Ideal Matchmaking

bk1234

We've heard several times in comments from Oktagon that they recognize the current mastery / tier matchmaking system isn't ideal. 

IMO, the current system is flawed because it reflects quantity over quality. It's fairly easy to get all the way to Platinum fairly quickly, but if you don't have a quality collection, you will not fare well there -- eventually losing interest in the game. 

So my question is, what would be the best fix for matchmaking?

tfg76

I think that matchmaking should be random within your Tier.

It's the Tiers that are the problem. You should be put in a Tier that reflects your recent event results. Players that do consistently well in results move up, if you do badly, you can go down. More prizes are awarded at the top, less at the bottom. This seems like a fairly obvious system that is used in most other competitive endeavours, and allows new players to play against other new players, and experienced players to continue to be challenged. 

khurram

Win percentage of a player. High scorers in an event may face others with the same score, or close.

bk1234

Win % maybe -- but not score -- score is determined by how much someone plays, not necessarily how well. 

Also I think potential to score should be taken into account -- which is what they tried to do with mastery. Collection size, ratio of mythics and rares, something like that should matter. 

Mainloop25

There should also be more than 3 tiers. 

tfg76

Well, I think "potential to score" follows naturally from the fact that you've recently done well in events. Modelling that based on card pool seems unnecessarily complex.

khurram

bken1234 said:

Win % maybe -- but not score -- score is determined by how much someone plays, not necessarily how well. 

Also I think potential to score should be taken into account -- which is what they tried to do with mastery. Collection size, ratio of mythics and rares, something like that should matter. 

I disagree that score is determined simply by how much someone plays. Not in events like HoD. You have to win and complete the objectives. People who play well score well.

What is potential to score and how would you determine it. If score is determined by how much someone plays then potential to score must be determined by how much time someone has to play, and not necessarily how well. 

Matthew

https://en.wikipedia.org/wiki/Elo_rating_system

theobserver

I see people talk about Elo everytime this is raised. Do you really support a system where you lose points/rankings if an ai loses a game with your deck against another player?

It sounds counter intuitive to assume players want this. Just making sure you know what you're asking for. 

asm0deus

Level up players with experience points when playing/winning. Up to level 60 like planeswalkers. And every 10 levels players have a choice to move up tiers. Where every tier has significantly better (25%-50% better) rewards. Same could be done with coalitions where we have tiers for coalitions to better determine the rewards we could be getting for coalition events. Where we would be competing with other coalitions from the same tier level only. 6 tiers for players/6 tiers for coalitions.

FeralSkald

Hidden math.

Assign a value to cards and total deck. Do not calculate opponent until 'fight' is selected to avoid deck alteration.

Example numbers: 20 per masterpiece, 10 per mythic, 5 per rare, 3 per uncommon, 1 per common. 

So a deck running 1 MP, 5 mythics, 3 rares, 1 uc. = 88
Quick check for a deck within a +/- 5 variance. Randomize all decks in that range as potential matches.

Have no idea if that would work, or is even codeable but it would equalize opponents, as well as adding a strategy element of not necessarily always running your best cards.

Kardynal

theobserver said:

I see people talk about Elo everytime this is raised. Do you really support a system where you lose points/rankings if an ai loses a game with your deck against another player?

It sounds counter intuitive to assume players want this. Just making sure you know what you're asking for. 

I think you're right that an absolutely strict translation of the Elo system to MTGPQ wouldn't be appropriate, and it would need to be adapted in certain ways.

But I do think that the general principle of matching winners against winners, and losers against losers, as is the case in paper MTG swiss tournaments, is a good way to go.

I think winning human players should play against decks which the AI Greg plays well, to give the most challenge, and make event leaderboards more meaningful.

Of course, with the current state of the AI Greg, there's sometimes a marked difference in decks which play well and those which the AI Greg plays well; most notably cycling.

bk1234

khurram said:

bken1234 said:

Win % maybe -- but not score -- score is determined by how much someone plays, not necessarily how well. 

Also I think potential to score should be taken into account -- which is what they tried to do with mastery. Collection size, ratio of mythics and rares, something like that should matter. 

I disagree that score is determined simply by how much someone plays. Not in events like HoD. You have to win and complete the objectives. People who play well score well.

What is potential to score and how would you determine it. If score is determined by how much someone plays then potential to score must be determined by how much time someone has to play, and not necessarily how well. 

Some people choose to stop at progression. In fact, I know quite a few top players in this game with HUGE collections and a 100% win rate who stop at progression, thus have lower scores. 

It wouldn't be fair for someone struggling to get 150 points to be paired against someone who chooses to get 150 points. 

This is also why collection quality to be part of this. 

Mburn7

Honestly, I like the idea of having tiers to separate out players and reward the "'better" ones. 
If I remember correctly, the purpose of the mastery system when they implemented it was to encourage people to use a more diverse selection of cards and reward those with better collections who put the time in to explore new combos and stuff.

Obviously this didn't work, because people abused the system to win more in lower tiers, and because the current system just congregates everyone in Platinum after a while (I think the current system with 2 more tiers after platinum would help a lot, as the really top players would get shuffled up).

I think a better system should be based on collection, but not based on usage (so that it can't be abused).  Every card you have gives points, the higher your score the higher your tier (and add more tiers).

