Get the Balance Right

(Violator) #1

One of biggest issues right now is team balance and endless stacked matches. Team selection currently appears to be completely random with the rule that if numbers are odd attackers get the extra player.

Here is a suggestion for how balance could work considering the stats that Echo could be or is giving us. I would base the balance on the following:
k/d ratio. This is pretty obvious but on the whole players with the highest k/d ratio tend to win maps in DB in combination (one single player rarely makes a difference).
XP earned per match. Players with the highest XP at the end of the match tend to be doing the objective / playing their class as well as fragging. They are more likely to play as a team rather than individuals (this game really doesn’t reward going it alone - a top rambo player may end up with a 3-1 k/d but did nothing for the objective and still lose the match. An entire team with a 3-1 k/d is clearly containing the enemy so will almost certainly win).
Win/loss ratio. This speaks for itself and I think is the most important factor. Certain players combined will overpower the other team 90% of the time. It usually only takes 2 or 3 to do this.

I would suggest giving every player a rating based on the above three figures. The figures should take into account other players in the match. (for example if player X and Y have won together say 70% of the time, ensure they are on opposite teams, assuming that the next highest combination of two players together is less than this by a certain margin say 10%). In terms of working out balance I would say ‘X + Y together win 70%’ is better than ‘X wins 80%’ and ‘Y wins 50%’ say.

There is also player level, but I don’t consider this represents a player’s skill as much as their persistence.

Most of the top players are willing to swap to even up but can’t due to the quirks in the lobby (e.g. the ‘Join’ button to swap teams gets locked out once the lock in starts, but prior to that you can swap from a 5v5 to make a 6v4…). You also can’t join any team if you were in spec mode the previous match.

(tokamak) #2

K/D is inaccurate, it has a defender bias, it’s skewed towards certain mercs and it favours players that don’t risk themselves at objectives. A team of attackers could build a great k/d ratio while losing terribly while a team of attackers that gets the job done could have a terrible K/D ratio.

I agree that W/L would be a great indicator but that’s only true in such a small player pool. Once the player numbers start increasing this indicator will quickly lose any significance.

The best indicator of how well a player is doing is xp/min

(Glottis-3D) #3

[QUOTE=tokamak;511823]K/D is inaccurate, it has a defender bias, it’s skewed towards certain mercs and it favours players that don’t risk themselves at objectives. A team of attackers could build a great k/d ratio while losing terribly while a team of attackers that gets the job done could have a terrible K/D ratio.

I agree that W/L would be a great indicator but that’s only true in such a small player pool. Once the player numbers start increasing this indicator will quickly lose any significance.

The best indicator of how well a player is doing is xp/min[/QUOTE]
we can normalize k/d, by deviding it into AK/D and DK/D (attack kd and def kd)

(montheponies) #4

no perfect way but I’d favour kdr over xp/min anytime…if you can combo kdr+wlr with some kindof weighting for the numbers then it would be a massive step forward from what we’ve got just now…as for teams being stacked favouring attackers, i really dont see that, seems to be random - could be wrong.

either way I’ve played a good few matches over the last couple of nights where the defensive bias was there on SW leading to;

R1. Fullhold (complete with spawnrape) - 15min of sheer pishness.
R2. Steamroller - over in a couple of minutes

Also as noted, allow specs to be kicked, as on 6v6 servers a single player makes a big difference…

(INF3RN0) #5

Not everyone is Freze when it comes to K/D.

(Violator) #6

Teams aren’t favoured to stack attackers with the exception that they always seem to get the extra player in the case of odd numbers, we have equally stacked matches where defenders pwn (usually Whitechapel (spawncamping), Bridge (bad placement of the EV on first obj) and Terminal (pump blown). My main point was not to balance based on individuals but on combinations of players.I think Toka has a good point that it will not be such an issue when we have a large playerbase. I would rank the importance of the attributes as

  1. win / loss
  2. xp / match
  3. k / d

The weighting of these would need to be refined over time. Also certain mercs are a lot harder to earn XP for (seems to be easier for support classes especially field ops) so this would need to be taken into account.

(tokamak) #7

And let’s not forget that killing and winning is already included in your xp score. This means that we already have the weighted score system we’re looking for. It’s called xp. If you believe kills don’t reward enough xp or any other useful task isn’t weighted properly then that’s still open to discussion of course, it’s tantamount that this system becomes an incredibly accurate representation of the player’s worth.

However, when you’re building a game where one specific merc is already called ‘Fragger’ then you’ve got to acknowledge that K/D isn’t going to be a fair representation of all the different playstyles that can be played in this game. Some excel at building a high K/D, some are mediocre at it and then there’s the objective crash dummies and revive mules. The latter will often have to sacrifice their K/D for doing their actual job well.

Even displaying K/D before the XP score in the tab is sending the wrong message. It makes the goal of the game, the objectives, seem like a mere after thought, a fancy hobby to indulge in once you get bored of counting bodies.

