nineball
Semi-Pro Squid
So a few days ago I read an article linked in an earlier post about Splatoon 1's ranking system and its flaws, and while most of the math stuff went over my head, the article did link a python program that someone made that came to the same conclusion as the article, and that python program inspired me to make a simulation of the Splatoon 3 Anarchy ranking system.
Also bit of a disclaimer: I am not that much of a math and statistics kind of person, I am just a programmer that likes playing with numbers and I might might make a few mistakes here and there.
How Anarchy Ranking Works
For quick reference, and because the wiki has this information split up in two pages, I also will explain in this thread how Anarchy's ranking system works. I have wrapped the explanation in an expandable spoiler tag for those who do not care and just want to see the juicy data.
Sources: https://splatoonwiki.org/wiki/Rank, https://splatoonwiki.org/wiki/Anarchy_Battle
The Simulation
Simulation source code here if you are also a nerd like me: https://gitlab.com/9ball/splatty3-rank-simulation
For the first simulation, we will have 100,000 players playing in back-to-back series over 1,000 games, which I think should be aproximately the amount of games the average dedicated player should be playing per season. Every player has an 50/50 chance of winning/losing each game, and medals are awarded completely at random.
It appears that most of the player-base landed themselves in S rank! This might not come as a surprise since its really well known that this ranking system has a really strong upwards bias, especially with the lack of a way to naturally rank down. This feature of the Anarchy ranking system leads to a problem belovedly known by the community as "rank debt", and it appears that these simulated players are no stranger to it! Debtors, highlighted in orange, make up the grand majority of the simulated population.
A simple solution to this problem that many have proposed is to simply just have people naturally rank down when they lose enough points, so I implemented my own rank down system in the simulation.
How my rank down system works is that if after a series a player is below 0p, they will be sent down to the middle rank of the lower bracket (e.g. A- 0p => B 350p). I also made it possible to rank down within a bracket (e.g. A+ => A) This is how a simulation of that system went with the same parameters as before:
Then for curiosity's sake, with 10,000 games.
That's a bit surprising! Its really interesting how adding just a simple rank down feature to anarchy can cause the ranks to spread out like that, and consistently across a large number of games too. A few tweaks to the reward system could make the spread more even but I'll leave that for another day.
Well, that's all I have to show for now. I did plan on poking at the simulation a bit more with some more parameters but those plans ended up not coming to fruition, though I might finish those plans at a later date.
Also bit of a disclaimer: I am not that much of a math and statistics kind of person, I am just a programmer that likes playing with numbers and I might might make a few mistakes here and there.
How Anarchy Ranking Works
For quick reference, and because the wiki has this information split up in two pages, I also will explain in this thread how Anarchy's ranking system works. I have wrapped the explanation in an expandable spoiler tag for those who do not care and just want to see the juicy data.
Sources: https://splatoonwiki.org/wiki/Rank, https://splatoonwiki.org/wiki/Anarchy_Battle
Rank
Anarchy has 11 total ranks, them being C-, C, C+, B-, B, B+, A-, A, A+, S, and S+. Each of the letters are split into groups (e.g C-, C, C+) with the exception of S and S+ which are each their own group themselves. I will be calling these groups "brackets" going forward.
Ranking Up
In order to go up a rank within your bracket, you must accumulate enough points depending on your current rank. There is no way to rank down.
When you reach the point threshold for the highest point in your bracket, you will have to participate in a rank-up challenge where you must win 3 times before you lose 3 times. Rank-up challenges do not cost more than a normal series, contrary to my own memory 🫠.
Series
A series is a sequence of games consisting of up to 5 wins or 3 losses. The more wins you get in a series, the more points you will earn at the end of the series. The earnings are exponential. These earnings are based on the bracket you are in.
Cost
The cost to begin a series is based on the letter rank in your bracket.
Yes, the individual letter rank determines point cost, which means that if you, for example, reach A+ and then tank yourself all the way down to 0p you will still have to pay the price that an A+ player would pay in the place that an A- player would be in order to enter another series while only gaining points at the same rate. Take this information as you will
.
Medals
The medals you earn in battle add to the points you gain at the end of a series. A gold medal is worth 5p, a silver medal is worth 1p, and no medal is worth, well, nothing. These points can add up to 15p.
Anarchy has 11 total ranks, them being C-, C, C+, B-, B, B+, A-, A, A+, S, and S+. Each of the letters are split into groups (e.g C-, C, C+) with the exception of S and S+ which are each their own group themselves. I will be calling these groups "brackets" going forward.
