How to choose a server location

Most of the time, tapping Smart Connect is all you need. When you have a specific purpose, such as online gaming, a service hosted in a particular region or a fixed egress address, knowing how nodes are scored and filtered helps you choose better. This article covers the three core metrics, how Smart pick scores nodes, and the principles for choosing by hand.

The three core metrics

Latency. The time a packet needs to travel from your device to the node and back, in milliseconds. It is dominated by physical distance and routing: light in fibre covers roughly 200 km per millisecond, so 1,000 km in a straight line means at least 10 ms round trip, and real paths are longer.

Load. How much of the node's bandwidth and connection capacity is in use. A heavily loaded node delivers poor throughput even when its latency is low.

Packet loss. The share of packets lost in transit. Loss triggers retransmissions and hurts real-time applications (games, voice, market data) in particular, often more than latency itself.

There is also jitter, the variation of latency. A steady 60 ms is usually better than a connection that swings between 30 and 150 ms.

How Smart pick scores nodes

Every TP VPN node continuously reports latency, load and packet loss, and the server combines them into a weighted score per node, where lower is better:

Metric	Weight	Normalisation
Latency	50%	Full penalty at 300 ms
Load	30%	Full penalty at 100% load
Packet loss	20%	Full penalty at 5% loss

A few important rules:

Only nodes in healthy or warning status take part in routing; offline, in-maintenance and full nodes are excluded automatically.
A new node that has not reported latency yet is scored as worst case so it is never recommended by mistake.
The lowest-scoring node is marked "Best", and each scene category is scored separately.
The app reconnects automatically when the network changes and retries with exponential backoff after a drop.

Because the score combines three metrics rather than latency alone, it occasionally sends you to a slightly more distant but steadier node. See Smart routing.

The nearest-node rule and its exceptions

For general use, the node closest to you usually has the lowest latency. But "close" should be measured from the service you are reaching:

Public Wi-Fi protection, everyday browsing: choose the node nearest to you.
A service hosted in a specific region: choose the node nearest to that service's servers. The data has to get there anyway, and letting the encrypted tunnel cover as much of the path as possible is often steadier than exiting near home.
Cross-border work and remote access: choose a node in the same region as the company servers and add it to favourites so it stays fixed.

Physical distance sets the floor for latency; no node can move data faster than light. What a node can improve is the congestion and stability of the path.

How the scene routes choose nodes

The Nodes tab offers four categories, each filtering and ranking nodes by its own strategy:

Category	Strategy	Coverage	Plans
Smart pick	Weighted automatic routing	All nodes	Every plan
Gaming	Latency first, UDP-priority forwarding, doubled jitter penalty	Hong Kong, Taipei, Tokyo, Osaka, Seoul, Singapore	Standard and above
AI services	Static egress on high-reputation ASNs, high-bandwidth nodes only	Los Angeles, New York, Tokyo, Singapore, Frankfurt	Premium and above
Exchanges	Packet-loss and latency thresholds, nearest exchange datacenter	Hong Kong, Singapore, Tokyo, Dubai	Premium and above

If your use case belongs to one of these, switch to that category and then pick the city nearest to your target. The routes are described in detail in Gaming latency and VPNs and Stable access to AI services.

Line types

The tag next to a node shows its line type:

Standard relay lines. Available on every plan.
IEPL premium lines. Tagged "IEPL"; they run over international Ethernet private lines with lower jitter and loss, available on Premium and above.
Dedicated egress IP. Tagged "Dedicated IP"; the egress address stays fixed, which makes allow-listing easy, available on Enterprise.

A lock icon means your current plan tier is too low; the node unlocks immediately after an upgrade, with no need to sign in again.

When to switch nodes

Latency is clearly higher than usual, or jitter is severe.
The app reports "Node unavailable" or the node shows "Maintenance".
Nodes in one city slow down during the evening peak; try a neighbouring city.
You have travelled and the node that used to be nearest no longer is.
The service you are using expects a particular egress region; switch to that region.

Switching needs no separate disconnect step: tap the new node in the list and the app handles the change.

Practical tips

Favourites. You can favourite up to 20 nodes; favourites and recently used nodes stay pinned at the top of the list.
Search. Search directly for a city, a country or a node code.
Default scenario. Set your usual category under Settings → Default scenario and combine it with Auto-connect on launch so the app connects with that scenario as soon as it opens.
Pinned versus automatic. A node you chose by hand stays in use until it becomes unavailable; tap Use smart routing to go back to automatic selection.