Sea of Thieves Proxy: Which Flow Goes Through the Tunnel and Which Does Not?
In a game set in a shared world the client does not open a single connection: account, store, matchmaking, in-game state and voice are separate flows. A proxy can carry only some of them. This page explains exactly where the split lies and how region assignment relates to your account.
Flow layersThe split between the account, patch, matchmaking, in-game state and voice flows.
02
TCP or UDP?The clear boundary between the packets the tunnel can and cannot carry.
03
Coverage matrixA comparison of which proxy type covers which flow and to what extent.
04
Region behaviourHow assignment is made and what actually depends on the account.
In titles like this, the sentence "the game won't connect" points not to one fault but to five independent possibilities. Account verification may be failing, the store client may be unable to download the patch, the matchmaking step may be producing no result, the in-game flow may be breaking up, or only voice communication may be down.
A proxy does not touch all five of these in the same way. It covers some completely and never touches others. The sections below first separate the flows, then show what passes through the tunnel according to transport protocol, and finally explain why region assignment and account region are different things.
Adopt this as your method: before trying to fix a fault, determine which layer it is in. When the wrong layer is chosen, every setting you change adds a new variable instead of solving the problem.
Which flows does the client open?
At the top is the account and session flow. Login, authorisation and profile requests run over TLS-protected TCP connections and follow the classic request–response pattern. This is the part a proxy carries most comfortably; when you define an exit, the address the other side sees becomes the exit's address.
The second layer is store and patch distribution. This traffic is handled not by the game process but by the platform client you run the game through. Files are pulled from a content delivery network over HTTPS; volume accumulates here, and almost all quota consumption comes from this layer.
The third layer is matchmaking: which world instance you are assigned to and which players you share a sea with are decided at this step. Most of the decision is made on the server side and the result is communicated to the client.
The bottom two layers are the in-game state flow and voice. Ship position, wave state, a fired cannon and player movement consist of small, continuously updated packets; proximity chat is a separate flow. The transport choice of these two layers is where the proxy debate comes to a head.
DIAGRAMThe flow layers the client opens
You can scroll the diagram horizontally to inspect it
The top three layers flow over TCP and can go through the tunnel; the bottom two commonly use UDP.
The TCP/UDP split: what does the tunnel carry?
When you want to reach an HTTPS destination through an HTTP proxy, the client has a tunnel opened with the CONNECT method, and the tunnel carries only TCP. How this method works is explained in detail in the HTTP CONNECT method . The bytes passing through the tunnel are encrypted; the proxy cannot read the content but does see which host you connect to.
For continuously updated position and state information, TCP's ordering guarantee is a burden rather than an advantage. Waiting for a lost packet to be resent is meaningless when the information it carries is already stale. That is why real-time flows commonly prefer UDP, and those packets cannot enter a TCP tunnel.
SOCKS5 protocol defines a method named UDP ASSOCIATE for carrying UDP. For it to work, the proxy server must offer this method and the client must be written to use it. Game clients generally open their own sockets directly and ignore the SOCKS5 definition on the system; the limits of the mechanism are SOCKS5 UDP support article.
The conclusion is summed up in one sentence: in this game a proxy generally covers the login, store and patch layers; it does not cover the in-game state flow or voice. For the general framing of protocol choice, proxy protocol selection guide is useful.
Why must the platform client and the game process be treated separately?
On the desktop the game is launched through a store client, and that client is not just a shortcut: it holds the account session, checks versions, manages downloads and starts the game. On the network side, that means a set of connections independent of the game process.
When you write an application-based routing rule, only the processes you add to the rule list are covered. Adding the game's executable to the list does not cover the store client's downloads. The reverse is also true: adding the client to the list does not push the game's own flow into the tunnel.
There is a second detail on the Windows side. Downloads coming through the store are often carried out not by the game's own process but by the operating system's download component. Since this component runs as a separate service, it falls outside application-based rules and is affected only by a system-wide setting. The steps on the system side are Windows 11 proxy settings article.
Note
After writing the rule, do not try to verify the exit from inside the game. Reading your address from an independent page on the same machine shows far more clearly which process is covered.
Why do patch downloads ignore the rule you wrote?
