When choosing a server in Shadowrocket, first confirm that it connects, then compare stability across repeated tests, and finally check real-world browsing and video playback. Assuming you already have your own provider and subscription, this guide covers the testing sequence in Home, how to control Global Routing, and what you can—and can’t—infer from locations and protocols. This site does not provide, sell, or recommend servers or subscriptions.
Understanding latency: response time isn’t download speed
Latency is usually shown in milliseconds (ms) and measures how long a particular test request takes to receive a response. It describes that test, not a permanent property of a server. Results can change when you switch Wi-Fi or mobile networks, use a different test method, or change the test target. Make sure test conditions match before comparing numbers in the list.
Connection setup, DNS resolution, the encryption handshake, server response, and ongoing data transfer are separate stages. A fast connection setup doesn’t mean the server has enough bandwidth to stream video continuously. Likewise, no response to one type of probe doesn’t prove that all app requests will fail. Interpret the numbers in light of the test method and what happens when you actually access the service.
Latency tests in Home: keep conditions consistent and compare repeatedly
These steps use the Home, Config, and Settings interface on iPhone / iPad as an example. Before you start, stop any large file transfers, stay on the same network, and make sure the subscription has finished updating. Don’t update the list while comparing servers, or the items being tested may change midway through.
-
Check the list
In Home, confirm that the servers from your existing subscription are listed. To add them again, go to Home, tap “+” in the top-right corner → Type → Subscribe, enter your own subscription URL, and save. The URL contains access credentials, so don’t include it in screenshots or public records.
-
Keep conditions consistent
Choose Wi-Fi or mobile data and don’t switch networks during this round. Go to Settings → On Demand and check the automatic connection settings to prevent them from changing the expected connection state during testing. Note the original settings and restore them afterward if needed.
-
Run the test
In Home, run Connectivity Test on the current list and wait for it to finish. If the app offers different test methods, use the same one for every comparison; don’t rank numbers from different methods together.
-
Compare repeatedly
Run two more tests on the few candidates you’re considering, for three rounds in total. Three rounds are a suggested comparison method, not an app requirement. Look for repeated timeouts and large swings in the results.
-
Verify the connection
Select a candidate in Home, connect, then test it with the same website or video. If a page is still loading after you switch servers, start a new request before judging the result so an existing connection doesn’t skew the comparison.
Repeatedly tapping the test button won’t fix the problem. If no results appear, first confirm that your regular network connection works, then check whether the subscription is valid and the server details are complete. If just one entry fails, inspect that entry. If every entry in the same subscription fails, investigate both the subscription status and possible restrictions on your current network.
Global Routing: distinguish server issues from rule issues
Selecting a server doesn’t guarantee that a particular website will connect through it. Home → Global Routing determines the routing mode, while rules and policies in Config may direct traffic elsewhere. Before comparing servers, confirm the actual route used by the target request. Otherwise, you may be measuring a direct connection or a route selected by another policy.
Config
RecommendedRoutes connections according to the currently active configuration. Check which rule matches the target domain and which policy that rule specifies; the server selected in Home alone doesn’t determine the final route.
Best for: everyday use and retesting rules after verifying an individual server
Proxy
Routes connections handled by Shadowrocket through a proxy. Use it temporarily to reduce interference from configuration rules while troubleshooting. It doesn’t guarantee that all device traffic will be handled, nor can it fix server-side issues.
Best for: temporarily checking whether a candidate server can access a target
Direct
Uses a direct route without forwarding traffic through a proxy server. This can help confirm that your regular network works, but page-loading results in this mode say nothing about a candidate proxy server’s performance.
Best for: checking your local network and ruling out basic connectivity issues
Scene in Global Routing applies routing settings by scenario; it isn’t another latency test. If you’re using a Scene, check which settings are actually active before comparing. Restore the original routing mode when you finish troubleshooting so temporary test settings don’t remain in place.
When checking rules in Config, start by understanding these four examples. They explain syntax; they aren’t instructions to replace your existing configuration. The example domains don’t point to services you’re meant to access. In particular, FINAL handles connections that haven’t matched earlier rules, so changing it can affect the route used by many requests.
DOMAIN-SUFFIX,example.com,PROXY
GEOIP,CN,DIRECT
IP-CIDR,192.168.0.0/16,DIRECT
FINAL,PROXY
DOMAIN-SUFFIX: Matches a domain suffix. This helps explain why related requests for the same website may be assigned to a particular policy.GEOIP: Matches against an IP geolocation database. It isn’t a location label in a server name and may not match how a website identifies a user’s location.IP-CIDR: Matches an IP range. In this example, private IP ranges use DIRECT.FINAL: Sets the final policy for requests that didn’t match any earlier rules.
