[Fixing Problems] Tier System

Kyokudai

Hello fellow MTGPQs.  I always see topics in General about problems in the game and have started feeling that the statements of the problem are in fact... part of the problem.  Mainly that they aren't specific enough to what the actual problem is, and this seems to lead to misunderstandings as they get handed over to the development team.  General also doesn't feel quite the right place for them but all the other forums seem very dead.  Here's hoping that isn't the case.

So, I was hoping to have a much more detailed breakdown of some specific issues so we can clearly state a problem and a solution. 

-----

Let's start with something simple that I think is actually a problem.  The Tier System

Problem Qualification:

1. System is lacking a well formed statement for it's purpose in the game (both explicit and implicit)
2. As a matchmaking component (if that is a purpose) it is far too coarse and un-focused to actually match
3. System is inherently flawed as it allows gaming of the system

Goals:

1. Establish what the tier systems purpose is
2. Fulfill the purpose the 1st goal provides
3. Eliminate player gaming of the tier system

Brainstorming:

1.  I believe that the system has ought to have a few purposes: a) Primarily it ought to be the primary mechanism to actually match up players together, b) Secondarily it ought to be used to organize the players into competition brackets (bronze, silver, gold, platinum)

Right now it is based on this mechanic called "card mastery".  Use a card enough and it adds it's value to your mastery field for certain color(s).  Reach a certain amount and you "level up" and go up in tier.  This is a flawed mechanic for several reasons.  It discourages card diversity: "Don't play around with those sub-optimal cards as you will level up into a tier your aren't prepared for".  Second it can inadvertently pull a player up in tier when they don't have a card base to compete.

2.  Fulfilling the purpose: I think we can do better than such a coarse knob.  What makes it so you can go up in a tier of competition?  Overall card collection (in my opinion).  As you have more in your pool, you have more options to tune your deck.  You no longer have to rely on sub-optimal uncommons or *gasp* commons.  But, we all know that not all cards are created equal.  In each rarity level, there is good and bulk, just like paper magic.  So, you need some weighted way of calculating it.

The term I was batting around in my head was maintaining color score for each color in order to give a collection score.  Something like this:

# Masterpieces * 30

# Power Mythics * 20

# Mythics * 12

# Power Rares * 10

# Rares * 5

# Power Uncommons * 2

# Uncommons * 1

= Color Score

Now, those numbers are completely bogus and the only criteria was weighting the contribution of each level toward what the color score would be.

As each color (including colorless) would have a score, it opens up a new realm both for matchmaking and bracket organization.  Now it can be based on a more gradient scale since there are numerical values being used instead of just broad levels.  Bronze could be scores of < 300 and silver can be 300 - 600 and so on.  It also means that matching can be more accurate, as it can be calculated on the fly.  E.g.  You choose Nahiri, a 2 colors PW: Red 100 + White 300 + Colorless 200 / 3 = 200 PW score.  You can match that up to  a mono-color PW of around 200PW score.  If you had used a mono color Gideon your PW score would have been 250 instead because your white collection score is much better than your red.

3.  Gaming the system: the problem takes care of itself.  At least the problem as it exists right now since you simply avoid all of the non-optimal cards to keep your tier low in order to easily? get the top prizes.

That's all I have to start off.  Hopefully others will find this interesting or engaging and maybe something will come of it, if only to give devs some new ideas they may not have considered. If you made it this far, thanks for reading the wall of text.

-Kyo

Laeuftbeidir

I do like the ideas, just one remark :
Stepping up in progression by the amount of cards owned might lead new players, that are buying big boxes to get a basic collection, to level to soon to Silver. A progression system based on ranking in events might be a better option.

You got me by the second sentence, by the way.

Kyokudai

@Laeuftbeidir I agree, that performance should be somehow part of the progression equation.  Probably should be a combination of collection scores and performance to help discourage gaming of the tier system.  The reason collection needs to be there I think, is that you don't have direct control over your collection like you do performance or Card Mastery as it currently is.

