MMR Matchmaking Planning Thread. Nerds welcome.

[Cape]

While this is being manually hosted, and before rating players is tested, public ratings and leaderboard should not exist, IMO.

[qan]

Teams should still be balanced in pairs in every arena. Use !st / !showteams on MMRBot to see the distribution.

I added an experimental option to create tiered teams for hosts: !mtt / !maketeamstiered This will group players together by general skill level, but at the cost of some balance. Still, ratings adjustment is based on the balance of the teams, so this isn't that big an issue. Technically any two teams can play one another and the ratings change will be relatively fair. Note that new unrated players who are far above/below 3000 will especially skew this.

Something's wrong with the adaptive class loader so it's not loading the new changes for some reason, but this will be in eventually.

[Cape]

Bot Note: Instances when multiple games run simultaneously; Bot should place all higher rated players in the same game.

[Cape]

As for dev time, work is slowing down for me a bit, fortunately, but I still don't have a ton of free time..

I recommend Terring on your free time

[Rab]

matchups seem to get less balanced when there are 4 teams, not entirely convinced it's matching 1v2 and 3v4

[Rab]

*thumbs up*

[qan]

Thanks for the heads up on that. I expanded the query to top 50. (This is all a temporary solution, though. Will need to tweak how we handle displaying ratings at some point.)

Base code still isn't in yet. BIET implemented Rab's basing alg into TSLBot, but this was right before I forked off MMRBot. There are some small differences between the two, so it means it's not just directly dropping the basing alg from TSLBot into MMRBot, but I don't think it would be too much of a challenge. IIRC it does still need testing, however.

We could have the bot select teams with recommended ships but still allow the captain to arrange teams as desired, especially for the early matches. Base is the place where matches between unrated players are bound to be the most lopsided, I think. Also, captaincy among unrateds would be a problem unless we use prior data (such as base stars) to assign captains. There hasn't been too much discussion on base recently, though. As for dev time, work is slowing down for me a bit, fortunately, but I still don't have a ton of free time.

For those asking about base ratings, yes, they're designed as split into 3 ships.

The biggest issue with base is that switches and subs are common, so unless people are willing to play without them, it needs a solid solution on how to handle those with regard to ratings adjustments (especially if using individual ship ratings). Rating changes are based on comparison of strength of the two teams at the start of a match, currently. I have a basic plan for how to handle subs/switches: essentially taking ratings snapshots before each sub/switch and applying ratings changes as a percentage of time played of the total round time for each ship. Not certain if that's mathematically sound, but it may be good enough for our purposes, so long as it can't be gamed.

[Rab]

there was a guy with 80 wb games on the mmr web page yesterday, now he's not there, how does this happen, are we sure the website's displaying everyone?

[Riverside]

According to whom? Why cant a bot just split 16 people into what it thinks are two even teams of 8 and assign the highest ELO player a cap (or even have people pm !cap to bot) who picks ships? Could even bake in trading so, say, Freq 1 could give Freq 0 a shark in exchange for one of their spiders. It doesn't have to be a grand multi-ship rating system.

Pirhips

personal preference is letting the bot maintain the matchmaking, and considering ways to move that forward

[Vehicle]

there are no player captains making line-up decisions for teams in mmr

The bot matches players to 2 (4, 6, etc) teams using individual ratings

According to whom? Why cant a bot just split 16 people into what it thinks are two even teams of 8 and assign the highest ELO player a cap (or even have people pm !cap to bot) who picks ships? Could even bake in trading so, say, Freq 1 could give Freq 0 a shark in exchange for one of their spiders. It doesn't have to be a grand multi-ship rating system.

[Riverside]

Like what 1800 level spider is a 900 level shark? And what captain wouldn't know that? If people want to play outside their main (and therefore rating-related) ship, that's up to the captain to take the handicap.

there are no player captains making line-up decisions for teams in mmr

The bot matches players to 2 (4, 6, etc) teams using individual ratings

[Vehicle]

Should just say screw ships and implement the current MMR system or a tweaked version that's purely based on W:L / amount of stomp. It's not going to make that much of a difference when you consider the ecosystem of players. Like what 1800 level spider is a 900 level shark? And what captain wouldn't know that? If people want to play outside their main (and therefore rating-related) ship, that's up to the captain to take the handicap.

I'd personally really like to see MMR basing but I do understand that it's much more difficult than a dueling system.

Edit: And do set times already. You'd get a lot more participation and potentially even people logging on during the week if you just said "Tuesday and Thursdays from 8-10 is MMR time". I don't see why not. And before you say "We're all volunteers blah blah," you and I both know you're gonna be online anyway.

[Riverside]

Can we get a base MMR bot? Have it go with TWBT ratings until 20 games or something.

qan , for basing, would it possible to create a temporary solution and separate the mmr basing rating of players into 3 ships?

Terrier, spider, shark

!p 5, 8

!p gives the default spider option (that includes players who wanted to shark or terr)

[beam]

https://trenchwars.org/mmr/

I noticed all the wbs are over 3k, and most of the javs too.

I thought mmr was meant to be zero-sum, am I wrong?

that's only the top 20 rated players with minimum 20 games played not a full list of everybody

[Rab]

https://trenchwars.org/mmr/

I noticed all the wbs are over 3k, and most of the javs too.

I thought mmr was meant to be zero-sum, am I wrong?