Suspect Tests: Stop having game limits, also I'm removing COIL

Zarel

Not a Yuyuko fan
is a member of the Site Staffis a Battle Server Administratoris a Programmeris a Pokemon Researcheris an Administrator
Creator of PS
#1
Hi, this thread is mainly for anyone currently running a suspect test.

Don't use maximum game-count limits. If you're using a game limit, you're using COIL wrong.

I've heard that it's common to have game limits. This is really really bad, and means that COIL is working exactly the opposite way from how it's supposed to work. To prevent further abuse, my plan is to remove COIL support from PS entirely. Future suspect tests should be run with a minimum GXE score and a minimum game-count.

From discussions with tier leaders, I am currently mostly convinced that there's no need for COIL at all - GXE is already a convergent score. It looks like having a tunable convergence speed is more of a frustration to players than a useful feature. Sorry, Antar. Feel free to use this thread to discuss any reasons a minimum GXE + minimum game-count might not be good enough for your use-case.
 

Zarel

Not a Yuyuko fan
is a member of the Site Staffis a Battle Server Administratoris a Programmeris a Pokemon Researcheris an Administrator
Creator of PS
#2
Josh (who doesn't have access to PR) asked outside of PR:
isn't the purpose of COIL to just represent how close you are to achieving reqs in a more comprehensible way?
I'm responding here in case anyone else wants to know:

No, definitely not. Imagine: "My GXE is 85 and I've played 35 games, and the suspect test requires a GXE of 82 and 50 games. So I'm 15 games away from achieving reqs." COIL is definitely less comprehensible than that.

The actual purpose of COIL is to make it so better players have to play fewer games to make reqs than worse players, and to be able to tune the B factor to adjust the number of games they have to play.

But in practice, the fact that very borderline players can get in by playing a very large number of games is distasteful to tier leaders. This can be fixed while preserving the other properties by tweaking the COIL formula, which, to be clear, I am willing to do if someone asks for that.

But unless a tier leader asks for it, I don't think the added complication of supporting COIL is actually worth it. Letting better players qualify in fewer games is nice and all, but it doesn't seem worth all the confusion about what a good number for a minimum COIL is, what COIL scores are reachable in how many games, how to tune the B factor, or all the other things people misunderstand about COIL.
 
Last edited:

DoW

formally Death on Wings
is a Pre-Contributor
#5
From my limited experience on the monotype council (before it was an official meta), the method used for determining the B-value was to determine which GXE you wanted qualifying players to need to get, and calculate from there. This therefore seems like it will do precisely the same job, while making it more obvious to the players what the council requires of them.

Apologies if this post does nothing other than stating the obvious, but it's something that hadn't been stated outright and I thought it was worth putting to paper the fact that (unless other councils determine the B value very differently) this won't change anything fundamental about the way reqs are chosen.
 

Zarel

Not a Yuyuko fan
is a member of the Site Staffis a Battle Server Administratoris a Programmeris a Pokemon Researcheris an Administrator
Creator of PS
#6
From my limited experience on the monotype council (before it was an official meta), the method used for determining the B-value was to determine which GXE you wanted qualifying players to need to get, and calculate from there. This therefore seems like it will do precisely the same job, while making it more obvious to the players what the council requires of them.

Apologies if this post does nothing other than stating the obvious, but it's something that hadn't been stated outright and I thought it was worth putting to paper the fact that (unless other councils determine the B value very differently) this won't change anything fundamental about the way reqs are chosen.
Yeah, I'm nearly sure that's not how Antar intended it to be used.

The minimum GXE you want qualifying players to get is your min COIL/40. The B factor just determines how many games they should have to play to qualify.
 

Zarel

Not a Yuyuko fan
is a member of the Site Staffis a Battle Server Administratoris a Programmeris a Pokemon Researcheris an Administrator
Creator of PS
#7
Also, allow me to introduce: Poor Man's COIL:

Your suspect test requirements can be something like:

GXE > 80
GXE + battle count > 130

This will also make it so better players can finish in fewer games, but without needing everyone involved to understand the math behind COIL.
 

Zarel

Not a Yuyuko fan
is a member of the Site Staffis a Battle Server Administratoris a Programmeris a Pokemon Researcheris an Administrator
Creator of PS
#10
make ladders great again plan
1) force good players to play more ladder games for suspect tests
2) ladder quality increases as more average players face good players (and learn from their experiences) and good players face good players
3) ladder suspect tests make sense again
Honestly, just not forcing good players to stop playing because they'll go over the game limit is probably already a huge benefit.
 

Users Who Are Viewing This Thread (Users: 1, Guests: 0)

Top