Choosing a location: test the network route and target service separately
Locations in server names are usually provided by the subscription provider. They don’t prove that the entry point, actual exit point, and physical server are all in the same place. The route to the entry point, inter-network links, and the route from the exit point to the target service all affect performance, so a geographically closer server isn’t necessarily faster.
How to compare locations: test interactive use and sustained transfers separately
Browsing and interactivity
- Use the same target page and compare the first response with the time it takes to load fully.
- Open several different pages on the same site and note whether they frequently stall.
- If the service has account-region requirements, check its own guidance.
Video and transfers
- Use the same content at the same resolution, and don’t download other files at the same time.
- Check whether playback continues smoothly after it starts, not just whether the first frame appears quickly.
- Retest during your usual hours to distinguish occasional congestion from persistent issues.
A location name is only a clue for narrowing down options. Keep candidates that meet the target service’s requirements and remain stable on your current network.
If a website says your location doesn’t meet its requirements, low latency won’t solve the problem. A service may determine eligibility using the exit IP, account region, and other factors. Selecting a location name in the client doesn’t change those server-side rules. Check the target service’s requirements instead of repeatedly switching servers to guess at the cause.
Protocol types: check settings and network conditions, not a fixed ranking
Shadowsocks, VMess, VLESS, Trojan, Hysteria2, and WireGuard use different connection methods and require different fields, but a protocol name alone doesn’t indicate speed. The same protocol can perform very differently on different routes and servers. Client settings only work when the server supports the corresponding protocol and parameters.
| Protocol | Check first | Limitations when choosing |
|---|---|---|
| Shadowsocks | Server address, port, encryption method, and password | The fields must match the server. Changing the encryption option won’t fix network congestion. |
| VMess / VLESS | UUID, port, and the transport and TLS parameters actually in use | Check each field against the existing configuration. Don’t assume entries with similar names use the same transport. |
| Trojan | Password, TLS settings, and the server-required name | If the handshake fails, check the parameters and device clock first. Disabling certificate verification shouldn’t be a routine fix. |
| Hysteria2 | Authentication, TLS, and UDP reachability on the current network | Built on QUIC; UDP restrictions on your network can affect availability, so don’t rank it by latency alone. |
| WireGuard | Key, endpoint address, port, and tunnel address settings | Uses UDP; routing and server-side tunnel settings also affect the result. |
In Home, tap “+” in the top-right corner → Type to check the type selected when adding a server. For existing subscription entries, use their actual parameters. Don’t change a Trojan entry’s Type to VLESS to try to speed it up, or replace the subscription port just because an example uses 443. The protocol, port, and authentication settings must all match the server.
Low latency but slow browsing: diagnose by symptom
Real-world access also involves DNS lookups, TLS handshakes, target-server processing, and content downloads. A webpage may load scripts, images, and video from multiple domains, and some requests may follow different rules. This can make the main content appear while the page keeps spinning. Refreshing the latency number alone won’t identify the problem.
- Low number, slow first load
- Check DNS resolution, the connection handshake, and the target service’s response. Retry the same page, then check Global Routing and the matching policy. Don’t mistake a cached page opening instantly for a performance improvement.
- Starts fast, then buffers
- Check sustained throughput, packet loss, and peak-hour congestion. Play continuously for a while at the same resolution, then compare another candidate. Don’t use a test’s response time to estimate video bandwidth.
- Test fails, website works
- Check whether the probe and the actual website request use the same route. A failed probe doesn’t mean every request will fail. Note the test result and consider it alongside your real-world needs.
- Results change after updating the subscription
- Return to Home, confirm that the expected entry is still selected, and check whether its address, port, or protocol changed. The same display name doesn’t guarantee that the underlying configuration stayed the same.
If you need to keep a record, note the network type, test time, Global Routing mode, results from all three rounds, and what happened during real-world access. Don’t record or share subscription tokens, passwords, UUIDs, or keys. The purpose is to control variables, not to create a permanent ranking of locations or protocols.
- First, rule out unusable entries: Set aside entries that repeatedly fail to connect and also fail during real-world access. Check their configuration before considering them for regular use.
- Next, compare consistency: Under the same conditions, favor entries that stay available and show less variation, rather than judging by their best result.
- Finally, test for your use case: Check page loading and interaction for websites, sustained playback for video, and actual access requirements for specific services.
- Keep a backup: Keep a tested alternative from your existing list. After switching, recheck the rules and the route used by the target request.
Download Shadowrocket and see the initial setup steps
Get Shadowrocket for iPhone / iPad from the App Store for a one-time $2.99 purchase. The developer is Shadow Launch Technology Limited; check the App Store listing for system requirements. The Mac (Apple silicon and Intel), Windows, Android, and Linux versions are free; get their installers from this site’s download page. Shadowrocket is closed-source commercial software. Purchasing the app doesn’t include a network service; you’ll need your own provider and subscription.