Wi-Fi and the neighbors' routers

[Fairman]

Wi-Fi and the neighbors' routers

Wi-Fi signals share a limited set of channels. In an apartment building, dozens of networks compete for the same few. Interference causes retransmissions, adding random delay.

On a speed test, the connection looks fine. The average bandwidth is high. But the latency varies: most packets arrive in 5 milliseconds, some take 80. That variation shows up as occasional sluggishness in the platform.

Evenings are worse, when everyone is online. Microwaves, cordless phones, and Bluetooth devices add noise.

The cure is a cable. Ethernet is not affected by neighbors, doesn't vary with the weather, and delivers consistent timing.

If wireless is unavoidable, using the 5 GHz band, choosing a less crowded channel, and placing the router close to the computer improve things. A modern router with good management of airtime helps.

The check is a ping test to a nearby server over several minutes, in both wired and wireless modes. The difference in the maximum and the spread of values is more informative than the average.

Most traders know this advice and ignore it. The first time a stop is delayed by a wireless hiccup tends to change that.

A small test removes the guesswork. Run a continuous ping to the broker's server for thirty minutes on Wi-Fi during the evening, then repeat on a cable. Chart both. The cable line is typically flat with a few small bumps. The Wi-Fi line often shows spikes and occasional dropped replies.

Even a modest USB Ethernet adapter, if the laptop has no port, makes a noticeable difference. And if running a cable truly isn't possible, powerline adapters or a mesh system with a wired backhaul usually beat a single router on the far side of the apartment.

What does a ten-minute ping test look like on Wi-Fi compared with a cable?

[Amanhasnoname]

I live in a block of flats and had exactly this problem without knowing it. My speed tests always looked great, but I kept getting fills that felt late. When I finally ran a ping to my broker's server for ten minutes on Wi-Fi in the evening, most replies were around 9 ms but the maximum was 140 ms and there were a couple of dropped packets.

The same test on a cable straight into the router showed a maximum of 14 ms with no drops at all. The average barely changed, which is why I'd never noticed before. It's the spikes that hurt, and they always seemed to land when the market was moving fast and I was clicking.

My trading desk is two rooms from the router, so I ended up running a long flat cable along the skirting board. Powerline adapters were my first try, but the house wiring made them nearly as jittery as Wi-Fi. A cheap flat cable fixed it for good.

[Shadow Trader]

Speed tests show 200 Mbps and everything feels fast. Then a trade is sent and takes two seconds. The reason is rarely bandwidth. It's jitter and packet loss.

Wi-Fi shares the airwaves with neighbors, microwaves, and everything else. Packets get retransmitted, and every retry adds delay. Average ping can look fine while the worst-case spikes, which are exactly what hurts an order, go unseen in a casual test.

A wired Ethernet connection eliminates most of that. The difference isn't visible on a download speed test, but it appears in a continuous ping test run for ten minutes: the maximum and the variation between pings tell the story.

Other contributors include a router overloaded by streaming devices, a cheap network adapter, and ISP congestion in the evening.

Tools like a continuous ping or an MTR trace to the broker show where loss happens, and whether it lies in the home network or further along the route.

It's the kind of improvement that costs a few dollars for a cable and changes the reliability of every trade.

What shows up when a continuous ping runs for ten minutes on the current connection?