The new ladder website - current state

"

unsane wrote:

Something seems broken. 13th Zulinn has 32 points in 7 (solo) races.

Similarly 14th samao has 26 points in 5 (solo) races.

Both players had one really good, 1st in class race, so the issue probly stems from there.

Reverted back, dividing total solo race points on the amount of races participated was fail.

I'm thinking of reducing points for RNG-'fun' races by 3 times so they have less impact on solo races ladder. Also with new points formula

S7 solo ladder looks like this


Tried counting best five solo races results - doesn't works because there is like 5 people on first 5 position with 2500 points. Need a good way to evaluate races results that appear less than 2 or 3 times in whole season.

"

Kheffin wrote:

If points are the current way that is working, maybe remove bursts and add a minimum number of races needed? Bursts are worth less and are more swingy than the other solo races.

Sorry cant understand this message :x

I think I misunderstood the "points." I just saw your formula that gives points based on ladder position; I thought you meant actual points, that is, a burst race gives 2 points for being level 5.

Burst races could then give a maximum of 10 points (2 points for 5, 6 points for first in class, 2 points for rank 1), it is worth a lot less than a 1 hour solo, which could give 23 points (10 for level, 10 for first in class, 3 for first overall). This would skew the average to be lower.

However, if I'm understanding correctly with your formula, this doesn't matter. I still think burst races should probably be valued less overall if they aren't already, but I don't know if others would agree.

The amount of points given is based on records, the fact that there is a lot of races of same type prevents random people to be on the top (like this #1 scion on position 255). And this set of races that happen once per season is a problem now. They make my position higher than people's with lots of records who play 10 times better than me. Its need to be fixed, but when I fix one problem, others appearing Kappa

I am not a statistician.

By-hand formulas like this don't personally strike me as all that useful or interesting. Example: exilestats.com. Any of these non-standard ranking systems someone comes up with is arbitrary. I prefer simple, cold, hard, normalized numbers.

So how do we do that? Again, IANAS, but with the data you have/access to, i don't think it would be very difficult to rank players based on: avg percentage of the all-time record for any given racetype+class combination. You could average these percentages for their overall "skill level".

What you would end up with in the overall ranking is something like:

99.91% Throzz
99.70% cwu
99.00% Helman
98.22% Morsexier
85.37% jstq
...

Sounds pretty hard :) Might try to put this system into points distribution formula

or maybe it's not


Did some great improvements on files structure so now I need to upload 6 times less files than before. Now when you load a specific race, you load all characters and overall table at once. If you switch between characters, you browse same loaded file so its open instantly.
Added all seasons.

looks fantastic.

Its funny watching me go from places 21, 2, 4, 16 etc etc. I'm sure we can refine it as we continue.

Boof at #1? No further refining necessary imo

"

boof wrote:

Boof at #1? No further refining necessary imo

If you want a reliable list you need level exp caps and sorted by class and divide by the amount of participations one time with and without rips.


Somebody reached level 35 4 times in 30 events will receive a worse ranking than somebody reaching 32 10 times in 20 events.


You can give each level a linear score.

Then you generate a sequence for each user.

You set up something like that:

If you count rips too set them to 0.

Totalscore = Capscore1 * x1/(Total-Rip) + Capscore2 * x2/(Total- Rip).....

Create a list after you collected results. Either by level or an XP area.
The less levels you include the more punishing it will be for certain players.

If you record from 1-42 you are fine for 1-3h soloraces.
If you record from 30-35 it won't include most players and good racers in a 85min race will be punished because each time they died it will consider them the same as level 1 character.
Dying would improve the score in the riplist.



Your last formula will just favor constant racers that's why Helman is lowranked
TagPT is high ranked because his results barely differ.

You can easily break this in a simply race.

Example Strand Race:
A player only killing hillock in every race will have 500 pts added each time he reached 155 exp.
Somebody having a level 3 record and the rest is level 2 will receive less points.

"

Hilbert wrote:

Somebody reached level 35 4 times in 30 events will receive a worse ranking than somebody reaching 32 10 times in 20 events.

I want exact opposite - one best result matter more than 50 average results.
Still trying to understand your formula