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

Duplode

Let's get it rolling! On the subject of cars, I would like to begin by making a pitch for two of them:

  • Superkart: I like the Kart, and would enjoy seeing it again at ZakStunts. If it comes up at a insufficiently twisty track during the season, we can always slap a -10% penalty on it. If you want to try it for yourself, head to the current race at Cas' competition.
  • Toyota Sprinter Trueno: It would be good to give one of the Toyotas a chance to shine. I feel the Trueno is a more interesting car than the Corolla, as due to its low real top speed it is closer to the LM002 than to all the other mid-slow cars. There is one problem: the Trueno has extra off-road grip similarly to the Lotus, but exaggerated to the point it becomes more bug than feature. However, as the Trueno was never used in a competition, it would be straightforward to prepare and distribute a bugfix release -- I will do just that if we decide to use it next season.
As both of my candidates are slow cars, it sounds sensible to bench the Skyline this time if either or both of them are approved for 2019.

There are all those other cars to consider, of course, and perhaps I'm overlooking some great candidate (you can find all of them, or at least the released ones, at Southern Cross). So please have your say  :)

Cas

Trying different cars each year is very good, so we get to know them better and we don't forget they're there. How are handicap levels set at the start of the season?  Do they continue from the last race of the previous one?  In that case, how are newly incorporated cars assigned handicap by default?
Earth is my country. Science is my religion.

Duplode

Quote from: Cas on September 23, 2018, 03:52:34 AM
How are handicap levels set at the start of the season?  Do they continue from the last race of the previous one?

It worked like that from 2011 up to 2017. Carrying the bonuses over is one way of ensuring all cars get a chance (as there are more cars than tracks in a season, a car which starts with a relatively weak bonus may end up not being used before the year is over). This year, though, the bonuses were reset at the beginning of the season -- I can't say much about that, as I only noticed that now  :D In any case, this might be a good opportunity to revisit this matter -- we have been using bonuses for long enough that there might be some meaningful statistics about their evolution to be squeezed.

Quote from: Cas on September 23, 2018, 03:52:34 AM
In that case, how are newly incorporated cars assigned handicap by default?

I believe the protocol was/is to do laps at a handful of test tracks and use that to figure out reasonable starting values.

dreadnaut

#3
Quote from: Duplode on September 24, 2018, 05:11:46 AM
This year, though, the bonuses were reset at the beginning of the season -- I can't say much about that, as I only noticed that now  :D In any case, this might be a good opportunity to revisit this matter -- we have been using bonuses for long enough that there might be some meaningful statistics about their evolution to be squeezed.

Actually, 2018 coefficients follow from last year as well, but there were both large custom changes (including the Ford Ranger getting 20 points to start with) and a rounding effort to stabilise the numbers.

When I made it possible to updated coefficients through the website, I discovered that the numbers had drifted quite a bit over time —a mix of small mistakes and car rotation, see chart attached. At that point, I decided to round the coefficients down to 12 points per car (192 total, down from 199 from the 2017 season).

Calculations are now automatic, which should prevent further drifting. Custom, per-track bonuses can still change the total, but the UI shows some warning signs if that's happening.

Quote from: Cas on September 23, 2018, 03:52:34 AM
In that case, how are newly incorporated cars assigned handicap by default?

Let's say "feelings" :)  On one side, we want the seasonal cars to be used in at least one track of the season, which means they would get slightly higher bonuses than they deserve. On the other hand, it's not nice to mess up with the coefficients too much, because other cars might end up unused for too long. But it also depends on the tracks coming up and the other cars coefficients: if we introduce a car of a specific class I would give it coefficients close to the other cars of the same class.

More than about maths, it's about keeping things interesting: a) rotate all cars b) increase uncertainty in chosing the right car c) make of many tracks a memorable event (the powergear race, the slow and tricky race, etc)

Duplode

I didn't know the bonuses are now updated automatically. That's cool!

Quote from: dreadnaut on September 24, 2018, 12:49:36 PM
b) increase uncertainty in chosing the right car

Looking at car choices over recent races, this seems to be working as intended  :)

Cas

I think the car bonus system is something amazing that ZakStunts has and indeed, I've seen almost every car used in races, so it works very well. There is only one goal that seems very hard to achieve: car diversity on the same race. It looks like, while cars from race to race change a lot, the trend is to use only one car or two per race. I can think of a few ideas that could help in this regard, but all of them are risky, that is, they could lead to undesired, unexpected results as well.

