Thinking for a new car rating system in ZakStunts

[Mark L. Rivers]

I try to express my thought (awful? genial? mad? Please say it here!) about a new car rating system in ZakStunts.

At the beginning of the new system, Zak attributed after hard tests the bonus percentage for every car. The result was very good, but sure the percentages could be tuned race after race. In this way we would have obtained a more and more balanced car rating near (for how it's possibile) to the perfect rating. Obviously, a car can be more favorite than another one in order of the track design. But thinking about the general features of every car (speed, grip, acceleration, brakes) we can say that averagely:

- LM002 is usually the car wherewith we obtain the highest time.
- Indy is usually the car wherewith we obtain the lowest time.
- Using a Porsche 962 we obtain a lower time than using a Countach
- And so going on.

Powergear is sure a very important variable, but paying attention during the design, we can avoid that this feature can twist the rating.

I think that the original idea with the new system was just to individuate the ideal car rating, a sort of universal rating for Stunts cars, a rating that could permit to everyone in many cases to race and win with his favorite car.
Anyway applying race by race different percentages, trying to induce to race with cars usually not used (increased bonus for unused car, a higher bonus for car unused from more races, hard lower bonus if the same car occupies the full podium), has lead us  very far from the idea about a universal car rating (if this  really was the starting goal as I thought).
Currently we usually see races where only two cars can really obtain good times, often the cars with a very high bonus. But this affects the races in a bad way.
Maybe some tracks would be wonderful to race with a certain car. But everyone will have to race on that tracks only with the cars that will have a high bonus.
I think that the concept at the base of the system (race and try to win with the car you prefer thanks to a more and more balanced car rating) is currently quite far.

So, and this is my first major proposal, I think that we have to aim to a quite fixed rating, tuning it only after 3 or 6 months, basing the eventual modifies of the rating only after seeing the average results of the car used in those months, and not after every race.
In this way it will be the design of the track, and not the floating car bonus attributed in that moment, to determinate the best car to use.

Moreover, and this is the second major proposal, I think that, to obtain a more fixed rating, the bonus system should be different than the current. Let me trying to explain.
Currently the time realized is lowered or increased in order of the positive/negative bonus of every car. For example, in Z87, my 48,55 with Indy becomes a time higher than a minute, and my 1.08,90 with Ferrari becomes about 51 seconds.
As it has been already discussed in another topic, this system has a "bug", that is drivers, in hard maxouting phase, could not to see decreased their time on the scoreboard even they had lowered their race time for 5 or 10 centiseconds.

But we can use another system:

1) We assume that the absolutely slowest car is the LM002
2) We fix the car rating of the LM002 at 1. This means that times realized with LM002 will not modified. For example, realizing 1.05,30 with LM002 we ALWAYS see on the scoreboard 1.05,30.
3) We attribute a higher rating for the other cars that ALWAYS increases their times (for example, if the rating for INDY is 1,50, then realizing 45 seconds with it, we'll see on the scoreboard 1.07,50 (67,5 seconds, that is 45 x 1,50).

So:

Indy >>>  Race Time 45,00  >>>  Scoreboard 67,5
Indy >>>  Race Time 44,90  >>>  Scoreboard 67,35

No more situations where, for example, a driver have to improve for 0,20 seconds to see on the scoreboard an improvement for only 0,10 seconds.


Finally a minor proposal in order to tuning furtherly the system:

If ZakStunts'system could relieve the highest speed reached reading a replay (dstien could help us? ), the system could attribute a higher rating than fixed car rating when it relieves that powergear has been reached, 245 Mph (or higher) for Indy, Ferrari and Corvette, and 203 Mph (or higher) for Acura.

So, for example:

Indy >> Race Time 45,00 - Highest speed relieved 185 Mph >> Rating 1,5 >> Scoreboard 67,50
Indy >> Race Time 30,00 - Highest speed relieved 245 Mph >> Rating 2,5 >> Scoreboard 75,00


OK, I'm ready to read your comments...

[CTG]

Car bonus system is too complicated and useless - I already told that several times. One track, one car is my real world.

[Chulk]

I think it's way too complicated. Don't get me wrong, idea is good but too hard to implement, and even to explain, see how long it took for you!

[Mark L. Rivers]

Well, sincerly I don't think that

1) tuning the car rating after 3 or 6 races instead after every race
2) keep the LM002 with a fix rating assigning a higher rating to all other cars

is really too complicated. Maybe it's possible these concepts are not good, ma sure not complicated...

[zaqrack]

thanks for the proposals Mark, I have only browsed them through, I'll read them detailed later, and of course I'd like to see a nice debate on it.

The reason for introducing the percentages was not to create an universal car system, where every track can be raced with any car - this is at least impossible, as tracks can vary a lot.

The main reason was to modify ZakStunts into something special- and this was only a first step. Sure as CTG said, the easiest way is to have one track and one car / month - the traditional way of Stunts racing. But I just got bored of it during the past 8 years, and also, all the other competitions work that way - why not be different? There has to be someone trying new ideas all the time - the community would be dead now, if all the competitions would only use the same rules all the time.

