TSLB Drafting

[Rab]

There are different ways to do TSLB drafting, and it's not clear which is best suited to this league. What do we want from it? Are we happy enough with how it currently works? Discuss...

Season 1 - Attempt 1
- Initially the terrs would be distributed between the teams, then the sharks would be distributed between the teams, then the spiders would be distributed between the teams.
- Imagine it like a captain picking an average shark over a top-tier spider.
- That made some very imbalanced teams.

Season 1 - Attempt 2
- In response to imbalanced teams, drafting was changed to distribute players between the teams based on their overall rating.
- Overall ratings don't guarantee that everyone gets a complete team (1 terr, 2 sharks, 4-5 spiders).
- But it does generally work 'okay'.

Rab Draft
- I wrote a proof of concept to see if I could do a better job by trying to make the bot pick like a real captain would.
- Draft is a snake draft (like TWDT).
- Each captain behaves the same way.
- Captains try to get the best rated players available, but also prioritise certain ships. (requires per-ship ratings)
- As rounds of picking progress, your standards lower. (e.g. in round 1 you won't pick a mid-rated terr, but in round 3 you might have to).
- Have awareness of which of the teams is your opponent, and respond to what they pick (so you don't get out-sharked for example).
- Almost guarantees a complete team (1 terr, 2 sharks, 4-5 spiders).
- Almost guarantees the maximum value from your picks (assuming ratings are accurate).
- Knows what role people were picked for, so can include that in the draft results output.

Discussion:
- Rab-draft is effective, *if* your objective is to pick the most competitive teams possible with the players available.
- Some people want to play certain ships which aren't necessarily their best ship. If in Rab-draft someone chooses not to play the ship they were picked for, you can end up with a worse result than the current draft.
- I saw some people trying to specify what ship they wanted to play, saying things like "!p terr". This feels nice to an individual. But to a bot trying to create 2-4 solid teams it's a real pita.
- If there's only enough players for 2 teams, would captains prefer to pick manually?

[Jessup]

I don't quite understand everything about your snake draft idea Rab but I do like how you are thinking when it comes to ratings progressing as a season progresses. I am a BIG fan of a sliding rating scale on performance as the stats collect. I think a rigid frozen rating system is far from ideal and causes harm because it is way too static.. Over rated and under rated players will continue to wreck chaos into matchmaking if we can't adjust ratings based on performance as a season progresses.

[ogron]

Proposals for TSLB matchmaking have already been submitted for months, there's no point in discussing them when Qan won't implement them.

It's the same crappy sorting from last season.

It was a big priority for me, but unless I can get a coder to replace Qan, nothing can be done.

[IxaDor]

I like this concept Rab, thanks for your input. I think if enough of us come forward, we might be able to get it implemented next season. Static/mysterious ratings are horrible when you know you can perform well in another ship but have 0 chance to play in it because of them, that and you have 0 chance to change them, even if you have improved since they were created. I don't know if this is the end all solution but it is a great step forward. I am excited and hope this gets some attention from Qan. Have a great day. Talk to you more later. Bye!

-IxaDor

[Rab]

There's no lack of attention from qan. The thing holding up progress is deciding what the right thing to build is.

I was gearing up to implement Rab Draft myself when Ogron decided he didn't want to terr anymore and announced out of the blue a ship retirement thread (since retired). These things weren't accounted for (and I think can't be accounted for) so I had no choice but to wait.

- People were saying they wanted fairer teams. This means less flexibility.
- Then people said they didn't want to play their best ships. This means more flexibility.
- I don't see a way to deliver fairer teams, and more flexibility, at the same time. So I'm trying to get you guys to think about it and decide what's most important to you.

In the absence of a clearly best answer, the best action is no action.

The draft can only be as good as the ratings, perhaps the current draft would be fine with dynamic ratings. Either way it impacts the draft, so we can't change one without considering the other.

[Dwopple]

I think the Rab Draft idea is a solid step in the right direction. However, I think if the goal is to prioritize best rated players, AND per-ship ratings you have to introduce another variable into the draft algorithm. This might require some politicking from the players themselves. Examples of variables are: hidden variables like # of games played in X ship OR similar to the !p terr command. In the latter example, you could introduce ship preference requests i.e. !p 5,8 and set a limit for ship preference voting to influence the draft, but ultimately leave it to the (C) to decide...

Not gonna lie i'm a little tipsy right now so not 100% this makes sense. Either way, I think an extra variable is needed for the algorithm.