Mayak Blog

VPN for gaming: packet loss on your provider’s path and how to check it

The game loads, the lobby opens, and then in the match your character snaps back, stutters or never spawns at all. Websites and video work fine. You turn on a VPN — and the match runs smoothly. That is what packet loss on the way to the game server usually looks like. Below: how to check for it on a computer, in the game itself and on a phone, how to tell that it is your provider losing the packets, and which VPN for gaming helps in that case and which only adds delay.

In short. A VPN for gaming is not an accelerator but a different road to the game server. It helps when packets are lost on your provider’s path: through the tunnel they take another route. When the path is healthy, the tunnel only adds delay: in our measurement the tunnel itself added 1–2 ms, and the detour through an exit on a good route 10–30 ms (details in the ping piece). Checking for packet loss takes a couple of minutes — with a command on a computer, a counter in the game or a test in the browser. The key is to measure with hundreds of packets and compare in pairs: with the tunnel and without.

We make Mayak — a private network with an Android app. Our tunnel runs on AmneziaWG: on the main path game traffic travels in it as UDP, and the ping to every country is visible right in the country list. You can check for free whether the tunnel changes anything on your network. 3 days free after you confirm your email, no card needed.

What packet loss looks like in a game

An online game swaps small packets with the server dozens of times a second: your inputs, the characters’ positions, shots. A lost packet never arrives. In the game it looks like this:

  • You snap back, your character “teleports”. The game on your device does not wait for the server: it draws the movement at once, as if the server will agree, and then checks with it. When the server’s answers are lost or arrive in a bunch, the character is put back where the server last saw it. That is how movement prediction works in fast-paced online games.
  • The character does not spawn, an ability fires late, a hit “did not count”. Most likely the same thing: a message from the server or to it was lost or late.
  • “Unable to connect to the game server”, a timeout in the middle of a match. There is so much loss that the server stopped hearing the player.

Ping can look decent all the while. The ping command times only the packets that came back and reports the lost ones on a separate line, as a percentage. So “ping 40 and the game is unplayable” is not a contradiction. High ping and packet loss are different troubles: what ping is made of and how much a tunnel adds to it we covered in a separate piece on high ping in games. This one is about loss.

Why the game loads but the match falls apart

This is the complaint that reached us: the game loads, but in the match the hero does not spawn and keeps snapping back; with a VPN — even with an exit in Russia — everything is smooth. We did not measure that case ourselves, but the picture is familiar and the explanation is simple.

Loading and the match travel differently. Login, the store, updates and loading usually go over TCP: it asks again for every lost piece, so with loss the download is slower but gets there. The match itself in fast-paced games usually goes over UDP: nothing is asked for again, because a character’s position from a second ago is of no use to anyone. What is lost in a match is gone for good — and turns into stutter.

Why a VPN helped. With the tunnel on, the provider no longer carries the game’s packets to its server but encrypted packets to the exit’s address — along its own route to that address. If the loss sat on the old route, the new one may not have it. An exit in Russia is even handier here than one abroad: the detour to it is shorter, and it changes the road just the same.

Hence the main sign of loss on the path: websites, video and downloads work while the match falls apart. Checks that watch whether websites are reachable do not see it: a website needs a short exchange, a game needs a long, steady stream.

How to check for packet loss

One rule for every method: measure with hundreds of packets and compare in pairs — the same server, neighbouring minutes, with the tunnel and without. We learned this the hard way: on 31 August 2026 a run of 10 packets showed “10 % loss”, and the same path over 60 packets showed none. One lost packet out of ten is not yet a loss rate; loss in every run in a row is.

On a Windows computer

Open the command prompt (Win+R, then cmd) and send a hundred packets to the game server, if the game shows its address, or to any reliable address in the same country:

ping -n 100 address
pathping -n address

ping reports how many packets were lost at the end. pathping goes further: it finds every hop on the path and sends 100 queries to each, so it takes a few minutes. In the result every hop gets its own loss percentage. Look at where the loss starts and whether it carries on to the end. If a percentage appears at some hop and holds all the way to the destination, packets are being lost from that point. If loss shows at only one intermediate hop and not after it, that router is simply not answering queries addressed to itself while forwarding other traffic just fine. That is what Microsoft’s pathping documentation says in so many words.

On Linux and Mac

ping -c 100 address
mtr -rwc 100 address

mtr is the same as pathping, in one table; it is installed separately, from the system’s packages. You read it the same way: what matters is loss that carries on to the last line.