2008 was a test season - (so as 2007 was a failed attempt with the special rules) to see, how you adapt to the car percentage bonus resulting in an unusual style of racing. My main goal is to add more aspect to the game, than just racing and RH - and to turn ZakStunts into a racing AND strategy game in the near future. I want ZakStunts to require more skills from the winner than just RH, and add the need of strategic skills, but avoiding the need of luck as much as possible. 

My initial plan for 2009 is to maintain the car bonus (modifications like yours is welcome, and will possibly implemented of course), and to allow each pipsqueak to be able to use each car only once/season. This could work very well with your proposal of fixed set of percentages.
This way pipsqueaks have to evaluate the track and THINK before sitting down to race and RH - and that's the way I'd like to see ZakStunts develop.

[Chulk]

Well, sincerly I don't think that

1) tuning the car rating after 3 or 6 races instead after every race
2) keep the LM002 with a fix rating assigning a higher rating to all other cars

is really too complicated. Maybe it's possible these concepts are not good, ma sure not complicated...

Clearly, I didn't get the idea... that's not complicated (the idea at least). And combined with

use each car only once/season.

it could work great, considering some cars are more versatile than others, and for example, Indy can be good in almost every race but Carrera or GTO can't...

PS:

... Maybe it's possible these concepts are not good, ma sure not complicated...

I guess you meant 'But' 

[Mark L. Rivers]

PS:

... Maybe it's possible these concepts are not good, ma sure not complicated...

I guess you meant 'But' 

That's right! 

[BonzaiJoe]

I'm not too enthusiastic about those suggestions... Basically I think people are forgetting that the car bonus system is incredibly difficult to set up perfectly, and it's already running quite well compared to what one might expect. For example, the idea about a separate power gear ranking would be impossible to implement because this factor depends completely on how long the power gear part is.
About having the LM002 as reference instead of the Indy: fine, but it only solves a microscopic problem and leaves the real problems unsolved.
If we do as you suggest and only change the ratings every 3-6 months, we will have the problem of some cars with a permanently too-high bonus, and some cars not used at all. Some cars will still be good for specific tracks, but some cars would never surface and some would always surface. This part of the system isn't working so badly.
Please consider these things and tell me if they are not problems, or if you have solutions for them.
Maybe I'll try to think of some other ideas too

[Duplode]

Well, let me give my two cents to this most important debate... although I like the original concept of car bonus system quite a lot (and both of its goals of giving pipsqueaks more freedom with car choices and making races more strategic), I agree with many of points raised by Mark (and by Akoss too, on the renormalization topic). The main issue is that, even though the concept is very good, with the current implementation it only comes to full fruition at very rare occasions. The strategical element, as Zak pointed out, has not much emphasis - making a good, or innovative, car choice early on is more often a question of instinct rather than strategy. Regardless of that, I do enjoy quite a lot the initial days of exploring the track with several different cars. However, the only race up to now where those multiple alternatives were a crucial element to the very end was Z82. In contrast, all other races were essentially one-car contests and/or powergear-dominated. And speaking of powergear, there is another serious issue the "free-for-all" model induces: restrictions on track design. I find it very sad, for instance, that Zak can't place a loop in the middle section of his tracks in order not to make the race an automatic Indy/Acura party (and even l/r corks are looking dangerous lately).

Looking from that perspective, allow me to comment your positions, and present mine while doing so:

1. In spite of the varying degrees of criticism and of the multiple argumentations, nobody seems to be completely satisfied with the current implementation.

2. Even if doing a multi-car system is so troublesome, I can't really agree with CTG on reverting to vanilla rules. Innovation is good, and even with the problems it has not really harmed the contest. And besides, we'll have USC for classic racing anyway...

3. Same goes to Akoss, even if I agree with many of the points he raised in that older post (and yes, it's too bad we can only perform real analysis on "boring" one-car races under the new system!  )

4. I do not believe Mark's proposal (the first one) alone would work fine either. Taking into account what we have seen on this season, controlling which cars are to be used by track design exclusively is an extremely difficult task, so using semi-fixed bonuses would most likely aggravate balance issues as pointed out by BJ. Even if perfect control by design could be achieved, the final result would be to restrict even further Zak's options for track creations and make races essentially one car - and so we would likely be better with vanilla rules. Standardized powergear adjustments are impossible as well, not only due to the dependence on PG section length but also because extracting speed information from a replay will require knowledge of source code, as what actually happens on a replay is only calculated at runtime, using nothing but the stored keystrokes in the .RPL file. Mark's second suggestion (fixed 0% bonuses for LM002) is a must, however.

5. BJ's comment is spot on: under the current model, the car bonus system won't get much better than what we already have. A large part of the balance problem arises from issues inherent to the game, and no matter what is tried to adjust the bonuses (slowing the percentage evolution, changing initial rates, etc.) they will eventually reach an equilibrium configuration and those problems will resurface.

