The difference in routing algorithms

[Wapping]

Just for a bit of fun and using near enough the same routing preferences for each, I asked three reasonably popular app’s to create me an A to B route from Chelmsford to Norwich. It’s a bit under 100 miles, so a decent enough morning’s ride.

The black line is MyRoute.

The red line is Beeline.

The magenta line is Kurviger.

With all the fuss over the XT, it would be interesting to see what it spat out.

 

[stever1]

Just for a bit of fun and using near enough the same routing preferences for each, I asked three reasonably popular app’s to create me an A to B route from Chelmsford to Norwich. It’s a bit under 100 miles, so a decent enough morning’s ride.

The black line is MyRoute.

The red line is Beeline.

The magenta line is Kurviger.

View attachment 450218

With all the fuss over the XT, it would be interesting to see what it spat out.

I take It the blue line is the XT ?
It looks the same as the MRA line.

 

[Wapping]

I take It the blue line is the XT ?

I don’t have my XT to hand.

 

[Wapping]

I don’t have my XT to hand.

Now I do.

It’s the green route, created in my XT, shoving the route creation algorithm slider thing across to maximum:

The dullest is arguably the MyRoute offering. That is not surprising.

Beeline and Kurviger are similar, at least in parts.

The XT’s is different but, not to bad as an alternative. It maybe puts the lie to “It is crap”.

All of them coukd be fine tuned, fairly easily.

 

[BillWright]

Would the MRA route change when you alter the vehicle type, with the bike route meant to be enjoyable. Or is that same option available on all the platforms?

 

[Wapping]

Would the MRA route change when you alter the vehicle type, with the bike route meant to be enjoyable. Or is that same option available on all the platforms?

Possibly.

Everything I read on the MyRoute forum, suggests that the app performs best in car mode. As I rarely, if ever create A to B routes alone, it’s not something I worry about too much.

 

[zerax9]

It would be better if you could tell us the distance and estimated journey time for each route. I notice that often the routing is based on fastest time, regardless of distance... so a 1 min difference may actually involve an extra 20 miles !!

 

[Wapping]

It would be better if you could tell us the distance and estimated journey time for each route…..

Please… you forgot, please.

To do that I have to convert the track imports from the XT, Kurviger and Beeline into routes, or I don’t get a time estimate. Here you are:

From left to right.

Green, Garmin XT, 106 miles, 3 hours 11 minutes

Black, MyRoute, 86 miles, 2 hours, 16 minutes

Magenta, Kurrviger, 99 miles, 3 hours, 21 minutes

Red, Beeline, 93 miles, 2 hours, 32 minutes

Please note:

A. The conversion from a third party track into a route in MyRoute is not always exact. I have tidied up the obvious glitches in the Kurviger offering and that from the XT.

B. The time estimates are all those given by MyRoute, following conversion of the tracks into routes.

C. I had to select and match the algorithm setting of each of the four as closely as I could to each other. This is not always possible as the choices offered up are not uniform across all four app’s. In essence, they were: Fastest time, excluding highways.

D. It’s all meant to be nothing more than a bit of fun and an example of the different results that are possible, even over an 80 or so mile (reasonably straightforward) line between Chelmsford and Norwich. I chose those two places, simply as I just about familiar with the roads. This helped me to assess whether the offerings were totally mad or not.

E. Whilst not bad, each had faults. This is maybe not surprising as I moved from one app’s offering into another app. Whilst GPX exchanges are meant to be perfect (using tracks helps) you can get anomalies. Very often this is simply down to each app having different base maps, so the original app’s track’s trails of breadcrumbs do not always match the second app’s roads. It might also explain why some owners complain about their XT, when running third party created routes. The device has simply done its best with what it was given.

F. Likewise, the XT offered up a sliding (increasing / decreasing) scale of ‘Adventure’ for the A to B route from Chelmsford to Norwich. I slid this to maximum ‘Adventureness’ (if there is such a word) which did make a difference. If bods are not altering the default position of the slider, it might explain, in part at least, why they are not happy with the A to B route the XT’s algorithm offers up. That said, there are several other variables (not least preference settings) which can also change results. I rarely use ‘Take me from A to B’ route creation, so it’s not something I have played about in too much.

G. That there are considerable difference in the routes offered up, that is not too surprising. That said, Beeline and Kurviger were not a million miles apart for long sections. I could have played around more in Kurviger’s default setting (it offers up assorted ‘twistyness’ settings) which night have bought the two closer together.

H. All of the offerings would have worked. In other words, had the rider followed the routes blindly, he would have ended up in Norwich. But the rider might have started shouting. The level of shouting might well have varied and / or been reduced if the fellow had played about with the route offered up. But that is something else entirely.

