News:

Herr Otto Partz says you're all nothing but pipsqueaks!

Main Menu

Cars and rules for 2019

Started by Duplode, September 23, 2018, 02:25:00 AM

Previous topic - Next topic

dreadnaut

Quote from: Cas on November 13, 2018, 10:02:02 PMIf the cars recover more slowly but go down just as fast as before, this could happen.

The problem I have is that to keep the number of points constant, I need to add as many as I remove :(
Do you think it would help to "flatten" the bonuses at the beginning of the season, bringing the cars closer to each other?

Overdrijf

#16
It might help to reset the bonuses to a value closer to the average for that car. Resetting the bonuses to be closer to each other mostly benefits the faster cars.

The gap the cars need with each other to be considered for use is still the same. If the LM002 is in the current setup about as fast as the Indy when the LM002 has 30% bonus while the Indy has 10% bonus, that stays the same.

The thing you could think about flattening out is their difference from their average. Cars that haven't been used since the beginning of the year, ready for their time to shine in january or februari might have a big enough bonus that under a changed system they'll see use several times in the new year, because their bonus goes down more slowly they don't get dropped to the bottom of the usefulness list, but maybe to halfway. But it's cars that have recently been used that could get the real short end of the stick. If their bonus is now exceptionally low a slower change in bonuses could see them rise slower and remain unused for over a year (unless of course track makers correct for this with the manual bonus adjustments). It won't take three times as long for these currently very low bonused cars to get used with a 1% bonus compared to 3%, since they don't need to get as far past their average to be considered, since other cars sit closer to their average as well, but it would take longer.

The same does not go for cars that get bonus changes from being used in the new year. It's a starting issue, not a permanent downside. A vehicle's drop from being used is proportionally smaller, so their time to rise back up is about the same. A bit faster even if the idea accomplishes its goal and we see more tracks where multiple cars share the podium.

Duplode

Quote from: Overdrijf on November 14, 2018, 07:25:37 AM
Resetting the bonuses to be closer to each other mostly benefits the faster cars.

In particular, it benefits power gear cars, as often the tiniest of margins suffices to turn a track into a PG freeway. The 2008 season illustrates this potential problem.

Cas

Yes, there is this issue the guys point out. In my terms, normally, all cars could have a fixed handicap if they were similar in their handling, only with different top speeds and an acceleration proportional to the top speed. But that's not what happens. Some cars are capable of things that make them very different from a track to another and PG is the single most noticeable one of these features. So, say, Porsche Marcy Indy should get not one, but two fixed handicap values if one wanted to do this. But then how would one judge which would apply?

I think the current car-shuffling system in ZakStunts does a decent job in getting us try all cars. It could be better, but it's really hard to make it better without, well, making it worse XD   I see this and for this reason, I'm just as much in favour as I am against about changing it and will support any outcome. If you guys want, I can set up an experimental shuffling system different from that in ZakStunts during next season so we can mess up there and not break anything here. For example, in the past, I was thinking of setting up fixed handicap values, but obtaining them from the results of at least two different tracks (one favourable for PG and one not favourable), making an average. This would make PG cars non-suitable for some tracks, but very good for others. Another thing that could work is simply banning a race's top-car from the rest of the races in the season.

Well, anyway, I'm at your service
Earth is my country. Science is my religion.

Overdrijf

#19
I did a little Excelling.



I wanted to identify which bonuses might be good to change if the amount of change per month was to go down. So I wanted to identify the cars that are currently far above or far below their typical bonus values.

To that effect I took the average coëfficients of the current cars over the last three years and their standard deviations, based on all the values the coëfficients have had in that time. If the amount of change was to go down to +2% for a month of no use I'd consider changing the bonuses for anything that per January first sits outside of a single standard deviation. If it went down to +1% I'd consider changing everything outside of half the current standard deviation. In the final two columns I took the current bonuses (before the end of ZCT208) as an example to see how much change roughly both of those options would require.

In the second to last column you can see the amount of cars outside of 1 standard deviation is pretty manageable. Yellow is too low, purple too high (the shades of green are above and below average but within the limits). If the current bonuses were in effect on January first in this scenario I'd consider lowering the values for the LM002 and the Ranger and raising the one for the F40. But overall the values seem pretty suited to carrying over without or with minimal change.

