It’s an issue with all track toys and extreme track toys. I’ve sent idk how many tickets in regarding this. Every drivatar in track toy races seems dialed to 11. Drives me nuts.
For Track Toys '16 BMW M4 GTS and '18 Lamborghini Huracán Performante are Street Scene safe vs Unbeatable bone stock. Stock BMW has issues with sidewalks in Edinburgh Street so it’s good to adjust racing line accordingly. Lambo is all-arounder, also only AWD TrackToy IIRC. Anyway, that AWD explains a lot.
Stock ‘12 Porsche 911 GR2 RS is also very good for most situations.
There are some that turned out to be quite situational, like '16 Dodge Viper ACR. In stock configuration it has very poor acceleration after about 145 – 150 mph which renders it uncompetitive on high speed races. It was very good on more technical courses in Edinburgh, where top speed and acceleration past 120+ mph doesn’t matter that much.
As rule of thumb, cars that are safe in Street Scene / Unbeatable tend to be that also in Road Racing.
Well known Problem with Unbeatable A.I. especially in Streetscene. One of the Reasons i reduced the Experience Level of the A.I. to “Very experienced”. Less Rubberbanding, ramming , Break Checks and “Super Mario like Nitro Speed ups”. They definitely beatable at some Point and in some Tracks but they should rename the A.I. at this Level to “Cheating A.I.”.
1 Like
Well known according to who?
Not every car is equal but it’s not outright cheating. I did test run with selected bone stock cars from D-class to S1 vs Unbeatable. Results are here.
Some screenshots here: Pages 7 to 60.
For more recent cars, Lexus LFA is Street Scene safe and new Gunter Works Porsche Carrera 2. Ferrari Portofino can’t make it in the Monument Wynds if there are AWD’s ahead of it in starting grid, or at least I couldn’t. Highland Charge was a real challenge with it too.
I asked for event type and routes to see if I could repeat the results and see if it’s a glitch. It can happen in Street Scene that Drivatars can get ~6 seconds boost to their normal performance. Glitch can be identified by missing traffic and it can keep happening from race to race. So far it has been fixed by re-establishing connection to Horizon life. What connection has to do with single player, don’t ask me.
How outright has it to be to be outright?..
Glitch is very rare to occur, say at least that only very few players here have brought it up to begin with. Definition of glitch in this context is “game not working like it should” but I’m sure you know that already and are posing intellectually dishonest one liner that contributes nothing to solving OP’s problem just to flap your gums.
AI is outright cheating though, even on Unbeatable. That is very evident from the fact that you can take something like Dodge Hellcat, lot’s of horse power but weights as much as a tank and handles like one too. It has the looks and power and… is garbage as sports car but players can win with it. You can actually take almost every car there is and win with it against Unbeatable, despite PI difference, despite driving RWD vs AWD.
So yes, that’s Drivatar AI outright cheating, for players benefit. There.
For opposite, I just did playlist events vs. Unbeatable with stock cars. Pagani Zinque Roadster, Porsche 911 Carrera 2 Gunther Werks. For Street event I used '92 Honda NSX with body kit and it was boring. Did Street Scene run through with 992 Toyota Supra 2.0 GT Twin Turbo, won every event.
Didn’t see cheating AI anywhere.
There’s this thing with this whole Cult of Cheating Drivatars. Just like some UFO cult, I have never seen single screenshot, single video, no proof for it at all. I have found some genuine things but then nobody ever cares to replicate them so we could submit a ticket and get those inconsistencies fixed. Oh wait, cults don’t really care about serves themselves. Then there is this whole social aspect, members buying whatever garbage story from other members and those evil facts would ruin that all. How awful of me!
1 Like
This alone doesn’t guarantee that the glitch is unlikely to occur in certain cases.
Do you expect the players themselves to fix the glitch?.. Regardless, one probable workaround has been provided:
I don’t see how your reasoning is related.
Probably related with the car selection.
The AI’s performance seems to be directly related with the Suggested Driving Line, despite which seldom really reflects the car’s real performance.
When it’s underrated, alright, easy game; but when overrated…
High power RWD cars with tricky handling (not bad handling potential (or good even…) but hard to handle), mostly heavily tuned ones, tend to suffer the “overrated” problem.
2 most observable:
- Leader Speed Boost (to the 1st place when you reach the 2nd, ignoring the car’s real performance)
- Collision Immunity (just ram the Drivatars and see how invincible they are)
One thing I keep seeing mentioned from ClawedScroll is using “stock” vehicles in all of his tests. Why “stock” instead of being built properly to each PI (D, C, B, etc.)? If you are choosing “stock”, the A.I. will adjust accordingly.
