Drivatar difficulty and behavior discussion

The AI isn’t actually “there” on the track, or in the world like the player’s car is depending on the distance between them and the player’s car. This is why the world and its in-game physics don’t effect them unless the player’s car is within a certain radius and they become “real”, so they can be interacted with e.g. rammed; ramming the player, etc.

They’re digital phantoms that give the illusion they are in the world and interacting with it and the player, but only when they’re in close proximity to the player’s car. This is to cut down on computing and rendering cycles. This is why they only exist 100% of time on the mini-map, or the world map (if you pause during a race) if they aren’t in visual range ahead, or behind you.

So, this begs the question: Could all 11 Drivatars be rendered in real time (always there even when out of range of the player’s car) and be forced to abide by the same in-game physics and limitations players do?

Modern consoles and PCs are powerful enough to do this because they do it in other (non-racing) games, but I’m wondering if there is something besides rendering and AI routines that are preventing PG from doing this in the Forza series?

I’d be willing to take a hit in the number of AI in a race if it meant a more level playing field and just better AI that doesn’t rely on artificial speed boosts and other things to be “competitive”.

No they can’t be rendered over large distances using their full physics, but some ideas will work. For example there is a mistake on Goliath where the Ai has to brake on a straight, and if you are behind them there you can pass them all, but at a distance they do not brake there at all. If just that one thing worked it would help a lot.

The easiest way to fix the Ai over a distance is to have them stick to a race time, so that they cannot go faster than a certain speed. Even if they pass through fences they will remain at the speed that a player would finish by hitting the fences.

S2 done. Don Joewon Song AM Vulcan AMR.

Track i believe is set to cloudy, its whatever the current default is for Bahia de Plano Circuit

Now with regards to having to go out of my way to get it done I usually tune cars myself. If not happy I go to Don, Rock, Bedlam, Jezza etc and the car goes in my “roster”.

So for the AM Vulcan this is me turning up an running.

So there are no exploits glitches Ai trickery being done.

This thread is a bit of a head scratcher to me. People seem to complain they can’t beat the AI one minute but then say they can be beaten but only if you ‘break’ them.

They say unbeatable should be unbeatable but then say they have beaten them but don’t consider it winning?

That you should be able to beat AI in any car when this has never been the case in any previous forza game. You always had to have an appropriate car/ build to suit the track at hand. E.g. Road America power build (win) Handling build (good luck with that).

The AI do not copy your tune, then speed boost - they build the car however they want (usually power focussed) to get to a similar PI. Nothing more.

Hitting things does not give a speed boost to lead drivatars, it slows you massively, while they happily carry on laughing their little AI socks off.

I’m not a programmer, just some one who has played a lot of Forza games over the years and can only speak from my own experience. You have to be quick to be beat the unbeatable level AI, you have to be clean & you have to have an appropriate car/ build for the track at hand.

You can not expect to ‘run what you brung’ and beat them every time. Crashing in to AI doesn’t slow them, it slows you, so pick your overtakes, avoid contact (some is inevitable but if it’s door to door, you’ll probably get away with it).

I’m not an elite driver who is getting involved in this discussion to make out they’re fine because I can beat them - if I enter a HLC on the forums I’m mid table if I’m lucky.

Not in FH4 you didn’t. Unbeatable was easy in FH4 I drove it every day using S2 cars. I don’t want it that easy, just hard enough to win. The game is supposed to chose matching cars for the choice of car that you pick, you should always be able to win with a good tune. If you pick a Reliant Robin you want other Reliant Robins to match. I think you should already know that.

The underlying issue and why this thread has 38 pages is Forza - both series - Is far enough into its life cycle where it should have more advanced AI that isn’t reliant on cheats like artificial power boosts and rubber banding.

Other racing games put players on equal footing. So, even if a player is beaten by the AI, no matter the difficulty, it’s a fair race and it’s on the player to get better. This is thrown out the window with how the Drivatars are designed to not operate on the same level as the player for the reasons we’ve gone into in previous posts. Most of it isn’t speculation, by the way. If people have even a cursory understanding of how AI and video games work in general, you can see how the Drivatars aren’t really in the world with the player until they hit a certain radius that activates the world physics for interactions, etc.