(montheponies) #8

I completely agree - personally i wouldn’t have KDR on the scoreboard. It’s also a reason why I’m frustrated with persistent stats in any shape or form, it tends to lead to perverse behaviours, like folk leaving servers before a loss is registered or changing team to ensure they maintain the all important KDR. Tough one to crack though as it’s the norm nowadays - we’re all guilty of being statwhores to some degree.

As for using XP, then yes it would work but you would need to really skew the number for a kill to be significantly higher than any other function - using xp/min could then work.

(Violator) #9

I’m all for removing k/d from the scoreboard, its fine for TDM but not objective-based modes.

(Glottis-3D) #10

i am not against k/d in the scoreboard.

and i agree that with proper XP-counter we could only use XP/min to balance teams.
k/d can be 25/5 for a sniper. but his XP/min can be very small, because he just waits for enemies to come.

i dont think, that current XP-counter is good. winning team always/usually has sevaral times more XP.

(tokamak) #11

What would you prefer?

(spookify) #12

There should be a complex calculation! SD should be able to come up with a complex formula…

Excluding player rank.

Calculating SPM, overall acc, win loss %, and other things to come up with an amazballs rank!
Or use none of these if they dont work.

I only got to advanced trig, calc, stats2 and physics 2 in college so I blow at math…

If I worked at SD I could contact the math or physics or stats department of a university in England, London where ever u are and be like hey we need your help! They would flip over the chance to create a formula from numbers u can provide.

The SD Oxford Ranking System or SORS or SDORS

(Glottis-3D) #13

several times more XP can mean two things:
team A is several times better than team B
OR
XPcounter is very biased.

i’d start with removing/lessen bonus xp like like “long defence +200xp” or “Round win +500”

and i would probably go as far as giving small XP for killing small XP players, and giving more XP for killing high-XP players.

current system is ideilogically biased - to encourage OBJ play. but is not a good representation of players’ skill all around the game.

(spookify) #14

Ok for real it just came ti me however I have got the full details 100%

Step one - no rank will be giving until a specific about of game play or special number of mercs are unlocked. (sort of like lol where u cant play ranked until that time)

Step two - receive your rank or rating

  • almost all newly ranked or rated players should be very close with only a few differences based on acc, SPM and all that.

The base of this ranking system is the players ranked play level and a divisible number.

Lets use LOL

But first lets say there are a max of 20 ranks
Bronze 1 to Diamond 3

Again I am laying in bed on my phone so I dont have the math BUT:

A persons base calc number should be 1 right? We have 20 mercs and can join ranked at level 20

20/20 is 1!!!
Add additional calculations for acc, SPM and win loss and

1* .40 5.6052% = 1.16 * 100 = 116 player rank

Math
Level * acc * SPM * winloss%

Max rank
20 diamond 3

20 * .45 * 7.30 * 67% = 4,401 rank

(tokamak) #15

I do love the idea of making the xp for kills context-sensitive.

But as for skill, how would you define it? With the goal of the game in mind (finishing/defending that last objective) what about the current xp doesn’t represent the player’s contribution to that goal?

(Glottis-3D) #16

[QUOTE=tokamak;511871]I do love the idea of making the xp for kills context-sensitive.

But as for skill, how would you define it? With the goal of the game in mind (finishing/defending that last objective) what about the current xp doesn’t represent the player’s contribution to that goal?[/QUOTE]
you, see. k/d and w/l already included into XP.
but was that balanced inclusion? i am not sure.

if we exculde (for rank-determination purposes) those two and have separate stats for all players
k/d
w/l
xp/min

then with some statistics we can form multipliers for each stat so that to make them balanced.

(Anti) #17

[QUOTE=tokamak;511842]And let’s not forget that killing and winning is already included in your xp score. This means that we already have the weighted score system we’re looking for. It’s called xp. If you believe kills don’t reward enough xp or any other useful task isn’t weighted properly then that’s still open to discussion of course, it’s tantamount that this system becomes an incredibly accurate representation of the player’s worth.

However, when you’re building a game where one specific merc is already called ‘Fragger’ then you’ve got to acknowledge that K/D isn’t going to be a fair representation of all the different playstyles that can be played in this game. Some excel at building a high K/D, some are mediocre at it and then there’s the objective crash dummies and revive mules. The latter will often have to sacrifice their K/D for doing their actual job well.

Even displaying K/D before the XP score in the tab is sending the wrong message. It makes the goal of the game, the objectives, seem like a mere after thought, a fancy hobby to indulge in once you get bored of counting bodies.[/QUOTE]

We already shuffle teams between matches based on what we call a ‘Public Skill Rating’, an aggregate rating of XPM across multiple matches. We use this system very much for the reasons tokamak mentioned.

It’s by no means ideal though, we have a number of things we still need to do. One issue is that we have to shuffle in the lobby at the lobby’s start so that players can pick their Mercs based on their side. This causes an issue where people decide to leave/join the session after the shuffle, messing the balance up - ideally we’d shuffle on match start but this would impact Merc choice a lot.

