CSR Rebalancing

Games Other Halo
1

I, like many of you, have been playing for the past week and a bit and have already noticed the numerous flaws and issues with the CSR rating system. So I’d like to make some suggestions. First of all however, I do agree that there needs to be two types of CSR, one for FFA Gametypes and one for Team Gametypes. This is because the CSR system needs to measure both an individual’s effort and a team’s effort, and in most cases, both.

My suggestion is to have it as is follows, have a consistent invisible ranking bar that fills according to how you play your games, this should count your W/L and your K+A/D.

Have all players start at 0, you rank up if you hit 100 and you rank down if you drop below 0. Each rank-up resets the bar to 0 before adding on excess points accumulated from the last game you played.

The below examples assume all players start with 0/100 Points at level 25.

FFA Gametypes

Each game played is worth a base of 10 points, this is multiplied by your position:

Top 1-25% (1st-2nd place in an 8-Man FFA) - +10x Multiplier
Top 26-50% (3rd-4th place in an 8-Man FFA) - +5x Multiplier
Top 51-75% (5th-6th place in an 8-Man FFA) - -5x Multiplier
Top 76-100% (7th-8th place in an 8-Man FFA) - -10x Multiplier

So if you’re a player who places 1st or 2nd, you’d get the base 10 points multiplied by +10x, meaning you’d be at 100/100, in other words you’d rank up to 26 and begin the next rank at 0/100.

If you’re a player who places 3rd or 4th, you’d get the base 10 points multiplied by +5x, meaning you’d be at 50/100, or halfway up to your next rank at 25.

If you’re a player who places 5th or 6th, you’d get the base 10 points multiplied by -5x, meaning you’d be at 50/100 on your previous rank, ultimately ranking down once to 24.

If you’re a player who places 7th or 8th, you’d get the base 10 points multiplied by -10x, meaning you’d be at 0/100 on your previous rank, ultimately ranking down once to 24.

Team Gametypes

Each game played is worth a base of 10 points, this is multiplied by your K+A/D Spread and given a bonus via a win or a loss.

Positive K+A/D Spread: +7.5x Multiplier
Negative K+A/D Spread: -7.5x Multiplier
Win: +75 Points
Loss: -75 Points

So if you’re a player on the winning team that got 8 kills, 2 assists and 9 deaths, that’s 8+2/9 = 1.11, or positive, so you’d gain the +7.5x Multiplier, then you’d add the +75 for winning on top of that.
So 10x7.5+50 = 150, you’d rank up to 26 and begin again at 50/100. If you were to win another game in this fashion, you’d level up twice to 28 as your bar would fill to 100/100 twice.

If you’re a player on the winning team that got 5 kills, 4 assists and 15 deaths, that’s 5+4/15 = 0.6, or negative, so you’d have a -7.5x Multiplier, then you’d add the +50 for winning on top of that.
So 10x-7.5+75 = 0, your experience bar would not move and you would remain at 25.

If you’re a player on the losing team that got 15 kills, 1 assist and 8 deaths, that’s 15+1/8 = 2, or positive, so you’d gain the +7.5x Multiplier, then you’d take off 75 points for losing.
So 10x7.5-75 = 0, your experience bar would not move and you would remain at 25.

If you’re a player on the losing team that got 2 kills, 0 assists and 12 deaths, that’s 2+0/12 = 0.17, or negative, so you’d have a -7.5x Multiplier, then you’d take off 75 points for losing.
So 10x7.5-75 = 150, your experience bar would drop by 150 points so you’d delevel twice down to 50/100 at rank 23, if you were to lose another game in the fashion you’d delevel twice again down to 0/100 at rank 21.

Other Information

Minimum Amount of Games required for a 50 in FFA Gametypes: 49
Minimum Amount of Games required for a 50 in Team Gametypes: 33

Thoughts? I’m open to any criticism.

2

The linear scale will mean players at the top and bottom will inevitably trend towards infinity.

Also, the lack of accounting for the estimated strength of your opponents or teammates will make it easy to boost/grind games out.

I suggest you research systems like ELO, Glicko and TrueSkill and understand how they work first.

3

I won’t comment on how your numbers will work, but overall having a team based ranking system will make people work as a team to win, and not run around selfishly trying to get all the kills. It was a dumb idea to have CSR measure score, and it could be seen from a mile away. I guess just not by 343…

> Also, the lack of accounting for the estimated strength of your opponents or teammates will make it easy to boost/grind games out.

Bout to say that as well.

4

> Also, the lack of accounting for the estimated strength of your opponents or teammates will make it easy to boost/grind games out.

Whatever the system, it definitely needs to compare each players skill, if both a CSR5 and CSR49 get beaten by CSR50 with the same stats for that game then CSR5 should not be penalised anything like as harshly.

Your system doesn’t address this and its absolutely crucial.

5

> > Also, the lack of accounting for the estimated strength of your opponents or teammates will make it easy to boost/grind games out.
>
> Whatever the system, it definitely needs to compare each players skill, if both a CSR5 and CSR49 get beaten by CSR50 with the same stats for that game then CSR5 should not be penalised anything like as harshly.
>
> Your system doesn’t address this and its absolutely crucial.

Why would a CSR 5 with a 49 and 50?

In a team setting, the 5 sucks and holds down the whole team because he isn’t as good as everyone else. Or, you inv him to your party to boost him, in which party restrictions greatly matched you with players who were doing the same.

