Field note · Updated · Jay Fan
A calmer number on a tunnel belongs to the new path.
A VPN can change the path and the number you see. It does not prove the old path was fixed, and it is not a substitute for measuring loss or NAT. Write the direct row and the tunnel row as two paths. This desk has not published a tunnel-versus-direct game comparison, a before-and-after delay figure, or a loss mark taken on both.
Symptom
What does a tunnel change before you call the old path fixed?
A game VPN, or a system VPN left on while a game runs, encapsulates traffic and hands it to an exit the operator chose. From the house, the next hop is the tunnel. From the game server, the session often arrives from that exit rather than from the address your ISP assigned you. The route in between can be a different set of networks than the direct path. A delay figure read after the tunnel is up is a figure for that arrangement. It is a new observation. It is not a corrected copy of the observation you had with the tunnel off.
The number can move in either direction, or it can look the same, and none of those outcomes repairs the path you left. A calmer match on the tunnel means the tunneled path was calmer for that sitting. A worse match means the tunneled path was worse for that sitting. A match that feels identical means you did not see a difference you can point at. In every case the direct path is still the row you recorded with the tunnel off. If you never recorded that row, you do not have a comparison. You have one path and a story about another.
Exits are not neutral labels. A nearby exit and a distant exit are different destinations for the first half of the trip, even when the game still displays the same region name. Some clients show the exit city. Some only show that a tunnel is connected. If the client will not show the exit, write “not shown.” Do not invent a city from the product’s marketing map. If you changed exits inside the sitting, the tunnel row is already two paths. Mark it void and start again with one exit held still.
Split rules matter as much as the on/off switch. A product can send the browser through the tunnel and leave the game on the direct path, or the reverse. A number you read in a browser, or on a public echo, can then describe a path the match is not using. Before you cite any figure, write whether the game itself was inside the tunnel. If you cannot tell, the cell stays blank. A blank is unread. It is not a claim that the game followed the browser.
Tunnels also tend to replace more than the route. Many of them answer names themselves, so a resolver swap rides along for free. Some of them translate addresses a second time, so a console NAT label read under the tunnel describes the tunnel’s translator. Either side effect voids a different note if you mix the questions. The DNS note is the lookup. The console NAT note is join, voice, and the label. This page stops at the path change.
This desk has not measured the split. There is no paired game session, tunnel off and tunnel on, with the server name held. There is no loss mark compared across those paths. There is no NAT label compared across them. The August 19 web-edge record timed ICMP and TLS toward public and publisher websites while a tunnel stayed up. It has no direct-path control, and it did not time a match server. Citing it as proof that a tunnel repaired a game path borrows a file that never made that comparison.
Decision table
When does a VPN help, and when does it only move the bottleneck?
Use one row for the sentence you can defend. These cells are a decision guide, not a timed session from this site. If your own log has an empty cell, you did not read that column. An empty cell is not a zero, and it is not a pass.