The picture of "a proxy is defined but the download goes over the normal line" during an update has three typical causes. The first is the scope problem above: the process carrying out the download is outside the rule. The second is IPv6 bypass; if your exit is IPv4-only and the destination is reachable over IPv6, the operating system prefers IPv6 in most setups and the proxy is bypassed entirely.
The third is local network exceptions. The "do not use a proxy for these addresses" list in the proxy settings may overlap with some of the hostnames the distribution network uses. Reading this list once after setup makes hours of later diagnosis unnecessary from the start.
The common symptom of these three causes is insidious: no error appears, the download completes and the setup looks like it is working. Yet the address the destination sees is your real address. The simplest way to verify scope is to compare your quota consumption before and after the download starts.
If you want to protect your quota, the preference is usually the opposite: deliberately leave patch traffic outside the proxy. On a pool billed by data transferred, a single update can consume your entire monthly budget in one go.
Which setup covers which flow, and to what extent?
Seeing the scope as a matrix speeds up the decision. The rows are the setup types, the columns the flow layers; the cells are relative weights showing how far that setup covers that layer. It is not a measured value but a map for comparison.
The first conclusion from the table is that across the three layers flowing over TCP, there is little difference between setup types. The difference appears mainly in the in-game flow column, and even there no option reaches full coverage. This is a limit arising from the nature of the protocol.
The second conclusion is that while a rule at the gateway level gives the widest coverage, it also brings the highest complexity. Routing every device at home through the same exit makes it harder to work out which device is affected when something goes wrong.
The third conclusion is on the concurrent connection side. When the game client and the store client run together, the number of open connections is higher than expected; the interruption that appears when your panel's limit is reached looks like a faulty setup. The meaning of the limit is concurrent connection limit article.
DIAGRAMThe coverage weight of setup types across flow layers
You can scroll the diagram horizontally to inspect it
The cells are not measured values but relative coverage weights for comparison; in the in-game column no option reaches full coverage.
Choose an exit for your work around Sea of Thieves
A stable static exit is preferred on the account and verification side, and addresses from different countries for regional display checks.
Choose whichever you need from our residential proxies, datacenter proxies, IPv6 and ISP solutions. Every plan comes with unlimited options, 99.9% uptime, rotating proxies, sticky sessions and 24/7 support. Ideal for web scraping, ad verification, SEO monitoring and digital data collection.
ISP ProxyStatic Turkish IPs registered to an ISP
ISP-registered static Türkiye IPs; they combine datacenter speed with the reputation of a real carrier. Ideal for long sessions and low-ping use.
Region assignment, account region and exit location
These three concepts are often confused. Region assignment is a server-side decision about which region's infrastructure the game session will run on, generally made automatically based on the characteristics of your connection; the client offers no manual region selection interface. Account region is the country your platform account is registered in and determines store behaviour. Exit location is simply which address your traffic appears to come from.
The relationship between the three is one-way. Changing your exit location does not change your account region; the account region is defined on the platform side and has its own rules. A proxy is not a tool for disabling regional access restrictions, and this page was not written for such use.
The only thing exit location really affects is the length of the route. The closer the exit is to you and to the destination, the smaller the cost of the added hop. But note: at best that cost approaches zero, it never goes negative. Because a proxy adds an extra hop, it increases round-trip time in most setups; the does a proxy lower game ping article covers this expectation in detail.
When choosing an exit country, staying consistent with your account's usual country of use produces the least friction, both in interface language and in extra verification requests. Which locations are offered is proxy locations on our page.
DIAGRAMThe effect of exit location on route length
You can scroll the diagram horizontally to inspect it
The values are relative weights showing how short the route stays when viewed from Turkey; they are not measurements, and none is shorter than a direct connection.
Latency budget and measurement discipline
Do not look at a single number when evaluating an exit. Total time consists of three parts: the distance between you and the exit, the distance between the exit and the destination, and the load the exit server is carrying at that moment. Since the third part changes during the day, a one-off measurement is misleading.
Always take the measurement in two states: with the proxy on and with it off. The difference is the real cost of the added layer, and hiring an exit without seeing that cost sets the stage for slowdowns whose cause cannot be worked out later. For a simple start, ping test is enough.
The second measurement is stability. Rather than the average, look at how often the worst results occur. In a game set in a shared world, an occasional long stall is more disruptive than a consistently slightly high value. Tools that report only the average hide this variance; note the highest values during measurement as well.
