Cars and rules for 2023

[Daniel3D]

Alright. Sometimes i hate being sort of right because I'm not opposed to change. There are two saying's in English (probably of American Origin) that are stupid.

• Don't change a winning formula.
• Don't fix if it ain't broken.

Both mean that nothing changes.
The Dutch have a saying that makes more sense to me.

• stagnation means decline.

You always have to keep searching for a better way.

Don't change for the sake of changing, but if there is a better option. Go for it.

Back on topic.
"Good talk, let's keep it as it is" at least for the top 12 points.
----\/----EDIT------\/------
After reading the post's several times and looking at this graph.
You cannot view this attachment.

I have a idea to make the lower part more exciting for the rest.
But I don't know what impact it will have.
The points after position 12 get very low in the current system, if you graph the points you get on a scale the line disappears after position 12.

Would it be more interesting to follow exponential regression and (make position 10=3, 11=2,5 and 12=2) from 13 onward a further regression following the graph as close as possible.
That would make the difference after position 12 smaller but also worth fighting for.
I personally don't bother to improve much at the moment because I never made top 12, so the points I get are meaningless to me.

The exponential regression doesn't matter much for the top 12, but it may be fairer as a whole.

[Duplode]

Would it be more interesting to follow exponential regression and (make position 10=3, 11=2,5 and 12=2) from 13 onward a further regression following the graph as close as possible.
That would make the difference after position 12 smaller but also worth fighting for.
I personally don't bother to improve much at the moment because I never made top 12, so the points I get are meaningless to me.

That's an interesting idea! Under the current rules, those small fractional points are tiebreakers which, in practice, are separate from the rest of the scoreboard (given the three discards, twelve 13th places bring less points than a single 12th place). I don't feel there's a strong reason to keep it like that, if a change would help with motivation.

(One thing that larger fractional points would reduce is the value -- symbolic or otherwise -- of reaching 12th place/"full" points. I'm not sure how much that actually matters, though.)

As for the points for each position, right now I'm inclined towards a slightly different system: keep things as they are up to 12th place, and after that have a ratio of 0.7 between consecutive positions, which is approximately what we'd get with an exponential starting from 7th place (rather than 1st, as in that graph). Here is what it would look like, rounding to two decimal places:

Code Select Expand

12 1
13 0.70
14 0.49
15 0.34
16 0.24
17 0.17
18 0.12
19 0.08
20 0.06
21 0.04
22 0.03
23 0.02
24 0.01


[Duplode]

An additional note about this idea. While starting the exponential points from 10th place isn't my favourite option at the moment (non-integer point assignments larger than 1 feel weird to me), if we were to do that it might make sense to choose an exponential such that its slope at 10th place matches the linear slope from 1st to 10th:

You cannot view this attachment.

The ratio between consecutive positions for that would be exp(-1/3) ~ 0.717, which happens to be pretty close to the 0.7 I had suggested earlier. The resulting point assignments, rounded to two places in the table below, look sensible to my eyes:

Code Select Expand

10    3
11    2.15
12    1.54
13    1.10
14    0.79
15    0.57
16    0.41
17    0.29
18    0.21
19    0.15
20    0.11
21    0.08
22    0.05
23    0.04
24    0.03
25    0.02
26    0.01


[Daniel3D]

This looks like a real improvement to me.
Thank you for working on it.

[alanrotoi]

Why not top 20 with integer numbers? 20-19-18-17-16-15-14-13-12-11-10--9-8-7-6-5-4-3-2-1. There won't be any difference or advantage for the top but you'll have a point fight in lower positions of the scoreboard. Top 20 or top 18 or top 15 as you wish but extending the points scale would add interesting new fights. Those who race just few minutes or almost without RH could be benefited with a more inclusive point system.

[Duplode]

Why not top 20 with integer numbers? 20-19-18-17-16-15-14-13-12-11-10--9-8-7-6-5-4-3-2-1.

I'm afraid I don't like this very much. One big concern is that it would raise too much the cost of missing a race. To illustrate the issue, I have added a 2021 simulation at the end of this post.

If we want full points for more people, perhaps the best thing to do is multiplying everything by 5 (so 60 points for a win and +5/+10 LTB) and adopting @Daniel3D 's suggestion of linear points down to a certain point of the scoreboard and exponential points after that. In particular, I think switching to an exponential at 8th place can, after rounding and tweaking the numbers a bit, give a nice compromise between our suggestions:

Code Select Expand

Pos.   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20
Pts.  60  55  50  45  40  35  30  25  21  17  14  11   9   7   6   5   4   3   2   1


Here is the 2021 simulation with the 20..1 extended linear system. Look at the last column to see who would lose positions with the change (also, you can scroll the table to see what happens down to 20th place).