 

[er-minio]

It maybe puts the lie to “It is crap”.

The adventurous routing functionality (admittedly, I used it just 3 or 4 times since I have the XT but...) I think it's one of the few things that work well in its software

 

[John Armstrong]

F. Likewise, the XT offered up a sliding (increasing / decreasing) scale of ‘Adventure’ for the A to B route from Chelmsford to Norwich. I slid this to maximum ‘Adventureness’ (if there is such a word) which did make a difference. If bods are not altering the default position of the slider, it might explain, in part at least, why they are not happy with the A to B route the XT’s algorithm offers up. That said, there are several other variables (not least preference settings) which can also change results. I rarely use ‘Take me from A to B’ route creation, so it’s not something I have played about in too much.

Another query

For the XT, what is the result if you have the adventurous setting at a minimum, so assuming that you have the "fastest" option it should give you the least time? This I think is where you see the strangest results as the resulting route, as per JB's example is nowhere near what is logically the "fastest" route.

 

[Wapping]

Another query

For the XT, what is the result if you have the adventurous setting at a minimum, so assuming that you have the "fastest" option it should give you the least time? This I think is where you see the strangest results as the resulting route, as per JB's example is nowhere near what is logically the "fastest" route.

Good question.

I’d have to play about with it a bit to find out. That said, the differences might be amplified or reduced, depending on where you are going from (A) to where you are going to (B). What I mean by that is that in Alps, by way of extreme example, there may be less choice of roads for the algorithm’s sliding scale to work with. In north east France, with its lots of little roads, there might be lots of choices for the algorithm to use.

 

[Blue]

Have been experimenting with Chat GBT a bit with routes of late entering a question about a route between the two places this is what it came up with obviously needs a bit of playing about with.

Could you find me a route from chelmsford to essex taking in interesting places in between and show it on google maps

ChatGPT said:​

Thought for 35s

Nice — I’ve put together a scenic, interesting driving route starting in Chelmsford that takes you through a mix of gardens, historic towns, castles and classic “Constable country” riverside scenery, finishing in Colchester. I’ve also included a Google Maps directions link you can open to view the route and get turn-by-turn directions on your phone or desktop.

---

Suggested route (Chelmsford → Colchester) with highlights​

1. Start — Chelmsford — the county town; good place to begin with cafés or a quick walk in the city centre.
2. RHS Garden Hyde Hall — expansive, award-winning RHS garden on a windswept hill: great for a morning stroll among themed gardens. RHS
3. Hylands House & Park (Chelmsford) — grand neo-classical villa with parkland, good for photos, a picnic or a short walk. Hylands Estate
4. Coggeshall — charming medieval market town with timber-framed buildings and the “wool church” (lots of local history). Wikipedia+1
5. Hedingham Castle (Castle Hedingham) — one of the best-preserved Norman keeps in England; worth stopping for the keep and the grounds. Wikipedia+1
6. Dedham / Dedham Vale (Constable Country) — picturesque valley immortalised by John Constable’s paintings; riverside walks and classic countryside vistas. Wikipedia
7. Flatford (Bridge Cottage / Flatford Mill) — iconic Constable locations on the River Stour; great for photos and short riverside walks. flatfordandconstable.org.uk
8. Mersea Island — a pleasant coastal detour on the Blackwater estuary; oysters, sea views and a relaxed fishing-village vibe. visitmerseaisland.co.uk
9. End — Colchester — historic Roman & medieval city (Colchester Castle, museums, lively café scene) — a great place to finish the day. Wikipedia+1
10. 
    
11. 
    
12. Open this route in Google Maps
    
    Click / tap this link to open the route (Chelmsford → RHS Garden Hyde Hall → Hylands House → Coggeshall → Hedingham Castle → Dedham → Flatford → Mersea Island → Colchester) in Google Maps:
    
    

    Chelmsford to Colchester

    www.google.com
    
    
    (Opening that link will show the full driving route and let you get turn-by-turn navigation or modify the order of stops on the fly.)
    
    
    
    Quick tips
    
    • If you prefer coastal scenery first, you can reverse the route (start Chelmsford → Mersea → Colchester → then inland).
    • Allow time at each stop: gardens and castles typically take 1–2 hours; the small towns and Constable sites are easy 30–90 minute stops depending on how much you walk. (Check opening times for RHS Hyde Hall and Hedingham Castle on the day you travel.) RHS+1
    • If you want a shorter day trip, drop Coggeshall or Mersea to keep the route compact.
13. 
    

