API server for OAA, calculating MMR and giving people fake internet points.
npm i to install the node dependencies
Copy .oaaserverrc.example to .oaaserverrc and edit the values to be correct. To test locally against OAA running in tools mode, simply edit the URL override at the top of bottlepass/server.lua to point to the IP address where your test Bottlepass server is running.
Bottlepass is designed to reward simply playing the game. You only get Bottlepass experience if you finish the game, and it's designed to not heavily favor winning teams. It's supposed to take months of dedicated play to reach max level.
It takes 100 xp to level up your Bottlepass, and the level cap is 100 with an option to prestige and reset to 0. At the end of each game with 10 players, each connected player receives:
- 1 xp for every 2 minutes the game lasts.
- If the game lasts at least 25 minutes then there is a “Full Game Bonus” which is random between 20-30.
- 50% xp bonus for hosting a lobby
- +5 First game of the day bonus
- +5 first win of the day bonus
- 100% Bonus XP to everybody in the lobby if someone in the lobby is playing one of their first 3 games
MMR is hard. No system is perfect, and no system truly works well without matchmaking. This is our system that we built anyway.
Our MMR system is a version of the ELO system, modified to fit the 5v5 format. The basic principle is to predict odds that a team will win, and correct their MMR based on the results. The more favored a team is, the less they gain on a win and the more they lose on a defeat. Conversely, the projected loser doesn't lose as much, and will get a large MMR jump on an upset.
The amount of MMR distributed to each player is based on a few things. First, both teams' odds, as measured by their K values, determine how much MMR is at stake (as detailed below). Then, each player's MMR outcome is based on their current MMR compared to their teammates' value; high MMR players in lower tier teams have less at stake. Even games with even teams (such as organized scrims) have the greatest MMR movement.
ELO revolves around a predictive function that, given two competing scores, gives the expected probability the first score wins. When a match is completed, the expected and actual (0 on a loss, 1 on a win) outcomes are compared. The difference between them is then multiplied by the "K value" of the participant to obtain the MMR change (note that, for the loser, that will be a negative number).
The K value is essentially the rate at which one moves through the rankings. Our K value system is taken from chess, where there's a large K value for lower MMR players then it scales downwards until "master" level, where it hits the lowest value and stops changing. This makes it easier for a player to reach their actual skill level, whereupon they stabilize.
These are the current K-values, per MMR bracket (do note these are subject to change, as we tune the system):
| MMR Range | K value |
|---|---|
| below 500 | 80 |
| between 500 and 1000 | 40 |
| between 1000 and 2000 | linear 40 to 20 |
| 2000 or above | 20 |
The ELO function is 1 / (1 + Math.pow(10, (score1 - score0) / 400))
In LaTeX: 10^{-\frac{\mathrm{YourScore}-\mathrm{TheirScore}}{400}}
Scores are multiplied by 2 to make up for the additional pass through the ELO function