Team mate and AI setups

[joetoml1n 42]

In real life, setups between teams mates are transparent. Both sides of the garage will have access to the setup data for each car. Practice run plans often are mirrored, so they can get back to back comparisons between different setups. Amongst other things, this can include tyre pressures, suspension setups (camber, ride height, roll stiffness, torsion bars etc..), wing levels - all of which are modifiable in game. So therefore, it'd be nice to be able to at least see what setup your team mate is running. As a player, we could then do something different to find out the relative strengths and weaknesses, helping to find our optimum setup. Also, for less experienced players, it'd give them a decent starting place for a setup (Yes I know they have the 5 stock setups, but I'm unsure if they're track specific or just 5 generic setups that cover all tracks).

Also, as an extension of this, it'd be nice to get some basic info on AI cars. In real life, teams have photographers on contract who's job it is to get shots of competitor cars on track, and spy shots in the garage if possible! So some basic setup info, such as wing level would be readily accessible. In fact, even the most casual viewer sat at home on the TV can spot that. Obviously the game doesn't adjust aero cosmetically (which is a shame, it'd be a neat touch if it did that too), so we have no way of knowing what aero levels the AI are running. So if there was a screen that said this kind of thing, that'd be beneficial. This is not something that wouldn't be readily accessible in real life, so it should be implemented into the game IMO. The figures don't have to be exact either, maybe the screen could say something like Front wing 8-10, Rear Wing 9-11? This would give you a range, somewhere in which their setup sits. It'd be a good ball park figure at least! Maybe in the future take it one step further and have this as an upgradeable feature in HQ - allowing you to be more accurate? Or get other basic info on other setup attributes, like maybe anti roll bar or stiffness. This could be based on hypothetical in game team radio (ie, the driver supposedly said to the engineer there's too much oversteer so want to go softer on the rear). Maybe I'm over complicating things here and rolling too far with this - but I do think an implementation of a team mate setup screen and basic competitor setup screen should be possible and would be true to real life.

 

Your thoughts please?

[marioho 686]

This touches on one of may main hopes for the future – a better availability of data to base our setups upon.

But I think there's not enough variability between the AI setups to make this specific take useful. If Codemasters were to improve on that too, and give each team a base setup for each track considering their car characteristics, then that would be great and halfway up there! As of now the AI setup is as follow for all their cars:

I don't think it changes from track to track.

Introducing more variability I think would be a prerequisite to fully fledge you suggestion. On the other hand, having a better feedback from our own performance is one thing that could be implemented right now. Even basic data like average lap temps and tyre wear broken down by tyre – as of now when you get back to the garage all you have is the overall wear of the entire set. 

[RS Phil 214]

5 hours ago, marioho said:

This touches on one of may main hopes for the future – a better availability of data to base our setups upon.

But I think there's not enough variability between the AI setups to make this specific take useful. If Codemasters were to improve on that too, and give each team a base setup for each track considering their car characteristics, then that would be great and halfway up there! As of now the AI setup is as follow for all their cars:

I don't think it changes from track to track.

Introducing more variability I think would be a prerequisite to fully fledge you suggestion. On the other hand, having a better feedback from our own performance is one thing that could be implemented right now. Even basic data like average lap temps and tyre wear broken down by tyre – as of now when you get back to the garage all you have is the overall wear of the entire set. 

I totally agree. We need a telemetry program while all the data are available in the game.

It’s so weird to have a unique setup for AI while they proposed 5 preset setups for the player! They should at least use them according to track and stop discussions about AI difficulty differences between tracks.

[joetoml1n 42]

10 hours ago, marioho said:

This touches on one of may main hopes for the future – a better availability of data to base our setups upon.

But I think there's not enough variability between the AI setups to make this specific take useful. If Codemasters were to improve on that too, and give each team a base setup for each track considering their car characteristics, then that would be great and halfway up there! As of now the AI setup is as follow for all their cars:

I don't think it changes from track to track.

Introducing more variability I think would be a prerequisite to fully fledge you suggestion. On the other hand, having a better feedback from our own performance is one thing that could be implemented right now. Even basic data like average lap temps and tyre wear broken down by tyre – as of now when you get back to the garage all you have is the overall wear of the entire set. 

