Clan stack issue

[Beherith]

Zydox, now think a bit before you say stupid things.
If he builds dfens in the back, he aint gonna deal much damage right? If he makes static dfens on the front, and kills alot of stuff with it, then hes still better than his enemy who uselessly throws alot of stuff at it so that the player can deal alot of damage.

If he makes alot of eco-since you cant directly deal damage with eco- hes gotta make alotta units, air, nukes, bbs or other stuff to deal alot of damage.

[imbaczek]

http://research.microsoft.com/en-us/pro ... l/faq.aspx

read this. it's state of the art and far from perfect (sometimes far from good enough) at the same time.

[BaNa]

http://research.microsoft.com/en-us/pro ... l/faq.aspx

read this. it's state of the art and far from perfect (sometimes far from good enough) at the same time.

interesting system, while i like the sound of it, i would

a. make the uncertainty of player skill increase with time not played

b. use ingame info to differentiate between teammates (damage dealt mebbe)

[hunterw]

if you wanted to get all fancy and write some autohost LUA script that would parse the damage dealt log of each completed game, calculate the percentage of time that any given player was leading the graph, 2nd place, 3rd place, etc, then accumulate a rating derived from that, that would work pretty well.

certainly i have seen useless simbasers win overall damage dealt, but they lead it for maybe 5% of the time at the very end of the game and are terrible the rest of the way through. if someone leads damage dealt for 25% of the game near the beginning and flatlines, they are awarded more rating points than that simbaser, just as it should be.

[Jazcash]

I can't agree more with 95% of the Spring community that SMURFS are the issue balance is screwed in a lot of games.

[Teutooni]

Dox, I think you are onto something here.

Damage dealt per average damage dealt weighted by time could be a nice approximation of skill, although it could be "cheated" by going on a dgun rampage / combombing at the very start.

Beherith is right, any damage dealt will do, especially in the beginning. If you position early LLT's so that the enemy simply has to throw all his resources at them in an effort to kill them, you have certainly earned the damage dealt. A line of HLT's and MRPC's sure doesn't help the team much in an offensive sense and might leave his allies in a Xv1 situation, though.

Here's where I think damage received comes in handy. If one player gets swarmed he might not deal much damage but receives a ton of punishment otherwise divided to the team. If one player makes stumpies and another one makes a huge porcline and the enemy attacks, largely dealing damage to the stumpies in front of the porc but gets killed mostly by the towers, I think both the stumpie player and tower player should get some credit for beating the attack - one soaking the damage, one dealing it.

So,

skill = (Damage dealt by player x / average damage dealt) * (1 / game time) + (damage received by player x / average damage received) * (1 / game time))

This should only be applied to the winning team, though. If you lose, your rank doesn't change. If you win only because of the help of others your relative rank goes down. So stacking teams lowers your rank!

[JohannesH]

A few players are starting to agree now though that their is not enough variety in ranks. What's needed most is an extra rank between gold star and Veteran as there is a giant split there. Possibly also a rank after Veteran, something like 1500 hours ingame.

This would be good in some games, but in others not. If the scale is widened only in the end, that means vet players get stacked up with complete newbs even more, and lose cause the difference between ranks is still bigger in the lower ranks.

This isn't such a problem when there's no newbs in the game though... Or could a 6,5 rank be added for example?

[pintle]

I can think of several (by no means 50%+ but still a few) 1v1s I have played where the damage dealt graph does not correspond to who won the game.

I'm no statistician, I won't engage in arguing the merits of various models, just offer my 2 cents:

Smurfing is a far bigger problem with the AH current balancing than Noobstars/clanstack.

As a definitive measure of skill is very hard to reach (AFAIK top level RTS 1v1 players don't often get much more than a 50% win ratio) timebound ranks, accepting that experience is only a benchmark guide, may be the most simple solution to stratifying players, however crudely.

