I wouldn't test with wireless, but if wired is dropping, then upon drop you could run a traceroute (I think in Windows tracert, I normally do this in Linux). tracert is not perfect, as it uses ICMP protocol (which differs from your application), but it can show route failures at particular hops. You'd name your end target and see hops along the way. Hops which have ICMP disabled won't show a hop, but if ICMP does go further, then a further node along the path will still show up. You might find a target destination which is reliable and high bandwidth and expected to not fail, e.g., a google.com tracert or ping, since there is no way a backbone would fail along that route. If you were to go to a specific game server route, then it might not be testing public routes and might instead be part of the ISP at the game server.
You could open a terminal and set to infinitely ping (not ping rate, that's a "flood ping", but having it ping every second without ending). You could also set up a tracert to that address, but not actually hit the enter button until a disconnect. Hops along the route which work guarantee that part of the route is good. If it gets to your modem, then that part is good. If it gets from modem to the local junction box, then it is entirely not part of your household.
If a named address, e.g., google.com, cannot be resolved, then you have a DNS issue (assuming ping to dotted decimal addresses still work; if all route is cut then DNS has to die too). A DNS lookup shows one of Google's addresses is 142.250.72.46, so before you have a failure, you could try both of these to verify DNS and ping (adjust for infinite ping that never quits):
ping 142.250.72.46
ping google.com
tracert google.com
Note the address is IPv4 (not entirely important, but it doesn't test IPv6).
You could also find the address of some specific destination and do the same. Watch ping to the dotted decimal address (e.g., 142.250.72.46). If that still runs when connections fail, try using the named address (google.com) since that also tests DNS. Simultaneously run tracert...this is a series of pings to each route hop until getting to the destination. You might find that consistently some particular hop has a sudden loss of packets, but earlier hops do not lose packets. Should some particular hop be losing packets during a loss, then the ISP could use this to look at that hop. If this succeeds you will want to save a copy of the tracert and the time.
Incidentally, some of the most common failures are from rain or weather changes at some part of the route hop. The more interesting failures are sometimes due to cracks in shielding of above-ground wiring, such that aircraft radar can interfere...but only when overflying the failed location. In Colorado the Air Force academy training for AWACS provides their schedule to the Comcast cable company. When a node is found to fail only during an overflight, they've found old/failed shielding.