| What you can point at | A tunnel can help | The tunnel only moved the bottleneck | Read next, on the path the complaint is about |
|---|---|---|---|
| The region will not connect on the direct path, and the same server name connects only after the tunnel is up | Reachability. You found another way for that session to exist. | You still have not shown the direct path improved. “Connects only on the tunnel” is the whole finding. | Keep the server name in the row. Do not upgrade the finding into a loss or NAT result. |
| The same server name is reachable either way, and the delay figure looks different | Only as a different path you might choose to play on. The calmer row is the one you are actually using. | The direct path was not rewritten. A worse tunnel row is the same kind of fact, pointed the other way. | Hold the server, the exit, and the house still. Otherwise the difference has too many authors. |
| A loss mark, a voice hole, or a position snap is present on both paths | It did not. The symptom survived the path change. | The exit changed and the hole came with you. The tunnel is a weak explanation. | Name the shape on the path you will actually play. Flat, jump, or hole. |
| The figure jumps only while a backup, a call, or an install is sending | It did not address the queue. Small game packets still wait behind the send, in front of the tunnel. | You relocated the far side of a full uplink. The queue stayed in the house. | The evening-queue note. Pause the transfer with the tunnel held still. |
| The house is quiet and the same server is worse at the same evening hour on the direct path | Only as a different route around that hour, tested in a different sitting. | A calmer tunnel hour does not describe the direct evening path, and it does not identify which network was busy. | The ISP evening checklist, with the tunnel off. |
| An invite fails, voice is one-way, or the console test prints a strict label | Rarely as a path test, and not as a reading of the console’s own label. The label under a tunnel is the tunnel’s translator. | You added another address translation and then judged the old one. | The console NAT note, tunnel off, one vendor change. |
| The server picker, the resolver, or the cable changed in the same sitting as the tunnel | Nothing you can cite. The row is void. | You cannot tell which change the number followed. | Keep the void row. The next sitting changes one thing. |
Two paths can be ugly for different reasons in the same hour. A tunnel that reaches a region the direct path cannot reach can still show a loss mark once you are in. Write both facts. Reaching the server did not answer the hole, and a hole on the tunnel did not answer the direct path you have not played.
What do you write in one sitting, and what stays blank?
Copy the header into your own notes. One sitting is three passes kept close together: direct, one tunnel exit, direct again. Fill a row only after that pass. The first body row is a label so the columns have a meaning. It is not a household test, and it is not a result from this desk. The three passes start empty on purpose. A cell you leave empty is unread. Do not turn it into zero later, and do not let a spreadsheet turn it into zero for you. A word such as “quiet” means you looked. A blank means you do not know.
| Pass | Game and server shown | Tunnel | Game inside it? | What else changed | Delay shape | Loss mark | House sending? | Void? |
|---|---|---|---|---|---|---|---|---|
| Example shape | Picker, realm, map, or host | Off, or one exit if shown | Yes, no, or not shown | None, or name the second change | Flat, jumpy, calmer, worse, or unread | Quiet, ticking, or not shown | Idle, a named send, or unread | Yes if two things changed |
| Direct | ||||||||
| Tunnel | ||||||||
| Direct again |
What this desk has not measured, so the blank rows cannot be mistaken for our results:
- No game session compared with a tunnel off and a tunnel on.
- No before-and-after delay figure, and no published millisecond reading of either path.
- No loss mark, voice hole, or position snap compared across those paths.
- No console NAT label read on the direct path and again under a tunnel.
- No ranking of tunnel operators, exits, or products, and no claim that a particular exit is the right one for your line.
Troubleshooting steps
How do you compare one tunnel without spoiling the row?
- Write the complaint in one sentence before you connect anything. The region will not load. The delay figure is the complaint. A loss mark is moving. Or a party will not form. Only a path question belongs on this page. A party that never forms belongs on the NAT note. A hole in a match that already connected belongs on the delay-and-loss note, after you know which path you were on.
- Play the direct pass first, with the tunnel disconnected. Record the game, the server or host the client shows, and whether the house is idle. Read the shape in words: flat, jumpy, or a hole. If the client shows no loss mark, write “not shown.” Leave the delay cell blank if you did not look, rather than copying a number from a different match. Do not start a backup “so the line is realistic.” A send is a second variable. It belongs in the evening-queue note.
- Connect one exit and confirm the game is inside the tunnel. Hold the same server name. Hold the cable or the radio still. The cable comparison is the Wi-Fi note. Hold the resolver still if the product lets you. If connecting the tunnel also changes how names resolve, write that beside the row so you do not later cite it as a pure path test. Menu names move between products and versions. Trust the screen in front of you.
- Play a comparable moment and fill the tunnel row. Same mode, same server name, house still in the same state. If the server name changed, you changed destinations. That row can still be useful as a reachability note. It is not a comparison of one server across two paths. If you cannot tell whether the game used the tunnel, stop and leave “not shown” in that cell.