In the last column we shift to 0.5 standard deviations, the stricter limits I would consider if the change went down to 1%. Yellow and orange are too low, purple and pink too high. I this scenario based on the current values I'd consider changing the bonuses for every car except the GTO, the Lancia and both IMSA cars. Overall the current values would not be very suited to carrying over, and I'd consider looking at averages over longer periods to be able to make more precise adjustments.

The up and downside discussed earlier still apply: bigger chance to see more cars viable per race, but also a shifts towards cars being used on the tracks they're most suited for, with the biggest offenders being good powergear tracks that would see more use of powergear cars and less use of regular fast cars like the IMSA's.


As for the selection of cars for next year: I'd like to voice support for the Xylocaine as a fun fast car, good for rollercoaster tracks and mad tricks, and for the GT3 as one of the fastest of the slow cars, good for technical racing with slow sliding corners but better suited than most slow cars for combining this with faster elements.

I'd also take this chance to promote my own cars some more, but of course I'm too much of a gentleman for that. 8) Although people looking to try the latest batch out can visit the Race for kicks event that's about to enter its last week. ::)

Cas

Great work!  I would like to make some experiments with car balancing next season in R4K, as I said, but I'm not sure. Being OWOOT, using slow cars would mean tiny tracks and the popularity of the tournament is not like that of ZakStunts, so not having to consider different cars simplifies it for participants, which might otherwise prefer not to race. I do have an idea I still have to work upon, though.
Earth is my country. Science is my religion.

dreadnaut

#21
Let's try this!

I'll modify the bonus increase from "cars not in-the top 6 gain +3%" to "cars not in the top 6 gain +1%", and bring the current bonuses closer to their average. I will also reduce the total points in circulation, from 12 points per car to 10 (from a total of 192 to 160).

The hypothesis is that by reducing the speed of variation, in each track more cars will be closer to their "useful threshold". This might not change which car wins a race, but we will see more variety of cars through the scoreboard —which should be more newbie friendly and more interesting. We'll see how it goes and re-evaluate at the end of the year.

Custom cars:
I propose to remove: Melange XGT-88, Nissan Skyline GT-R, and McLaren Honda MP4/4

The custom cars for 2019 woul then be:

  • Ferrari F40
  • Ford Ranger
  • SuperKart
  • BMW M3 (from the DTM pack)
  • Audi V8 (from the DTM pack)
Let me know what you think!

Overdrijf

#22
O, wow, I'm honored. This is pretty cool.

I'm potentially maybe even a little too honored. I like the cars that are staying, the Ranger will never be a personal favorite of mine but it has a good niche of its own and the F40 is a fast slow car, which offers an exiting driving style and a nice contrast to more regular fast cars. One could argue though that the two DTM cars are pretty similar and the kart sits in the same kind of niche, just at a more extreme point. An argument could therefor be made to replace one of the DTM cars with for instance one of the cars made to bridge the gap between IMSA and INDY.

I'm not making that argument, mind you. I designed these things to be pretty much my ideal cars, and I will love driving them. I'm quite happy with the feedback so far and I hope others are enjoying them too. I'm just saying that if anyone does wish to advocate this viewpoint I understand why.

As a counterargument: it's fun to use new cars, keep things fresh. (EDIT: I'm also happy with the inclusion of the BMW, as it's my slight personal favorite of the three DTM cars.)

(Alternatively you could try adding the DTM cars under a single coefficient. I have no idea if that's feasible under the Zakstunts system, it's experimental and you're already experimenting with the coefficients, it could lead to one of these cars being generally favored over the other(s), and it would give both less individual screentime. I'm sure you've already toyed with this idea because I mentioned how the idea of a 3 car pack was inspired by this idea on release.)

(In related news: I posted a graphics update for both the DTM cars (a few days ago) and the Superkart (in the past hour) in their respective threads (kart, dtm). It changes nothing but the 3D model, and some text in the case of the M3. That probably makes the M3 the only car of these four for which the Zakstunts system will make a distinction between the 1.0 and 1.1 versions.)

afullo

For me it's ok, we can start the season with these modifications!

Overdrijf

As for the modified bonus system: I'm curious to see it in action. I think the bonus points awarded by track designers will help a lot for the transition. For example: The Jaguar sits at -19 and would take a year and a half just to get to the positive numbers again (which it usually needs to compete) at 1% per track, but with just one 5% boost somewhere in the year it should become quite manageable. (Especially considering that this is one of the biggest deviators, we just had 2 Jaguar tracks, it should be getting a break.)

