Use Policy Groups in Clash Verge Rev on Windows: Latency Test Walkthrough

What policy groups are in Clash Verge Rev

A policy group is a named bucket inside your YAML proxy-groups: section that lists other outbounds—individual nodes or even nested groups. When Mihomo resolves traffic, rules point to these names, not to raw server lines. That indirection is what lets providers ship tidy labels such as Proxy, Auto, or Streaming while hiding dozens of underlying endpoints.

Clash Verge Rev surfaces those same buckets on the Proxies page. Each card typically corresponds to one group from the active profile. The visual layer does not invent new logic; it exposes the strategies your subscription author declared. That is why two profiles can look wildly different side by side even on the same app version.

Understanding policy groups matters because many newcomers conflate them with the high-level Rule / Global / Direct toggles at the top of the window. Those toggles choose how aggressively the core applies your rules: array before it ever reaches a nested selector. Policy groups decide which concrete outbound wins once a rule already aimed traffic at PROXY or a custom constant. Keep both layers in mind while testing.

Where to click on Windows after launch

On Windows, start from the left rail. Select Profiles first and press Use on the configuration you want so the runtime reloads config.yaml (or the merged equivalent). A greyed Proxies list almost always means no active profile or a failed download—fix that before chasing phantom latency bugs.

Move to Proxies. The layout groups cards vertically: global favorites may float at the top if your build pins them, while the remainder mirrors the subscription order or alphabetical sorting depending on the theme pack. Expand a card by clicking its title row; leaf nodes appear underneath with radio or list semantics based on whether the group is select or url-test.

If you enabled System Proxy or TUN elsewhere in settings, keep those states stable while testing. Flipping TUN midway changes how Windows routes packets and can invalidate a latency run that began on a different stack. For a controlled experiment, stay on one mode for several minutes.

Group types you will actually see: select, url-test, fallback, relay

Select groups are manual. The UI waits for you to click an outbound, and Mihomo keeps honoring that pin until you pick another row or the profile reload wipes state. They are ideal when you insist on a specific city for banking or latency-sensitive games.

url-test groups automate movement. The core periodically performs HTTP latency probes against url: targets, ranks healthy proxies, and promotes the fastest candidate within optional tolerance: millis. The Clash Verge Rev badge might read Auto or show a small chart icon depending on localization. When users say “automatic line selection,” they usually mean this strategy.

Fallback resembles url-test but prioritizes the first alive upstream in list order, which is helpful when providers want deterministic failover instead of constant benchmarking. Relay chains nodes sequentially for obfuscation-heavy topologies; you rarely “latency test” the inner hops individually from the GUI because the chain behaves as one logical pipe.

Seeing the type at a glance prevents frustration. Clicking furiously inside an auto group expecting a permanent manual lock will not work unless the author nested a select child underneath for that purpose.

How latency testing maps to the lightning button

Most Verge Rev builds expose a speed test or lightning icon near each group header and sometimes beside individual nodes. Pressing it issues an immediate probe through the same mixed port your browser would use, measuring round-trip delay to whichever URL the developer configured—often http://www.gstatic.com/generate_204 or a provider-specific endpoint.

Rows update with millisecond integers when the handshake completes. A blank or dashed cell means the attempt never finished before the client-side deadline, not necessarily that the server vanished forever. Retry after disabling aggressive Windows Defender scanning on foreign executables or after pausing VPN-within-VPN experiments that double-NAT the path.

Run group-level tests when you want a holistic snapshot and per-node tests when you suspect one city is degrading. The UI sorts visually but does not always commit the same ranking to disk; url-test groups still obey YAML timers for persistence.

What url-test in YAML adds beyond the button

Power users eventually open the profile editor or external viewer to read blocks such as:

• interval: — seconds between automatic retests for the group.
• tolerance: — allowable delta before swapping winners, reducing flappy jumps.
• lazy: — defer probing until the group receives traffic.
• url: — the probe target; mismatches here explain odd divergence from what the UI button pings if a skin overrides defaults.

When the subscription renews, authors sometimes shrink interval to chase realtime load balancers; that behavior can thrash selections on hotel Wi-Fi. If you maintain a local patch, consider raising the interval slightly for stability, knowing you inherit maintenance cost.

If a url-test group never switches even though the UI shows fresh numbers, check whether another nested select parent still pins an outdated child. The engine evaluates from the outside in; a stale outer choice blocks inner brilliance.

Switching nodes manually without fighting automation

For select groups, click the row you want. The highlight moves immediately, Mihomo records the preference in runtime state, and new flows should adopt the outbound on the next connection. Long-lived TCP sessions might linger; closing the tab or toggling the target domain forces a clean handshake.

