In Shadowrocket, latency readings help screen connections; they don’t measure bandwidth or confirm that a destination website is reachable. This guide is for users who already have a provider and subscription. First, distinguish the test methods, then keep your network and Global Routing settings consistent while testing existing servers repeatedly. Finally, choose primary and backup options based on exit region, protocol requirements, and your actual tasks. This site does not provide, sell, or recommend servers or subscriptions.
1. What Latency Tests Do—and Don’t—Measure
Latency is usually shown in milliseconds (ms) and indicates how long a particular probe takes to complete. Before interpreting a reading, check the test method and destination. A TCP connection test to a server port and a request sent through a proxy to a test URL follow different paths. Don’t rank them together just because both results are shown in ms.
A TCP connection test measures how long it takes your device to establish a connection to a specified server port. A reachable port doesn’t confirm that proxy authentication succeeds, that a destination website is accessible, or that there’s enough throughput for large downloads. A request sent through a proxy involves more steps, but its result still applies only to that test URL under the network conditions at the time—not to every website.
| What You’re Measuring | What It Can Tell You | What It Can’t Prove on Its Own |
|---|---|---|
| TCP connection time | Whether a TCP connection to a specified server port can be established, and how long it takes | Successful proxy authentication, working UDP, or an accessible destination website |
| Request time through a proxy | Whether one request to a specified test URL completed | That other websites will behave the same way or that long transfers will remain stable |
| Actual page load | Whether the page’s DNS lookup, connection, and resource loading completed | Sustained download speed or how other apps’ connections perform |
| Sustained transfer performance | Throughput and interruptions over a period of time | That bandwidth will be the same at all times |
A test timeout doesn’t mean a server is permanently down. The test URL or server port might be unreachable, the network might restrict a particular type of traffic, or protocol settings might be incomplete. For connections that rely mainly on UDP, a TCP probe to the same port isn’t enough to determine whether the connection works. Try connecting with the relevant protocol and test access to your actual destination.
2. Keep Test Conditions Consistent Before Comparing Results
Results from different times, networks, or test methods aren’t directly comparable. First, check your existing subscription and candidate servers on Home, then review Global Routing. Use Config to inspect your configuration and Settings to check related options. Change only one variable at a time so you can tell whether a difference comes from the server, routing rules, or your current network.
-
Check your subscription
On Home, confirm that your Subscribe entry has updated and that the candidate servers are still listed. To add one, go to Home → + → Type → Subscribe, enter your existing subscription URL, and save. Subscription links contain access credentials, so don’t include them in public screenshots or test notes.
-
Keep the environment consistent
Use the same Wi-Fi or cellular network throughout a test round, and pause large downloads. Go to Settings → On Demand and check the connection conditions to avoid automatic changes to the connection state during testing. Record your existing settings first.
-
Check routing
On Home → Global Routing, check the current mode. For everyday testing, leave Config selected. To isolate the effect of routing rules, you can temporarily compare with Proxy. After each change, confirm which server is selected, then restore your original settings when you’re done.
-
Repeat the tests
Use the same latency test method and URL for three rounds, about ten seconds apart. Record each server label, transport protocol, number of successful attempts, and time per round. Don’t record a timeout as 0 ms.
-
Test with a real task
Connect to a candidate server, open the same destination page, and perform the same task. Note the time to first content, ongoing loading, and any reconnects. Repeat the same process with the next server; don’t compare results from different websites.
3. Why Is a Website Still Slow When Latency Is Low?
A single page load can involve a DNS lookup, connection setup, a TLS handshake, and requests for multiple resources. A latency test may cover only one of those steps. The route from the server’s exit point to the destination website, available bandwidth, congestion, and packet loss can all affect subsequent transfers. So low latency and slow loading can happen at the same time. To troubleshoot, first determine whether the delay is at the start or during sustained transfer.
- Nothing appears for a long time after you tap: Check DNS resolution, connection setup, and the destination site’s response. Try another known-accessible destination with the same server to see if the problem is limited to one site.
- The page appears quickly, but images or video keep loading: Check throughput, packet loss, and server load. A successful short probe doesn’t rule out congestion during sustained transfers.
- The test returns a number, but the connection fails: Check the address, port, authentication details, and TLS parameters in your subscription. A reachable port doesn’t mean the proxy protocol handshake succeeded.
- Switching servers makes no difference: Check whether the request is actually using that server. A rule may select DIRECT or another member of a policy group.
- Only some page resources fail: A page may load content from several domains. If the main page and its resource domains match different rules, check the routing result for each connection.
Traffic limits, service periods, and concurrent connection limits set by your existing provider can also affect availability. Check these in your account or service details; don’t try to mask exhausted traffic or failed authentication by repeatedly switching protocols. Bandwidth tests consume data. A few checks under consistent conditions are enough—there’s no need to keep running tests in bulk.
4. Choose a Region Based on the Exit Point and Destination—not Just Distance
A region in a server name is usually a label supplied by the provider; it doesn’t prove the entry location, exit location, or full route. Your device may connect to an entry point first, then pass through a relay before reaching the exit point. A shorter geographic distance can mean lower propagation latency, but the actual route, network peering, and congestion can change the result.
Choose a region for the task at hand
Everyday browsing and interactive tasks
- Start with candidates in your existing subscription that succeed across several tests with little variation.
- Retest using the first load and continued interaction on websites you use regularly.
- Test at the same time of day for the same task, and keep one verified backup option.
When a specific region is required
- First, check which regions the destination service allows and any account requirements.
- Confirm the actual exit region instead of relying on the region shown in the server name.
