sc2 match making style

(Ragoo) #181

lol, I just wanted to read through this thread since I’m pretty active @ SC2 but then I saw waaaalls of text :smiley:

Wanna see my last games?..

Bf7nL.jpg710x380

:C so close to Master league and I start to fail so hard :stuck_out_tongue:

(Linsolv) #182

The thing is, folks, all ladder-based game systems like StarCraft or League of Legends (or whatever your preferred DotA derivative) have a non-ladder, “free-play” mode. You can go into the free mode and while it will to a certain extent track ELO to keep you interested, it won’t affect “your” ELO. What that means is that if your friend wants to try out unlocks, he can go into “unranked” play, and BAM. Your unlocks are all there, and he’s not messing you up. Drunk? Unranked. Sleepy girl? Unranked. Playing on a laptop in a subway station with a Verizon USB connection? Unranked.

What you’d want to have to avoid the whole system of “BAWW HOW DO YOU RANK TEAMS?!” is that you have the matchmaking server track the team as a unit. Each player (you’d use log-in accounts, like Steam or PSN or XBL accounts here—for a simple “enter name” system it wouldn’t really work) in a team is tracked, and if you change players, you have a different team.

If your team is brilliant, then you rise quickly even though you have to play a couple matches against nobodies.

Then, if you want to play as a premade, you join as a team. It’s not that complicated.

If you want to play alone with randoms, then I’d need to do an absurd amount of stat-tracking… which, come to think of it, a server could do. After tracking most every stat you can think of (Kills, Deaths, Times you Healed/Buffed/Resupplied, Times you were Healed, Times you were Buffed, Times you were Resupplied, objectives completed, and finally wins and games played) you’d look for correlations in win percentage and the above traits amongst as large a player-base as you could find.

Then, you track based on a weighted percentage to assign a player ELO based on performance, and wether or not you are playing a “high percentage” play-style that just got beat this time, or if you were dicking around in the corner while your superman team went out and beat everyone for you.

(Coolaguy) #183

[QUOTE=Linsolv;269659]The thing is, folks, all ladder-based game systems like StarCraft or League of Legends (or whatever your preferred DotA derivative) have a non-ladder, “free-play” mode. You can go into the free mode and while it will to a certain extent track ELO to keep you interested, it won’t affect “your” ELO. What that means is that if your friend wants to try out unlocks, he can go into “unranked” play, and BAM. Your unlocks are all there, and he’s not messing you up. Drunk? Unranked. Sleepy girl? Unranked. Playing on a laptop in a subway station with a Verizon USB connection? Unranked.

What you’d want to have to avoid the whole system of “BAWW HOW DO YOU RANK TEAMS?!” is that you have the matchmaking server track the team as a unit. Each player (you’d use log-in accounts, like Steam or PSN or XBL accounts here—for a simple “enter name” system it wouldn’t really work) in a team is tracked, and if you change players, you have a different team.

If your team is brilliant, then you rise quickly even though you have to play a couple matches against nobodies.

Then, if you want to play as a premade, you join as a team. It’s not that complicated.[/quote]
Yep.

[QUOTE=Linsolv;269659]If you want to play alone with randoms, then I’d need to do an absurd amount of stat-tracking… which, come to think of it, a server could do. After tracking most every stat you can think of (Kills, Deaths, Times you Healed/Buffed/Resupplied, Times you were Healed, Times you were Buffed, Times you were Resupplied, objectives completed, and finally wins and games played) you’d look for correlations in win percentage and the above traits amongst as large a player-base as you could find.

Then, you track based on a weighted percentage to assign a player ELO based on performance, and wether or not you are playing a “high percentage” play-style that just got beat this time, or if you were dicking around in the corner while your superman team went out and beat everyone for you.[/QUOTE]Or… you could just track ELO (through wins/losses vs opponents of varying skills)… as it represents ELO-skill without having to rely on a sub-set of secondary statistics that can (read: will) be exploited.

Your summary is pretty much the jist of the way that any reasonable company would implement ELO. Better yet, completely obscure it from players to have it be a background feature of matchmaking. Net result: Positive experience for the players and the community.

