Have portal creators ever ridden in the mountain?

Simple question after having accomplished a bunch of portal climbs.

Though portal is definitely a nice idea, I have to ask this question. Passes in the mountains are difficult, that’s clear. But their slope is relatively regular. When I see gradients of 17% followed by a descent in zwift portal I first had to laugh… then to cry. I had to ask myself whether the developers have compared the climb they’ve created with the reality… or maybe they just have never been climbing passes IRL? For riding on “real climbs” I still prefer using other apps or even re-creating the climb myself, to make it much more realistic regarding gradients…

Am I the only one or do you agree?

Wouldn’t know; Portals are places the 99.999% of us can only dream to get to ride in the first place, so of course everything has to be taken with a grain of salt.

As far as we’re aware, they’re just GPX or some elevation data pulled from some source they will never speak of, but likely not produced internally.

So yes it’s a very safe bet that it is far from reality.
But this is just the case when it comes to “sim games”

Happens all the time even in sim racing games where they take LIDAR scans of tracks and things still don’t seem all that correct if you’ve ever been there before. (Most of this is true because race tracks vary month to month, year to year very greatly as they’re worn in)

So the reality is… there is no solution here; they’re working with whatever data they’re collecting. But as just stated, even the best of the best radar scanning for track geometry in sim racing… doesn’t quite do it.
All we know for certain, Zwift isn’t going and scanning these locations let alone putting sensors of any type near the real routes themselves. If they did; I imagine portals would be very different (and we’d probably be paying for unique locations)

43%!!! :man_shrugging:


Ok thanks for your answer. That’s also what I thought: taking some GPX data and incorporating it in Zwift. Sad they don’t use any type of smoothing and just throw it in Zwift without any “consistency check”


Yes, I’ve done a number of them in real life like Col Aravis. It’s not really the same, but it depends on the climb. I used the climb portal a few times then never again.

That is why it would be nice if community authors could create and upload real climbs with real scenery.

Imagine the other ways up to Alpe d’Huez being linked into Zwift ADZ - which by the way also doesn’t really reflect the real thing except in the number of bends (I’ve done the real thing a couple of times).

1 Like


Q: What is Climb Portal?

A: Climb Portal is your go to spot for a fun and challenging climb. We will have a rotating selection of real world climbs that will be simulated on Zwift with an all new look and feel.

I have not done any of these IRL, but this

is surely not right.

I trust the information at Zwiftinsider

“Zwift’s Climb Portal provides easy access to a rotating schedule of GPS-accurate virtual versions of famous real-life climbs.”

Perhaps anybody can compare the profiles?


Maybe it’s just me, but I’m afraid that I don’t understand what you’re getting at. You’re saying that it’s wrong that some climbs in the climb portal have a small “down” section part-way through the climb because real climbs don’t do that? Because some certainly do in real life.

Some corners on hairpins on real climbs have very varied gradients on one cross-section of the road at each corner, e.g. it’s ridiculously steep on the very inside, and the gradient gradually slackens off the further you go towards the outside of the hairpin.

Could you be more specific about the problem you’re seeing, please?


Of course a mountain climb is not regular at 7.2% during the whole climb. Of course, when the next k is at 7.2% average decline, there are sections at 4% and some at 10%. But on portal, in a k at 7%, you face gradients of 15% several times. I’ve done a lot of passes in France (Platzerwasel for instance) and I promise it is not as irregular. Of course some sections are steeper than other, but it is definitely not 12% for 100m, then -2% for 100m, then again 13% for 100m, 0% for 100m and so on.

Climbs are much smoother in reality. @Jurgen_Grusdat’s picture is exactly what I mean. In a 10% section you will face gradients of 12, maybe 13%. But you won’t face gradients of 20%, and even less 43% LMFAO.

In my opinion, zwift has taken elevation from a GPX route following the streets without any smoothing parameter. You may know Flamme Rouge. If you don’t apply any smoothing the profiles will look rather irregular with gradients that we see in the portal. Zwift should definitely apply some smoothing to avoid sections facing for example “50m at 20% and 50m at -4%” and smooth that as a much likely 8% section

1 Like

This picture is very interesting because it’s just before a curve. And that’s generally exaxctly were gpx maps completely miss the correct slope

1 Like

Yeah - that one HURT.