Code Select Expand

Räcer Races Pts. 20..1 Pts. 12..1 Position Change
1 Alan Rotoi 12 189 117 0
2 Duplode 12 178 106 0
3 KyLiE 12 152 80 0
4 Overdrijf 12 141 69 +1
4 Zapper 12 141 69 +1
6 dreadnaut 12 130 58 +1
7 CTG 6 118 70 -3
8 Heretic 12 115 43 +1
9 afullo 12 105 33 +1
10 Akoss Poo 5 90 50 -2
11 Cas 12 89 18.19 +2
12 Ryoma 7 80 28.06 -1
13 Shoegazing Leo 12 78 9.37 +5
14 Friker 4 60 28 -2
15 Marco 6 51 15.28 0
16 Stan 286XT 12 50 1.57 +6
17 KaoS 9 49 3.44 +3
18 GTAMan15 5 46 10.16 -2
19 Frieshansen 3 40 16 -5
20 Igor 7 35 1.38 +3

By the way, here's what we'd get with the cutoff-at-8th system I have suggested above. Note there are no position changes down to 13th place, and the changes that do happen are quite a bit milder:

Code Select Expand

Räcer Races Pts. Pts. (12..1) Position Change
1 Alan Rotoi 12 585 117 0
2 Duplode 12 530 106 0
3 KyLiE 12 400 80 0
4 CTG 6 350 70 0
5 Overdrijf 12 345 69 0
5 Zapper 12 345 69 0
7 dreadnaut 12 290 58 0
8 Akoss Poo 6 250 50 0
9 Heretic 12 224 43 0
10 afullo 12 181 33 0
11 Ryoma 7 151 28.06 0
12 Friker 4 142 28 0
13 Cas 12 127 18.19 0
14 Shoegazing Leo 12 97 9.37 +4
15 Marco 6 96 15.28 0
16 Frieshansen 3 80 16 -2
17 GTAMan15 5 69 10.16 -1
18 Seeker1982 2 56 10.08 -1
19 KaoS 9 55 3.44 +1
20 Stan 286XT 12 53 1.57 +2


[alanrotoi]

Those who missed 6 and 7 races were who lost the most. I think if you miss half of the year or more you are aware about not fighting for the points. If this is the only weakness I still support it. Afullo, Heretic, Leo, Cas, Stan... They raced the hole season and the system does not recognize it yet.

[Daniel3D]

If we want full points for more people, perhaps the best thing to do is multiplying everything by 5 (so 60 points for a win and +5/+10 LTB) and adopting @Daniel3D 's suggestion of linear points down to a certain point of the scoreboard and exponential points after that. In particular, I think switching to an exponential at 8th place can, after rounding and tweaking the numbers a bit, give a nice compromise between our suggestions:

Code Select Expand

Pos.   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20
Pts.  60  55  50  45  40  35  30  25  21  17  14  11   9   7   6   5   4   3   2   1


I like this line. Since we have often more than 20 participants I want to make the same point for the points after 1.
I suggest 0.95 -> -0.05 linear. That way even those who are averaging around 25 or so can have their own battle because differences are smaller in points.

[zaqrack]

Howdy

Let me drop my 2 cents here...

It is not a coincidence ZakStunts had most of the time a linear scoreboard.
IMHO there are two key - and contradicting - factors, which always drove the point system:

1) Value regular participation as much as possible (which if I remember well is also the reason why LTB was born in the first place ... then came quiet days to make it even more important)

2) Do not discourage regularly participating pipsqueaks if they skip a few races (hence ignoring the worst races)

Most of the time we struggled with a low amount of regular pipsqueaks and hence the point system was always somewhat biased for participation. The one exception was 2003 when we had plenty of pipsqueaks, so we switched to nascar-style, as it was not really a problem for most to be present throughout the season.

I still feel that race winners are awarded in other ways well enough (hall of fame, etc.) and we you can never encourage enough the "middle class", who are enthusiastic and very often the backbone and driving force of the community, even though they cannot even dream of finishing on the top.

That said, I have not been really around recently. If you feel the situation has changed, do not hesitate to change the points system. Just wanted to share the key ideas that drove my thoughts back in the early ZS years and hint that -at least for me- the key driver for the system was always the incentive to keep the community active and engaged.

The points for the 12+ positions were a technical compromise. Any change to this part can just improve things.

[Overdrijf]

I have another option to consider. I had some time to math like a matherfucker during my train rides and I was thinking: could we make something that gives everyone the same relative differences to fight over? So basically: if pipsqueak 1 improves from place 5 to place 1 and gets double the score for their troubles, then pipsqueak 2 improving from place 10 to place 6 also sees their score double.