In FFA, my first question still stands.

6

> The linear scale will mean players at the top and bottom will inevitably trend towards infinity.

I don’t really follow what you’re getting at here.

> Also, the lack of accounting for the estimated strength of your opponents or teammates will make it easy to boost/grind games out.

That is definitely a problem, if only the system paired players of equivalent skill together.

> I suggest you research systems like ELO, Glicko and TrueSkill and understand how they work first.

TrueSkill uses a system which is pretty much identical to the traditional ELO system, and to be honest I’ve always felt has over complicated things.

Disagree with me if you like, but surely if the system is doing its job it should be pairing you with players of your skill level already right? So why does the system need to additionally work out if you should win or lose based on the levels of your opponents?

The answer is pretty obvious: The CSR System (and the TrueSkill System by extension) simply does not work, it takes random players and sits them together, works out which players should do well and which ones shouldn’t and then based on the results, alters the ranks of the players.

Now while this sounds good in theory, it doesn’t work as the players on the higher end of the scale are expected to win, so when they do, they don’t rank up, case-in-point, myself. And of course when the players on the lower end of the scale are expected to lose, they do, and they don’t rank down.

My complaint is pretty simple, if you win games repeatedly, you should still be ranking up because you do not belong at that level and you should not be playing against the players you’re playing against.

I mean I don’t mind playing against weaker players in SWAT repeatedly and beating up on them, gets me a lot of nice overkills, killtaculars and killtrocities for any montages I might put together in the future.

By all means keep feeding me big multikills and win after endless win, wrecking the K/Ds and W/L ratios of all those on the enemy team and thuroughly frustrating them at every turn, corner and endgame screen while I tear through player after player in a merciless rampage, causing hours and even days of work and effort on their part on their stats to go to waste in a matter of minutes as I continuously spawntrap, spawncamp, out-maneuver, out-strafe and out-think them in every single scenario, situation and systematic response they dare to try.

But still I feel that the system is a little unfair on those players, no?

7

> > The linear scale will mean players at the top and bottom will inevitably trend towards infinity.
>
> I don’t really follow what you’re getting at here.

If I’m one of the best players in the world, there simply won’t be many players around my level of play. So MM will have to put me against players that are known to be worse than me. But beating these players - in your system - is worth just as much as if I was playing someone at or above my level.

So, at least on average, I’m going to be facing a lot more players that are worse than me. So I’ll be gaining points and keep going up and up. The range will get stretched out.

Flip for the opposite end.

> > Also, the lack of accounting for the estimated strength of your opponents or teammates will make it easy to boost/grind games out.
>
> That is definitely a problem, if only the system paired players of equivalent skill together.

Matchmaking problem - not the skill ranking system. Balancing the conflicting goals of skill match, connection, and low waiting time.

> > I suggest you research systems like ELO, Glicko and TrueSkill and understand how they work first.
>
> TrueSkill uses a system which is pretty much identical to the traditional ELO system, and to be honest I’ve always felt has over complicated things.

No, TrueSkill is more similar to Glicko actually in that it has a notion of variance.

> Disagree with me if you like, but surely if the system is doing its job it should be pairing you with players of your skill level already right? So why does the system need to additionally work out if you should win or lose based on the levels of your opponents?

See above.

> Now while this sounds good in theory, it doesn’t work as the players on the higher end of the scale are expected to win, so when they do, they don’t rank up, case-in-point, myself. And of course when the players on the lower end of the scale are expected to lose, they do, and they don’t rank down.

Wrong. If you win your mean skill will always go up - just much less if you were expected to win anyway. If you lose your mean skill will go down - and again much less if you were already expected to lose.

8

> If I’m one of the best players in the world, there simply won’t be many players around my level of play. So MM will have to put me against players that are known to be worse than me. But beating these players - in your system - is worth just as much as if I was playing someone at or above my level.
>
> So, at least on average, I’m going to be facing a lot more players that are worse than me. So I’ll be gaining points and keep going up and up. The range will get stretched out.
>
> Flip for the opposite end.

If you’re one of the best players in the world, you’ll hit the cap at 50, like you should do.

I’m sorry but just in case you were unaware, I’m talking about Halo 4, I thought that’d be obvious as we’re on the Halo Waypoint forums but I suppose I failed to mention that I’m talking about Halo 4, which has a level cap at 50 and a lowest level at 1. So you can’t level above 50 or below 1. Just making that clear.

> Matchmaking problem - not the skill ranking system. Balancing the conflicting goals of skill match, connection, and low waiting time.

Yes, it is a Matchmaking problem, they do need to fix the issue, sadly no skill ranking system will work as intended until this issue is fixed.

> No, TrueSkill is more similar to Glicko actually in that it has a notion of variance.

Ah I see, I had no idea.

> Wrong. If you win your mean skill will always go up - just much less if you were expected to win anyway. If you lose your mean skill will go down - and again much less if you were already expected to lose.

No, if you win and the system expects you to win, you will not level. Case in point, myself. Take a look at my last ten games if you’re still unsure.

1-25 within 5 games, then I went up to 28 within 6 games following that, simply put, the leveling system suddenly slowed when I hit 25 due to the absence of stronger players to pair me up against.

I’ve also heard countless stories of people stuck at various levels for ten or twenty games just due to being paired against lower level opponents. Look around on this subsection of the forum and I’m sure you’ll find plenty on just the front page alone.