Are we sure that's the AI setup? That is the balanced/default setup from the 5 built in selections, but are we sure all of the AI are running this and at all tracks? Where did you get that info from? A telemetry app? Could it be that the setup info for AI isn't broadcast by the telemetry? In which case, the app is just defaulting to the, well, default. I have noticed in the telemetry app I use, that some parameters for the player aren't visible for AI either..

 

IF, and it is an IF of course as we don't know, the AI are running that setup at all tracks, there would surely have to be a lot of manipulation of performance and laptimes by the game (especially on higher/faster difficulties), which would be very unfair. For example, I haven't noticed the AI being significantly slower in a straight line than me (in fact mostly the opposite), in Monza for example. Something that they should be if they were running 5-6 wings compared to my 1-2.. In order for them to match my pace, their car performance would have to be altered significantly, which would be so bias! 

 

AI using different setups isn't farfetched, in fact it's widely known to be used in other games. I would expect a AAA developer and game to be able to accommodate this.. Especially as car setup is a huge factor in F1 IRL.. Hopefully Codemasters can confirm whether AI do use custom setups, tailored track to track or if they use just the 5 pre-sets, or something else entirely..

@BarryBL Can anyone from CM shed some light please?

[marioho 686]

Yes, that's with a telemetry app. We can test with another. But the app is registering all the AI data - braking, throttle application and so on. Maybe it defaults to displaying this when the specific stream is empty or the game broadcasts a shell of a setup, but both are a bit far fetched I think.

When I get on the PC I'll see if I have AI laps recorded from another app.

[jrod22145 98]

I’m pretty convinced the ai do not use custom setups.  I’m currently doing Monaco in driver career mode at 103 AI and they consistently reach 175+ mph through the speed trap.  Running 7-11 wings I can only reach 166 mph using Rich/Overtake no matter how good of an exit I get out of Portier.  I had the same problem while doing the Full F2 season that precedes starting your F1 career.  In the F2 race, the ai would run away from me down the front and back straight but I’d have to lift off or even brake through the tight twisty bits to keep from running over the AI.  Sector 3 was a nightmare!  I had no chance of overtaking anyone much past the first lap, which I guess is realistic, but not for the right reasons haha.

 

I truly believe the AI’s pace is adjusted, based on difficulty, by increasing their overall speed through the corners and traction on corner exit, not by making them more intelligent leading to better setups and lines through the corners. 
 

Another good example I have of this also comes from my F2 season.  At Austria the feature race started out dry but it started raining with about 10 laps to go.  For about the first 4-5 laps I was doing fine, running in 3rd holding about a 1.5-2 second gap to the car in front and about a 3 second gap to the car behind.  With 4 laps to go, drs was disabled and this is when my problems really started.  By T3, the cars behind me were now on my rear bumper.  Out of T3 the ai were accelerating like it was dry, no wheel spin, no short shifts.  They just sped off like a rocket.  By the end of the next lap I was out of the points.  Ended up pitting for inters with 2 laps to go while the AI carried on finished the race on drys.  
 

Overall, the scenario wasn’t altogether unrealistic (not putting), but the amount of traction and speed they had on those last couple laps was unmatchable and I most certainly wouldn’t have made the finish on dry tires running at the pace they were.

[joetoml1n 42]

1 hour ago, jrod22145 said:

I’m pretty convinced the ai do not use custom setups.  I’m currently doing Monaco in driver career mode at 103 AI and they consistently reach 175+ mph through the speed trap.  Running 7-11 wings I can only reach 166 mph using Rich/Overtake no matter how good of an exit I get out of Portier.  I had the same problem while doing the Full F2 season that precedes starting your F1 career.  In the F2 race, the ai would run away from me down the front and back straight but I’d have to lift off or even brake through the tight twisty bits to keep from running over the AI.  Sector 3 was a nightmare!  I had no chance of overtaking anyone much past the first lap, which I guess is realistic, but not for the right reasons haha.

 

I truly believe the AI’s pace is adjusted, based on difficulty, by increasing their overall speed through the corners and traction on corner exit, not by making them more intelligent leading to better setups and lines through the corners. 
 