I was first thinking about using powers of 2, everytime the total score gets half as low the steps between the points are also halved, going something like: 16, 14, 12, 10, 8, 7, 6, 5, 4, 3.5, 3, 2.5, 2, 1.75, 1.5, 1.25, 1, 1.875, ugh, decimal system. This is getting ugly.

So then I figured we can approximate powers of 2 using a factor 10 as if it was 8, using steps of 5, 2 and 1. This score system is almost as good with relative differences as pure powers of two, and it's easily and infinitely scalable. Everytime you go down another order of magnitude you just put a zero in front of the last series and there's your next numbers.

I found 3 different ways of bridging an order of magnitude using 5's, 2's and 1's. The first is obviously too steep: 10, 5, 3, 2, 1, 0.5, 0.3, 0.2, 0.1, 0.05... This one takes only 5 steps to cross an order of magnitude. So gaining 5 places gets you 10 times the score. And as you can see gaining just one place can gain you as much as twice the score.

The other two are more reasonable. There's one that crosses an order of magnitude in 10 steps and one that does it in 16. Here is a graph showing the two, including those nice awkward "baby's first Excel graph" lines to help visualise where and how much these lines deviate from a nice curve/exponential regression. (Also no axis lables, and an awkward empty bit because the axis starts at 0, I'm a total graph bad boy) Starting points are arbitrary, but I chose to start both of them at 3 steps above 10, to give the podium positions the highest absolute point gains.



As you can see the 10 point system (the grey line) actually maps pretty close to Duplode's hybrid system (the light blue line). Starting at 25 points the number 12 gets 1 point, the number 22 gets 0.1 point. It looks reasonable at first glance. However, I feel like it may be too steep. 10 steps up to get 10 times the score feels reasonable, but up to twice the score for 2 places gained (4 to 2 for instance), that's big. (Though not nearly as big as going from place 13 to place 11 in the current system, of course.) (Full order of this system: 25, 20, 15, 10, 8, 6, 5, 4, 3, 2.5, 2, 1.5, 1, 0.8, 0.6, 0.5, 0.4, 0.3, 0.25, 0.2, 0.15, 0.1, 0.08...)

So I would probably prefer the 16 step system (the orange line). Starting at 16 for first place the number 12 gets 3.5 points, the number 20 gets 1 point and the number 36 gets 0.1 points. And that's the part that feels unsatisfying about this one. In a bit of a slow month there might be 14 people on the track, so someone gets 2.5 points for showing up, that feels weird. It also starts using half points very early, doesn't feel extremely elegant. However, you double your score by gaining 4 to 5 places (edit: called it exactly five before this edit). That feels good. Those 2.5 point for showing up and ending 14th? They're not going to completely invalidate your hard fought battles for places 20 to 18, for which you still got 1 and 1.4 points respecitively. Ending 3rd while your rival got the win is also a sensitive blow but not a season ender, as it's the difference between 12 and 16 points, you can still make up for that. (Full order of this system: 16, 14, 12, 10, 9, 8, 7, 6, 5, 4.5, 4, 3.5, 3, 2.5, 2, 1.8, 1.6, 1.4, 1.2, 1, 0.9, 0.8, 0.7, 0.6, 0.5, 0.45...)

(If you start at 12 points rather than at 16 place 12 gets only 2.5 points, place 18 gets 1 point and place 34 gets 0.1 point, which may map onto our average attendance numbers better.)

So that's probably what I would do, the 16 step 521 system (patent not pending). It makes the experience of fighting for second to last place more similar to fighting for first place, it's infinitely scalable, and the point differences you get for positions feel good. Because the score for every position is higher than under the current system we could furthermore consider giving a leading time point per 200 hours rather than 250. (But not if 1st place gets only 12 points, then I'd leave it at the current number.)

Leading time wise I would scrap the rule that says you can't keep any hours you gain that go over the limit while you can cary over hours that fell under the limit, even to another season. That still feels weird to me.

We could switch to a new LTB system altogether where we try something wild like just keeping track of everyone's leading hours and hand out a thirteenth race worth of points at the end of the year or something, but we just got a new leading time bonus system (and honestly that suggestion I just made up sounds pretty bad, overpowered mostly). So I would prefer to keep the latest version and tweak it a little over scrapping if altogether, also because we have seen good leading time battles this year, the system clearly motivates people.