In the game itself

A counter in the game beats any command: it measures exactly the stream that bothers you. In CS2, according to gaming guides, this console command puts the share of lost and out-of-order packets on screen:

cl_hud_telemetry_net_misdelivery_show 2

In Dota 2 it is Display Network Information in the advanced options: next to ping it shows loss separately for incoming and outgoing traffic. Many mobile games show only ping — then judge by the stutter and compare with the tunnel and without. We have not checked these counters in the games ourselves.

On a phone

Android has no ping command at hand, so on a phone there are two routes: the counter in the game, if there is one, and a test in the browser. Online tests such as packetlosstest.com send a stream of small packets over WebRTC, which behaves like UDP, and count loss separately towards the server and back. By default that is 15 packets a second for 10 seconds, i.e. 150 — as the test’s own page says; the server is picked from a list that includes Germany and Lithuania. The test measures the path to its own server, not to the game’s, but it is easy to repeat with Mayak and without, on Wi-Fi and on mobile data.

How to tell that your provider is losing packets

Packets get lost in four places: in your home network or on the radio link, at your provider, at the junctions between networks along the way, and at the game itself. A VPN can help only with the second and third, so rule out the rest first. In this order:

  1. Is it the radio? On a mobile network with a weak signal, latency jumps by hundreds of milliseconds before any tunnel is involved: on 28 July 2026 we saw idle jitter of up to 1058 ms at a ping of 84. On Wi-Fi, move closer to the router or test over a cable. Loss that disappears next to the router is not your provider.
  2. Is something downloading? A game update, a torrent on another computer, high-definition video fill the line, and the game packet waits in the queue. In our measurement ping was 76 ms at idle and 650–670 under load (the write-up).
  3. Is the game down? If it is bad for everyone at once, check the game’s news: the problem is on its side.
  4. Smooth on another network? Same game, same server, but mobile data instead of home internet — or a friend on another provider — plays without stutter. Then the problem is your provider’s path, not the game and not the device.
  5. Where does the loss start? In pathping or mtr: if the percentage appears at the hops right after your router and carries on to the end, it is the provider’s network. If only at the last hops, it is the game’s network.
  6. When? Loss that comes in the evening and goes away at night looks like congestion: some stretch of the path runs short of capacity at peak hours.
  7. Smooth with the tunnel, not without it? That is nearly a diagnosis: the road through the exit has no loss, the direct one does.

What to tell your provider. “The game lags” goes nowhere. Specifics work: the server address, the time, the pathping or mtr result and the hop number where the loss starts. With a ticket like that, support has something to check.

When a VPN for gaming helps and when it gets in the way

With a tunnel, a packet first goes to the exit server and from there to the game. To your provider it is one encrypted connection to one address, and it carries it along its own route to that address. Valve explains its own relay network for games with the same argument: through it a faster route than the direct one can be found (Steamworks documentation). A tunnel works the same way — it changes the road.

  • It helps when packets are lost at your provider or at a junction along the way, and the road through the exit avoids that stretch. The only way to find out is a pair of measurements: with the tunnel and without.
  • It does not help when the loss is on Wi-Fi or the radio link (the tunnel rides the same radio), on the game’s side, or when mobile internet is limited to allowlists: then games may not work with a tunnel or without one (what we know about that).
  • It gets in the way when the direct path is healthy. Then the tunnel only makes the road longer: in our measurement on 16 September 2026 the tunnel itself added 1–2 ms, and the detour through an exit on a good route from Russia to servers in Germany and Finland 10–30 ms (the tables).

The worst case for a game is UDP inside TCP. Some connection methods carry all traffic inside TCP. Then a single loss holds up every packet behind it: TCP waits for the resend, and the game gets stale data. Mayak has such a path only as a backup: if the “Backup” mark is next to the status, the connection goes through a bridge over TCP. It works, but for a match it is the worst option — try another network or another country. The “Via Russia” mark means the connection goes through a Russian entry point and the leg is longer than the direct one (what the marks mean).

A tunnel changes the road, not the game’s rules. If a game is closed to your country or you want to change your account’s region, that is a question for the game’s rules, and breaking them can cost you the account — see, for example, what the Steam agreement says. We do not give advice like that.

Which VPN for gaming to choose: five signs

A “gaming VPN” is not a special kind of thing but an ordinary tunnel with five things set up right for games. You can check them before paying.