Not many players run around in “stock” cars all the time, especially if they do any ranked races. I would like to see you pick a car or class (extreme track toys maybe) that has been mentioned to be an problem and build a car properly to fit, then do your tests and see what the results are. I bet there are a few differences.
Well, for Cross-Country races there is more than enough evidence about “cheating AI”. In the sense of unreachable race times which would decimate the #1 Rivals time.
Try to achieve this lap time in D500 winter…you won’t.
In Dirt, Road or Street I’ve never encountered it this extremely. There are some cars which are brutally fast off the line (Maverick for example) and I think it is safe to say the AI pace in Street is higher than in Road for whatever reason.
But so far I have to encounter a race in those 3 categories which isn’t winnable. In most cases you can win against “unbeatable” AI with every car (obviously broken ones like the Reliant Robin or Rimac not considered).
Cross-Country on the other hand can be outright unwinnable as my posted picture shows. The reasons for this are widely known (draw distance of obstacles, snowy/watery surface and so on).
1 Like
Like I wrote, it’s very easy to identify glitch in single player by traffic cars disappearing. Very few reports of that.
Another method is taking screencaps from end race results. Exiting, letting game connect, racing the same race again with same car and setup, screen capping results and comparing to previous screencap. Assuming player finished second or third in first race and top three in second attempt, if there’s a ~6 seconds boost for Drivatar time, there is a chance there is a glitch. So there would be evidence and case for ticket.
So far your method has been “umm… people feel like it so it must be true”. What the heck are devs supposed to do with that? I sure as heck wish Playground Games doesn’t “fix” anything based on feedback like that.
Oh, but it is, very. Like understanding the whole principle FH4 Drivatar AI is built around.
LOL no and it has nothing to do with racing line. See below.
Yeah, that happens. Usually doesn’t mean anything what comes to winning the race though.
There are tables there. I also provided link to screenshots. I have done my best to provide data and I’m not responsible for action or inaction of others.
EDIT: Quotes you pasted from users Jerakeen and grolchie aren’t data thread but entirely different topic related to Horizon season. You could have just write that not everybody reads long texts or if you have such an issue with yourself, instead of inserting quotes from unrelated thread from users whom probably even know their responses has been taken out of topic / context. So garbage in, but at least you practice what you preach.
My storage space at TrueAchievements is full but screencaps are published in my Xbox activity feed.
'15 Chevrolet Camaro (S1 818 stock and S1 900 upgraded).
Street Scene, Ambleside Ascent, Autumn (rain), Unbeatable, no aids, manual.
Stock: - 1st. 2:26.864
Drivatar 2nd: 2:27.249
S1 900: 1st. 2:09.897
Drivatar 2nd: 2:13.602
Win margin:
Stock: 0:1.615
S1 900: 0:4.295
Road Racing, Lake District Sprint, Spring, default blueprint (Autumn, Track toys), Unbeatable, no aids, manual.
Stock: -1st: 2:21.475
Drivatar 2nd: 2:23.644
S1 900: 1st 2:06.210
Drivatar 2nd: 2:08.880
Win margin:
Stock: 0:2.169
S1 900: 0:2.670
For Ambleside Ascent Unbeatable difficulty poses a challenge, strategical one. Drivatars can finish it couple secs faster but I leave winning strategy with car like that '15 RWD Camaro to be discovered for those who… like to play the game.
S1 900 is just a bunch of upgrades without any tuning and has limited stop speed at Lake District. With S1 900 Drivatar finishing in second position wasn’t even visible in minimap at some point. Huge boost happened just to keep it somehow in the race.
'15 Camaro setup share code: 936 177 983
I answered your question because benefit of a doubt, though I wonder why you didn’t tested it yourself if you are so curious.
Challenge by human players in ranked is something else compared to Unbeatable. Only worth scenario in single player compared to ranked is Unbeatable stock for two reason.
- Learn to work on routes, work arounds, best racing lines, which depends of car and isn’t always suggested one, with something that isn’t what you necessarily want but what you have
- Drivatar AI in FH4 compensates for bad tunes even on Unbeatable and thus may give misleading understanding of real competitive potential of car/tune for ranked.
So just FYI: Watering down Unbeatable difficulty would negatively impact / eliminate potential competition in ranked from players that are new to game.
Cross Country is broken for sure, there’s no question about it.
There is also issues in Street Scene with Peel vehicles that are still broken: Picture here.