Ranking Up
In order to go up a rank within your bracket, you must accumulate enough points depending on your current rank. There is no way to rank down.
| Rank | Breakpoint |
|---|---|
| C- | 200 |
| C | 400 |
| C+ | 600 |
| B- | 350 |
| B | 600 |
| B+ | 850 |
| A- | 500 |
| A | 800 |
| A+ | 1100 |
| S | 1000 |
| S+ | ∞ |
When you reach the point threshold for the highest point in your bracket, you will have to participate in a rank-up challenge where you must win 3 times before you lose 3 times. Rank-up challenges do not cost more than a normal series, contrary to my own memory 🫠.
Series
A series is a sequence of games consisting of up to 5 wins or 3 losses. The more wins you get in a series, the more points you will earn at the end of the series. The earnings are exponential. These earnings are based on the bracket you are in.
| Rank Bracket | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| C | 20 | 45 | 75 | 110 | 150 |
| B | 30 | 65 | 105 | 150 | 200 |
| A | 40 | 85 | 135 | 190 | 250 |
| S | 50 | 105 | 165 | 230 | 300 |
| S+ | 50 | 105 | 165 | 230 | 300 |
Cost
The cost to begin a series is based on the letter rank in your bracket.
| Rank | Cost to Enter Series |
|---|---|
| C- | 0 |
| C | 20 |
| C+ | 40 |
| B- | 55 |
| B | 70 |
| B+ | 85 |
| A- | 110 |
| A | 120 |
| A+ | 130 |
| S | 170 |
| S+ | 180 |
Yes, the individual letter rank determines point cost, which means that if you, for example, reach A+ and then tank yourself all the way down to 0p you will still have to pay the price that an A+ player would pay in the place that an A- player would be in order to enter another series while only gaining points at the same rate. Take this information as you will
.Medals
The medals you earn in battle add to the points you gain at the end of a series. A gold medal is worth 5p, a silver medal is worth 1p, and no medal is worth, well, nothing. These points can add up to 15p.
Simulation source code here if you are also a nerd like me: https://gitlab.com/9ball/splatty3-rank-simulation
For the first simulation, we will have 100,000 players playing in back-to-back series over 1,000 games, which I think should be aproximately the amount of games the average dedicated player should be playing per season. Every player has an 50/50 chance of winning/losing each game, and medals are awarded completely at random.
| Rank | % of players | % of players (bracket) |
|---|---|---|
| C- | 0% | |
| C | 0% | 0% |
| C+ | 0% | |
| B- | 0% | |
| B | 0% | 0.01% |
| B+ | 0.01% | |
| A- | 0.08% | |
| A | 2.14% | 21.90% |
| A+ | 19.68% | |
| S | 67.92% | 67.92% |
| S+ | 8.16% | 8.16% |
It appears that most of the player-base landed themselves in S rank! This might not come as a surprise since its really well known that this ranking system has a really strong upwards bias, especially with the lack of a way to naturally rank down. This feature of the Anarchy ranking system leads to a problem belovedly known by the community as "rank debt", and it appears that these simulated players are no stranger to it! Debtors, highlighted in orange, make up the grand majority of the simulated population.
A simple solution to this problem that many have proposed is to simply just have people naturally rank down when they lose enough points, so I implemented my own rank down system in the simulation.
How my rank down system works is that if after a series a player is below 0p, they will be sent down to the middle rank of the lower bracket (e.g. A- 0p => B 350p). I also made it possible to rank down within a bracket (e.g. A+ => A) This is how a simulation of that system went with the same parameters as before:
| Rank | % of players | % of players (bracket) |
|---|---|---|
| C- | 0.0% | |
| C | 0.01% | 0.03% |
| C+ | 0.02% | |
| B- | 0.17% | |
| B | 2.90% | 12.06% |
| B+ | 9.00% | |
| A- | 7.00% | |
| A | 18.89% | 62.68% |
| A+ | 36.79% | |
| S | 23.14% | 23.14% |
| S+ | 2.10% | 2.10% |
Then for curiosity's sake, with 10,000 games.
| Rank | % of players | % of players (bracket) |
|---|---|---|
| C- | 0.0% | |
| C | 0.01% | 0.02% |
| C+ | 0.01% | |
| B- | 0.15% | |
| B | 2.80% | 11.65% |
| B+ | 8.71% | |
| A- | 6.81% | |
| A | 18.89% | 62.41% |
| A+ | 36.71% | |
| S | 23.57% | 23.57% |
| S+ | 2.34% | 2.34% |
That's a bit surprising! Its really interesting how adding just a simple rank down feature to anarchy can cause the ranks to spread out like that, and consistently across a large number of games too. A few tweaks to the reward system could make the spread more even but I'll leave that for another day.
Well, that's all I have to show for now. I did plan on poking at the simulation a bit more with some more parameters but those plans ended up not coming to fruition, though I might finish those plans at a later date.