Another good example I have of this also comes from my F2 season.  At Austria the feature race started out dry but it started raining with about 10 laps to go.  For about the first 4-5 laps I was doing fine, running in 3rd holding about a 1.5-2 second gap to the car in front and about a 3 second gap to the car behind.  With 4 laps to go, drs was disabled and this is when my problems really started.  By T3, the cars behind me were now on my rear bumper.  Out of T3 the ai were accelerating like it was dry, no wheel spin, no short shifts.  They just sped off like a rocket.  By the end of the next lap I was out of the points.  Ended up pitting for inters with 2 laps to go while the AI carried on finished the race on drys.  
 

Overall, the scenario wasn’t altogether unrealistic (not putting), but the amount of traction and speed they had on those last couple laps was unmatchable and I most certainly wouldn’t have made the finish on dry tires running at the pace they were.

That all sounds too similar. I completely agree that from my experiences, the AI is just boosted in performance as the difficulty rises. I play on broadly the same difficulty (105 currently) but the AI have noticeable boosts in certain areas, or sometimes a lack of negative effects from conditions. 

It looks to me like poorly optimised AI. IMO, 110% should be the "default" AI level, and anything lower selected should make the AI more susceptible to errors, less accurate on taking the racing line, less confident on the throttle etc etc. Nothing significant, as the differences are around 0.1s per 1%, but that should be how the AI is adjusted in my opinion. However it looks like that when the difficulty goes over a certain %, the AI is adjusted up. This creates an unfair and advantage in certain key areas. They don't act as human anymore, very few mistakes (and I'm not even talking about long term race effecting mistakes here, more just snatching a brake here and there, running wide on an exist kerb, understeering slightly wide etc), and very few devotions from optimum. We already know they have a basic tyre model, which again makes the situation worse. Leads me to think about other things, like : do they get effected by powertrain/gearbox wear like we do? Yes occasionally they take engine penalties, is that because the engine they were using earlier in the weekend is worn, or because it was scripted (either based on number of laps, number of races etc). Leads on to the question of do they suffer reductions in power outputs like we do? And again, yes im aware they have in race issues sometimes and retirements, which we don't, but until they point they don't seem to be effected by reduced power output?

Edited August 25 by joetoml1n

[NinjaElmo 19]

I can confirm that all the AI seem to use the exact same setup every single time. Looking at things with my own telemetry program, they always had 5-6 wings and 5-7 ride height, whether it was a dry Monza or a very wet Singapore.

Which makes things a bit weird that they even have setups, when they don't change them at all...

This would surely cause an issue with mixed conditions races, where the player speed change because of conditions is largely setup based, whereas the AI just get slowed down a certain percent. I had a mixed conditions race at Singapore, where I designed my setup ready for rain (higher ride height, lower pressures), but the AI were still cruising past me with full traction out of slow corners, despite their 'worse' setup. Then when the track was dry, I couldn't pass them on the straights even with overtake on, high fuel mix, and DRS, because I was running 8-11 wings to their 5-6. 

[RS Phil 214]

Definitely wrong from CM if AI use the same setup while 5 preset setups are given to the players but this leads me to another question. Does the AI really race or are the time scripted so they match a difficulty level? It’s totally unrealistic to have such a high wing setups for Monza. We should easily overtake them in the straights, even in highest levels but they are fast, very fast.

is the AI affected by setups? And then are they affected by tyre wear? My guess, according to what has been said, is that all this is scripted (speed, wear, ...) and this could explain why they have so many difficulties to overtake, for example.

Edited August 26 by RS Phil

[joetoml1n 42]

11 minutes ago, RS Phil said:

Definitely wrong from CM if AI use the same setup while 5 preset setups are given to the players but this leads me to another question. Does the AI really race or are the time scripted so they match a difficulty level? It’s totally unrealistic to have such a high wing setups for Monza. We should easily overtake them in the straights, even in highest levels but they are fast, very fast.

is the AI affected by setups? And then are they affected by tyre wear? My guess, according to what has been said, is that all this is scripted (speed, wear, ...) and this could explain why they have so many difficulties to overtake, for example.

Yes I agree, I feel they’re heavily scripted in terms of pace, which for P for example, is fine - as their involvement  in that session is a non issue. But for a racing situation, they should also have to “drive” the race and be bound to the constraints of their car setup and track conditions etc. It definitely feels like they’re designed to do a specific laptime for any given difficulty level - rather than an increased intelligence or consistency. 

