Two connections can both read 45ms in the scoreboard and play nothing alike. One feels crisp. The other teleports enemies, eats your inputs, and drops you from the lobby mid-fight. The scoreboard number is identical because ping only counts the packets that made it home. The packets that vanished never show up in that reading at all. That gap between what your ping tells you and what your connection is actually doing is why so many players spend a weekend “fixing lag” and change nothing that matters.
Packet loss and high ping are two different faults with two different repair paths. Treat them as the same thing and you will tune the wrong setting every time. I have spent 20+ years in IT operations, including years on telecom infrastructure in Saskatchewan, and the single most common mistake I watched people make was reading one number, assuming it explained everything, and never checking the other. Here is how to tell the two apart in about five minutes, and the exact order of fixes for each.
Slow versus broken: the distinction that decides your fix
Keep one sentence in your head and most of this gets simpler: high ping makes a game feel slow, packet loss makes a game feel broken.
With high ping, everything you do still happens. It just happens late. You fire, the server registers the shot a beat after you pulled the trigger, and the target has already stepped behind cover. Frustrating, but consistent. The connection is a slow road, not a broken one.
Packet loss is different in kind. The data does not arrive late, it does not arrive at all. Your character snaps back to where it was a second ago. An ability button does nothing. An enemy stutters across three positions instead of moving through them. Because games run on UDP, which has no built in retransmission, a dropped packet is simply gone, and the game guesses to fill the hole. That guessing is what you see as rubber-banding and warping. This is also why a speed test can look perfect while your game falls apart: speed tests run over TCP, which quietly retransmits lost packets, so the loss never surfaces in the result.
| Symptom | High ping | Packet loss |
|---|---|---|
| How it feels | Slow but predictable | Broken and jumpy |
| Shots register late | Yes | Sometimes not at all |
| Rubber-banding / warping | Rare | Common |
| Random disconnects | No | Yes, at higher loss |
| Shows on a speed test | Sometimes (as latency) | Usually hidden |
| In-game ping number | High | Can look normal |
That last row is the trap. You can sit at a clean 40ms ping and still be dropping 4% of your packets, because ping never measures the ones that disappeared.
What ping actually measures, and the numbers worth caring about
Ping is the round-trip time for a packet to reach the game server and come back, measured in milliseconds. Lower is better, and the thresholds that matter in 2026 competitive titles look like this:
- Under 20ms: Excellent. No perceptible delay, ideal for ranked Valorant, CS2, or Marvel Rivals.
- 20 to 50ms: Good. Comfortable for any competitive play.
- 50 to 100ms: Fine for casual sessions, noticeable in fast shooters.
- 100 to 150ms: You will feel input delay in Warzone, CS2, and Battlefield 6, and you will lose close duels.
- 150ms and up: Effectively unplayable competitively.
Researchers at Worcester Polytechnic Institute have spent years measuring exactly how latency degrades first-person shooter performance, and the pattern they document is that aiming and shooting tasks stay tolerable up to roughly 100ms and fall apart past it, while movement is more forgiving. Their work on latency and player performance is a good reminder that “high ping” is not one problem. It hurts precise, timing-sensitive actions far more than it hurts general movement.
What packet loss actually measures, and why 1% is the line
Packet loss is the percentage of packets that never reach their destination. The severity bands are steep:
- 0%: Normal. What you want.
- 1 to 2%: Already noticeable in competitive play. Expect occasional rubber-banding.
- 3 to 5%: Serious. Warping, missed inputs, hit registration problems.
- 5% and up: Frequent disconnects, near unplayable.
The reason even small numbers matter so much: a study from the Institute of Information Science at Academia Sinica in Taipei found that once packet loss crosses roughly 1%, the average time players stay in an online session drops by more than half. People do not consciously measure loss. They just feel a game turn hostile and quit. One percent sounds trivial. It is the point where a match stops being fun.
There is a nuance worth knowing. Some WPI research on Unreal Tournament found that within the ranges typical of the era, packet loss had a smaller effect on the actual scoreline than latency did, because modern netcode is engineered to hide occasional loss. That does not contradict the point. It means loss is often invisible to your score right up until it is severe enough to feel awful, which is exactly why you have to measure it directly instead of trusting your kill count to warn you.
Diagnose which one you have in five minutes
Read the game’s own overlay first
Most current games expose both numbers separately, and this is the fastest read you can get. In Valorant, open Settings, then Video, then the Stats tab, and enable the Network Round Trip Time and Packet Loss graphs. In Warzone, go to Settings, then the telemetry menu, and switch on Packet Loss, Latency, and Network RTT. In CS2, enter cl_showfps 2 or net_graph 1 in the console. In Fortnite, enable Net Debug Stats under Video settings. You are looking for two independent values: ping in ms and loss as a percentage. If ping is high and loss is zero, you have a latency problem. If ping looks fine but loss is above zero, you have a loss problem. You can absolutely have both at once.
Run a route test, not just a speed test
A one-shot speed test hides intermittent loss. To catch it, watch the connection over time. On Windows, the free tool WinMTR (or PingPlotter) run against your game server’s IP for ten minutes while you play will show loss hop by hop, so you can see whether packets die inside your house, at your ISP, or deep in the route. For a quick local check, open Command Prompt and run ping 8.8.8.8 -n 100. Any “Request timed out” lines mean loss on your path.
