Since people are talking about the LLL format again, and the model is doing fairly well these days, I’d like to offer some possible variants on the random-deck theme, using the model to rein in the usual level of matchup variability. Since these would require use of the model to run, I’d be happy to run a tournament that used these.
Format names are a work in progress.
- Deathstrike Deathmatch: Decks are randomly chosen for each match, weighted by the probability of their being the best deck (possibly with a cut-off, since there’s still a large spread on this). This favours the stronger decks, so we could expect a lot of Black starter and Necromancy, but the spread is different for first and second player. No mercy.
- One Hand Tied: Decks are randomly chosen for each match, weighted by the probability of their being the worst deck (again, possibly with a cut-off). This would favour the lesser-seen specs, such as Ninjitsu.
- Midori’s Balanced Bouts: Decks are randomly chosen for each match, such that the resulting matchup is considered to be roughly fair, before accounting for player skill levels. In theory, this would result in matches that come down to player skill much more than deck selection.
- Midori’s Final Balanced Bouts: Decks are randomly chosen for each match, such that the resulting matchup is considered to be roughly fair, taking player skill levels into account. In theory, this should make for lots of interesting matches. If the players are unequal in skill, we could expect the stronger player to get some odd decks to work with.
All formats exclude mirror matches, for the sake of variety.
We can expect some weirdness around decks the model doesn’t understand well, but it means we can see decks that have never appeared in a tournament before.
Thoughts?
3 Likes
I personally don’t really want to see a tournament of black and necromancy everywhere. Otherwise, I’m game.
4 Likes
I agree, I’d be more interested in the other formats. All those Bargains in CAMS20 were the most stressful matches I’ve ever had. I thought it was worth adding for completion’s sake, however.
1 Like
#4 is the most interesting to me, but what do you do if there’s players you have insufficient data on?
Same thing, it’s just more likely the match turns out to be lopsided in retrospect. Without more data, that’s the best I can do. I could update the model between rounds to try and help with that.
1 Like
I favor option 3, then 2, then 4, then wayyy down 1.
5 Likes
Since most of the variants are about increasing deck variety, I’ve added a rule that mirror matches are excluded.
1 Like
I think next tourney should be Option 3 here.
1 Like
If I’ve got time, I’ll put up some sample matchups for Option 3, to show what participants could expect.
1 Like
It has been awhile since we ran a random decks tourney, I’d be down for that
1 Like
Still open to better names. I should’ve called option 4 Final Balance Buster 
Revisiting this, I think #2 sounds a lot better to me now. I just think it would be really fun to see underrepresented specs and also having to struggle with jank.
my ranking of preference is pretty even across 3 2 and 4, as before. Just voting against 1, really.
1 Like
Yeah I’m with @zhavier, slight preference for 3 and 2 over 4 but they’re very tight and a clear step above 1. We see plenty of tryhard decks, I want weird for XCASS events.
@charnel_mouse how would you feel about running XCAPS21 with one of those formats?
Comfortable, I think. I’ll see how easily I can generate deck pairings first, but if that’s straightforward I can plug pairings in once I’ve done the Swiss pairings. Will probably need help on running things in general, I’ve never done it.
1 Like
I’m happy to walk you through the spreadsheet Zhav made to help w/ matchup pairing/posting/tracking and how I use it
1 Like
Some example pairings for Option 3:
| P1 |
P2 |
| [Balance]/Blood/Finesse |
[Balance]/Law/Present |
| [Truth]/Disease/Present |
[Ninjitsu]/Anarchy/Future |
| [Blood]/Discipline/Past |
[Finesse]/Fire/Necromancy |
| [Peace]/Blood/Fire |
[Truth]/Demonology/Disease |
| [Finesse]/Blood/Necromancy |
[Fire]/Balance/Strength |
| [Disease]/Discipline/Fire |
[Necromancy]/Balance/Fire |
| [Blood]/Finesse/Law |
[Finesse]/Blood/Strength |
| [Blood]/Discipline/Past |
[Past]/Balance/Finesse |
| [Balance]/Blood/Ninjitsu |
[Blood]/Feral/Necromancy |
| [Future]/Peace/Truth |
[Balance]/Blood/Strength |
| [Law/Truth]/Present |
[Blood]/Law/Ninjitsu |
| [Fire]/Past/Truth |
[Necromancy]/Fire/Present |
| [Past]/Balance/Peace |
[Peace]/Feral/Finesse |
| [Growth]/Bashing/Disease |
[Feral]/Discipline/Truth |
| [Law]/Fire/Growth |
[Bashing]/Present/Strength |
| [Peace]/Demonology/Growth |
[Bashing]/Peace/Strength |
| [Law]/Demonology/Feral |
[Disease]/Bashing/Fire |
| [Bashing]/Blood/Future |
[Law]/Bashing/Necromancy |
| [Peace]/Future/Past |
[Fire]/Ninjitsu/Truth |
| [Bashing]/Demonology/Disease |
[Strength]/Disease/Growth |
For Option 2, we can have separate deck sample lists for Player 1 and Player 2, since I already calculate Nash weights for them separately, as well as overall weights.
I still need to work out how to sample them, but here are the ones with smallest weight for each.
Player 1:
[Feral]/Future/Past
[Future]/Law/Truth
[Feral]/Law/Ninjitsu
[Balance/Feral]/Peace
[Future/Present]/Disease
[Future]/Fire/Truth
[Past]/Fire/Truth
[Future/Present]/Truth
[Peace]/Feral/Future
[Past]/Disease/Truth
Player 2:
[Law]/Ninjitsu/Present
[Discipline]/Law/Present
[Feral]/Finesse/Past
[Fire]/Discipline/Present
[Discipline]/Demonology/Present
[Discipline]/Growth/Present
[Discipline]/Necromancy/Past
[Discipline]/Future/Present
[Strength]/Bashing/Disease
[Peace]/Discipline/Fire
I’m all for creating some fair deck match-ups and interesting matches. Option 3 seems to have a distinct advantage for doing that because of the way your model considers the sides of the match-up in relation to each other (if I understood some of the data thread correctly); option 2 won’t do that [reference needed :)]
If people still wanted to specifically see decks in use that are normally too ‘bad’ to see much use, is it possible to have an option 2.1 where you pick one side of the match-up as above and then the other side to be a fair opponent?
Just seems a shame not to use the model to its full fair-match-making capacity in option 2
Hmm, maybe. I’ll draw up some expected opponents for the above Option 2 draws.