For url-test groups, manual clicks may be ignored or rolled back unless the author exposed a dedicated manual branch. Common patterns include embedding a Manual select sibling that lists the same proxies or using duplicate groups—Auto for lazy users and Proxy for tinkerers. If your profile lacks that fork, duplicate the YAML locally or ask the provider for a variant; brute-forcing clicks against auto logic wastes time.

Windows-specific hotkeys differ by build, but right-click context menus occasionally expose Copy node name or Jump to rule shortcuts. Those quality-of-life extras help when you cross-check against logs.

How rules, MATCH, and nested groups interact

Even perfect latency nods fail if rules: point elsewhere. A streaming domain might bypass your Auto group via a dedicated RULESET entry. Before blaming the selector, open the Connections or Logs panel (if your build bundles it) and confirm which outbound handled the flow.

The closing MATCH rule is the safety net. If it targets Proxy, every unmatched flow inherits whatever node that group currently advertises. Changing Auto without editing MATCH still influences most casual browsing because providers structure YAML that way on purpose.

Nested groups add recursion: an Auto group may contain regional select children. Latency tests bubble statistics upward, yet selection semantics remain per card. Document your mental model on paper if the subscription uses more than two layers; screenshots help when reporting issues to community chats.

Troubleshooting misleading metrics and sticky selections

Timeouts immediately after import

Verify system time, disable alternate security clients that inject TLS inspection, and confirm your provider tokens are valid. Batch tests that all fail usually mean DNS to the probe domain is blocked, not that every node died.

Low latency but sluggish pages

The node might rate-limit HTTP/2, suffer congested peering, or sit on overloaded CPUs. Switch regions, enable TUN mode if split-DNS is leaking, or cross-test with a wired connection to isolate Wi-Fi jitter.

Auto group reverts after manual clicks

That is expected for url-test. Clone the group as select in a user patch or rely on any DIRECT bypass rules for emergency offline work rather than fighting the scheduler.

UI differences between minor Verge Rev versions

Icon placements shuffle between releases. If you cannot find the lightning control, expand the overflow menu on a group header or consult the release notes for the week you installed.

How this differs from setup guides and macOS tray tutorials

If you still need download links, installer signatures, and subscription import taps, read Clash Verge Rev setup for Windows and macOS first—this article assumes those boxes are checked. On macOS, ClashX Pro Rule versus Global explains tray-centric mode toggles that Windows Verge Rev exposes differently; the underlying policy group concept stays parallel, but the pixels are not identical.

For classic Win32 users migrating from archived Clash for Windows forks, expect clearer Meta-core defaults here yet familiar vocabulary. The mental model ports forward even when spacing in the Proxies grid changes.

Need ecosystem-wide proxy basics while you tweak selectors? Skim subscription maintenance so auto-updates do not yank your manual YAML edits unexpectedly.

Practical workflow checklist before blaming the provider

1. Reload the profile and confirm the timestamp in the corner matches your last save.
2. Run a batch latency sweep from the group header, not only from a collapsed row.
3. Pick a manual node in any available select branch and load two different sites—one domestic, one international—to validate Rule behavior.
4. Capture a screenshot with timestamps if you open a support ticket; include Mihomo version from Settings → About.
5. Revert experimental flags if you enabled custom external controllers that might race the UI state.

Summary

Policy groups in Clash Verge Rev on Windows are the selectable cards under Proxies, mirroring proxy-groups from your active profile. Latency tests visualize reachability for those outbounds, while url-test groups add timed automation, tolerance, and probe URLs that may differ slightly from on-demand button checks. Knowing whether you are inside a select or url-test strategy tells you whether clicks stick or the scheduler will move again on the next interval.

Compared with all-in-one consumer VPN apps that hide outbound logic behind a single connect slider, that transparency is the point: you see the skeleton of your traffic engineering. The downside is babysitting—when providers ship complex trees, you must actually read the cards. Many “lightweight” VPN clients also obscure probe methodology, which feels simpler until you cannot explain why Netflix jumped regions after dinner. Generic wrappers rarely expose nested YAML, so when auto selection misfires you get mystery meat routing instead of a tweakable tolerance: knob. Telemetry-heavy store apps can be equally opaque about why Windows reports “connected” while DNS still leaks outside the tunnel. Staying inside the Clash ecosystem keeps policy vocabulary consistent whether you script Mihomo on a router or click through Verge Rev on a laptop.

When you want a maintained Windows GUI on top of that engine—without hunting forum mirrors for unsigned bundles—use one place for Verge Rev and companion cores so updates stay aligned. Download Clash for free and keep policy groups, latency probes, and manual overrides under the same roof.