How to Fix Packet Loss in Online Games

You peek, the enemy is not there, and half a second later you are dead anyway. The killcam shows them seeing you long before you saw them. Players blame aim, servers, or luck — but when the world teleports and your inputs feel delayed in bursts, the usual culprit is packet loss: some fraction of the data leaving your PC never arrives, and the game has to guess or rewind. The good news is that packet loss is measurable, and most of the time the measurement alone tells you which side of your router the problem lives. This guide walks through diagnosing it in the right order, with the numbers that separate normal from broken.

What packet loss actually is

Online games send data as small packets, typically dozens per second to the game server. A packet that fails to arrive gets discarded, and either the server or your client has to extrapolate. A few losses per minute are invisible. Losses that cluster — 2% or more sustained, or bursts where several arrive together missing — produce the classic symptoms: rubber-banding, characters sliding, shots that register late or not at all, and voice chat that cuts out in syllables.

Two neighboring metrics matter alongside it. Latency is the round-trip time of a packet; jitter is how much that time varies. You can play well at 80 ms with 2 ms jitter. You cannot play well at 40 ms with 30 ms jitter, because bursty delay behaves like loss — packets arrive too late to matter. Diagnose all three together.

Step 1: Measure against a real target, not a webpage

Open Command Prompt and run a long ping against something stable:

ping -n 100 1.1.1.1

That sends 100 packets, about 100 seconds of data. Read the summary at the end: Lost = X (Y% loss). The numbers to keep in mind:

• 0–0.5% loss: effectively normal; look elsewhere if symptoms persist.
• 0.5–2% sustained: noticeable in shooters and rhythm games; worth fixing.
• Above 2%, or bursts of 5+ in a row: actively broken; keep diagnosing.
• Jitter (the spread of times): under 5 ms is good; over 20–30 ms is a problem even with low average ping.

Then repeat the test against the game's own region if you know an IP or hostname for it, because the path to your game server is not always the path to Cloudflare. For a deeper view, pathping (built into Windows) or the free WinMTR tool traces the route and reports loss per hop over minutes — this is how you find out whether packets die inside your home, at your ISP's first router, or further out.

Read the timing summary too, not just the loss line. If times look like 12ms, 13ms, 14ms, 12ms, the connection is stable. If they alternate wildly — 12ms, 180ms, 11ms, 240ms — you have a jitter problem even though the loss line reads zero, and everything in the bufferbloat and Wi-Fi sections below applies. Run the test twice: once with the house quiet, once while someone streams video in another room. A clean test that collapses under load is the signature of bufferbloat, and it is the single most fixable cause on this list.

Step 2: Rule out your own hardware first

Wi-Fi is the single most common source of game packet loss, and the tests above will show it. Wireless retries, interference, and roaming between access points all discard packets. Two quick experiments separate Wi-Fi problems from everything else:

1. Move the gaming PC or console to a wired connection (or run the long ping again on a laptop plugged in by ethernet in the same spot). If loss disappears on wire, the fix is cabling, not your ISP.
2. If you must stay wireless: use the 5 GHz band with a strong signal, keep the device in line of sight of the router, and pause any large transfers during the test. Mesh nodes and extenders add a wireless hop — gaming on the node's backhaul is a common hidden cause.

Also check the wired path itself: a failing ethernet cable or a negotiated 100 Mbps link instead of gigabit shows up as intermittent loss. In Windows, Settings → Network & internet → Ethernet shows the link speed; it should read 1 Gbps on any modern equipment. Swap the cable before blaming the network card — C7 CRC errors in SMART-style logs are the networking equivalent of a bad SATA cable.

Step 3: Catch the uploads

Bufferbloat is packet loss caused by your own queue: when something saturates your upload, every other packet — including game traffic — waits in line. This is why the game turns to rubber-banding precisely when a cloud backup starts, someone uploads a video, or a torrent seeds. Two fixes, in order of effectiveness:

• Find and throttle the sender. Every cloud sync tool has an upload-limit setting (OneDrive: right-click tray icon → Settings; Steam: Settings → Downloads → limit bandwidth). Capping uploads to 70–80% of your measured upload speed keeps the queue short.
• Enable SQM/QoS on the router. Routers with Smart Queue Management (OpenWrt's sqm-scripts, Ubiquiti's Smart Queues, many ASUS and FRITZ!Box models) keep latency flat under load. Test the difference on the free Waveform bufferbloat test: an A grade means latency under load barely moves; an F means your router queues hundreds of milliseconds of packets during uploads — game-breaking even at 0% "loss" when idle.

Step 3.5: The house full of other devices

A single gaming PC is rarely alone on a home network, and the household's collective behavior shows up in your ping test:

• 2.4 GHz congestion. The 2.4 GHz band has three non-overlapping channels and every neighbor's router fights for them. Smart bulbs, wireless cameras, baby monitors, and older laptops all chatter there. Games belong on 5 GHz (or wire); if your router broadcasts one combined SSID, split the bands so you can pin the gaming device to 5 GHz.
• An aging router. Consumer routers accumulate memory leaks and thermal problems over years. Symptom: loss that appears after days of uptime and vanishes after a reboot. If a weekly power-cycle reliably fixes your evening loss, the router is asking for replacement or a firmware update — check the manufacturer's page once, update once, and stop rebooting as maintenance.
• Cable-line splitters and modems. On cable internet, every cheap splitter between the street and the modem degrades signal. Unexplained loss that survives wired tests and SQM, especially with slow upstream, is worth a technician visit — ask the ISP for signal-level readings on the modem's diagnostic page.

Step 4: The things that look like fixes but are not

• Changing DNS servers. DNS resolves names once per session; it has no effect on in-game packet flow. It can fix "cannot connect", never rubber-banding.
• "Gaming accelerators" and VPNs for loss. A VPN adds a hop and usually adds loss. The exception is genuinely broken ISP peering to a specific game region, where routing through a different path sometimes helps — but measure before and after with the ping test above, and only keep it if the number actually improved.
• Lowering in-game graphics settings. That fixes frame times, not network. If symptoms correlate with heavy firefights on screen, read our stuttering guide instead — stutter masquerades as lag surprisingly often.

Step 5: When it is the ISP or the server

If loss appears only in the evening (roughly 7–11 pm), only on certain routes, or only in one game's region, you are looking at congestion or peering outside your home. Collect the evidence before calling: two WinMTR exports showing loss starting at or beyond your ISP's first hop, taken at the problem hour, plus a wired test proving your LAN is clean. ISPs can fix congestion — channel changes, line retraining on DSL, upstream repairs on cable — but only with data. If the game itself is the common factor, check the server status page and regional selection; switching from an auto-selected region to the nearest fixed one removes a transatlantic detour, and persistent loss on one provider's route is a support ticket, not a home fix.

The short version

Run a 100-ping test: under 0.5% loss is fine, above 2% needs fixing, and jitter above 20 ms ruins games regardless of ping. Go wired to eliminate Wi-Fi, cap background uploads or enable router SQM to kill bufferbloat, and use hop-by-hop traces to prove whether the problem is yours, the ISP's, or the game's. Fix the things the measurements indict — and skip the DNS myth entirely. For the other half of "feels laggy", our input lag guide covers the display-and-peripherals side, and the wired-versus-wireless arithmetic lives in this comparison. The full troubleshooting series is in the blog index.

Menu names and setting locations reflect Windows, common router firmware, and the tools named above as of 2026 and may move with updates. Thresholds quoted are practical rules of thumb from community diagnostics, not vendor guarantees.