There’s not always a better/faster route through Cloudflare’s network. Sometimes the best route can be the public internet. So in this instance it appears that the public internet is faster… by default I believe we go with a 90-10 split to test routes (we’ll route 10% of traffic over alternate routes to see if we can find a faster route). If those routes aren’t faster than you’d see a negative split for that traffic. Since you’re at slightly higher than 10% of traffic that could mean it is a relatively low amount of traffic or that for a small period of time/ a small percentage of traffic there was a faster route.
@eva2000 yes, we should probably add more statistics and it;'s something that is being worked on/ tested. It’s an interesting problem to display data in ways that are most meaningful and which best represent the overall performance impact as seen by the average user.
In this case for the same domain in the last 48 hours I’d argue the statistics have flipped the opposite direction at a global level (will let the OP share them if they wish) but still by region there is variance with one being ~= -34% and the other ~= +70%. Obviously in the -34% we send very little traffic over argo while in the other region we send a great deal of it.