The third variable is how many people share the exit. As the number of users on the same address grows, the queue on the server side grows too and latency variance becomes pronounced. A dedicated exit removes this variable but raises the cost; which side you choose depends on how long your sessions run and how much interruption you can tolerate.
Tip
Repeat the measurement in three different parts of the day and take the worst result as your basis. Since a single voyage lasts several hours, a session that starts at a quiet hour ends at a busy one; so the binding number is the one taken in the time slot when you actually set sail.
Setup and verification: how do you order it?
Advance the setup layer by layer. First write the rule on a single machine, then verify your exit address from an independent page, then confirm you can log in with the platform client, and open the game last. Skipping a stage hides which layer the fault is in.
Stage
What it verifies
If the symptom is wrong
Exit address check
That the rule is active on the machine
If the address does not change, the rule was written out of scope
Platform login
That the account layer goes through the tunnel
If login stalls, it is an authentication or quota problem
A small download
That the store layer is covered
If quota does not decrease at all, the download is outside the rule
Opening the game
That the matchmaking step works
If it stalls here, the block is on the network side
In-game session
The behaviour of the state flow
If there is a problem only here, the cause is on the UDP side
Connection details take the same form with every provider: proxy.example.com a hostname such as, 8080 a port number such as, and if required username with password fields. The real values are in your panel; the port number itself does not determine the protocol.
When completing verification, check the leak side as well. In browser-based tests the WebRTC interface can expose your real address independently of the proxy setting; measuring this behaviour once is enough.
What is a proxy really good for in this game?
The realistic framing is this: here a proxy is not a performance tool but an access and visibility tool. It does not improve in-game smoothness, does not determine which region matchmaking assigns you to, and does not change your account's regional status.
It does, however, genuinely help in a few scenarios. Using a known, fixed address when leaving a corporate or shared network is one. Verifying how a store page or announcement content looks from another country is a second. A third is measuring a provider's stability and latency behaviour.
In teams where several people use the same exit, one habit helps: recording in a single place which machine uses which exit label. Without such a record, it becomes hard to separate who is affected when a fault occurs, and everyone starts retrying the same settings. A simple table prevents both this repetition and arguments over quota.
Make the decision with one question: does changing my exit address make a difference to this task? If you cannot give a clear yes, leaving the setup as it is is the best option; an unnecessary layer only increases the number of variables you have to measure. Compliance with the game's and the platform's terms of service is, in every case, the user's responsibility.
Frequently asked questions about Sea of Thieves and proxies
01Does the in-game connection go through a proxy?
In most setups, no. The in-game state flow commonly uses UDP, and CONNECT a tunnel opened with it carries only TCP. SOCKS5's UDP method exists to close that gap, but the client has to use it too.
02If I change my exit country, does the region I am assigned to change?
Region assignment is a decision made on the server side, and the client offers no manual region selection. Your exit location only determines which address your traffic appears to come from; it does not change the assignment or your account region.
03Why does the patch download not go through the proxy?
There are three typical reasons: the process carrying out the download falls outside the rule's scope, your exit is IPv4-only while the destination was reached over IPv6, or the exception list in the proxy settings covers the distribution hostnames.
04Does voice communication work with a proxy?
Proximity chat is a separate flow and, being real-time, commonly uses UDP. For that reason it does not enter a TCP tunnel. When voice does not work, checking whether the network allows UDP is a better starting point.
05Does a proxy improve the game's response time?
Do not expect it to. The added hop lengthens round-trip time in most setups. The rare cases where your default route is unnecessarily long are the exception, and they can only be identified by comparing two measurements taken with the proxy on and off.
06Do I need separate rules for the store client and the game?
If you use application-based routing, yes; the two processes open separate connections. A system-wide setting covers both, but then patch traffic also goes through the same exit and, on a metered pool, costs grow quickly.
07Does the concurrent connection limit cause problems in this game?
It can. When the game client and the store client run together, the number of open connections is higher than expected. The interruption seen when the limit is reached looks like a faulty setup; choose the limit in your panel according to how you use it.
08How do I know the setup is correct?
Proceed layer by layer: first read your exit address from an independent page, then the platform login, then a small download, and the game last. The stage at which you get stuck tells you directly which layer the problem is in.