    ---

    
    If you’d like, I can:
    
    
    • Trim this to a half-day or full-day itinerary with estimated drive times and suggested lunch stops.
    • Convert the stops into a printable Google Maps list (so you can open it in the Maps app directly).
    • Replace the endpoint with Southend-on-Sea instead of Colchester if you’d rather finish at a classic seaside resort.

 

[Wapping]

Whilst it’s different from established A to B routing mediums, AI is starting to learn route creation and bashing many different bits of ‘stuff’ into a reasonably coherent offering.

We, amongst the last generation familiar with paper maps, are living in interesting times, where the science fiction of Star Trek (touch screens, ‘communicators’ like a flip phone, voice command “Hey Siri”, vast data storage ‘Clouds’ and heaps of ‘information’ about a place at the touch of a button} have become mainstream without us noticing. Virtual reality, will be the next big thing, enabling us to ‘walk’ into say a town or hotel and ‘see’ what it’s like.

 

[zerax9]

Please… you forgot, please.

It was just a suggestion for an improvement... I was trying to help... lesson learned.

 

[Wapping]

I asked ChatGPT for a motorcycle route between Chelmsford and Norwich, avoiding highways. I am sure if I’d refined the request it would easily have improved on it. The answer came back in around a secind or two.


Suggested Route Overview & Highlights

Below is a rough outline of a route with waypoints. You’ll need to adjust locally as road conditions or closures may vary.

Route (approximate, ~100‑120 miles depending on detours):

1. Start in Chelmsford
2. Head NE via B1008 / B113 toward Great Dunmow / Felsted
3. Pass through Stebbing, Finchingfield (a picturesque Essex village)
4. Continue toward Sudbury (crossing into Suffolk)
5. From Sudbury, take quieter roads through Hadleigh, Lavenham
6. From Lavenham head toward Stowmarket, bypass via rural roads
7. From Stowmarket go toward Diss, via smaller country roads
8. From Diss, take back‑roads through Harleston or Hingham into Norfolk
9. Then head north toward Wymondham and into Norwich

Highlights / notes:

• Finchingfield is often called one of the prettiest villages in England — well worth a short pause.
• Lavenham has beautiful medieval timbered houses and character.
• Between Diss and Norwich there are gently rolling roads through Norfolk countryside, avoiding the busier A11 where possible.
• You’ll cross from Essex → Suffolk → Norfolk, enjoying a variety of rural scenery (farmland, hedgerows, small villages).

You will use mostly A and B roads (single carriageway). You may need to tolerate short stretches of more primary A roads to connect segments, but the idea is to avoid motorways and high-speed dual carriageways.

It was just a suggestion for an improvement... I was trying to help... lesson learned.

 

[Simon_100]

Whilst it’s different from established A to B routing mediums, AI is starting to learn route creation and bashing many different bits of ‘stuff’ into a reasonably coherent offering.

We, amongst the last generation familiar with paper maps, are living in interesting times, where the science fiction of Star Trek (touch screens, ‘communicators’ like a flip phone, voice command “Hey Siri”, vast data storage ‘Clouds’ and heaps of ‘information’ about a place at the touch of a button} have become mainstream without us noticing. Virtual reality, will be the next big thing, enabling us to ‘walk’ into say a town or hotel and ‘see’ what it’s like.

Stone me you're dead right. For many years in the 'eighties we drove up to Norwich from Brighton at least one weekend every month and the choice was always A12 via Ipswich of M11/A11 via Cambridge - amazing how many new roads have been built since then

 

[JohnnyBoxer]

Another query

For the XT, what is the result if you have the adventurous setting at a minimum, so assuming that you have the "fastest" option it should give you the least time? This I think is where you see the strangest results as the resulting route, as per JB's example is nowhere near what is logically the "fastest" route.

Thank you
I preset my XT to shortest distance
No motorways
Put a destination in, some 15 miles away
Slid the Adv Routing & pressed calculate
It took me on a jaunt of some 20 miles via the main A roads, instead of across country on 2 lane C&D roads & the odd B road
I ignored it and set off and the actual cross country route was 13 miles (7 miles shorter than Garmin’s ‘shortest’ route & 8 mins quicker to boot)
Along my route of ignoring it, it tried to re route me back, the other way to the nearest A road and double the Mileage
That’s the reality
Going to try it again as a round trip in the unit and drop in some selected waypoints

 

[Wapping]

For the XT, what is the result if you have the adventurous setting at a minimum, so assuming that you have the "fastest" option it should give you the least time? This I think is where you see the strangest results as the resulting route, as per JB's example is nowhere near what is logically the "fastest" route.

I have played around in my XT, using the Chelmsford to Norwich route, simply by way of reference.