So maybe a performance parameter would have a dampening effect on the tier up based on collection, but if your collection reaches a certain point it will cause a tier up even with the dampening effect.

AngelForge

I don't like this color system, sorry.
(Though it feels better than the current system)
You tried to cover up the fact that cards of a rarity differ widely in power, by adding the "power" attribute, but I don't think that will be enough.

In my opinion, an Elo system is the way to go, when it comes to matchmaking.
You measure the performance of players in PvP events and match them accordingly.
(I would exclude training grounds for the performance measurement and matchmaking due to have a place for fun decks).

The tiers i would use for an initial ranking (bronze get fewer ELO points than platinum) and then throw it over board.

Mastery for cards should count towards the chance to draw a duplicate. If you draw a card in a booster that is mastered you get ONE more chance to pull something else. This would increase the chance to get cards you don't have a bit but wouldn't eliminate the RNG aspect.

Event prizes would be given out on the ELO ranking of participating players.
Here you have to make sure, that high ELO players don't just join the events with really good prizes.
A way to do this is streamlining prizes or punish inactivity (in other words, you should put your high ELO to a constant test, if you want to keep it there or you start dropping once you stop playing).
Here, it is also a chance to give a benefit to activity (or just avoid the punishment for inactivity).

Given this, it could of course happen, that someone with a really good set of cards drops down to a low ELO, due to inactivity, and when she/he starts playing again is messing up some other low ELO but active players, because they are matched up against that player that starts playing again.
No system is without flaws, of course and you have to compromise somewhere.

Kyokudai

Completely separating the color mechanism from matchmaking...  Hmm, this seems mostly reasonable and definitely desirable.   

@AngelForge I think you uncover some of the difficulties that I have to imagine the Devs have uncovered.  We are not directly competing with each other, therefore a direct competition rating system like ELO doesn't work, in my opinion, for prize support in an individual event.  I imagine this is why we have the ribbon system, because it is much closer to a golf system where the number of strokes taken on the course is what matters.

Why do I think that ELO doesn't work for prize support

ELO ranks players in a competition based upon the winning player taking points from the losing player.  It takes into account varying skill levels through the mechanism of a very good player will take very few points from their opponent if their opponent is lower rated than they are.  Since we are not directly competing, with the AI essentially competing on our behalf, we are leaving our performance in the hands of it.  If we lost points due to an AI improperly piloting our deck then our deckbuilding strategy would drastically change to a setup that the AI can actually run intelligently.  If we change it from typical ELO to only gaining points things get better a bit.  But now everyone is only going up and never down.  In the grand scheme of multiple events this would somewhat correct especially if you allow for tiering up / down based upon your ELO, but for within an event this is crippling since you are matched by ELO and won't get pairings like in a swiss tournament.  So if you are middle of the pack, or worse, lower in the pack, you have no hope to get to the top for prizes, since you have 5 chances and will probably not accrue enough points being matched by a similar ELO level.

Whew... I feel like that rambled on a bit and I apologize. 

That's quite the dilemma we have going on.

I actually find the analogy of golf type of competition more appealing in this kind of game anyway, and it makes sense since our *performance* is being graded against each other instead of *winning and losing*.  You have 5 chances per node to get as many ribbons as you can.  (It would be best to give all possible attempts up front instead of accruing more later in the competition timeframe, enabling somewhat more fair competition than to allow those that can play 12 hours a day an advantage over those that only have 2 hours).

I think we opt to not use ELO for everything. 

I think it has a great place in determining the matchmaking. 

If we use ELO for the Tiers or brackets and allow you to move up and down between brackets (after event is over) then that's good too.  Initial tiers can be set by color / card mastery to start, then you let the system correct itself over time.  Tiers are important so that beginners aren't competing against veterans to have a hope to get a rare prize.