For example, what if, like in the currency market, cars coeficients would float during the race?  Say, using a car that hasn't been used on the race when at least other two different cars have been used already will signify an extra little bonus to the first pipsqueak using it. I know, very complicated, but this idea could lead to something simpler if analysed. Also, what if a pipsqueak would get a lower bonus on a car if he used the same car on the previous race, but pipsqueaks who didn't will get the normal bonus?

Another idea, longer to apply, but simpler. Suppose initial bonuses are assigned to cars from simple tests on tracks for the first race and then, on every next race, instead of trying to make the cars cycle, bonuses are recalculated according the the diversity obtained on the previous race and seeking a greater diversity?  The problem with this is that a whole season of racing would be sacrificed for this calculation. But, I could use R4K for this purpose next year... I don't know.

Now, beside any of these crazy ideas that are just brainstorming (feel free to use your creativity on them and change them wildly, ha, ha), one thing I'd like to propose is a field in ZakStunts where one can enter a lap and a car and obtain the lap with bonus applied or maybe like this: you click on one result on the scoreboard and that can take you to a menu where you can select one of the available cars and ZakStunts will calculate what lap you need to reach that pipsqueak with the car you selected. That tool would make it a lot more comfortable for pipsqueaks and would make them more eager to pick different cars instead of following the one that everybody's using. What do you think?
Earth is my country. Science is my religion.

Overdrijf

One thing you could try to get more cars viable per track is lower the car coefficients gain and loss. Not used is +1.5%, or +1%. This also lowers the drop from being used, keeping cars closer to the bonus where they are useful.

A side effect of this, which could be seen as both a positive and a negative, is that cars will tend to get used more on tracks where they are relatively fast. If the Rambo Lambo is 20% below it's ideal bonus when a track with ice turns comes up it may not get used, if it sits 10% below its ideal bonus it may be worth a shot.

dreadnaut

Quote from: Cas on September 25, 2018, 04:03:08 AM
There is only one goal that seems very hard to achieve: car diversity on the same race.

Is that actually a goal? Whatever the calculations behind it, that can only work if you artificially push or force single people to use cars that are not the best¹ car. The current system works by forcing "everyone", which has a pretense of fairness, but any similar mechanisms while a race is moving would target pipsqueaks in possibly unfair ways.

[1] "best" intended as handling + bonus for the track

Quote from: Cas on September 25, 2018, 04:03:08 AM
Now, beside any of these crazy ideas that are just brainstorming (feel free to use your creativity on them and change them wildly, ha, ha), one thing I'd like to propose is a field in ZakStunts where one can enter a lap and a car and obtain the lap with bonus applied

Mmmh, sounds like the RPLinfo page, or is it a different thing?

The "intersection chart" is something I sometimes make for myself, but it could be useful in general. Or what about just showing the top result converted for all cars?  "To lead with car X, you need at least 1'12!"

Quote from: Overdrijf on September 25, 2018, 09:01:01 AM
One thing you could try to get more cars viable per track is lower the car coefficients gain and loss. Not used is +1.5%, or +1%. This also lowers the drop from being used, keeping cars closer to the bonus where they are useful.

An equivalent solution is also reducing the total number of bonus points that circulate, which is currently 12×16 = 192. But if cars circulate faster, fast cars (Indy, MP4, IMSA) would become too strong: the only reason they are not an option is because they have large negative bonuses. Bring them closer, and they would always be faster.

Quote from: Overdrijf on September 25, 2018, 09:01:01 AM
If the Rambo Lambo is 20% below it's ideal bonus when a track with ice turns comes up it may not get used, if it sits 10% below its ideal bonus it may be worth a shot.

The track designer can assign or remove up to 20 points to any cars. Leo choose to keep values as they are, but we could have had the Lambo at 41% on this track. Because we trust the maths, but sometimes a track is begging for a specific car :)

Overdrijf

Quote from: dreadnaut on September 25, 2018, 04:23:13 PMBut if cars circulate faster, fast cars (Indy, MP4, IMSA) would become too strong: the only reason they are not an option is because they have large negative bonuses. Bring them closer, and they would always be faster.

