> I'd like to know that else can affect AI car in the way of separating their attitude from one another.Unfortunately, we just don't have the tools to do this very well. Braking Range and Error Chance seemed promising at one stage - and they do 'work', but use anything more than very small values and the human eye soon spots that those tweaks (especially Braking Range) tend to look false, or staged, on the track. To make our situation even worse, there's nothing in the MD that is car specific at all, so really we're left to do what we can with power and grip in the PFs.
I use a pretty sophisticated version of PRBlancos original PFCalculator which uses some little tricks to create more of an impression of 'driver personality' from the real world data (by giving 'exciting' drivers slightly more power and offsetting that with a slight reduction in grip, with some gentle Error Chance tweaking thrown in for good measure, like 500 or so). Conversely, a 'boring' but steady and reliable driver gets a little more grip and loses a bit of power in my model. Generally speaking, I think it works pretty well, without seeming forced - but it can never be quite as convincing as I once hoped it might be. You might want to try a set of my recent PFs one day and see what you think. Most people seem to notice a significant improvement, compared to what we might call the 'standard issue' (ie the default 2001 PFs), but I certainly wouldn't consider them to be perfect; because, given what we have to work with, it simply can't be done.
Another little thing I have built into my PFCalculator is the ability to use car profiles. Basically these borrow a concept from Microsoft Project, which allows each car to be given a profile reflecting gradual changes and characteristics over a full season, ie a car with an 'end loaded'
failure profile will tend to suffer most of its retirements later in the season, a car with a 'bellcurve'
power profile will tend to be at its most powerful in mid-season and a car with a 'sawtooth'
grip profile will tend to be slightly better in the odd or even races in the season and so on (there are seven different profiles in my model in total). This gives a new layer of subtlety to your results because there are essentially 22 cars x 3 settings (power,grip and failure) x 7 possible profiles so it's very difficult for the human eye and brain to spot them in action. Of course, it's not quite what you and I would really like to see (which would be a single variable or group of variables governing the personality of a particular driver) but it's a reasonable imitation in some ways - and you'll always find that the results over the course of a season are very close to what happened in the real world, but with enough variability to keep things very interesting, competitive and enjoyable; whether you are in a top car or even in one of the less capable ones (which I have always felt have often been a bit overlooked in most peoples' PFs).
Of course, as I'm sure we are both aware,
where this conversation ultimately leads is to the topic of SDI's retirement. Had he released GPxPatch v4.60, the release he seems to have been working on before he left us, we might just have got where we needed to be, because there were some fantastic enhancements planned, but alas, it seems like that won't be happening now; which is a terrible shame in my view. I mean, that's not a criticism of SDI, because he'd done more than enough for us already to earn his place in the hall of fame alongside Mr Crammond, but I can't help thinking that the one final release could have secured the future of GP4 for another twenty years - and without it, I'm not so sure.
One remaining hope however, might be the work being done on the physics modeling currently. I'm not involved and I don't know anything about it, but from what I've seen so far it's very promising - although whether it will ever become something we can use to create another Ayrton Senna remains to be seen as it's inherently more focused on the behaviour of the cars, rather than the drivers.
Anyway, those are my thoughts; so if you'd like a set of PFs let me know. Just give me the year you require and the race % you use and I'll gladly cook you up a set. (Has to be pre-2003 however. I'm just not into modern F1 so I don't have the RW data to work with).
> Klaus Feldmann stated that he believes the slip-stream thingy doesn't work for the AI driversHe's done more on this than I, but I agree completely with what he says. That said, I'm not at the point where I don't want to use the slipstream function at all, but I keep its effect quite small, at about 20 or less, because on an older track, like the Osterriechring for example, it's just too damned difficult for me to overtake otherwise and I feel it destroys the 'flow' too much at the default 50 level.
> I still haven't toyed around with the pit-stop times to learn how does it actually work in-game.Me neither, as I tend to do short races myself. (But for setting the pit lap numbers and the windows I highly recommend TomMks work. I use his formulas when I set them for other people because the default windows are often too short IMO).
> It's kinda sad that SDI didn't make many of the other MD 'Desc' values to be separate for all the drivers cos it would make incredible stuff while simulating the AI races, such as simulating the setup leaning more to the wet conditions (the thing that allowed Panis to win at Monaco'96), different gear ratios for them all and all of that.Well yes, BUT while I'm not smart enough to know for sure, I've a feeling that doing so might not have been possible, because nothing in the MD is car specific.
For me, as a general rule that I've mentioned once or twice in the past, PFs are for calibrating all the CARS relative to each other (including the human car via the Bellini patch) but MDs are for calibrating the TRACK relative to the entire field of cars. They are two completely different things IMO and best kept totally apart from each other. I don't think enough people understand the importance of this (assuming I'm correct that is!).
> I mean the game is very good at simulating the AI races but if now we have only grip (or driver skill) and power (engine values) to toy around and some other like the error chances and braking values (I've experimented with that a lot when it came out and this one works quite badly in-game cos the AI cars usually dont go for the risky/bold moves to overtake the extremely early/late braking ones and lose time behind them, and also these separate braking values all based/depends on the track MD values themselves which can be different for all the circuits so them need to be constantly altered/tuned up for each track which makes usage of this one quite unpleasing).My view on this is quite simple and I hate to even say it out loud really, but GP4's AI, while it is often brilliant at defending a race position, absolutely SUCKS when it comes to adopting an attacking position. Of course, it's still probably the best there is, but objectively, it's really not that good at overtaking when compared to a human. I'd love to think I was wrong about that, but I've never seen any evidence to the contrary yet - and that, on the bottom line, is why no amount of tweaking will ever create a convincing Gilles Villeneuve or a [insert favourite "crazy" driver here]!
> And btw as for the error chances - I think there are no error chances for the first lap for all the AI cars, cos even if I set the high error numbers they are getting into troubles only in turn 1 beginning their second lap.I agree. Like car failures, the error chance does seem to be disabled during the first lap - and I would argue that it's not strictly an error chance anyway, ie its more of a "tendency to run wide" kind of a thing. Useful though and in small amounts I like it.
> So yeah, it would be fun to see more features to make cars and drivers act different with some of them saving the fuel and tyres better or doing better while the wet weather.Yes, again I concur, but I don't think we have the tools.
> Hope SDI make something, but I just had such hopes after the previous versions of GPXPatch being released and it's been a long time since the latest one came out... I'm not even talking about somebody creating a tool altering the AI defensive aggressiveness which is coded within .exe cos probably making this one to be separate for each driver is not possible at all.There he is again! I always get the feeling that SDI can 'see' when his name is mentioned in a thread - and if that's the case I'd really like him to know how much he is missed here. Like you I really felt we were getting close to something incredible when 4.52 came out (and especially after I read his notes on v4.60) and for it never to come to fruition now would be a real tragedy in my view. But then, I guess we all have our own lives to lead, so much as I feel a real sense of disappointment I am eternally grateful for Rene for building the foundation for what we still enjoy - and who knows, maybe one day he might release the source code for others to take GPxPatch forward; like a proud father letting his child run free in the big bad world to stand on it's own two feet. That's my hope anyway.
Cheers.