 

 

1 hour ago, NinjaElmo said:

I can confirm that all the AI seem to use the exact same setup every single time. Looking at things with my own telemetry program, they always had 5-6 wings and 5-7 ride height, whether it was a dry Monza or a very wet Singapore.

I’m still not 100% that they are running that setup at every track, I’m still leaning towards the telemetry not outputting that data, so the application is just filling the blanks with the default. I just find it hard to believe Codemasters would have overlooked this? Especially as setup is king in F1 and an integral part of the experience. If they are using the same setup then the whole My Team, Career and GP modes are fundamentally flawed. That being said, there are some things that give weight to your hypothesis that they do in fact use the balanced setup everywhere. Like you said, running high wings levels at Monza, their speed would have to be falsely adjusted up to keep the laptime and straight line speed, but their corner and braking speed would also still be higher, due to the higher wings. 


It’d be very useful to get some comments from CM on this. Does anyone know if we can tag people who might respond? Is there a contact method we should be using rather than the forum?

[NinjaElmo 19]

2 minutes ago, joetoml1n said:

I’m still not 100% that they are running that setup at every track, I’m still leaning towards the telemetry not outputting that data, so the application is just filling the blanks with the default. 

I was messing around with reading F1 2020 telemetry in Python, it's all entirely my own code. I can assure you that I'm not defaulting anything in my program (I would get an error if there was no value), and that according to the F1 2020 UDP specification post, the packets are sending that setup info for each participant (so each AI and the player). 

There is the potential that the game is defaulting that information when it sends it, which would cause all telemetry programs to show the same result. Given this post seems to say that the AI doesn't change setup, I would think it's just the case of the AI all using the same setup though, unfortunately...

I agree that it's a weird thing to overlook, because it's just going to cause issues all over the place. I understand that their implementation is a fairly decent work-around to try to simplify things, but it makes designing a setup feel pretty pointless if the AI are essentially on rails.

[Davidj7077 64]

David Grecos answer here sounds as if the AI is using 4 at front wing and 5 at rear wing also in Monaco.

[joetoml1n 42]

1 hour ago, NinjaElmo said:

I was messing around with reading F1 2020 telemetry in Python, it's all entirely my own code. I can assure you that I'm not defaulting anything in my program (I would get an error if there was no value), and that according to the F1 2020 UDP specification post, the packets are sending that setup info for each participant (so each AI and the player). 

There is the potential that the game is defaulting that information when it sends it, which would cause all telemetry programs to show the same result. Given this post seems to say that the AI doesn't change setup, I would think it's just the case of the AI all using the same setup though, unfortunately...

I agree that it's a weird thing to overlook, because it's just going to cause issues all over the place. I understand that their implementation is a fairly decent work-around to try to simplify things, but it makes designing a setup feel pretty pointless if the AI are essentially on rails.

Oh interesting, thanks for confirming. Sorry I wasn’t trying to question your coding, just trying to find solutions to this as it most definitely is a strange one! It’s also something that would have to have been a deliberate act, which makes it more baffling! I don’t want to take that one line from David Greco as gospel, it’s not exactly a comprehensive answer, so I would still like clarification (although I’m aware I’ll likely never get it), as I can’t understand why they would make it like that? Especially when the game is fundamentally about car performance and making improvements, but the AI runs default everywhere - not playing to their relative strengths and weaknesses and not tailoring to tracks. It does add a lot of context to questions you’ll see floating around regarding AI performance at particular tracks or in particular situations.. 
 

[NinjaElmo 19]

12 minutes ago, joetoml1n said:

Oh interesting, thanks for confirming. Sorry I wasn’t trying to question your coding, just trying to find solutions to this as it most definitely is a strange one! 

Haha all good, I didn't take any offence, just thought I needed to point out my perspective and why I don't think changing telemetry program will have different results.

It is definitely a weird, and as you pointed out, deliberate, choice for CM to make. My guess is that the AI 'setup' has zero impact on how they actually drive, but because of the way CM has designed their telemetry output, each participant needs to have a 'setup', so those are just the values CM has given to any driver that is AI. If there wasn't the big AI performance issues on some tracks that you pointed out, it wouldn't be too much of an issue. But it seems like the tracks that are massively affected by driver setup (Monacco, Singapore, Hungary) always seem to be the ones that people find issues with the AI, which is explainably the case if the AI are on rails.