- Disconnect and play the direct path once more. The return pass is what tells you the first direct row was not a one-off. If the return pass disagrees with the first direct pass, write both and do not average them into a story. If anything else changed during the sitting, including a game update, a matchmaker move, or a housemate’s upload, mark every row in the sitting void and keep them. The next sitting starts clean.
This page does not name a product to install and does not name an exit to pick. Whatever client you already have, the log is the point. We have not timed the common ones against a direct path, and the August file cannot stand in for that test. Command lines for a Windows gateway echo already sit on the methodology page. An echo to a public address is not your match, and an echo that happens to travel through a tunnel is a reading of that echo’s path only.
When to switch measurement method
When do you leave the tunnel and open a different note?
Leave for the delay-and-loss note when the session is up and the thing you can point at is a flat number, a jump that returns, or a hole. Say which path that shape was read on. A tunnel left on during that naming makes the shape a tunnel shape. Turn it off if the complaint people actually have is the direct path. Do not do both renamings in one sitting.
Leave for the evening-queue note when the figure moves with a send. A tunnel does not sit ahead of your uplink. Pausing the transfer is the matching change, and it should not share a pass with connecting an exit. If the house is quiet and the trouble is the clock rather than a send, leave instead for the ISP evening checklist. That checklist wants the tunnel off so the hour you record is the direct path’s hour.
Leave for the console NAT note when the invite fails, voice is one-way, or a strict label is the sentence you can defend. Read the label the console shows with the tunnel disconnected. A label collected under a tunnel answers a different translator. It will not explain a delay figure on a match that already formed, either.
Leave for the DNS note when the name itself fails before a match exists. A tunnel that also swaps the resolver is the wrong instrument for that row, because it changes the path and the lookup together. Finish the lookup with the tunnel off, or admit the row is a bundled change and void it.
Leave for the Wi-Fi note when the only question left is the last hop inside the house. A cable plugged in during the tunnel sitting voids both notes. Plenty of console sessions have no cable at all. Write that limit rather than inventing a wired pass you did not run.
Stay on this page when you still do not know whether the game was inside the tunnel, when the only fact so far is that a region connects on one path, or when a calmer number is being treated as a repair and the direct row is missing. Stay when two exits were mixed and the sitting needs to be marked void before anyone cites it.
Leave network checks entirely when the picture hitches and every network mark the game shows stays still. A tunnel cannot draw a frame. Valorant, Tarkov, WoW, and Genshin are where a specific HUD word matters. Use the guide for the label on that client. Do not treat those pages as a second copy of this path comparison.
Do not cite the August 19 file as a game-path result or as a direct-versus-tunnel result. It remains one web-edge record collected with a tunnel up. A private first sort, before you spend the evening connecting exits, is still Diagnose lag. Pick the symptom that matches the sentence you wrote. The tool does not see your tunnel, and it does not store the log.
Which questions keep getting filed under a repaired path?
Does a lower number on a VPN mean the old path was fixed?
No. The tunnel is a different path, often with a different exit. A calmer figure describes that new path. It does not measure the direct path, and it does not repair it.
Can a VPN help when a region will not connect?
It can, when the session exists only after the tunnel is up and the server name is still the one you meant to reach. Write that as reachability on the tunnel. It is still not a loss measurement on the direct path, and it is not a NAT test.
Should a loss mark or a NAT label be read while the tunnel is up?
No. Those readings describe the tunneled path, and a tunnel often adds another translator. Loss and NAT have their own notes. Read them on the path the complaint is about.
If the number changes and you then turn the tunnel off, what did you learn?
You learned the two paths can differ, if the server name and the house stayed still. You did not learn a lasting repair. The direct path is the row with the tunnel off. Keep that row.
Has this desk published a VPN-versus-direct game comparison?
No. The August 19 file is a web-edge record collected while a tunnel stayed up. It has no direct-path control and it did not time a match server. The tables here are a decision guide and a blank log. An empty cell is not a zero.
Editorial log
November 5, 2026. First publication by Jay Fan. No tunnel-versus-direct game comparison is attached. The table is a decision guide. The log is blank until you fill it. Corrections: [email protected].