O no, the point is not to bring the bonuses closer together, the idea was to make them change more slowly. The Indy over the last three years has had coefficients between -48 and +3, so it generally sits at around -20 or so. If the coefficient changes get smaller you'd still expect it to sit around -20, it just goes up and down less. Because it's the same for all cars, the Indy gets used when its bonus gets around -10 to -5, rather than -5 to 3. So far it's all just change for the change, it works out exactly the same as the current system. But the point of the whole idea is that there will be more weeks when the Indy is relatively near that top of its coefficient curve. Under the current system two months prior to winning the Indy has 6 points bonus less than in the month it won, that's a difference of roughly 12 seconds over a whole race. 12 seconds is a lot, the Indy was not going to be competitive that month. But if the bonus for not getting used was just one percent then 2 months prior to the race in which the Indy won it already had only 2% less bonus, a difference of 4 seconds. That might have made it at least potentially competitive. And it's the same for every other car, including the slow ones.

O course, I don't know if more potentially competitive cars would actually lead to more cars being used per race, or mostly to a longer period at the beginning of the month where everybody tries out a bunch of cars before all settling on the same one anyway. That would be anticlimactic. But it's an idea to consider...

Cas

I think Overdrijf is right. I don't mean (and I think neither does he) that we have to make this change, but that this change would result in greater diversity of cars per track. In theory, the smaller the deltas applied to cars after each race, the slower the "favourite car of the race" will change, but the most precise the handicaps will eventually be, until they get to the point that, on every race, every car could be considered. Of course, if this were done, there's another problem: some tracks are simply better for a car than for another, so when a track favouring, say, the Jaguar, came up, it would cause its handicap level to be drifted from average, but, if changes are small and tracks are always very varied, this should work. Maybe I should make this experiment next year with a R4K season (more than Race For Kicks would be like "Race for Science" XD). Of course, the handicap bonuses that will work fine in OWOOT won't in freestyle and vice-versa.

This is just an analysis. I think ZakStunts is great as it is. But you know, some "what ifs" lead to good ideas sometimes, even if they are entirely different from the initial proposition, so I think they have to be stated.

About the scoreboard position calculator, the idea is that one can say "if I used this car, what lap do I need to reach each of the current scoreboard positions given the current bonuses?". I don't think it's the same as RPLInfo. The idea is that I don't have to create a replay.
Earth is my country. Science is my religion.

Duplode

Quote from: Cas on September 25, 2018, 10:37:57 PM
About the scoreboard position calculator, the idea is that one can say "if I used this car, what lap do I need to reach each of the current scoreboard positions given the current bonuses?". I don't think it's the same as RPLInfo. The idea is that I don't have to create a replay.

I guess you want something like what CTG, once upon a time, used to do by hand.

Cas

Yes, more or less it would look like that. I mean, you have beside each scoreboard line, a small icon or button. You click it and you get a menu with things you can do on that lap. One option could be "get here with another car". That leads you to a page in which you select a car and then you get the driven lap for the same resulting lap.

Another, nicer, way to do it would be this: beside each of the cars in the bonus list below the scoreboard, there's a radio button. There's an extra line that says "Show each lap with its actual car", which is the default. If you select any other, you update the page maybe with a button and the scoreboard is recalculated with the driven times all equal to the time it would take to achieve each of the resulting laps. Some colour or mark could indicate this is an estimation and not the actual scoreboard. This option might be unavailable if one is not logged in.
Earth is my country. Science is my religion.

Overdrijf

Another thing worth considering is plain increasing the number of cars. The current number of 11+5 cars is a good number, it lets every car in on a race roughly ones a year and the specific numbers of the bonus system have been balanced around it. But on the flip side quite a few interesting custom cars have come out over the years, and I don't want to suggest discontinuing some of the original cars.

dreadnaut

Let's pick this up again:


  • I can probably build the table suggested by Cas, although I would leave it for later since it does not need to be ready for the new season starting.

  • Increasing the number of available cars would be nice, but right now it's a pain because of the way the database is set up. I can change that, but it takes time and testing to make sure that stuff still works. (ZakStunts requires a bit of effort to maintain, because it always need to keep backward compatibility with previous seasons)

  • An easy change would be to "slow down" the coefficient variations. Everything right now is defined by a single value: the number of points an unused car recovers, which is 3. The cars on the podium lose a share of the total points recovered by all the other cars. I can change that number to 2 or 1, and everything will move more slowly. We can try this for the next season if people are interested.

Cas

I'm not sure about the slowing down. That is, it'd be interesting to test, but could cause the last five or six races of the season to be all-slow. If the cars recover more slowly but go down just as fast as before, this could happen. I'm OK with both doing it and not doing it. I'll agree with what most of you guys want :)
Earth is my country. Science is my religion.