How is the Coding Rank calculated?

[Neumann]

I don’t agree with Magus and Neumann’s view that being 1st is equally hard regardless of N. This is mathematically false. Counter example: if there is 1 person he will always be first. I’m pretty sure if you had a contest with a million people winning would be a little harder than if you had 1000.

Of course it would be, but we are talking about N between 1000 and 3500.

I see your point :
“I finished first in a multiplayer contest with 1000 contestants. If there had been 3000 more contestants, there is big probability that at least one person among these extra 3k had beaten me”.

Maybe you would have finished 3rd or 5th ? This a little ranking shift (for the top players at least) that shouldn’t impact earned points imo.

[Agade]

Oh I agree the difference is small, which is why I’m suggesting a minor change in the spirit of marchete’s orginal quality suggestion.

[MrAnderson]

The formula for every puzzles should be something like this:

BASE^((N - rank - 1) / N)

For the BASE, it could be something like this:

• Contest : 5000 (3 puzzles maximum; i don’t think we want to change it)
• Multiplayer puzzle : 2000 (or 2500)
• Optimization puzzle : 1000
• Golf puzzle : 200 (5 languages only, i don’t think we want to change it)

CG can mess with the exact weightings based on whatever balance they want to strike between the game types, but I think the overall concept is sound. It also has the benefit of being simple to understand; finish in the xth percentile and you’ll score y points, every time. The only extra complication we might need goes with Neumann’s point; while a huge sample size doesn’t make a significant difference, an abnormally small sample size does. I wouldn’t be opposed to the min function to reduce the points for low-participation events.

[Magus]

BUT there is one contest that had a very low number of contestants, in which being high ranked was a little bit easier : SF2442 (211 people)

I wasn’t aware that SF2442 rewarded points at all. I think it’s a special case. It was a private contest but some people heard of it. It don’t think it should reward any points.

[Palmipedus]

Hi all,

It’s nice to have a good and fair formula but don;t you think that the problem come from the fact we take the best 3 of all time, why not having an ATP like ranking? (and the race one too)

It’s just a ranking on the last 12 month, like most of the problem will be fix, and the ranking will be more dynamic. If someone stop playing then he will slowing lose of his point, until he starts playing again.

[d-s-x]

Ranking formula for Optimization (code size) seems a bit off. At least I can’t get same number as in https://www.codingame.com/home > RANKING STATISTICS > Code golf

Is there a special case when some players have solutions of the same size?

[GoofyGas]

How does the ranking for clash decay over time?

[OEtzi007]

Just looked at the CodinPoints-System and am trying to understand, how it works. I thought going through the example of Clash of Code should be simple enough.
Formula: min(N,5000)^​((N-​C+​1)/​N)
When I look at my score I see the following:
CP: 3829
Rank 67 out of 1,171,429 (Top 0.1%)

Applying the formula I get a result of
N=1,171,429
C=67
CP=4997,601220833
How does this agree with the formula? Am I missing something?

[eulerscheZahl]

Rank 67 out of 1,171,429 (Top 0.1%)

Your N is wrong, for Clash of Code you only count active players (clashed within the last X months). Check the points of the first in clash of code to get N.

[OEtzi007]

Ok if I use my current rank (today 76) and my current CP (3733) I can compute backwards the N value. I get a value of N= 4317 (1/200 of one million(total number of CoC players), this does not fit at all). This seems quite low and I can’t figure out how to check the points of the “first in clash of code”, what do you mean by that?

[OEtzi007]

Ah okay the first (best ranked) Code-Clasher has exactly 4317 CodinPoints, so this is the number of active players, I see. Thank you for your response!

[foasolution]

What is the main issue?