Prizes would be determined by ribbons as they are now, that way each event provides everyone with a possibility of earning that top of bracket prizes.  ELO can be used for tiebreaking if need be.  Of course your performance in each event (winning and losing) adjusts your ELO rating.

There is some gaming of this system, as you can opt to do poorly for a while in order to drop from platinum to gold, and then compete in an event seriously to clinch a #1 victory.  No system is perfect.

@AngelForge - That idea for color mastery = GENIUS!  I want to see that happen (assuming mastery no longer tiers me up inappropriately)

AngelForge

@Kyokudai
Thanks for your post!

You're right, an ELO system is for players who directly compete against each other and I wasn't aware of that. I called my idea ELO like, without really having ELO in mind, but a mixture of ELO and Swiss system, I guess
I learned that the swiss system just relates your win and loss ratio. I wouldn't find that enough, because you somehow have to include and your performance in the events (which is also related to secondary objectives).

I didn't mentioned it in my post, but I wanted to keep the ribbons for the progression rewards and you should still use them to measure your overall performance.
So, your ranking depends on the performance in the events (after thinking about it, I think you can leave inactivity out).
For that, it would be the easiest to streamline all PVP events to have the same ribbon gain and length.
That could be an event with 4 ribbons per win plus 3 (2/1) from the secondary objectives, having 8 battles per every 3 nodes, so 168 maximum points.
But instead of adding up the score, you adjust the ranking score you have earned so far, with your actual performance.
So, there would be a maximum score you can have. 
But when you use exactly this ranking to match players against each other, you will always get players who perform similar than you.

That the AI plays your deck, may be a disadvantage, but I think many players enjoy the idea to build a deck that the AI can play well. Since your ranking would also be based on how good the others perform, a top ranked players would have build decks that are well played by him and the AI (or was just very lucky over a longer period).

I hope what I wrote is understandable. English is obviously not my mother tongue and I feel like I sometimes jump to muuch in my description.

Edit:
It came to my mind, that a pure swiss system is probably the best way to go, because if you play a strong deck that is not caring about secondary objectives your performance would be very good even though you easily win all your battles. A system, based on your event performance, would put you lower than you should be because you make as much points as others who try to achieve the secondary objectives but lose once in a while by trying so...
To avoid this a pure Win/Loss ratio is probably the way to go. 

Kyokudai

I like it when lines of logic converge together.

@AngelForge I think that you are correct that within an event, stick with the pure swiss style for matching.  You fill face decks with approximately the same wins / losses as you, since you are node limited anyway.

So what we need to see happen to make the tier system and matching better.

1) Event matching where you are competing against player decks: match based upon win/loss ratios within the tier

2) Use an ELO-like system for the tiers, allowing you to move up and down based on long-term performance instead of the weird color mastery system (this could theoretically allow for more tiers than the current amount)

3) Use card mastery like @AngelForge suggested where a mastered card causes a single re-roll of that card slot when obtained from a pack

It would be interesting to see what would happen if the win/loss ribbon amount wasn't weighted so heavily in the node gains.  But that's kind of a different topic

Laeuftbeidir

IMO, the card mastery system could remain as a base for the reward system.. For the matchmaking system I like the suggestion above

Kyokudai

@Brigby Could you communicate my previous post on this thread onward?  I think we centered on a really good suggestion.

ErikInVegas

If D3 kept stats then they should already know.

As a first order estimate I'd expect win ratio is proportional to:  

(log(PW1.level)+log(PW1.CardsUsableOnThisNode))/
(log(PW2.level)+log(PW2.CardsUsableOnThisNode))

That would give extra weight to multicolour PWs.

It should give better matches regardless of rules or restrictions.

Since people are playing the AI I'd adjust the center point so that more than half the people win more than half the time just so more than half the people are happy more than half the time.  (by playing them against an AI PW with less cards not a lower level PW because that's too obvious)