Aside: Monday Night Combat matchmaking and Game Design

I’ve been playing quite a bit of console Monday Night Combat lately, and I am compelled to remark on the failures (read: editorial) of its matchmaking system in conjunction with its game design. (I still consider it incredibly fun, from a gameplay perspective, but I can’t help but feel there was a significant opportunity lost with the way that the game was designed and balanced in conjunction with its broken matchmaking.)

  1. Monday Night Combat requires and rewards player skill. Many features of the game are non-obvious (i.e. Ejectors, Annihilators, Bot Spawning, Juice, etc.) The tutorials and instruction provided by the developers are completely inadequate. The only reliable source of information is forums and youtube. (Note: In-game player progression is hampered by the conditions of matchmaking and other issues of game design to be outlined subsequently)

Result: Exceptionally sharp learning curve. Reduced Monday Night Combat playerbase.

  1. There are no mercenary playlists in Monday Night Combat. The most common experience is for a team of solo pub players to face off against organized teams of players (who often are far more skilled at the game). Rather than having balanced matchmaking that shifts players amongst teams in between games or finds evenly-skilled opponents, Monday Night Combat matches the one or two pub players against the full party-based team. Like the games that inspired it (e.g. Dota), Monday Night Combat’s in-game player rewards for successes are cumulative and synergistic, dramatically shifting the advantage in favour of players who accrue minor initial advantages.

Result: Completely one-sided matches. Reduced playerbase.

  1. Monday Night Combat is a team-based shooter that relies heavily on coordinated teamplay. People on losing teams tend to drop from matches in-game. Additionally, there are no ‘bots’ to take over the roles normally filled by players. The result is that the imbalance is swung even further to the detriment of the losing team.

Result: The exodus of losing players combined with the lack of adequate replacement is game-breaking. If the game would simply end promptly in the defeat of the losing team, then the problem wouldn’t be as dramatic. However, opponents tend to spawn camp the losing team’s spawn and farm kills off of them. (Leaderboards don’t help matters :slight_smile: )

Net Result 1: I’m very happy that SD has chosen to have players ‘replaced’ by bots once they leave games. I anticipate that the bots will help to give players a fighting chance at salvaging their game experience for the match.

Net Result 2: Brink could face similar broken matchmaking game design if it doesn’t incorporate either a mercenary playlist or some variation of TrueSkill / ELO.

(Linsolv) #184

@Coolaguy:

You’re right. You could just track a Win/Loss ELO per-player.

But then you get Atavax complaining that “ELO helps noobs” or “ELO helps cheaters.” So I decided to get some Brink-style Sabermetrics involved in our hypothetical algorithm.

I should’ve made clear that my post was 100% a response to things that people were complaining about. Had someone whining that they didn’t want to ruin their ELO? Fixed. Had someone whining that W/L isn’t as easy to maintain with pubs? Fixed.

RE: Monday Night Combat commentary

I haven’t played Monday Night Combat. I will not be addressing any of your postulates—I’ll take them as fact.

But based on my understanding, I have read that the players who have experienced Brink thus far have immediately, intuitively played in a team-based style that is not matched or recreated by any other game thus far. I’m planning to reserve judgment as far as if the game will be imba for PUGs vs Premades.

(Coolaguy) #185

[QUOTE=Linsolv;269730]I should’ve made clear that my post was 100% a response to things that people were complaining about. Had someone whining that they didn’t want to ruin their ELO? Fixed. Had someone whining that W/L isn’t as easy to maintain with pubs? Fixed.[/QUOTE]Those are actually pretty elegant solutions, provided the latter isn’t subject to exploitation, of course.

[QUOTE=Linsolv;269730]But based on my understanding, I have read that the players who have experienced Brink thus far have immediately, intuitively played in a team-based style that is not matched or recreated by any other game thus far. I’m planning to reserve judgment as far as if the game will be imba for PUGs vs Premades.[/QUOTE]It is only prudent to reserve judgement, as a consumer, until the complete product is released. However, the pessimist in me is skeptical.

That would be a remarkable achievement, indeed. The results could have far reaching implications in influencing ‘gamer’ behaviour: one worthy of a GDC talk at the very least and/or a psychology paper. That’s the optimist in me.

Best to wait and see.