Measuring…
The verdict for this game appears once the first full second of packets is in.
Comparing with other tests to this server…
Measuring…
The verdict for this game appears once the first full second of packets is in.
Comparing with other tests to this server…
Are you sure you want to delete all results?
| Date & Time | Duration | Avg Ping | Avg Jitter | Packet Loss | Actions |
|---|
Hunter pounces that land from nowhere? Charger hitting you around a corner you had already turned?
Left 4 Dead 2 runs on Source, which means it exposes more of its networking to players than almost any modern game - and it also means the numbers it exposes are routinely misread. Source separates several things that everyone else lumps together as lag: packet loss, choke, and the interpolation delay the client adds on purpose. Knowing which one you are looking at is the difference between a real fix and copying console commands off a forum.
What this test sends: 30 packets per second of 150 bytes, matching Left 4 Dead 2 rather than a generic ping. verified
Valve's Source servers run L4D2 at 30 tick by default; visible in net_graph and settable by operators.
Run the test above for at least 60 seconds. Here is what each number means for this game specifically.
Source reports loss and choke separately, and they mean different things. Loss is packets that were sent and did not arrive - your line. Choke is the server holding packets back because it cannot send at the rate requested - the server. This test measures the first, which is the half you can actually act on.
The client interpolates between received snapshots, which means it deliberately renders slightly in the past so movement is smooth. That buffer covers a certain amount of unevenness and no more. Jitter beyond it shows up as the stuttering special infected that this game is known for.
The default server rate here is well below a competitive shooter's, so a tick is around 33 milliseconds and the practical resolution of the game is coarser than in CS2. That is fine for the game it is - but it means a pounce or a charge is decided on a comparatively slow grid, and your position on it matters.
What you see in game, and which number on this page explains it.
| What you see in Left 4 Dead 2 | Look at | What it means |
|---|---|---|
| Pounce lands with no visible approach | loss | Frames of the attacker's approach never reached your client. |
| net_graph shows choke, not loss | server | The server is holding packets back. Your line is not the cause. |
| Special infected stutter as they move | jitter | Unevenness beyond what the interpolation buffer can absorb. |
| One player's sessions are always worse | server | A local server puts that player's connection in charge of everyone's game. |
Valve's Source engine came from an era when players expected to be able to see and tune their own networking, and Left 4 Dead 2 inherits all of it. The on-screen graph reports several distinct quantities that modern games hide behind a single connection icon, and because they are visible, they get blamed - usually the wrong one, and usually with a console command copied from somewhere.
The distinction that matters most is between loss and choke. Loss is the familiar one: a packet was sent and did not arrive, somewhere on the path between the server and you. Choke is not that at all - it is the server deciding it cannot send at the rate the client asked for and holding data back. The symptoms overlap, the causes do not, and no setting on your machine changes the second one.
Interpolation is the other commonly misunderstood value. The client does not draw the newest snapshot it has received; it draws slightly behind it, so it always has two known positions to move an entity smoothly between. That deliberate delay is what makes other players' movement look continuous instead of teleporting between updates. Players who set it as low as possible in search of an advantage generally get stuttering instead, because they have removed the buffer that was absorbing their jitter.
Underneath all of it sits a comparatively slow server rate. At thirty ticks a second, each tick is around 33 milliseconds, and the game's decisive moments - a pounce, a charge, a smoker's tongue - are resolved on that grid. This is not a flaw; it is appropriate for a co-operative game from that generation. But it does mean your client's picture is coarser than in a modern competitive shooter, and that a lost packet removes proportionally more of the story.
If the chart above shows packet loss or high jitter, work through these in order.
The Symptom: Some sessions play far better than others.
The game can put you on a dedicated server or run the session on one player's machine, and the two behave differently. A local session means somebody's home connection is the authority for everyone. If your campaigns are inconsistent, this setting is the first place to look.
The Symptom: You suspect a network problem but cannot name it.
Source's built-in graph shows loss, choke and the interpolation delay as separate numbers on screen while you play. Together with a measurement of your own line from outside the game, that is enough to identify almost any connection problem here without changing a single console variable.
The Symptom: Pounces and charges landing impossibly.
Special infected attacks are short, decisive and resolved on a coarse tick. Wireless bursts of loss remove exactly those decisive moments from your client's picture, which is why they feel like they come from nowhere. A cable is the fix and it is not close.
The Symptom: Bad campaigns when someone else is streaming.
Source's traffic is small and continuous, and a saturated upload queue delays it in bursts. Router QoS that prioritises interactive packets removes an effect that looks identical to a bad server.
Loss counts packets that were sent and never arrived, which is a property of the path between you and the server. Choke counts packets the server held back because it could not send at the requested rate, which is a property of the server. They need completely different responses.
It is the interpolation delay: the client deliberately renders slightly behind the newest snapshot so it can draw smooth movement between updates. It is not lag being added by accident, and setting it too low makes movement stutter rather than making you faster.
Because they are fast, decisive events resolved on a comparatively coarse tick, and your client is drawing interpolated positions slightly in the past. Packet loss removes frames of the approach entirely, which is what makes a pounce appear to come from nowhere.
It is different rather than automatically worse. A local session depends on one player's home connection and machine for everybody; a dedicated server does not. For a group spread across a continent, dedicated is usually the more consistent choice.