  1. Game traffic travels as UDP. WireGuard and AmneziaWG, which Mayak runs on, carry any IP packet as it is: game UDP stays UDP. Methods that put everything inside TCP are worse for games — why, see above.
  2. An exit close to the game server. The closer the exit is to the server, the shorter the second leg. For CS2 and Dota 2 you can estimate it: matches in these games go through Valve’s relay network, and Valve publishes its map openly. On 24 September 2026 it had no points in Russia; the ones closest to Russia are Stockholm, Warsaw, Vienna, Frankfurt and Amsterdam. So for these games it matters more how close the exit is to a Valve point than the fact that it is abroad. By the country list of 24 September 2026, Mayak has exits in the Netherlands, Poland, Germany and Russia.
  3. Ping to countries is visible in advance. In Mayak it is the “Ping” button above the country list: it measures the response time to each country’s server. The ping in the game will be higher by the leg from the exit to its server (why the app shows two different pings).
  4. The game can go around the tunnel. If the direct path is fine for the game and the tunnel is needed for other apps, it only gets in the game’s way. In Mayak: “Settings” → “Split tunneling”, mark the game and reconnect.
  5. The path is honestly labelled. If the connection moved to a backup path, it should be visible on screen: in Mayak, “Backup” or “Via Russia” appears next to the status.

On a phone and on a computer. Mayak’s own app exists for Android and Android TV. There is no app of our own for computers. Your account issues a config file for AmneziaWG: on Linux it works for us, while on Windows and Mac we have not checked it with a live connection — third-party apps on those systems do not accept our disguise parameters yet, and the tunnel may not come up at all (more in the help centre). The subscription link from your account is taken by the third-party Happ app for computers, but we have not checked that with a live connection either (every route by system). Free or paid is a separate conversation.

Our measurement: loss to Valve’s relays

On 24 September 2026, from 11:20 to 12:20 Moscow time, we sent 500 packets each to four nodes of Valve’s relay network — in Warsaw, Stockholm, Frankfurt and Amsterdam — by two roads: directly and through a live AmneziaWG tunnel to each of our exits. The measurement was taken from a work machine that physically sits in Moscow and reaches the internet through Kazakhstan — not a subscriber’s phone on a Russian carrier.

PathWarsawStockholmFrankfurtAmsterdam
Direct140⁠–⁠172 ms, 0⁠–⁠0.2 %148⁠–⁠183 ms, 0⁠–⁠0.4 %165⁠–⁠197 ms, 0⁠–⁠0.2 %175⁠–⁠206 ms, 0⁠–⁠0.6 %
Via the Netherlands225 ms, 0.4 %238 ms, 0 %211 ms, 0 %210 ms, 0 %
Via Poland194 ms, 0.2 %215 ms, 0.2 %216 ms, 0.4 %221 ms, 0.2 %
Via Germany181 ms, 0.2 %203 ms, 0 %164 ms, 0 %172 ms, 0.8 %
With the Russia exit209 ms, 0.2 %183 ms, 0.2 %201 ms, 0.2 %209 ms, 0 %

We measured the direct path again before each tunnel, and within the hour it shifted by 30 ms on its own — hence the range in the first row. The second leg — from the exits themselves to the same nodes, 300 packets each, at 11:15 Moscow time:

ExitWarsawStockholmFrankfurtAmsterdam
Netherlands23 ms, 0 %35 ms, 0 %8 ms, 0 %7 ms, 0 %
Poland1 ms, 0 %21 ms, 0 %23 ms, 0 %22 ms, 0 %
Germany18 ms, 0 %40 ms, 0 %1 ms, 0 %8 ms, 0 %
Russia49 ms, 0.7 %23 ms, 0 %41 ms, 0.3 %47 ms, 0 %

How to read it. There was almost no loss anywhere that day: the worst through the tunnel was 4 packets out of 500 (0.8 %), directly — 3 out of 500. From the European exits to Valve’s nodes not a single packet out of 300 went missing; from the Russia exit 2 packets were lost to Warsaw and 1 to Frankfurt. So on our path the tunnel had nothing to fix, and it added no noticeable loss of its own. Ping through the tunnel is the road to the exit plus the leg from the exit to the server, and the first part does not follow distance: the Poland exit sits a millisecond from the Warsaw node, yet the road to it from our machine turned out longer than to Germany, so Warsaw via Poland came to 194 ms and via Germany to 181. That is why the country is better picked by the ping in the app and by the game itself rather than by the map.

