Inconsistent routing, Braintree MA

[rogerl50]

Consider the two following permalinks, grabbed about 30 seconds apart:
1. https://www.waze.com/livemap/?zoom=12&l ... &to_seg=-1

2. https://www.waze.com/livemap/?zoom=12&l ... &to_seg=-1

The second link is to a route with the same destination as the first, but a starting point partway along the recommended route from the first link.

Magically, a new best route shows up, better than both of the routes recommended in link 1, that did not show up before. I have obviously noticed this in practice as well.

Any ideas?

[CBenson]

First thought is that, despite the name, the permalinks don't actually permanently record the routes, just the start points and the end points. Each time you open them the routes are recalculated based on the current conditions. So what we see might not be what you saw.

Given that I'm not seeing a significantly better route in the second link. Waze typically routes over regular streets only at the beginning (or end) of routes. So it is not uncommon for a route that starts latter to use some local streets near the start point that the longer route doesn't take.

[RallyChris]

Permalinks will change over time, because they are not snapshoting the roads but the start and end points. When I'm viewing your permalinks, the "first" route is the same.

My theory is that you got a response from two different servers, and the traffic data was not synced the same on both servers, or you hit it at the magical time change between average speed slices in the data.

If this product didn't use traffic data, you could make a better assumption that the route would be the same no matter when you request it.

[rogerl50]

All I know is that every time I use this route, up Burgin Parkway, the system routes me straight through. And every time I make the left turn to use the route I like, the system immediately gives me a route shorter than the one it started with. And it's not a map problem, because I can get it to give me that route by changing the starting point.

How are we supposed to get the system to improve if we can't post routings? This is ridiculous. Perhaps everything you've said in the posts above is true, but the fact remains that I see this behavior not infrequently - when routing from point A to point B, I get a suboptimal route at C, which is on the preferred route. When I do something different at C, I immediately get a better route to B than the one I originally had. It cannot be that a) every time this happens to me, there's a map error at C, or b) every time this happens to me, traffic conditions have changed, or c) every time this happens to me, I get a more friendly server after I do something different at C.

How would you suggest I go about getting to the bottom of this? I can paste screen shots if that would help. But the current situation is impossible. I really like Waze, but I fail to see how it can succeed if it's impossible to really have a conversation, supported by data, about its routing choices.

Roger Lipsett

[rogerl50]

Given that I'm not seeing a significantly better route in the second link. Waze typically routes over regular streets only at the beginning (or end) of routes. So it is not uncommon for a route that starts latter to use some local streets near the start point that the longer route doesn't take.

I don't get that. How is that a good routing choice? It will not give me the best route. And, even if it's so, why doesn't Waze reroute when I get closer to the intersection in question?

[bgodette]

And it's not a map problem, because I can get it to give me that route by changing the starting point.

That's a massively faulty assumption that many people unfortunately make. The only way to say it's "not a map problem" is to have edited every junction along the route and adjacent to it, and to have correctly Typed the segments. Simply checking to see if there's green arrows along the route is not enough as there's currently no visual way of distinguishing between soft and hard allowed/restricted turns without QW'ing each junction.

That you're getting different results based on starting point*, is a big indicator that the junctions involved have not been fully edited and that there's still soft allows/restricted. Soft allowed/restricted turns cause all sorts of odd routing due their apparently low penalties. And yes that means soft-allow has a penalty vs hard-allowed, or at least that's what testing seems to show.

*caveat: really long routes > 200miles are still unstable, especially if there's data coverage gaps, see this post.

[rogerl50]

[Soft allowed/restricted turns cause all sorts of odd routing due their apparently low penalties. And yes that means soft-allow has a penalty vs hard-allowed, or at least that's what testing seems to show.
this post.

Can you direct me to a piece of documentation that describes what these are? First time I've heard the terms soft and hard allowed. And is there no way to distinguish from looking at the map (old or new editors) which is which?

In my case, however, are you saying that kind of issue might cause different routings when I'm close to, but not yet through, the intersection in question as opposed to when I'm far away?

Thanks.

[bgodette]

[Soft allowed/restricted turns cause all sorts of odd routing due their apparently low penalties. And yes that means soft-allow has a penalty vs hard-allowed, or at least that's what testing seems to show.
this post.

Can you direct me to a piece of documentation that describes what these are?

Here (spend 4+ months reading) and indirectly in this video (gray turn restriction arrows). IOW there isn't any, but in any case each turn has *at least* four possible states, each with a different cost associated with it. Those states are "restricted by editor", "restricted from no turns having been seen from user drives", "allowed from user drives", and "allowed by editor", there may be others, but at a minimum those exist.

"Restricted from no turns having been seen" is the default state of all turns of all junctions from a basemap import, and is potentially also the default state of newly created junctions from editor action or paving. There's testing going on to determine the latter.

First time I've heard the terms soft and hard allowed. And is there no way to distinguish from looking at the map (old or new editors) which is which?

Until they add the gray arrows to the editor you can only test for existence by QW'ing a junction. If any segment turns orange there was one or more turns originating from that/those segment(s) that was not manually set by an editor, including reverse connectivity.

In my case, however, are you saying that kind of issue might cause different routings when I'm close to, but not yet through, the intersection in question as opposed to when I'm far away?

It certainly can.

[rogerl50]

I was hoping to figure this out from context, but...what does QW'ing a junction mean?

[gettingthere]

They are shortcut keys.

Q = disable all turns
W = enable all turns

Keep in mind that you should only enable all turns when all turns, from every segment are permitted at an intersection.