EDIT: Looking at the graph I spot another option: use the 10-step version but flatten out the top of the curve a bit. The top positions now give less relative gain than anything else, but still the biggest absolute gain. Maybe something like: 18, 15, 12, 10, 8, 6, 5, 4, 3, 2.5, 2, 1.5, 1, 0.8, 0.6, 0.5... It now has a less extreme upper range, matching Duplode's curve more closely, it avoids the problems of the 16 step curve where the last person in the race still gets a bunch of points and where the half points start too early but it still has a smooth curve going down the ranks and aside from the top everyone else still fights over the same relative differences, even if they're kind of big differences. Although this modification does break the elegance, so at that point why even use this system?

[Argammon]

Very interesting discussion about the point system!

I have a different and much simpler issue though. Should the track designer not gain any leading time for, say, the first week?

The background of my question is that I am planning a dual-way track in February. To make sure one of the ways isn't completely useless, I will have to playtest quite a bit. This would essentially give me an unfair advantage, though the experts will surely find the tricks rather quickly. What do you think? 


PS: There will be one simple and one more difficult way. The more difficult way will hopefully be a bit but not much faster.

[dreadnaut]

@Argammon With public replays, the advantage of the track designer disappears rapidly. There might be also small changes applied by the executive committee, just to mess with the designer's muscle memory

Points system: should we start with the smallest possible change that a) keeps a linear progression b) makes the 13+ scores more meaningful?

What about 12 -> 1 step 1, 0.95 -> 0 step 0.05, as initially mentioned by @Daniel3D?

[alanrotoi]

No please, stop building dual-way tracks.

[Duplode]

Points system: should we start with the smallest possible change that a) keeps a linear progression b) makes the 13+ scores more meaningful?

What about 12 -> 1 step 1, 0.95 -> 0 step 0.05, as initially mentioned by @Daniel3D?

While I'm okay with the general plan here, a ~5% gap between positions next to each other is too small, be it 1 over 20 or 0.05 over 1. Acknowledging performances with raised scores is good, but that should come along with proportionate gaps so that there's also extra incentive for fighting for positions. For that reason, if we are to extend the points system beyond 12th place, I think it is important to have some non-linearity in that region of the scoreboard. (See also @Overdrijf 's argument about relative differences in the post above, which is closely related to the point I'm trying to make here.)

That being so, here goes another compromise proposal: Keep the 12..1 linear system from 1st to 12th as it is, and use the following table beyond 12th:

Code Select Expand

Pos.  12    13    14    15    16    17    18    19    20    21    22    23    24    25    26    27    28    29    30
Pts.  1     0.85  0.7   0.6   0.5   0.45  0.4   0.35  0.3   0.25  0.2   0.16  0.12  0.1   0.08  0.06  0.04  0.02  0.01

Notes on this table:

• The point assignment is loosely based on an exponential with a 0.85 ratio between adjacent positions, with tweaks for the sake of simplicity and memorability.
• For the moment, 30th place looks like a sensible place to stop, given that the largest number of non-ghost pipsqueaks ever in a single ZakStunts race is 24. We can always review and extend the system if the number of races grows a lot.
• Optional addition: change 24th place to 0.14, shift everything below one position down (so that 30th gets 0.02), and have everyone from 31th on get 0.01. While I'm not sure how good an idea that actually is, I feel this system extends far enough down the scoreboard for a small participation reward like that to be affordable.

Leading time wise I would scrap the rule that says you can't keep any hours you gain that go over the limit while you can cary over hours that fell under the limit, even to another season. That still feels weird to me.

Discarding excess hours after LTB is won, though, sometimes adds an extra dimension to the LTB fight: if you get +1 early in the race, there's extra pressure to defend the lead until the +2, in order to see as few hours as possible being wasted. Not doing those discards would be equivalent to merely summing the hours over the season and dividing by 240, which is arguably less interesting in terms of jeopardy and entertainment. As for cross-season carry-over hours, while that is a bit weird indeed, it is defensible on the grounds of motivating up-and-coming pipsqueaks who couldn't quite get 240 hours over the previous season.

(On the points systems, later I'll run simulations for your 16-step system, or a slightly modified version thereof. It doesn't seem we'll get a consensus for adding non-linearity to the upper part of the scoreboard, but still I'd like to see how much milder would the changes be relative to the system I suggested earlier.)

[Overdrijf]

(On the points systems, later I'll run simulations for your 16-step system, or a slightly modified version thereof. It doesn't seem we'll get a consensus for adding non-linearity to the upper part of the scoreboard, but still I'd like to see how much milder would the changes be relative to the system I suggested earlier.)

That's a good idea, I should just do that myself. Will be up this evening. (You're still free to run your own analysis and build variants and such of course.)