What this measurement does not say. These are ICMP service packets to Valve’s nodes, not game traffic; from a server, not a phone; over one day, without the evening peak. Through the tunnel ICMP rides inside its UDP packets, so the first leg here is measured the way game traffic would travel it, while the direct path is not: some networks treat service packets differently from game ones.

What we checked

  • Valve’s relay map — with an open request to Steam on 24 September 2026, for CS2 and for Dota 2 (both got the same map version). Dota 2 has 33 points with relays, CS2 the same plus nine in China; there is none in Russia. That the nodes answer ping we checked with three packets to each of the four — a check that they can be measured, not a measurement.
  • Latency and loss through the tunnel — with a live AmneziaWG tunnel on 16 September 2026, 50 packets per “direct / through an exit” pair, with a false-echo control: not a single packet was lost through the tunnel in six pairs, the tunnel itself added 1–2 ms (details).
  • How many packets you need — on our own run of 31 August 2026: 10 packets gave “10 % loss”, 60 on the same path gave zero.
  • Radio and queueing — with measurements on a mobile network in Moscow on 28 July 2026: jitter up to 1058 ms on a weak signal, 650–670 ms of ping under load.
  • How Mayak carries traffic on each path — from how the app is built: the direct path and the path via Russia use UDP, the backup bridge is a tunnel inside TCP.
  • How to read pathping — from Microsoft’s documentation, not from memory.
  • Loss to four Valve nodes — on 24 September 2026 in the daytime, 500 packets directly and through a live tunnel to each of the four exits, with a false-echo control, and 300 packets from the exits themselves (tables above).

What we do not know

  • The case this article started from. We did not measure it: we know neither the game nor the provider. The explanation above is the most likely one given the signs, not a proven one.
  • Servers of mobile games. Where the servers of Standoff 2, PUBG Mobile and other mobile games stand we do not know, and we have not measured to them.
  • Where Dota 2’s Russian region plays its matches. Valve’s relay map has no points in Russia, but how Valve assigns matches to servers we do not know.
  • The counters in CS2 and Dota 2. We took the command and the setting from gaming guides and did not check them in the games.
  • Online loss tests. packetlosstest.com is someone else’s service; we have not compared its numbers with ours.
  • How games behave through Mayak. We have not tested a single game through our app; our measurements are service packets to servers, not game traffic.
  • Why a provider loses packets on a particular path. We see the fact of the loss, not its cause: congestion, equipment settings and an outage look the same from the outside.
  • Computers and iPhone. The config file is checked on Linux; on Windows and Mac it is not. On iPhone Mayak works via the subscription link through a third-party app.
  • Mobile internet allowlists. We have not measured games under such limits and promise nothing.

Short answers

Which VPN is best for gaming? One that carries the game as UDP, with an exit close to the game server and ping to countries on screen. “Best” is not in general but for your path: compare a pair of measurements with the tunnel and without.

What is a gaming VPN? An ordinary tunnel set up for games: UDP, a close exit, the option to send everything else around it. None can shorten the distance to the server: it too only changes the road.

A VPN for gaming on PC? Mayak has no app of its own for Windows. On Linux the config file from your account works; on Windows and Mac we have not checked it (more).

A VPN for gaming on Android? Mayak’s app is for Android and Android TV. A game that does not need the tunnel can go around it: “Settings” → “Split tunneling”.

Is there a free VPN for gaming? Mayak’s trial is free and needs no card. How free services differ from paid ones is a separate write-up.

What is packet loss? The share of packets that did not reach the server or come back. In a game it is stutter, snapping back and disconnects — even with normal ping.

How much loss is normal? On a healthy path, zero over hundreds of packets: through the tunnel in our measurement not one of 50 was lost in six pairs. One lost out of ten is not yet a diagnosis; loss in every run in a row is.

Packet loss in Dota 2 or CS2 — what to do? Turn on the loss counter in the game, rule out Wi-Fi and downloads, run pathping or mtr. If the loss starts in your provider’s network and goes away with the tunnel, pick an exit closer to a Valve point (where they are).

Why won’t games connect to their servers? There are four causes: an outage of the game itself, the game’s rules for your country or account, loss on the path, and limits on mobile internet. A VPN can help only with loss on the path.

Does a VPN lower ping? Usually not — it adds to it. Ping gets lower only when the direct path is roundabout; the numbers are in the piece on high ping.

Try Mayak

A tunnel that carries the game as UDP on the main path, ping to every country right in the app, and measurements we publish as they are — together with what they do not say. 3 days free after you confirm your email, no card needed.

The account is created right in the app. A confirmed email opens the access — 3 free days.