- Then compare latency and stability among candidates that meet the requirements.
Meet the destination service’s requirements first, then compare results under the same conditions. A region label or low latency doesn’t guarantee that the service will be accessible.
A website may also determine your region using account details, payment information, or past sessions. Changing your exit point doesn’t change your account region, and it doesn’t guarantee that the destination service will accept the connection. If you see a region-related message, check the service’s account and access requirements instead of cycling through a large number of servers.
5. Check Protocol Compatibility Before Comparing Performance
Shadowsocks, VMess, VLESS, Trojan, Hysteria2, and WireGuard are different connection methods, not a ranking from fastest to slowest. Different protocols in the same region may use different servers or routes, so don’t attribute every difference to the protocol. When using an existing subscription, keep the parameters supplied by the provider. Don’t change Type or the port just to get a lower latency reading.
| Protocol | Check first | What to consider when comparing |
|---|---|---|
| Shadowsocks | Server address, port, encryption method, password, and additional parameters | Check TCP and UDP availability separately; it depends on the server and configuration |
| VMess | UUID, transport method, and relevant parameters when TLS is enabled | Transport-layer settings and an incorrect device clock can both affect the connection |
| VLESS | UUID, transport method, TLS, and any other parameters required by the server | You can’t convert an existing server just by changing its protocol name |
| Trojan | Password, TLS, server name, and server certificate requirements | A reachable TCP port doesn’t confirm that TLS and authentication succeeded |
| Hysteria2 | Authentication, TLS, and UDP network reachability | Based on QUIC; UDP restrictions on the network can affect connection setup and persistence |
| WireGuard | Key, peer, public endpoint, and tunnel address configuration | Uses UDP; check the handshake and actual traffic rather than relying on a TCP probe |
A port number doesn’t identify the protocol. TCP 443 is commonly used for HTTPS, while UDP 443 can be used for QUIC, but servers can use other ports too. Use the complete configuration provided with your existing service. Changing the port to 443 doesn’t guarantee a faster connection, and disabling certificate verification isn’t a fix for connection failures.
6. Use Global Routing to Rule Out Misleading Results
The server selected in Shadowrocket isn’t necessarily the route used by every request. Global Routing controls how routing is applied, while rules in Config—such as DOMAIN-SUFFIX, GEOIP, IP-CIDR, and FINAL—determine what matches and which policy applies. Check the request’s route before interpreting a server test, so you don’t attribute DIRECT traffic to a proxy server.
Config
RecommendedApply rules from the current configuration. The action depends on the matching policy and may be DIRECT, a proxy policy, or another configured action. If no earlier rule matches, check how FINAL is set up.
Best for: everyday rule-based routing and checking which policy a destination domain actually matches.
Proxy
Route proxy-capable connections handled by Shadowrocket through a proxy. This can help reduce rule-based routing as a variable during a short comparison. It doesn’t take over all system traffic on your device or guarantee that a destination site is reachable.
Best for: a brief comparison to see whether routing in the current Config is causing a failure.
Direct
Send relevant connections directly to their destinations. Don’t use this mode to evaluate a proxy server. If direct access works but proxy access fails, that narrows the troubleshooting scope; you still need to check the proxy route and configuration.
Best for: a direct-connection comparison to distinguish local network behavior from the proxy route.
Scene applies routing settings based on a scenario. If you’re using Scene, first check which scene and routing conditions are active. Don’t change the scene, rules, and server all at once during a test. The rule snippet below illustrates matching logic only. Its example domains aren’t real services, and it isn’t a replacement for your complete Config.
[Rule]
DOMAIN-SUFFIX,example.com,DIRECT
IP-CIDR,192.0.2.0/24,DIRECT,no-resolve
FINAL,PROXY
In this snippet, example.com and its subdomains match DOMAIN-SUFFIX first and use DIRECT. 192.0.2.0/24 is a documentation-only example network, and no-resolve means this IP rule won’t trigger an active DNS lookup. Other requests fall through to FINAL and use PROXY. PROXY in an actual configuration must refer to a valid policy. GEOIP matches against an IP geolocation database; you can’t determine its result from a domain suffix alone.
7. Keep Reviewable Primary and Backup Options
Make your final choice easy to review: record the test date and time, network, protocol, Global Routing mode, repeated test results, and performance on the actual task. After an existing subscription updates, server labels, addresses, or parameters may change. The same label doesn’t guarantee the same connection conditions, so retest when anything significant changes.
- Primary option: It handles your usual tasks, performs reliably across repeated tests, and meets the requirements of your existing service and destination site. Don’t choose it based on one lowest reading.
- Backup option: It has been verified with an actual task and, ideally, has a known route or transport difference from the primary. Different names alone don’t prove that two options are independent.
- Switching order: Check your local network and subscription status first, then switch to a verified backup. Change only one variable at a time.
- When to stop: If all candidates fail at once, check your local network, authentication, data allowance, and subscription status before trying more servers based on latency.
- What to include in your notes: Keep server labels and results, but don’t share complete subscription URLs, passwords, UUIDs, or private keys. Remove sensitive details before contacting your existing provider.
This process helps you choose based on the current network, time, and task; it isn’t a permanent ranking. When your network or needs change, repeating the same short tests is more useful than chasing a one-time lowest latency.
Get Shadowrocket and Browse the Guides
The iPhone and iPad version is a one-time $2.99 purchase on the App Store. The developer is Shadow Launch Technology Limited; check the App Store listing for system requirements. Versions for Mac (Apple silicon and Intel), Windows, Android, and Linux are free. Find all platform options on the download page, and see the guide for steps to import an existing subscription.