Releases: philihp/openskill.js
v3.1.0
v3.0.1
v3.0.0
- Breaking Changes Change the way alternate models are used. Rather than sending in a string/name of the model, you now import the model function itself and pass that in. This enables people to extend the library and experiment with models that support partial play.(#296) c477d12, as an example of how your calls would change, see this test
- Breaking Changes Removes support for node 11-13 (#286) 5bb0a94, this version will also support Node 14/es2020. Next major version will not be until sometime after 2023-04-30 when 16/es2021 is the current LTS compile target.
- Adds Typescript types (#240) 8e46c2a
v2.1.0
Thank you @JLaferri! New options:
tau
(defaults to 0): Additive dynamics factor, which keeps a player's rating from getting stuck at a level. Normally, a player's sigma will only decrease as we gain more information about their performance. This option will put some pressure on this back up. This default will change to be sigma/100 with v3, to be more congruent with TrueSkill, but higher may make your rating system more exciting.
preventSigmaIncrease
(defaults to false): for a tau > 0, it is possible that a player could play someone with a low enough rating that even if they win, their ordinal rating will still go down slightly. If your players have no agency in matchmaking, it is not desirable to have a situation where a player goes down on the leaderboard even though they win.
rate([team1, team2, ...], {
tau: 0.083333,
preventSigmaIncrease: true,
})
v2.0.0
v1.6.0
- feature: add
predictDraw
(#227) f15bc4f - feature: add
predictWin
(#216) b479417 - fix: typo, thurston should be thurstone (#214) a0696e0
- docs: how to predict (#229) 28dc441
- docs: Update link to Kotlin implementation in README.md (#215) 28da195
- docs: add lua implementation fad75fa
- docs: Add contributing doc (#204) 0e5e4d2
- docs: Readme link to python lib (#197) 8b5a7a7
v1.5.0
- fix: hide benchmark/package.json from libraries.io (#192) 730ff12
- fix: husky 7 configuration (#190) 6b19946
- feature: add explicit support for Node 16 (#171) b281401
- feature: hint to bundlers that this has no side effects (#152) 42c2a6f
- refactor: instead of reorder, use sort-unwind (#150) 8b16ab6
- tests :add some 5v5 tests 9186ccd
- feature: add win probability function for 1v1 (#140) 0211569
v1.4.1
- fix that occurred on using the Thurston-Mosteller model with tied ranks, due to the wrong epsilon used in the gaussian approximation functions (#133) 206b46f
- fix for Plackett-Luce to respect a custom gamma function
- added some tests of a series of 3 games run through the reference
rank-1.02
program and using their numbers as a basis for unit tests here