[marioho 686]

I did some googling yesterday and came up to this post:

Read the comments by KrazyLurt.

Could be that the AI is not on rails and is still actually driving, but that their performance is generated differently. Setups are kind of a medium for the player to interact with the various mechanics simulated by the game, like drag and grip and downforce. Maybe Codemasters just cuts this middle man regarding the AI and goes directly to the raw numbers, stipulating beforehand the parameters for their driving and thus bypassing one resource hungry step for the system to deal with 21 instances running on the game.

I'm throwing this up there because otherwise all their other telemetry data would be just a shell too, wouldn't they? @NinjaElmo is in a way better position than I as all I do is stare at the AI graphs and put on an engineer facade pretending I get it, but it would be odd to have all that data being generated just to set it aside and give them an artificial lap time.

By the way, @David Greco CM is the one to tag.

[NinjaElmo 19]

Yeah, I agree with you @marioho, it's probably somewhere inbetween to make things less resource intensive, and for the most part, it does a good job. Changing conditions and certain tracks do seem to cause issues, but then of course that is always going to be the case, since some people are gods at wet weather conditions or Monacco.

In terms of the telemetry output, basically I think it's done to make the lives easier for telemetry program developers. The game outputs setup data for all drivers, regardless of how many and whether they are AI or not. Telemetry developers just need to read how many drivers there are, then save the setup data for each driver that many times. If the game only sent that data for player drivers, it would make things unecessarily complicated for other developers designing telemetry programs. Essentially it's easier for everyone if CM just output a piece of data that is sometimes pointless (AI setups), than to sometimes not output that data because it's pointless.

Of course just how pointless that data is I'm still not certain, but I understand them sending the data even though it might not mean much.

[joetoml1n 42]

Yeah, Thanks for the info on that other thread @marioho

I think I understand why it's like this. I guess to provide a baseline that is scalable and adjustable to difficulty etc? It is frustrating though that it's not a level playing field. Does this mean that the 1-11 sliders for aero are track specific? It'd be a silly way of doing it if it were! But with 6-6 being optimum to match AI top end (according to KrazyLurt in the other thread), thats bang in the middle of a 1-11 scale. Could this mean the numbers are arbitrary, where anything lower than 6 is less downforce/drag compared to the AI, on a decreasing scale and anything over 6 is more df/drag again on an increasing scale - but only relative to AI. So 6 at Monza and 6 at Monaco would still be different? Rather than a rational scale of set df/drag levels for each position on the slider?

I think I saw a thread before discussing whether people thought they were track specific or not - can't remember the outcome.. I'm just thinking out loud again and trying bounce ideas of everyone else, to try get to the bottom of this.

Edited August 26 by joetoml1n

[NinjaElmo 19]

14 minutes ago, joetoml1n said:

I think I saw a thread before discussing whether people thought they were track specific or not - can't remember the outcome.. I'm just thinking out loud again and trying bounce ideas of everyone else, to try get to the bottom of this.

Possibly this question is the one you were thinking of? I do remember a discussion about this too, but don't recall their being a definitive answer. That whole thread is helpful for working out some of the finer details in how the game works.

As a side note, check out the RealSport F1 2020 Setups, specifically the ones by Jacob Hancox. I race on 100% distance, ~85 difficulty, and I've been finding the setups from that site super helpful in terms of making a good race setup and explaining the steps behind each change. The game could do a lot more in terms of explaining and showing setup effects, but I think with enough tweaking you can start to get a proper 'feel' for things.

[marioho 686]

@joetoml1n I wouldn't go as far hand in hand with KrazyLurt comments haha

What I found useful there was his take on the AI not working through the setup interface we players have.

They're quite old and pertaining to a very different handling model. One thing that was heavily changed from F1 2019 is the downforce mechanic for instance, so his impressions about wing angles would probably be revised for F1 2020.

I don't think they're track specific. I've seen these threads too and can't pinpoint the answer but I think it would have caused an impression in me reading that they're track specific.

@NinjaElmo thanks! That makes perfect sense. However I was asking about the other telemetry data, like braking and traction loss and wheel spin and so on. Looks like they're real, that that's what is happening to the AI, instead of being just a shell of a value as it is with their setup. That's just my impression though, never paid much attention to it.