Your connection is fine. Your game disagrees.
You're three deaths in and you pull up the speed test. Ping sits at 90ms, download looks healthy, upload is fine. And yet you keep getting killed by players who appear to teleport, shot by bullets that, on your screen, haven't left the gun yet.
Meanwhile, your video call with a colleague runs without a hiccup on the exact same connection. She's clear, you're clear, the conversation flows.
Same router. Same cable. Same moment in time.
So what's actually going on?
The answer lives in what each application needs from you, right now
Latency is the round-trip travel time for a packet of data: from your device, to a server, back again, measured in milliseconds. A 90ms ping means every single message your machine sends takes nearly a tenth of a second to get a reply.
For a video call, that's invisible. Human conversation has natural rhythm, people pause and breathe and think, and a tenth of a second of delay between your mouth moving and the sound arriving on the other end is indistinguishable from the ordinary gap between a thought and a sentence. Video calling apps also buffer aggressively. They store a small cushion of incoming audio and video, somewhere between 150 and 300 milliseconds worth, and play it back smoothly. The buffer absorbs the jitter. The conversation feels live. Nobody notices.
Online gaming is a different animal entirely.
A fast-paced multiplayer shooter runs its game world at a tick rate, often 64 ticks per second on standard servers, 128 on competitive ones. That means the server processes the state of every player 64 times a second, each tick a 15.6-millisecond window. Your inputs need to arrive, be processed, and their consequences sent back to you inside that window, or you've missed your slot. You can't buffer the future. The game has already moved on.
At 20ms ping, you're submitting inputs with comfortable headroom. At 90ms, you're consistently late to multiple tick windows. At 150ms, the game's predictive code is doing enormous amounts of guesswork about where you even are.
The teleporting player problem (and why it's not a bug)
Picture two players: Mia and Dev. Same game, bought the same day. Mia has a wired connection with 18ms latency. Dev is on Wi-Fi in a congested building, sitting at 110ms.
When Dev rounds a corner and fires at Mia, his client sends that shot to the server. The server receives it 110ms later, processes it, sends the result back. Dev sees the outcome 220ms after he pulled the trigger. In that same 220 milliseconds, Mia has potentially moved behind cover on her own screen.
The server has to decide who was where, when. Most modern games use a system called lag compensation: the server rewinds the game state to match what Dev saw when he shot, checks whether the hit was valid from his perspective, and awards it accordingly. Technically fair to Dev. From Mia's point of view, she got shot while she was clearly behind a wall. Both experiences are accurate. Neither is wrong.
The physics of light and copper just don't care about fairness.
This is why high-latency players feel like they're cheating even when they're not. The server is honoring their delayed reality. It's not a bug in the code. It's a consequence of the code working exactly as designed, which is somehow more frustrating.
Video calls have no equivalent of this problem because nobody needs to rewind time to figure out whether you said a word correctly. The audio either arrives or it gets concealed by a brief synthetic fill. The stakes are completely different.
What people get wrong about this
The most common mistake is treating bandwidth and latency as the same thing.
They are not even close.
Bandwidth is the width of the pipe: how much data can flow per second. A video call uses perhaps 3 to 8 Mbps. A multiplayer game uses perhaps 40 to 80 Kbps of actual data, which is almost nothing. Think of a game's network traffic as a stream of tiny urgent telegrams, not a river of video. Games don't need fat pipes. They need fast round trips.
You could have a 1 Gbps connection with 200ms latency and your gaming experience would be genuinely awful. A modest 25 Mbps connection at 15ms latency, though? Flawless. Speed tests that lead with download speed are useful for streaming and file transfers. For gaming, the ping number sitting in the corner of that same test is the only number that matters.
What most guides skip is this: Wi-Fi latency doesn't just run high, it runs inconsistently. A wired connection at 18ms stays at 18ms. A Wi-Fi connection might swing between 12ms and 85ms depending on whether someone microwaved something, how many devices just woke up, or the sheer chaos of a shared spectrum. That variance, called jitter, is arguably more damaging than a stable high ping. A steady 80ms is at least predictable. The game's code can account for it. Jitter is chaos, and game engines hate chaos.
So what can you actually do about it
First, go wired if you're serious. A Cat 6 Ethernet cable costs almost nothing and eliminates jitter almost entirely. That single change will do more for your online experience than any router upgrade you're considering.
Next, pick servers geographically close to you. Physics doesn't negotiate. Light travels through fiber at roughly two-thirds of its speed in a vacuum, so New York to London is approximately 70ms of pure physics, minimum, before any routing overhead. If a game offers server selection, choose the nearest region.
Finally, check your in-game ping display, not your speed test. Most games show your actual server latency in real time. Below 40ms, you're in excellent shape. Between 40 and 80ms is fine for most genres. Above 100ms in a fast-paced competitive game and you're fighting the network as much as you're fighting your opponents.
Video calls don't ask any of this from you because they were designed around human perception, which is forgiving and slow. Games were designed around machine precision, which is neither.
The connection isn't broken. One application is politely waiting for you. The other one already moved on, awarded someone else the kill, and logged the whole thing as correct.