Test for bufferbloat, the cause both readings can miss
Here is the case that fools people. Your idle ping is a clean 25ms, your loss is 0%, and yet the game spikes the moment someone streams Netflix or a cloud backup kicks off. That is bufferbloat: oversized router buffers that hoard packets under load instead of clearing them, and it is common. Measurements collected by the bufferbloat project and by router vendors put the share of home connections with meaningful bufferbloat above 70%, with latency-under-load spikes of 200ms or more. The free Waveform Bufferbloat Test is the reliable way to see it, because it measures latency while saturating your line rather than on an idle connection the way a normal speed test does. If your grade is a C or worse, bufferbloat is your real enemy, and no amount of DNS tweaking will touch it.
Ping still high after the free fixes?
WTFast reroutes your game traffic through a private network with fewer hops, which is exactly where inefficient ISP routing quietly piles on latency.
If it is high ping: fix in this order
Go wired. If you are on Wi-Fi, this is the highest-value change you can make. Wi-Fi adds anywhere from 5 to 30ms depending on interference and distance, and a direct Ethernet run to the router often drops ping by 20 to 40ms instantly. Cat5e is enough for gaming; Cat6 is fine if you are already buying cable.
Pick the right server region. Playing on a distant server adds 80 to 150ms for nothing. Valorant, CS2, and Fortnite all let you see or select region. Choose the closest one with a healthy population.
Turn on QoS. Router-level Quality of Service prioritizes your game traffic so a housemate’s upload does not push you to the back of the queue. On ASUS routers this lives under Adaptive QoS with a Gaming preset; Netgear puts it under Advanced.
Switch your DNS. Flush it with ipconfig /flushdns, then set your primary DNS to 1.1.1.1 (Cloudflare) and secondary to 1.0.0.1. DNS will not move your in-game ping much, but it trims connection setup time and occasionally shaves a few milliseconds if your ISP’s resolver is slow. See our DNS guide for the full picture.
Clear bandwidth hogs. Open Task Manager, sort by Network, and stop anything pulling more than a megabit or two while you game: paused Windows Update, background cloud sync, a Discord video call you forgot about.
If it is packet loss: fix in this order
Replace or re-seat the cable. A nicked or loose Ethernet cable causes steady 2 to 5% loss and is the single most overlooked culprit. Swap in a known-good Cat6 and re-run your ping test before touching anything else.
Check for the Intel 2.5GbE bug. This one bites a lot of modern gaming PCs and almost nobody suspects it. Intel’s I225-V and I226-V 2.5-gigabit controllers, which ship on many ROG, Aorus, and MSI boards, have a documented flaw tied to their interpacket gap timing that causes random connection drops and loss. Intel confirms the I225-V driver issue and hardware outlets tracked the I226-V connection-drop problem across Raptor Lake boards. The practical fix: install Intel’s current driver directly rather than the one Windows auto-pushes, then in the adapter’s advanced properties disable Energy Efficient Ethernet (EEE), and under Power Management uncheck “Allow the computer to turn off this device to save power.” Silicon made before Intel’s Q3 2020 revision cannot be fully fixed in software, but those steps stop the drops for most people.
Move off congested Wi-Fi channels. If wired is genuinely not an option, put your gaming device on 5GHz and manually set the channel to 36, 40, 44, or 48. In a crowded apartment those upper channels are usually the emptiest, and clearing interference eliminates a lot of Wi-Fi loss.
Disable offload features on flaky NICs. In the adapter’s advanced properties, set Large Send Offload V2 (both IPv4 and IPv6) to Disabled, and Interrupt Moderation to Disabled if present. On certain Intel and Realtek chips these settings trade a little CPU for a meaningful drop in packet loss.
Localize the loss, then escalate. Run WinMTR to 8.8.8.8 for thirty minutes. If loss shows up at hop 2 or 3, that is your ISP’s infrastructure, not your home. Screenshot the run and call them with a specific line: “I have consistent packet loss at hop 2 across a thirty-minute traceroute.” That sentence gets you past tier-one scripting and toward a line quality test faster than “my game is laggy” ever will.
Consoles: the short version
On PS5, wire it through the LAN port, set DNS to 1.1.1.1 and 1.0.0.1, and confirm your NAT type is 2 or better. On Xbox Series X, the network diagnostics screen reports packet loss directly under Advanced Settings; anything above 0% means start with a new cable and work up. The types of lag and how to diagnose each apply the same way on console as on PC.
When the problem is the route, not your house
Sometimes you do everything right and the numbers still stink. Clean cable, wired connection, zero bufferbloat, and you are still sitting at 80ms to a server that should be 30ms, or bleeding 2% loss that your ISP flatly denies. In that case the fault is usually not in your home at all. It is the path your data takes to the server. ISPs route for cost, not for latency, so your packets can zig-zag through five cities to reach a server one region away, and every extra hop is another place latency stacks up and loss can creep in.
This is the narrow situation a game proxy network is built for. WTFast runs your game traffic over a private network engineered for fewer, more direct hops, which sidesteps the inefficient public route where mid-path latency and loss usually happen. It will not repair a damaged cable, congested Wi-Fi, or bufferbloat inside your own house, so run every free fix above first. But once your hardware is clean and the trouble is clearly out on the route, this is the tool that targets that specific problem.
Losing packets your ISP won’t admit to?
WTFast’s gaming network avoids the congested public route where mid-path loss usually starts. Test it on your own game and server before you pay anything.
The whole game here is diagnosis before repair. Read both numbers, not one. Run a loss test over time, not a single speed check. Test under load, because bufferbloat hides on an idle line. Once you actually know whether you are fighting a slow connection or a broken one, the fix list above stops being a scattershot weekend and becomes ten focused minutes.