FORUMS


Matchmaker Auto-Balance



Posted: //
Feb. 21, 2017, 4:14 a.m.



Being always in a lottery is terrible.

Every vet game is a "certain win/loss" before it even starts, when one team has 4 T4 and 4 T3 and the other one gets 3-4 T3 and the rest noobships.


Suggestion:

Based on player number of hours + Kills/Deaths ratio (heal could be included but it would complicate the formula), assign a "score" to each player.

[ Score = Minutes played * Kill-Death ratio ( so 1:1 ) would be the exact number of minutes ]

Obviously, a player who played much more and has a high kill ratio already would get a bigger number.

Then, distribute those players into the two teams with the following algorithm:

  • Sort players by score (top to bottom)
  • Add the highest score to one team
  • Repeat {
  • Add the next one in order to the team that has "less"
    }
  • Touchup: If the difference is still significant ( more than 1.2 ), swap the two highest scores.

Overall, the game would distribute the 16 players to the two teams with somewhat similar scores.

Much more fun games instead of knowing before it even starts - Ooooh, we're going to win this one - Oooh, 5 T4's we're dead.

Simple math. High-school programmer could do it...
How about it ?


Never give up, never surrender!


Posted: //
Feb. 21, 2017, 4:23 a.m.



Just go like this:

Recruit: tier I II
Veteran: tier III IV
Legendary tier IV V

Problem solved.


Posted: //
Feb. 21, 2017, 4:23 a.m.



Time played dependancy: extremely bad.
Will allow creating twinks for randomplayers raping. Does not represent real skill oe even knowlege of game mechanics.

K/D ratio dependancy: exremely bad.
Skilled teamplayer or primary healer will have low K/D, also "death farm" can be executed.

How to balance squads using this system?

If you want to chave an imitation of fair matchmaking - use K/D and effectiveness on each ship in fleet, take in account time played on each ship and exp on each ship, assign higher "weight" to squads, decide -should MM put friends in same or different team.

So I state that IMO your idea is bad.


The best defence against logic is madness!


Posted: //
Feb. 21, 2017, 4:31 a.m.



I think/hope there is some sort of mm algorithm. But it still has its flaws. You are right most games are decided by the mm.
TDM for example often ends like 100/<50. I think the aim should be more or less 100/80.

I stated before and do it again now. The mm goes completely off when squads come into play. I don't mean that squads don't get matched with other squads but that even when not accounting the squad the single players in these squads don't get matched properly.
An Example: Three rank 30 and above players with T4 ships enlist for a match, than most of the time they get into the stronger team which creates massiv imbalance. My impression is as soon as squad mm comes into play the squad members got treated like three T2s with no experience and then get matched according to that. There are rumors about a bug but it is not acknowledged by the devs.
Known issues list:
https://www.greybox.com/dreadnought/en/forum/topic/38007/?page=1


Posted: //
Feb. 21, 2017, 5 a.m.



Well, then heals should be considered too, not just K/D

Doesn't really matter "how" that Rank/Score is determined, but there should BE one. As the guy above me said, most games and 100/<50 because of the lottery, either one team gets the good players, or the other. Then rapefest begins.

And in 90% of the games you know you're going to lose or win even before it starts, in the countdown. That is the main problem... kills all the fun of competing when it's SO obvious that you're going to either lose... or win.

~~
Pre-made squads make the matter even more. Simply the fact that they are on some sort of voice chat makes them MUCH better then random people just happening to be in a team.
Equal ships, those people acting and reacting together are most likely to win.

Pre-made should only be matched against other pre-made. Otherwise, it's completely unfair against randoms.


Never give up, never surrender!


Posted: //
Feb. 21, 2017, 8:22 a.m.



K/D is a really bad indicator in a game like this. it is extremely biased towards ships like snipers and vets and will produce bad results when it starts stacking teams with 'amazing' snipers/vets who actually suck and 'sucky' tanks, healers, and brawlers who are far better than the system accounts for. Ditto for time played, just because you have lots of ours doesn't men you're good.

As another post said
Recruit = TI, TII
Veteran = TIII, TIV
Legendary = TIV, TV

would fix a good deal of the issue.

an additional solution would be to then divide the 16 players in the following manner: balance the teams by weighting both a player's fleet composition and what they've played recently in order to get a more even distribution of each tier on each team. This would work in roughly the following way:
each team gets a balance of people who have the option of either tier (don't put all the people who only have TIII on the same team) each team ALSO gets a balance of what the game thinks they'll play. for example: If I have 4 TIV ships but the game sees I've only been playing my single TIII ship for the last 10 games it should count me as a probably tier III but when choosing to put me on one team or the other it should place me on the team that has the least players with TIV ships in the fleet.

by accounting for fleets in this way, as opposed to the current, stupid, aggregate score system. You attempt to balance both what a player will probably play and what they can possibly play: hopefully achieving better balance.
Or we could just eliminate the issue entirely by trashing the fool tier system but hey it's only killing the game slowly what do I know?


-Citizen Soldier -"Do not bring forth an argument that can be disproved with a 10 minute google search."


Posted: //
Feb. 21, 2017, 10:34 a.m.



First, the game already uses such an algorithm, check the patchnotes for the shipyard update 1.3.1, second you have some players that are very competetive or very experienced with probably 5000 games and 500-600 hours into it, but because of several progression resets they are not that high-ranked yet.

Another problem is that there are not always enough players, the matchmaking is a statistical system, so for it to work you need huge amounts of players which we clearly do not have.

Then you can use a T3 although you have several T4s in your fleet because you want the xp on it, and last but not least your team might not have the right ships for the enemy tactic, as the game cannot balance the avaiability of classes/ships in each players fleet of a match.

Then keep in mind, the matchmaker wants you to win 50% of the games and lose 50% of the games, so if your W/L is far above 1, you should actually lose more. Next problem with this is that the tier mixing means that well-performing low tiers get mixed with low-performing high tiers, which might create an imbalance, while well-performing high tiers get weaker teams, because the MM expects them to carry/needs to balance the match against them (50% winrate goal).

Then, some of your mates may be drunk, trying new builds, experimenting or just doing stuff for giggles. Balance this!


Recruit Engineer

This forum is restricted, posts cannot be made.