Of course I have to jump into this one. Here are my comments on the state
of the NASCAR 2
AI and on Adam's comments :
First, the AI for N2 is better than N1 - at least it is more aggressive.
Some of the tracks are provided
with a decent set of racing and passing grooves - some are not. There are
also some alternative
AI grooves provided by after market authors free of charge - like yours
truly. These can improve
the sim racing experience quite a bit - although they sometimes sacrifice
just a little in realism.
However, there are many other aspects of N2 that sacrifice realism - so
long as the sim racing
experience is more enjoyable and rewarding - this IMO is ok.
The problems that I think exist in the basic AI software model are these :
1. The AI cars should better choose the passing groove they use. They
often choose the high
groove on tracks where that is the sure way to the back of the pack. This
will cause the faster
cars to get caught in traffic they should be able to pass. Code that
chooses a passing groove
based on the traffic being passed would help here. In the LP file
modifications I make, I try to
midigate this by improving the high passing groove.
2. The AI cars should always return to the racing groove when not in
traffic. I have had problems
where the AI cars run the low passing groove lap after lap basically making
them impossible to pass.
This occurs when the LP files are modified to speed up the low passing
groove - which improves
passing. This is an unwanted side effect. I end up making the low groove
very slow at one point in
the corner to get the cars out of that line and return to the racing
groove. At Talladega, I have done
this coming out of turn 4.
3. The AI cars should draft together better. What I mean is the faster
cars should draft together passing
slower traffic rather than breaking up and racing each other in slower
traffic. This messes up drafting
at the big tracks and causes traffic jams on the smaller tracks. This is
little tricky to implement without
stopping passing between the faster cars altogether. However, it may be as
simple as making the AI
cars follow the faster car when all the grooves are tied up. All too often
one car will pass and the next
car in line will follow the car being passed rather than dropping in behind
the faster car and also slipping
by. The current implementation often gets the third car in behind the
first car passing now causing
the two faster cars to get side by side and causing a traffic jam.
4. The AI crews need to make better pit decisions. This is the number 1 AI
problem. I sent many
e-mails to Papy suggesting fixes to this. The bottom line is that the
leaders should pit together under
caution - say when 50%+ into a green flag run and NO ONE should pit under
green when the fuel will
make it to the end but the tires are worn. There should perhaps be an
addition parameter added to
the track.txt file that indicates fuel load limit along with the
recommented pit interval for the AI cars.
I agree to a point. However, I did send Papy a good algorithm for pit
decision strategy well over a
year ago and N2 was released with the same unrealistic behavior. It does
hinder sim racing
enjoyment when the leaders get seperated due to bad pit strategy - or worse
yet pit with 2 laps to
go at Dover, where the pit interval defined in track.txt still leaves fuel
in the tank. This fix is actually
very easy to implement - I sent everything but the code - and I am willing
to write the code if that
would help.
Agree. N2 really is a pretty good product. It had good hardware
compatibility - which is always tricky
with a wide distribution product like this and is predictable. It has some
bugs, but I have not seen or
heard of lots of random lockups etc. I think Papy did a decent job
qualifying N2.
For the most part - the AI etc. is not that bad. I personally would love
to get more feedback on
specific implementation changes suggested. I was dissappointed that none
of the pit decision
suggestions made it into N2. They were clear, simple and better. Nothing
is worse than racing
for 3 hours with a group of cars just to have one of them make an
unrealistic pit call and get out of
contention.
I never expected this to be fixed after N2 was released. Also, it is not
that bad ! I have reworked
other LP files for N2 and will make some available soon. My latest
Talladega AI is actually pretty
fun - although the AI cars run up the track a little to make room for side
by side low side passing. I
will make this available soon. I have worked on some others too - most
recently rev 2 of Dover.
Even without AI improvements, N2 is a good simulation. I have no intention
of boycotting N3
or other Papyrus products based on N2. I would like to see some more
feedback on AI changes
in the future. I would also be happy to prototype and test some AI
improvements if this will help - including
the pit decision algorithm.
I enjoy N2 and I am not in the business of taking things personally.
Boycotting N3 would only be to
make a point - and personally I am not sure what that point would be. I
hope some LP file alternatives
will make you feel better about this - but if you are just taking the bugs
in N2 personally, then I suspect
this will not help.
/THansen