Only workaround below D 407 is to check Drivatar lineup before race and if there are Peel vehicles in grid, go to Difficulty and keep swapping between Unbeatable, Pro, Unbeatable and see if game generates new lineup without Peel vehicles, which is convoluted and I doubt many players even know about that possibility.
Peel vehicles cut of from Drivatar lineup is D 407, which is VW Baja bug, which then isn’t competitive because most of its PI comes from lateral G’s. There’s also '67 VW Type 3 that that doesn’t appear to be competitive in Street Scene / Unbeatable. Haven’t tested in Road Racing or Dirt though.
Stock Maverick in Street Scene could achieve near top Rivals Leaderboard times at least against top B-class stock. Haven’t had issues with it during my latest Street Scene run throughs though.
What glitch are you referring to?..
Lack of logical connection for the conclusion.
“If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.”
Make the game what people feel like.
Where’s your comment?..
Regardless, they cheat.
…Pretty sure the very comments apply to your posts here.
I recommend you to try the “X 999” class. Or if you insist stock, try “2017 Koenigsegg Agera RS”.
Combining with this:
You can find bugs, trucks, SUVs, … running ridiculously fast comparing to those cutting edge performance toys (race cars designed for no race).
What do you mean really?
Easiest way to see how silly the AI gets is to do a race with only formula drift cars offroad, good luck winning.
No, it’s very clear. If there’s ~6 seconds faster time for fastest, winning Drivatar in results screen on Unbeatabe, that can be seen as evidence for a glitch.
Good that game isn’t developed on that principle, horrible things has happened when people thought they saw a duck.
People like who? For me the game works, for some it doesn’t. That’s like every game ever.
Not my responsibility, besides I have always gone after features, not what is written to the box. You got the developer studio wrong BTW, it’s Playground Games.
You somehow, totally not on purpose appeared to miss it from your quote pyramid. But here again.
'15 Chevrolet Camaro (S1 818 stock and S1 900 upgraded).
Street Scene, Ambleside Ascent, Autumn (rain), Unbeatable, no aids, manual.
Stock: - 1st. 2:26.864
Drivatar 2nd: 2:27.249
S1 900: 1st. 2:09.897
Drivatar 2nd: 2:13.602
Win margin:
Stock: 0:1.615
S1 900: 0:4.295
Relative performance against Drivatars increases with PI.
Oh, but now you are using my Camaro run. How surprising. Wait, it’s not. But LOL, X-class isn’t homogulated at all, it’s the wild class. People who play the game know this. And what comes to Agera, or any car, burden of proof for alleged cheating is on you. I played along because other user appeared to be honestly curious, posted results, end of story.
And players can do that too. I have run upgraded vans, off-road vehicles, whatever vs Drivatar with sports car or whatever and won with considerable margin. I’m not going to argue about “But it’s only ‘cheating’ when Drivatars do it!”
Oh, did that mess up your brain a bit? Sorry, but this ends here. All this “hundreds or screencaps don’t matter because they aren’t single screencap and data don’t matter because it isn’t in simple English so LOL didn’t read” just isn’t going to cut it for you. Find a doctor or whatever, I’m here because of game.
We are done.
1 Like
Maybe you forgot to take the handbrake off? Common mistake. 
As we cannot see what we cannot see, things can only be delivered at best effort.
Ideally, everyone. (cater as more players’ likes as possible)
What has it to do with your “nothing to do with racing line” assertion?
Which means, you cannot reproduce the move with the very car.
Once tried racing a Drivatar (“Highly Skilled”) driving stock “2012 Ferrari F12berlinetta” with the very car (Freeroam Head-to-Head), was able to constantly realize that it was pulling away in the straight line.
I must also emphasize: no Drivatar drives any “X 999” car as of 1.410.986.2.
It was never about the debate, it is always about the truth.
I feel the opposite… I find “power builds” tend to be “overated” while “grip builds” tend to be “underrated”.
The main bottleneck is not the physics simulation, but the AI’s decision making.
Making the AI drives as good as (or be better than) the top tier drivers demands heavy verification on the algorithm design. They didn’t bother, and simply made this cheating Drivatard thing…
“If a necessary feature has a high astonishment factor, it may be necessary to redesign the feature.”
It was never meant to be personal.
I do appreciate whose effort, but the fact remains unsusceptible.