Fans are just fed up with the way the AI is designed and how it’s behaving in this iteration of the game. More importantly, it’s not just average players who are voicing their displeasure about how unfair and overpowered the AI is, either. The AI has always been “bad”, but now it’s not even hiding how bad it is and these bad experiences aren’t just relegated to the harder difficulties and those who play at that level.

I also think people don’t get what this Ai is.

I am not putting the concept on a pedestal.

But it is in some way based on our performance.

So if the top 5% of times are wallriding and if that makes them quicker it will influence the lap times the ai will run.

Am I wallriding? No. My rivals times in said car on said track will tell you that.

If I was bringing a knife to a gunfight then I might have to use tricks.

Feedback i have seen given on one of one of the good seasonal tunes was “i didn’t have to corner bomb to win”.

The meta to win is pick a car suited to the track, build it without wasting pi, tune it well and then drive it well.

We don’t know that they are programmed that way, in fact I don’t think any game has ever been programmed that way before.

I have to say I don’t see them wallriding and if they are learning from the LB’s (dubious) they would surely be learning from the clean times with no contact.

I have no idea why someone brought up walls in their argument before - other than as another misguided way of suggesting everybody but them is cheating in some way.

I absolutely agree that having an appropriate build and tune for the car will give you the potential edge to win - after that it us up to the individual driving to actually be able to convert that in to the win.

I don’t mean the drivatars will wallride.

I mean their times are influenced by our best times.

I am not saying they will match our time.

While I people laugh about what Dan Greenawalt used to say one thing he said a lot was “big data”.

The system is definitely failing in various ways. The big data must include outliers that break it.

I brought up walls obviously because the Ai don’t hit walls, and the Goliath doesn’t have walls so it is a fair race. Nobody is testing Goliath for some reason so I can only suspect it is because to beat the Ai you need walls.

Its midweek. Not enough time for me to do multiple Goliath runs.

But enough time for 1.

A class my WRX 05. Going in I thought it will.not have the top speed needed.

I get to about half way and conclude I will come 2nd for that very reason.

Now I have heard about the ai getting a boost at half way.

What about them falling asleep halfway?

I did not see any major error by the ai but about halfway i caught and passed them. Was a battle from there but I won.

I am not calling this a test. Its a single run.

It is a dirty lap on leaderboard presumably drafting at the start 14.29.573

I would usually want more top end speed here and my wrx is more grip build but…

You can chase my ghost if you want to watch me bouncing off trees since there are no walls

I have mentioned it before but there is a mistake in that Goliath track where the blue lines turn red on a straight so the Ai slow down there. It gives you a chance to beat them, but they only slow down there if you are close to them.

Wow just wow

The best times on the LB’s should be clean - an occasional marginal wall tap - where you stop breathing for a moment before you realise you’re still showing as clean - fair enough. But wallride - shouldn’t be anywhere near the top times.

In all honesty it’s times like these where a dev’s input (I appreciate it’s unlikely) would really help as someone posts a theory one minute, then others post it as fact a few pages later. I know Jonk has played the game and set some fierce lap times on some tracks - he has in the past popped in sometimes to correct some misconceptions. I don’t know this time around if he’s just playing for fun or whether he was actively involved in the dev process on this one. Still he would probably be able to dispel some of the myths/ chinese whispers that occur.

There will be quicker dirty times in the big data is what i am trying to say

I think they need to kill drivatars and ho back to normal AI.

Did they have something called adaptive difficukty at one point?

A Drivatar is just regular Ai, and Playground Games made it sound better by describing it broken into parts that slot together like a jigsaw puzzle.

Deleted

I’m not convinced there is any difference between drivatars & normal AI in forza terms - despite the hype a few years back.

‘Adaptive Difficulty’ isn’t a term I’m familiar with but sounds suspiciously like marketing speak for rubber banding. It gives the impression of close racing (adapting to the players times) - one thing I’m not a fan of is the dramatic finish thing. Where the black bars appear from the top & bottom of the screen as you approach the finish - if it is a close race in a manual transmission you are suddenly relying on the sound to change gear if needed.