As others pointed out we also don’t normalize XPM by team, which might be a good idea. Right now we kind of let the aggregate rating across last 10 matches address the variance.

I think the biggest issue is that we still don’t have a shuffle/vote shuffle mid match and with the current small beta population this means there isnt a ready supply of players to come in and fill the gaps.

All of this is known however and on our list of things to improve, but any other suggestions you folks have would be good.

(spookify) #18

Bronze 1 Rank

1 * ((1.5 * 10 *5%) + ((((40% + 35%) / 2) * 100) * 10%)) + ((560 / 10) * 30%) + (((52% * 100) * 1.5) * 55%)))
Rank * (K/D) + (Acc and HS Acc) + (SPM) + (W/L)

(15.75 + 41.25 + 72.8 + 120.9) * 1 = 250.7 Player Rank

Multiplier of Importance:
K/D 5%
Acc 10% (and Averaged)
SPM 30%
Win/Loss 55%

Bronze 3 Rank

3 * ((1.5 * 10 *5%) + ((((40% + 35%) / 2) * 100) * 10%)) + ((560 / 10) * 30%) + (((52% * 100) * 1.5) * 55%)))
Rank * (K/D) + (Acc and HS Acc) + (SPM) + (W/L)

(15.75 + 41.25 + 72.8 + 120.9) * 3 = 752.1 Player Rank

Diamond 3 Ranks

20 * ((1.5 * 10 *5%) + ((((40% + 35%) / 2) * 100) * 10%)) + ((560 / 10) * 30%) + (((52% * 100) * 1.5) * 55%)))
Rank * (K/D) + (Acc and HS Acc) + (SPM) + (W/L)

(15.75 + 41.25 + 72.8 + 120.9) * 20 = 5,014 Player Rank

I am assuming my math is wrong but you get the point…

(tokamak) #19

I don’t mind special attention to accuracy or k/d or kills per minute or any of that. But these are all completely isolated statistics that even combined don’t paint an accurate picture.

So instead of trying to knead them all in some aggregate score it would be way more fun to have these things in separate tabs on some statistics website. “Most accurate player of the week” “Most suicidal player of the week” “Traveller of the month” “Trickjumper of the day”. That way players can all chose whichever statistics they personally care about and compete on that.

Just don’t assume that k/d or accuracy is something all players should aspire towards. If my Rhino has to fire his minigun at an empty space for twenty minutes straight in order to win a match then I’ll do. When I have to die 50 times within the spawn of a single match to get an EV repaired then I’ll do that.

Sure when I’m a merc with a bodycount motive then it will be awkward if my K/D ratio sucks, then I’m not doing my job. But if I were a medic then I’d rather see my revive/death ratio or health per minute than my K/D.

So there’s a suggestion for the score tab: Let players pick their main statistic to show on the score screen. Some may pick k/d some may pick xp/min but there’s also going to be hipsters that pick other niche statistics to express themselves in the tab. It gives personality to the game without standardising something that can’t be standardised.

(spookify) #20

[QUOTE=tokamak;511875]I don’t mind special attention to accuracy or k/d or kills per minute or any of that. But these are all completely isolated statistics that even combined don’t paint an accurate picture.

So instead of trying to knead them all in some aggregate score it would be way more fun to have these things in separate tabs on some statistics website. “Most accurate player of the week” “Most suicidal player of the week” “Traveller of the month” “Trickjumper of the day”. That way players can all chose whichever statistics they personally care about and compete on that.

Just don’t assume that k/d or accuracy is something all players should aspire towards. If my Rhino has to fire his minigun at an empty space for twenty minutes straight in order to win a match then I’ll do. When I have to die 50 times within the spawn of a single match to get an EV repaired then I’ll do that.

Sure when I’m a merc with a bodycount motive then it will be awkward if my K/D ratio sucks, then I’m not doing my job. But if I were a medic then I’d rather see my revive/death ratio or health per minute than my K/D.

So there’s a suggestion for the score tab: Let players pick their main statistic to show on the score screen. Some may pick k/d some may pick xp/min but there’s also going to be hipsters that pick other niche statistics to express themselves in the tab. It gives personality to the game without standardising something that can’t be standardised.[/QUOTE]

First off SD did say there was going to be an advanced stat page or web-site of some sort way back in the day…

Second, that is why there’s importance multipliers… Yes, Rhino will have a super low acc but are you going to be playing rhino all the time? On the flip side is VAL who will have a super high acc and hs ratio… Are you going to be playing him all the time? You have 20+ mercs hopefully 40+ mercs at release all with different acc levels… That is why acc is only 10% of the over all importance of the score… Acc and K/d are almost side stats

The rank should be based mostly Win/Loss and SPM with other stats bring up the tail end…

In my math above K/D and ACC make up 22.7% of the over all Score

In my math above SPM and W/L make up 77.3% of the over all Score

In Bronze 1 ^^^