Maybe its time to move the Ladder starting skill down.

[IceDreamer]

This is a fundamental problem with treating TrueSkill as a ladder system. Tokyto is drafting a new ladder system which doesn't use TrueSkill directly for bracketing, which would solve this problem.

The proposed solution of just lowering the initial rating would just defer the systems ability to estimate the skill level of newer players until much later -- which doesn't really solve the problem of mid-range players being afraid of getting matched against smurfs.

I have never properly understood this line of thinking.

- Players start with a 1500 mean rating
- 1500 mean rating with 1000 games (So, an ACTUAL 1500 player)... These players are good. They are powerfully skilled.
- SupCom is one of the hardest RTS games out there
- 99/100 players (Yes, I'm guessing or whatever, but the point stands) are absolutely terrible when starting in FAF
- AI trained players think they may have an advantage, but really they don't have that much of a head-start on a brand new player

All this leads to the inevitable conclusion that yes, the system wants to adapt a player's numbers to fit their place, but that idea just... It doesn't WORK in the majority of cases. People start out shit! They should therefore be presumed to be shit.

This has been a complaint that pops up again and again, since the start of FAF, year after year. There HAS to be a reason for that!

Oh, and the proposed ladder won't fix the complaint here: New players quit before they really begin because they get utterly obliterated in games that are completely unfair. The new system can hide the numbers, but it won't change the outcome of the games.


EDIT: Going by the maths and the idea behind it, TrueSkill and ELO seem to have been designed for the CS:GOs and Overwatches of the world. Games with LARGE player bases, either established or anticipated. Games where the primary limiting factor is mechanical skill. Games where there are going to be players, and not in small numbers, coming from a decade or more of honing that skill. Games where a lack of knowledge is both relatively inconsequential and very quickly overcome. Games where skills other than mechanical (Think gamesense, tactics, objective-based gameplay, watching a minimap) can all be transferred in abundance.... The sensible thing here is for a new player to be slotted in in the middle, and find out where they sit. You can't just assume everyone, or most, fall into the "Potato" bracket.

This doesn't seem to fit us.

[Sheeo]

- Players start with a 1500 mean rating

Players start with a rating of (1500, 500), meaning "0" (1500 - 3*500) displayed/used rating.

- 1500 mean rating with 1000 games (So, an ACTUAL 1500 player)... These players are good. They are powerfully skilled.
- SupCom is one of the hardest RTS games out there
- 99/100 players (Yes, I'm guessing or whatever, but the point stands) are absolutely terrible when starting in FAF
- AI trained players think they may have an advantage, but really they don't have that much of a head-start on a brand new player

All this leads to the inevitable conclusion that yes, the system wants to adapt a player's numbers to fit their place, but that idea just... It doesn't WORK in the majority of cases. People start out shit! They should therefore be presumed to be shit.

This has been a complaint that pops up again and again, since the start of FAF, year after year. There HAS to be a reason for that!

I don't think that the inability of trueskill to know your rating before you've played any games means trueskill is a bad rating system, or the idea of statistically deriving your skill "doesn't work".

Oh, and the proposed ladder won't fix the complaint here: New players quit before they really begin because they get utterly obliterated in games that are completely unfair. The new system can hide the numbers, but it won't change the outcome of the games.

I'm not sure which proposed ladder you're talking about here, could you be explicit?

I think the replacement ladder should properly bracket, and before initial placement of a new player into a bracket, it should only match against players already in the lower brackets. This should accomplish exactly what's being asked for, without needlessly messing with trueskill internals.

[Sheeo]

The proposed solution of just lowering the initial rating would just defer the systems ability to estimate the skill level of newer players until much later -- which doesn't really solve the problem of mid-range players being afraid of getting matched against smurfs.

Mid-range players already have many reasons to like FAF . They won't stop playing forever because of some fear from smurfs. Extreme newbies on the other hand are just evaluating the game - after some merciless beating they can decide that FAF ladder just isn't for them.

I'm not suggesting we allow that. But this isn't the way of fixing that.

[Sheeo]

This is a fundamental problem with treating TrueSkill as a ladder system. Tokyto is drafting a new ladder system which doesn't use TrueSkill directly for bracketing, which would solve this problem.

Please don't break the current matching system, I play ladder occasionally, not regularly and if I have to grind against noobs to get a decent game, I will never play ladder.
No ELO/ladder point matching, or at least have a couple placement matches to fix that shit if you do.

I don't believe I suggested not to use TrueSkill for matchmaking. I'm suggesting not to use it as the sole basis for placing you on a ladder ranking table, yielding exactly the problems described in this thread.

[RealityCheck]

Sheeo, I assumed changing the starting skill is as simple as changing two or three numbers. Is it actually much more complex?

[sasin]

This is a fundamental problem with treating TrueSkill as a ladder system. Tokyto is drafting a new ladder system which doesn't use TrueSkill directly for bracketing, which would solve this problem.

The proposed solution of just lowering the initial rating would just defer the systems ability to estimate the skill level of newer players until much later -- which doesn't really solve the problem of mid-range players being afraid of getting matched against smurfs.

Why not lower the mean but keep the deviation high? Mean of 800, deviation of 500, or some such.

[rtc]

EDIT: Going by the maths and the idea behind it, TrueSkill and ELO seem to have been designed for the CS:GOs and Overwatches of the world. Games with LARGE player bases, either established or anticipated. Games where the primary limiting factor is mechanical skill. Games where there are going to be players, and not in small numbers, coming from a decade or more of honing that skill. Games where a lack of knowledge is both relatively inconsequential and very quickly overcome. Games where skills other than mechanical (Think gamesense, tactics, objective-based gameplay, watching a minimap) can all be transferred in abundance.... The sensible thing here is for a new player to be slotted in in the middle, and find out where they sit. You can't just assume everyone, or most, fall into the "Potato" bracket.

This doesn't seem to fit us.

the Elo rating system was designed for chess, over 50 years ago, and is still used by FIDE today

[IceDreamer]

OK so I have been educated in some of the nuances of how the system works.

Long and the short of it is as follows:

- This is a problem right now
- Lowering certain values would, in the short term, produce the desired effect for new players
- Lowering those values would also promote smurfing
- In the long term, this action would have an impact on the overall TrueSkill FAF-planet of unknown scale, but certainly in a bad way
- The correct solution is the proposed ladder system, where "Brackets" are largely composed of hidden rating counts, but new players are able to be matched only against the bottom few brackets without the long-term negative effects on the underlying system.

[Blodir]

The real short term solution is to force match new players against low rating players, without touching trueskill values (which would come with unpredictable consequences)

[biass]

Prehaps a good time to push for ladder leagues / divisions if you're breaking making some new system?

Would be nice to have heard some stuff about it instead of not being able to provide feedback and then having it forced on us later