The matchmaking itself in matches I have always felt should be based on score in that event.  You should be facing decks +/- say 10 rankings of you (so if you're in first you face only decks of the top 10 players in that event, but if you're in 100th you only face decks from players ranked 90-110).  It would drastically reduce perfect scores (since perfect decks will face off against each other) and make it easier to make progression, since inconsistent decks face off against other inconsistent decks and not keep getting paired up against other OP decks.

This would also remove the issues with PW leveling, since you can't sandbag in a lower level walker if there is a good chance it faces off against nothing but level 60 opponents.

Matthew

Mburn7 said:

I think a better system should be based on collection, but not based on usage (so that it can't be abused).  Every card you have gives points, the higher your score the higher your tier (and add more tiers).

To me, bringing collection into the ranking system sounds needlessly complex. Plus, the way you have described it would mean that people who have limited collections would be rewarded poorly, even if they are able to do very well in spite of that limited collection.

Mburn7 said:

The matchmaking itself in matches I have always felt should be based on score in that event.  You should be facing decks +/- say 10 rankings of you (so if you're in first you face only decks of the top 10 players in that event, but if you're in 100th you only face decks from players ranked 90-110).  It would drastically reduce perfect scores (since perfect decks will face off against each other) and make it easier to make progression, since inconsistent decks face off against other inconsistent decks and not keep getting paired up against other OP decks.

This would also remove the issues with PW leveling, since you can't sandbag in a lower level walker if there is a good chance it faces off against nothing but level 60 opponents.

I'm more inclined towards this idea. If a reset occurred for every event, it would mean that each event would develop into its own microcosm of competition, with someone (or some few) coming out on top in the end.

You could even incorporate this into a broader level of competition that is spread out over "seasons", which could be weekly, monthly, quarterly, or whatever the desired length may be. That would be a cool way to drive competition, as well as keep players engaged with the game over a longer period of time. Furthermore, it would be a way to implement a reset button into the system that could serve as a way to prevent any one player from remaining dominant for too long.

FeralSkald

Matthew said:

Mburn7 said:

I think a better system should be based on collection, but not based on usage (so that it can't be abused).  Every card you have gives points, the higher your score the higher your tier (and add more tiers).

To me, bringing collection into the ranking system sounds needlessly complex. Plus, the way you have described it would mean that people who have limited collections would be rewarded poorly, even if they are able to do very well in spite of that limited collection.

Mburn7 said:

The matchmaking itself in matches I have always felt should be based on score in that event.  You should be facing decks +/- say 10 rankings of you (so if you're in first you face only decks of the top 10 players in that event, but if you're in 100th you only face decks from players ranked 90-110).  It would drastically reduce perfect scores (since perfect decks will face off against each other) and make it easier to make progression, since inconsistent decks face off against other inconsistent decks and not keep getting paired up against other OP decks.

This would also remove the issues with PW leveling, since you can't sandbag in a lower level walker if there is a good chance it faces off against nothing but level 60 opponents.

I'm more inclined towards this idea. If a reset occurred for every event, it would mean that each event would develop into its own microcosm of competition, with someone (or some few) coming out on top in the end.

You could even incorporate this into a broader level of competition that is spread out over "seasons", which could be weekly, monthly, quarterly, or whatever the desired length may be. That would be a cool way to drive competition, as well as keep players engaged with the game over a longer period of time. Furthermore, it would be a way to implement a reset button into the system that could serve as a way to prevent any one player from remaining dominant for too long.

Perhaps in a season setting, but not for single events. You'd just end up with people waiting till the last possible moment to play charges to keep their score low. Then unloading everything at the event end, so only their last few matches would be against high scores. 

Kinesia

Kardynal said:

theobserver said:

I see people talk about Elo everytime this is raised. Do you really support a system where you lose points/rankings if an ai loses a game with your deck against another player?

It sounds counter intuitive to assume players want this. Just making sure you know what you're asking for. 

I think you're right that an absolutely strict translation of the Elo system to MTGPQ wouldn't be appropriate, and it would need to be adapted in certain ways.

But I do think that the general principle of matching winners against winners, and losers against losers, as is the case in paper MTG swiss tournaments, is a good way to go.

I think winning human players should play against decks which the AI Greg plays well, to give the most challenge, and make event leaderboards more meaningful.

Of course, with the current state of the AI Greg, there's sometimes a marked difference in decks which play well and those which the AI Greg plays well; most notably cycling.

Ok. this is getting more interesting...


People play against people who usually win about much against people about the same level as them _but_ we also try and use the decks this level of people loses a lot too... Or something...

hawkyh1

I think something could be done with average mana per turn.
all cards are fueled by mana. better players on average get
more mana due to better gem converters, cascades etc.
together with other data like deck composition it might give a
good idea on player ability.

HH

Gilesclone

Make the tier based on % of mythics owned with lower ranks based on % of rares owned.

As it is now there are people in platinum with 30 mythics competing against others with all of them.

__Adam

I'd argue that anything much beyond 50% win/loss shows a flaw in matchmaking

wereotter

It ought to be, in my opinion, a multifaceted system. 

Tier one one should be how many standard legal cards you own. This would prevent people from sand bagging in lower tiers to try to get rewards as well as not punish returning players for mastery on old cards.

Tier two should look at your win/loss ratio within the previous tier and match you against players with simular ratios. So if you’ve done well against similar power decks, you get paired against higher win ratio decks later. Also this ranking ought to be referenced across events