Overdrijf

I did some more excelling. I picked up where I left off: getting car bonuses closer to their "fair" value now that the amount of change per month has decreased. The same color coding as above: The pink and purple colors indicate a car above their normal values, orange and yellow below. Green is around normal, with light green just below the average and dark green just above it.

The last race had quite some contenders which should mean that purely by chance we're on a good path. If your track is coming up in the next few months, you might consider giving a few manual bonus points to the Jaguar, the LM002, the F40, the Countach and/or the GTO. Those five cars might need a little help getting to their fair value.

The Indy is the only one that looks like it has too high a current bonus in the numbers, but the Indy is a weird car which has had a weird few years now, the years on which these numbers were based. I don't think it's actually that high above its actual fair average value. And while the Indy won last month that wasn't really a proper Indy race, so it's fine if its number comes up ones more in the nearby future. For these reasons I encourage restraint in going by my results and dumping too much of a manual negative modifier on the Indy.

For this month I tried to distribute the manual negatives in such a way that we would again have several contenders. Let's see if I succeeded.

P.S. bonus factoid: when there's 14 negative points to distribute between a portion of 50%, one of 30% and one of 20% the bonus algorithm will apparently go with 8, 4 and 2 points respectively.

dreadnaut

Quote from: Overdrijf on February 10, 2019, 10:36:53 PM
P.S. bonus factoid: when there's 14 negative points to distribute between a portion of 50%, one of 30% and one of 20% the bonus algorithm will apparently go with 8, 4 and 2 points respectively.

The points removed from the podium cars are rounded down, so with 14 point the cars in the top three positions would receive 14 * 0.50 = 7, 14 * 0.30 = 4, and 14 * 0.20 = 2. But those number sum up to 13, so the "rounding" point is always assigned to the car in first position.

dreadnaut

Quote from: Overdrijf on January 13, 2019, 03:36:29 PM
As for the modified bonus system: I'm curious to see it in action.

So here we are, six month in. Time to validate our hypothesis.

In the first six races we had four podiums with two different cars, sometimes three different cars in the first six. Looking back at 2018, were most races were dominated by a single car, the difference is impressive.

We'll see how the second half of the season goes, but at the moment it seems a resounding success :)

Cas

I agree!  So far, this experiment has given very positive results!  And there's always a "what if" when considering cars and watching the scoreboard... many sudden surprises!
Earth is my country. Science is my religion.

Overdrijf

#29
I'm not quite sure what I did wrong/imperfect in my calculations, maybe I should have tried balancing around the highest points the cars reach rather than the center, maybe I underestimated the addition of new fast cars, but I feel like throughout the year we've seen a bit more of the faster cars than entirely justified by complete randomness and a bit less of some of the slower cars. (The BMW made the top 3 in 3 races, though never in first place, the Indy, 962, Audi DTM and kart made the podium in 2 races each. Not that I'm complaining, I like fast cars...)

In related news I feel like now we've reached the point where that imbalance has mostlty fixed itself. The cars that sat at too low a bonus to be viable this year also seem to have risen enough to stand a chance next year. The Jaguar sits near the other fast cars, The two Lambo's sit near the top (though they are slow enough that they could be there for a while). The Ferrari GTO does not have a stellar position but it is set up for one of the next two powergear races. The F40 probably got the rawest deal after all, that one might still be stuck in traffic for a while.

So far 6 races featured multiple cars in the top 3, 5 did not. Last year there were 3 mixed scoreboards, in 2017 2 and in 2016 none. Part of that is luck, but the system change does seem to be doing what it was supposed to do.

For next year we have to consider what to do with the monthly bonuses. We can keep them as they are, plus and minus 10 percents which are permanent. This means these bonuses have a big effect on when cars are due. On the plus side: the cars that people like come up more often. We can lower the amount the track maker can hand out as bonuses but keep it permanent. This lowers the size of the effect, but also the impact track makers have on how their track is run. The third option is to make the bonuses temporary. If the Indy sits at -6 and you give it +10 the bonus becomes 4 for the duration of the race, but it goes back to -5 for the next month (or less if the Indy ends up in the top 3). This option is probably the fairest one, and it gives track makers control over their track without big lasting consequences. On the downside it may encourage trying to set up 1 car races, just give +10 to your favorite car and yolo it. On the upside again, if that's your favorite car, why not?