MM values for specific balance (Part 1)

Introduction

In my previous post i already pointed out some values which you can use as a MM balance. Therefore there are a lot more values who can have a different effect to the game. 
Basically you can go several ways when you start with a MatchMaker:

• Random with just a tier/class adjustment
• Basic values, like winrate or ratings
• Detailed values
• Conclusion TL;DR at the end of Part 2

Each of those attempts have positive and negative aspects like, increased queue time, worse matchups in a bad tier and vehicle balance situation or short queues and so on.

Random MM

This is basically the easiest out of all 3 MM based codes, as you decide only on to 1-3 values who are creating the matchup. The lower the amount of values the faster should be the MatchMaking. You might ask on what values you usually pick for this:

• Tier balance

Tier balance would be, that every team has the same amount of tiers per side, as for example 5x T9, 5x T8 and 5x T7 per side. This value ignores the type of vehicle and playerskill, but makes sure that each side has the same amount of different tiers spread over both teams.

• Class balance

If i stick with my above example of tiers, this value checks on the classes that are spread over each side. This does make sense, as a class imbalance already makes it harder for one side to fight the other team, as they could not benefit from for example map advantages or key points.
This leads typical to imbalanced matchups, where for example 1 side gets all the vehicles with a high vision range or only heavy armored vehicles and so on.

• Platoon balance

Since platoons can enter the solo queue, they need to have a balance here, since platoons add something indirect to the players that are in it. Usually platooned players are communicating with a VOIP program, which is leading to an advantage over solo players who usually talk or write with each other in a random matchup.
Platoons in general are a hard value of balance, as in a perfect matchup you have both sides with a equal amount of platoons, within the same tier, the same vehicle and most likely within the same skilllevel. This means platoon balance has a lot of influence from the above mentioned values.

Conclusion

If you pick a random MM, you want to use either all 3 values to have a decent balance or you focus on just 1 or 2, while 2 values would make more sense, due the specific types of each code. What you can do is, that you don't use tier balance, due the fact if you have a good balance between the different tiers and a small tierspread, the negative aspect of not having it, would be lower, since the vehicle balance kicks in.

Regarding high tier balance in AW, this is not the case, so you need this one actually. Same goes for Class balance, which is needed due the circumstances of high tier balance, which leads me to platoon balance and their overall influence in a battle.

Basic values

Basic values use a set of different values that determine the skill of each player to get a fair matchup. While this sounds reasonable, it highly depends on how your very own rating works and on how much this rating is representing the actual playerskill in a short amount of time or in the long run.

 

• Playerrating

This is most likely the key factor for having a good or bad code. We all know the ratings which are used in WoT to determine playerskill or give a certain direction on how you can manipulate the rating in your favour. Everyone knows it and everyone done it to a certain point.
Though i know the pro's and con's for it, having a rating is a natural approach to make a difference between players. Chess for example has a ELO system or even football aka the FIFA has a system to represent somewhat a countries strength due recent results.

Even regular sports have to a certain point a rating, which is the time. If you run the 100m in 11 seconds as a male from the USA for example, you are quite good, but not world class. If we stick the the USA and a world class runner, which runs the 100m in 9.7seconds, he would be considered to be world class, just because reaching those times, require a certain amount of training and "skill".


Those are of course different examples on why having a rating is nothing exclusive, but needed if you want to have the difference shown between actual playerskill. The question is on how detailed this rating works out, and if it takes into count that the vehicle/class balance can be really impactful, like for example using illegal drugs to get more physical power, if you are a athlete.
Essential for this kind of MM, is the playerrating and what it does take into count. If this rating doesn't pick up the right data for a player, it can be seen as a flawed rating, due the false displaying of the actual skill of a player.

• Winrate

This is more a simple approach of a value, since it doesn't reflect a lot with a small sample size of games. The reason why it can be seen as a flawed value, is the amount of games you play in a vehicle, next to the fact, that you can be lucky with your teammates more or less, without contributing anything to each of your games.
Another important factor is the amount of wins versus the same tier, higher or lower tier, tierdifference (power creeps) or even classdifference. You can have for example a 60% winrate vs lower tiers, while using a power creeped vehicle, but just a 30% winrate vs same tier.

We could go even further down the road, and say that you have been incredibly lucky with your teammates and you got carried in 40 out of your 70 games for example. This would give you a high winrate, while the actual "winrate" is a lot worse than you are actually.
That is a main reason on why winrate balance doesn't reflect a lot by itself in a small sample size, and even in a larger sample size the chance of getting carried might be rng based high.

• Vehiclerating

You could start using a vehiclerating, next to your playerrating to determine differences in skill. While the playerrating reflects the overall outcome of a player from all his statistics, a vehiclerating is specific and bound to each vehicle you are using.
The advantage of having such a specific rating in your formula is, that it reflects your very own performance at its best. Questionable is, on what values you define this rating.
Reasonable values :

1. Battles in a vehicle up to a max, so you can't stack it endlessly
2. Average spotting damage compared to its role
3. Average damage compared to its role
4. Average reputation earned on win / lose / draw
5. Average damage blocked compared to its role
6. Average assist damage compared to its role

I think there could be some more values, but i think those are a good start for a basic rating. Within the next point (detailed values) i will most likely list up more values which you could even add to a vehiclerating, but those are really specific and might be too much for a basic rating to start off.

Conclusion

Ratings are needed as a part of what determines actual skill between players in any sports. The big question is, if you can identify the right values to clarify "skill" while having enough knowledge on where to cap skill. The latter one means, what i wrote before when mentioning " Battles in a vehicle up to a max ", as you need to set a maximal amount of battles who are influencing the rating for a vehicle in this specific case. 

Overall Obsidian needs to identify, if they are using the right values for their rating and they should get immediately away from a winrate focused MatchMaking, as the value itself displays not the actual playerskill up to a certain point.