As for PYBs owning nubs... I was talking to you guys on TS a few nights back: in all sincerity play more 1v1s, you seem to be getting frustrated at the lack of challenge in 4v4+ games

I don't want some uber "no lifer" star rank that just means I get even shittier teams than I currently do

[zerver]

The best balancing mode is random. You just do !random a couple of times and ask the other players if the balancing is OK.

Unfortunately !random is flawed on some autohosts.

I first do !random and then !balance.
The autohost responds something like
Balancing according to mode: random.
The teams were already balanced.

As you see, this does not make sense. The teams can never be "already balanced" for a random balancing.

Other autohosts disallow the random balancing altogether, which is even worse.

[Jazcash]

The best balancing mode is manual from a host that knows most players.
This with a combination of "NOSMURFS" = Best balance.

[zerver]

The best balancing mode is manual from a host that knows most players.

This would be ideal, particularly if host the host does not play and therefore is not biased.

[Tobi]

The best balancing mode is random. You just do !random a couple of times and ask the other players if the balancing is OK.

Unfortunately !random is flawed on some autohosts.

I first do !random and then !balance.
The autohost responds something like
Balancing according to mode: random.
The teams were already balanced.

As you see, this does not make sense. The teams can never be "already balanced" for a random balancing.

Other autohosts disallow the random balancing altogether, which is even worse.

• re-balancing
  SPADS balancing algorithm is based on a random seed so that it can swap players that have equal ranks. However, this random seed DOES NOT change when doing !balance => you can say "!balance" tons of times, a battle with same players will be balanced in the same way (so you don't have to yell at someone because he did !balance until he is sure to win). If you want to change balanced teams, you have to use the "!rebalance" command (which requires vote by default). That's the only purpose of the "!rebalance" command: changing the random seed so that it tries to alter currently balanced teams (the random seed is also changed at the end of a battle, so that teams are changed between battles)

[monkerl]

Hey,

as you may know i'm in the eXe clan, usually playing with my teammates and I'm one of the people who love playing with their friends. Zeuchs Andre and me joined eXe for one reason, so we could play with competent players...that's the whole reason why there are clans (or am I wrong). Additionally, we play together with our pc's in one location and especially then it really sucks if your are not in one team, but can look the others on the screen...

my opinion is that smurfing is a much bigger issue as it was voiced before. If you want a more balanced spring, I'd say we need a completely new system of judgeing a players skilllevel. I think with useing the endgame statistics, you could do a lot, but every mod would need a different math, so it may not be the right path either...

[Jazcash]

But whatever you do, don't hide ingame time. I've spent so many sad sad days for my rank

[Teutooni]

Would it be hard to make a little hack to sum the ingame time of all smurf accounts detected for one player and set that sum to all those accounts? Perhaps keep the actual ingame of each account in store elsewhere just in case.

[123vtemp]

dmg dealt alone does not show a good player
there are many cases in which a player does not deal a tone of dmg and yet efficiently ends a game

for example .. if he takes out all come quickly he will not likely have dealt a one of dmg in warfare but will have ended the game

[Jazcash]

Not to mention nukes don't count towards damage. So you could win the game with nukes and have 0 damage dealt.

[RogerN]

Proposal: Rank based on damage dealt to immobile targets.

Granted, this wouldn't work for all mods. But for most traditional mods (i.e. anything with static factories/resource extractors) it gives a good picture of a player's ability to kill the enemy. Porcing won't help you, since you won't be killing the enemy's base. On the other hand, raiding and attacking the enemy base is promoted.

You could also include damage done to a commander.

[[TS]Lollocide]

Fail. Why does everyone spend ages rehashing similar balance ideas when:

A. Everyone will say 'Oh that can be abused' etc until its 10 pages long.

B. Nobody will code it because of A.

and finally.

C. Nobody will code it period.

[SwiftSpear]

For the millionth time. We're not changing the lobby ranking system.

I'm all for adding tournament prize ranks and stuff, or even a voluntary ladder, but for people who are just playing casually, they will always be ranked by ingame time and nothing else.