6. Considering everything we discussed, and particularly BJ's observation, it would seem the only path to make a multi-car system much better than it is now is through large-scale changes on the system, changes that go beyond the tuning of percentages or update schemes. That is the main reason I support Zak's plans of restricting usage of a car to one race per driver per season. It will enforce car choice diversity, versatility, introduce extra strategy factors and, most likely, prevent overt powergear dominance. Additionally, as Chulk implied, Mark's suggestion on semi-fixed bonuses fits a lot better with such a system than with the current one. I see only one real risk with the proposal: towards the end of the season, the likelihood of some drivers getting completely powerless at certain races due to early choices with consequences they would have next to no control would be quite large. This issue, however, can be possibly addressed in a rather straightforward way by allowing drivers to have a repetition wildcard for one of the final three races (only for the final races, else balance would be seriously endangered). Besides that, allowing use of tuned cars would by itself help avoiding the issue, by giving drivers more options (trying not to include rocket-cars or overdo IMSA-style, Zak might raise the car count to 14 with Melange, Speedgate and GT3. If, for instance, I can finish the long-announced Nissan Skyline in time for next season and it gets approved by the community, there will be 12 + 3 cars already).

7. Finally, an additional suggestion which could be implemented with the proposed 2009 system. As I mentioned before, the restrictions on track design freedom bother me quite a bit - and I suppose that applies to Zak as well. Therefore, I think it would be excellent if Zak would be able to say, on a few select races, "well, this is a perfect track for slow-car handling, so that's how it should be raced". In other words, I support the creation of car blacklists for specific races - that is, the possibility to declare Z95 can only be raced with, for instance, Audi, LM002, Carrera, Lancia, and Countach, and implement that by making the submission system reject replays from any other car. Of course, that would be a rule to be used with care, so as not to distort the system (only in certain occasions, and trying to balance blacklists so cars get approximately equal chances through the season).

[zaqrack]

Thanks for the detailed comment and the ideas Duplode. Especially the 14/15 piece car set and the restricted cars on some tracks could make ZakStunts really interesting. I'll surely consider implementing these into the 2009 rules.

[CTG]

Medal table of the cars in this system:

1	Porsche March Indy	2	2	3
2	Acura NSX	2	2	0
3	Ferrari GTO	2	0	1
4	Lancia Delta Integrale	1	1	1
5	Chevrolet Corvette	1	0	1
6	Audi Quattro Sport	0	1	1
6	Porsche Carrera 4	0	1	1
8	Jaguar XJR IMSA	0	1	0

No medals with Porsche 962 IMSA, Lamborghini Countach and Lamborghini LM002...

[Mark L. Rivers]

I'm not too enthusiastic about those suggestions... Basically I think people are forgetting that the car bonus system is incredibly difficult to set up perfectly, and it's already running quite well compared to what one might expect. For example, the idea about a separate power gear ranking would be impossible to implement because this factor depends completely on how long the power gear part is.
About having the LM002 as reference instead of the Indy: fine, but it only solves a microscopic problem and leaves the real problems unsolved.
If we do as you suggest and only change the ratings every 3-6 months, we will have the problem of some cars with a permanently too-high bonus, and some cars not used at all.

Well, I'm not forgetting that there's no way to set up perfectly the car ratings. Anyway I think that cumulating statistics we can get (maybe even after a year) a quite reasonable balanced rating that could permit us not to touch it for months and months. And if we could obtain this, we would have races where tracks would determine the car to use and not the bonus of the moment.

Maybe the idea about changing the rating after 3-6 months is not too good and we could go on tuning it after every race. But I would cancel the progressive bonus (+1% +2% +3%) and drastically reductions of it in case of victory (and more for multiple podium). This generates an instability that penalizes too much some cars privileging other ones.

I think that putting the LM002 as reference and increasing or decreasing linearly the rating of the other cars (in case of victory of LM002 obviously all other cars would decrease their rating) would permit a more accurate tuning in order to obtain the quite reasonable rating aforesaid. Without forgetting the main goal of this, that is improving more easily on the scoreboard not seeing dead races after a early hard maxouting.

About Powergear separate bonus. It's sure correct what you say about not to know how long the powergear part is. But maybe the separate bonus would permit to use PG cars a bit more frequently than now. Look current Indy bonus. If next tracks will be accurate in order not to reach an easy PG, how many races we will have to wait before to use the "king of the car"? Instead, with a normal bonus (not reaching the PG in whatever part of a track) we could use it almost in every track, avoiding the current long banishment.

About what Duplode suggested, races dedicated only to a restricted pool of cars, I had thought the same thing. But, as we can see currently, there are not many races where "similar" cars like Audi, LM002, Carrera, Lancia, and Countach are really all simultaneously competitive. Only one or two cars for races can usually win. So the problem would probably remains.

About Zak's idea, using a car only one time in the year could be nice. Anyway I think that there could be a problem with this rule in order of the tactic to choose. If I don't know which tracks I would have to race during the season (and obviously I can't know them) how will I able to choose the right car to race in the current track? Why should I choose a car instead of another if I don't know how the next tracks will be? Maybe we can get a better result if it will be established a rule that permits to a driver to use a second or even a third time a car he already used, knowing however that he will have in this case a little penalty (-1% -2%) on his final time.