Indeed - I should have made this clear, thanks dockb. To further clarify doctorkb's statement with my previous example:

Hwy 1 segments should be named "Hwy 1" in both directions. A ramp with signage that includes text such as "Hwy 1 E" would be named "to Hwy 1 E", as per current guidance for ramp naming.

e-sarge wrote:So, at this point, we are looking for any real-world examples where the road does not have cardinal directions, and BDP is clearly preventing the turn-around-and-go-back. Once/if we have some good examples, we could then try adding the cardinal directions to see how it changes the BDP.

So, what would your significance threshold be? 1 case in 5? 1 case in 10? 1 case in 100? For example, I've just tried 5 or 6 separate examples on Hwy 1 (which has no cardinal directions) through Vancouver using Route Checker in WME, and in each case, I was cleanly exited, looped around in the fastest way possible, and sent back down the highway in the direction I came from i.e. no BDP. At least at a first glance, if BDP is indeed invoked by having the same name on both directions, it's only in exceptional circumstances. Are we making mountains out of molehills?

How about this:

This is the route you want: Permalink.

You miss the exit, and are rerouted: Permalink

Two routes are now being suggested, with the second one being the next-exit-u-turn-back-on-the-highway route. To me it looks like BDP is at work here, because the #1 route is longer by 4 minutes (currently, at any rate, 11 vs 15 minutes) i.e. it looks like a BDP penalty is being assigned to the second route.

Or could there be other causes, like road type transition penalties or others? We could use this as a test case, whereby renaming the segments of Hwy 1 to include the cardinals (or adding Hwy 1 as alt-name to ramps/overpasses) we could see whether the routing order is reversed .

I suspect in my previous tests I was selecting ramp segments as start/end points to force routings in such a way no other routes were faster, even with the BDP penalty. It's tricky to pick start/end points that are appropriate.

Reading the current BDP wiki page, it is explicitly stated in the notes that BDP will apply to >1 segment u-turn routes on divided highways with the same name. This example supports that, I think. I may be starting to be swayed to e-sarge's point of view... or maybe more to dockb's alt-name solution for ramps and overpasses to maintain continuity, although that would be a lot of work...

Well in a case like this, if it makes a five minute difference, then yeah, I'd want to be routed around and back on the highway. We know BDP is a time penalty. For a long enough distance to the next exit, even with the penalty the return route will likely use the highway. It's for cases like the one above where it gets tricky. In a post last year, CBenson mentioned the BDP penalty was thought to be around 5 minutes. If it's still in this ballpark, then for most urban freeways, BDP is likely going to play its part when a wazer misses an exit, because the next exit is likely not very far away, and the difference between the freeway return route and the local road route isn't likely to be that great.

Whether or not it really matters is a whole other question and here you're likely to get lots of difference in opinion. For me personally, I use the waze app to save time, period. That's it. No other reason. So yeah, if it sends me on a route that is five minutes longer than getting back on the highway, it's not doing its job as well as it could. Seen that way, I would like to find a solution that disables BDP on these missed exits scenarios. Both the solutions presented seem workable i.e. rename the divided highways to include cardinal directions OR add the highway name to the alt-name of each exit's ramps and overpasses to ensure continuity of name.

@Nagamasa: Did you look at my example? It was precisely the same scenario as yours. At any rate, I am in agreement with you that the missed exit scenario is likely a common one, and the BDP penalty should be avoided if at all possible to ensure the most direct re-routing. Short of Waze tweaking the BDP algorithm to recognize this type of scenario, using cardinal directions on the divided highway segments seems the most direct solution, certainly easier to implement than the overpass/ramp alt-name solution. That said, to me it's just as much of a hack as the alt-name proposal: it's hacking the map to obtain desired results rather than fixing the BDP algorithm. But, as we have control over the former but not the latter...

One quick observation: if the BDP penalty is still on the order of about five minutes, then for those more rural situations where the exits are spaced far apart are actually less likely to be impacted because the alternate side-street routes will have t > highway route + 5 (or whatever BDP penalty is). That is to say, even with the penalty the u-turn highway route will still be the shortest and therefore selected as best routes.

Also agree that cardinality hack level is < alt-name hack level.

I'd suggest the example I posted in the first page of this thread is a good one to experiment with as it is consistent and I think fairly convincingly shows evidence of BDP-affected routing.

https://www.waze.com/livemap?lon=-122.5 ... t_text=Now

This is the missed exit scenario. The back-on-the-highway ("BOTH") route is 4 or 5 minutes shorter than the back roads ("BR") route, but is the #2 route on the Live Map list, indicating that BDP is very likely at work. A wazer in this scenario would be routed along BR route even though BOTH route is 4 minutes faster.

I propose to try the alt-name strategy for the ramps and overpass. Expected behaviour if BDP is nullified is that once the tiles update, the same Live Map route should invert the order of proposed routes, with BOTH #1 and BR #2.

dockb, you okay with me trying this?

codgerd wrote: I propose to try the alt-name strategy for the ramps and overpass. Expected behaviour if BDP is nullified is that once the tiles update, the same Live Map route should invert the order of proposed routes, with BOTH #1 and BR #2.

I have updated the appropriate segments so that each segment has "Hwy 1" either as its primary or its alternate name. As per the behaviour described on the BDP/USA page, this should be sufficient to prevent a name-discontinuity.

Tiles have updated.
https://www.waze.com/livemap?lon=-122.5 ... t_text=Now

PRE alt name routings:


POST alt name routings:


Behaviour is as predicted: BDP has been suppressed and the BOTH route is now offered as #1 route, with BR #2, with routing times consistent with what they were pre-edit (some real-time traffic currently affecting my live map routings by 1 min or so, but this is inconsequential...)

I think this is strong evidence that the non-cardinality of Hwy 1 is affecting routing for missed exit scenarios in circumstances where the BR route is less than the BOTH route + the BDP penalty. From what I can gather, this penalty is still thought to be in the vicinity of 5 minutes. Thus, we can expect this behaviour to crop up for many missed exit scenarios where the exits are spaced apart somewhere in the vicinity of the example's.

I'll leave this config for a day or two, then will remove the alt-names from the affected segments to confirm that the routing order inverts again.

I think this evidence is strong enough to suggest routing is being affected in a significant and systematic way, as per my original post on this thread. It means that re-calculated routes in missed exit scenarios can be up to about 5 minutes longer than the optimal route. Whether this is significant enough to reconsider the policy of not having cardinal directions on divided freeway segments I'll leave to the CMs. (In my opinion it is significant enough.)

doctorkb wrote:I'm not sure we want to go the way of cluttering it by including cardinals, though the work of alt-naming all the ramps and overpasses (and also keeping those alt-names there from under-educated editors) may be more than it's worth.

The alt-naming scheme, as pointed out by (I think) e-sarge, might also have the unintended side-effect of possibly suppressing BDP when it was actually required. At the very least, careful analysis of each interchange/exit would be required to make sure behaviour was as intended. I'm not sure we can pull this off without a huge concerted effort, and certainly not without mistakes being made.

(Is it naive to think we could petition Waze to make alterations to the BDP algorithm? One additional test in the BDP algorithm could effectively take care of our problem: The directionality of the last segment before the name-discontinuity section and the first segment after the name-discontinuity section could be compared. If these were, say, 160 - 200 degrees apart, the interpretation would be that this is a back-on-the-highway in the opposite direction type of route, and BDP could be suppressed.)