By default I have my device set to: Faster time, promoted recalculation, avoid U-turns, avoid unpaved roads.

To create the base route of Chelmsford to Norwich, I used the method shown in er-minio’s excellent tutorial:

This gave me a pretty direct route, up the A12 to Ipswich. Around the Ipswich ring road, then up the A140 to Norwich. In all, 80 miles in 1 hour 48 minutes. In short, what I’d expect.

I then touched the spanner in the top left corner and selected Route Preference. This brought up choices of routing for the route:

Faster time (this one was already selected as it’s my default setting)
Adventurous
Straight line
Shorter distance

I chose Shorter distance. This made no great difference at 80 miles in 1 hour 53 minutes. Not surprisingly, it is just about the same, simply as it’s basically just two roads between Chelmsford and Norwich.

I then touched the spanner again, then Route Preference again, then Adventurous. This brought up a route of 80 miles in 1 hour 48 minutes. In other words no different to the faster time route. But…..

I then touched the orange box in the upper right corner, next to Go! The device made a recalculation, producing a very different route of 106 miles in 3 hours 7 minutes. Have in mind that the standard ‘fastest route’ that the device offered up was 80 miles in 1 hour 48 minutes. The route is now 26 miles longer by distance but some 1 hour 19 minutes longer by time. That is a significant difference. The route now goes: Chelmsford > Great Dunmow > Saffron Walden > Haverhill > Bury St Edmunds > Diss > Norwich. It really isn’t too bad, going nowhere even close to Ipswich.

I then slid the slider control from maximum, one unit to the left. This made no difference.

I then slid it one more unit to the left. This altered the route a bit, giving 88 miles in 2 hours 24 minutes.

I then slid it again, right to the left. This gave 80 miles in 1 hour 47, with a route shape pretty much like the shorter distance option.

Conclusion / comment

A. Depending on the A to B journey, fastest route versus shorter distance, might make little to no difference. A paper map may well tell you why.

B. Adventure routing of the same A to B route, can make a significant difference. But, it might well depend on what roads are available for the device’s algorithm to chose from. On the Chelmsford to Norwich route, there are plenty of roads to the west of Chelmsford for the algorithm to work with. On a short A to B route, this potential may well reduce, obviously.

C. All the calculations and recalculations were quick and undoable.

D. Are the ‘Advenure’ routes any good? Arguably yes:

1. If you have the slider set to maximum and there are enough roads available for the device’s algorithm to work with.

2. Are they perfect? To a degree that might depend on how fussy you are. Some might make a silly detour, just to go around a roundabout. For what it’s worth, Kurviger can be very guilty of this, too.

3. If you expect the algorithm to take in every really small road between A and B, you might well be disappointed. But, if that is your expectation, then you’d be better off making yourself a bespoke route, rather than letting a dumb (but really quite clever) device do it for you. Or, learn how to alter routes from within the device itself, which really isn’t too onerous.

4. If you think your pre-XT Garmin device was better at it, then use that and pat yourself on the back.

My suggestion:

1. Try it for yourself. It’s free and you can’t break anything.

2. Maybe start with my Chelmsford to Norwich route and copy what I did, to see if your results are the same.

3. Try your own trial route.

4. Don’t moan…. The device can’t hear you.

PS My use of the words ‘you’ and ‘your’ is generic. It is not directed to you, John…. Or anyone else

 

[Wapping]

JB has kindly shared the troublesome route suggestion, spat out by his XT.

When I’m back from the wilds of Suffolk, I’ll take a look.

 

[Wapping]

JB’s route starts on the Market Square in Thirsk, finishing in Hovingham at the Bakery cafe:

A. Google

Three routes, of which the possibly most scenic is 18 miles in 31 minutes



B. Kurviger

Using the moderately twisty algorithm, it offers up:

19.95 miles in 35 minutes



C. MyRoute

Using the ‘avoid highways’ algorithm, it offered up:

18.94 miles

29 minutes



D. XT

Setting fastest time.

It offered up:

19 miles

29 minutes



I then selected the XT’s ‘Adventurous Routing’ and set it to maximum:

19 miles

28 minutes



In short it made no appreciable difference.

I then shaped the route, using the XT’s ability to shape routes.

18 miles

30 minutes



Conclusions:

1. I cannot definitively say why the XT didn’t offer up a more ‘adventurous’ route.

2. My guess (and this is a real guess) is that the algorithm struggles over comparatively short distances.

3. Why do I guess this? Because we saw that the XT offered up very different routes between Chelmsford and Norwich, but that is over 80 miles.

4. Shaping the route within the XT was easy.