It’s funny that this same issue was brought up in FM7 (where you can see the telemetry of the A.I. cars). If you watched the HP and g force numbers for the lead A.I., they were showing numbers outside of the abilities of whatever PI they were listed as. FH does not have the option to do this, so the “boost” the A.I. gets is unknown (but it’s still there). It’s especially noticeable when you run a 1 make/type race with a stock vehicle. The lead A.I. will walk you in a straight line, no doubt.
The discussion has derailed a bit. It’s not a question of are the “Unbeatable” A.I. beatable, it’s how PG programmed the difficulty level based on “boosting” attributes instead of increasing the A.I.'s ability to actually drive fast. Lower difficulty levels show that they are fast in a straight line, horrid around corners. It’s why power builds in the right hands are so good against A.I., yet grip builds can still compete fairly well.
It blows my mind how people praised this being the BEST racing game and most REALISTIC, while we have these kind of AI…
I just notice a race yesterday with the AI, though my car is completely STOCK, no Tuning. Unless the AI had Tuned it’s car…we both ran a 1 : One.
But the AI turn corner without ever need to slowing down and it wouldn’t undeersteer even one bit nor ever oversteer due to the stupid weird power band or just bad Stock Gear Ratio tuning in general.
You can’t even drive this car on the 1st - 3rd gear without spinning your tires.
So the AI just get further and further each corner, since it’s car is like a completely different car from mine. Even on the straight you can feel that sometimes even we both have the same exact car, the AI have better Top / Acceleration.
Sometimes the AI starting on the 1st Position, even the other AI couldn’t ever catch up to it…it will be MILES ahead…
BUT ! What’s amazing is, if you overtake the AI, no matter what position. While you were struggling to overtake it before but once you overtaken them. All sudden the AI as if having some sort of engine problem with their cars…they slow down significantly and you can easily pull far away from them…while before you struggle to even see their tail lights…
This is the best AI ever I’ve seen in all the racing games ( clearly being sarcastic )
Even Wreckfest does better than this…heck even GRID 2019’s AI is better than this…
(Quoted text for context, not necessarily attempt at direct engagement. No disrespect or ‘challenge’ to quote’s posters intended.)
Forgive me, not trying to sass anyone or be contrarian for the sake of it, but has anyone stopped to consider that perhaps this (or at least some degree of it), what we see on our end as broken, may actually be working as designed? Where Motorsport is a more technical product that is far less forgiving where skill is concernd, Horizon without question is designed and marketed for much broader appeal to a wider audience. It may be worth keeping in mind something someone said very succinctly in another thread (I think a recent Trial commentary), to remember that this game is also enjoyed by small children, older folks, and newcomers alike as well as the seasoned and experienced players already familiar with both Horizon and Motorsport and racing products in general. As such, the game is designed for a vast range of skill sets, including AI and it’s behaviors per difficulty level. “Unbeatable” AI ― in a game that pits grandfathers, their grandchildren, their top-tier Forza RC neighbors, and Mario Andretti against each other ― doesn’t mean ‘impossible’ AI. So where things that appear broken to us, like when AI blasts through cross country destructibles with no slowdown but then mysteriously fall off pace after being overtaken, …perhaps that’s actually by design. Cross country destructables with no AI slowdown penalty could’ve been simply an easier way to code an amped up difficulty for the only event category to send you barreling through a few stone walls before your ascent up a precipice to pull a Thelma and Louise. Or when the AI that launches out like gangbusters (requiring near-early onset arthritis olympics with the controller to catch up with them) uncharacteristicly falls behind aver being overtaken, maybe that’s more like a sort of built-in reward. I’m not saying that would necessarily the best or most elegant way to pull it off, but it may have been the easiest, cheapest way to deal with difficulty levels for the entire spectrum of skills.
I’m also not saying they didn’t eff anything/this or that up, one doesn’t have to look far for really obvious blunders…but there’s been a good deal of bluster around this and a number of other core, inner-plumbing issues, so-to-speak, that I suspect have actually been functioning as designed or very close to it, much more than we know and despite what we we’d like to see otherwise.
Regardless whether any of is that is right or wrong, hats off to ClawedScroll554 for his exhaustive work posted here and other threads…he didn’t have to do any of this, spend his time studying a myriad of issues and then take the time to report on it in such depth and detail – for OUR benefit – is truly commendable and deserving of some gratitude 
.
(And I don’t recall whether it was regarding his post lengths in this thread or another related one, but I promise you that you won’t see an effort like his come from PG, so those bemoaning the length of posts can go and get stuffed; this is a forum, a place for writing and reading - if you can only hand 150 or less, and can only manage ‘u r lolz omg jajajaja leet,’ find twitter.)
1 Like