paths, but different names don't prove that their underlying routes are independent.
- Primary record: Note the entry name, local network, test method, real-world results, and time of day.
- Test the backup: Connect and open the service you use. A successful latency test alone doesn't mean the backup is ready.
- Retest at peak times: Test during your usual hours. A good result during the day doesn't guarantee the same performance in the evening.
- Retest after changes: Check the egress and real-world performance again after a subscription update, rule change, or network change.
Use this order for your final choice: parameters match and the connection works → region and service requirements are met → requests use the intended egress → sustained use is stable → compare latency. This helps avoid repeatedly switching just to chase the lowest reading while overlooking whether the connection works in practice.