I cannot speak to any of the Col de or Alpe de climbs on Zwift 'cuz I’ve not ridden them but Old La Honda is in my yard and I’ve ridden it a few times in the last few decades. Rode it in the Climb Portal for the first time yesterday and it is a fairly accurate depiction of the 20 plus minute effort required. Turns and pitch are about right.
The only thing they get wrong and this is the case in nearly all of the climbs on Zwift is the gradient doesn’t get stiffer at the inside radius of the corners. This is why steering is such an advantage on the climbs because taking the shortest way around each bend doesn’t come with a penalty in grade. Other than that, OLH on Zwift is pretty close to OLH IRL without the pesky scenery or annoying chatter of your younger and more talented riding buddies as you struggle up the hill…

1 Like

Yeah this is going to be the obvious downside of how these are produced; maybe less apparent in my initial response.

They’re just taking XYZ points, and spreading out a “road width” from there.
You get some super messy results if the turns are too tight, and the radius of the turn is sharper than the width of the ‘road’ and you get super awkward clipping. (I think this is actually apparent in one of the portal climbs, but I don’t recall which).

That said this is entirely true for all of Zwift; and it’s technically fine… for our purpose at least; that the left and right sides of the roads are equal; not one side is steeper than the other as your example (which is definitely what we see outdoors). But in the gaming world with roads, and certainly in particular with Zwift… it doesn’t offer much (not to mention is massively more difficult to do / at minimum, far more time consuming).

It’s easy to say GPS correct, but you need whoever is working on that to go and ride the real climb and then say, right, GPS says one thing, but I rode it for real and I think certain parts need to be tweaked to match the real climb.

Taking from a GPX isn’t perfect - you have to correct them.

I’m fortunate to have ridden most of the big famous mountain passes from Albertville going south to Nice. Some of them have some pretty brutal sections but never seen 40% or more.

I’ve also found on my local very steep climb that the Garmin 1030 is slow to react to the gradient - it takes time for it to show the actual 16-18% section. Also noticed it overseas on Col Croix de Fer.

This is where we need user generated content and a download store to host user generated climbs with full scenery. The users will have more time to spend on a single climb to make it stunning.

Like you have the download store in MS Flight Simulator. It has things like airports, add on planes, scenery etc.

1 Like

Completely agree with you, and that’s what I’m wondering about: why Zwift does not correct the gradients (e.g. by smoothing the elevation)? To my mind that could (should?) be an important improvement for portal climbs

PS: regarding Garmin incline, I was a little disappointed. Gradient is slightly underestimated and it takes some time for “synchronization”. That’s the only moment where I miss my Polar… but that’s another debate!

I want to be the guy who Zwift pays to go out and ride the climbs IRL just to get the gpx file :star_struck:


Probably takes two riders doing that. Sign me up please…

1 Like

depending on how the (actual) road is engineered and laid out you may well see 40%… on the innermost 10cm of a switchback hairpin. irl you would ride further out, even a meter and there the gradient may be a more reasonable 15% or whatever. this type of corner design is common in british climbs, on mountain passes it’s often the case that a hairpin will be flattened to some extent to even out the gradient

i guess the point is you dont always get to choose what part of the road you ride on in zwift

On HR Alps they usually have two lantern rouge riders who ride alternate days. This year Adrian (some of you will know him, he’s been doing that for ages) and Paulo.

The stages are so hard that you wouldn’t be able to do that day after day. This year was even worse with the severe heat (40°C).

It would be fun to be the person doing the riding to test / check the GPS data. Provided you get me a car to go back down to make it easier.

I wonder if the GPS plot the use sometimes slightly cuts the corner and so it has unrealistic super steep gradients in places as you’re effectively cycling up the verge on the inside of the tarmac.


Not having ridden those climbs but just ones local to me in UK and in Mallorca, VERY steep VERY short sections, possibly nearing 40% do exist, but right in apex, which no-one would (could?) ride and it will be missed on normal gpx files. In fact I looked at some GPX files for a flat area near me and 20% gradients were shown