37 replies to this topic

[FiveDigits]

Introduction

Matchmaking Phase 3 will feature a variation of the Elo rating system.
We do not know exactly what will factor into the rating, but I assume the rating will measure the personal performance of each player. Teams will probably be represented by the average rating of their players or might get their own rating (in 8v8).

In addition to player rating several other factors have been considered to balance match-ups:

• Drop Weight: Either each team's tonnage would have a limit (e.g. 500 tons for 8 players) or the team's total tonnage would somehow be factored into its rating. That means a lighter team of high-skilled players would be matched against a team of lower-skilled players in heavier mechs.
  This approach is flawed in that it does not account for any qualities of a mech besides its weight. Speed, weapons loadout and special equipment (e.g. ECM) would not be included in the rating process. This is also contrary to the concept of Role Warfare because it promotes the notion of "bigger is better".
  
  
• Battle Value: In the table top game the concept of Battle Value is used to balance opposing forces. It factors in the mechs' engines, weapons, equipment etc. While this would be more suitable than a simple drop weight measurement it is still a set of static factors. Adjustments would always be required to reflect changes in gameplay (balancing) and the emerging meta game.

Proposal: Mech'tchmaking
Elo (and related rating systems) have proven their worth for skill-based player vs. player matchmaking.

My suggestion is to expand those rating mechanics to the actual BattleMechs for matchmaking in MW:O. Just like players' Personal Rating (PR), each mech variant (HBK-4SP, AS7-D-DC, ...) should get its own Mech Rating (MR). When the system is first introduced starting ratings could be assigned based on statistical data of the variants' performance in the past or all mechs could start out at the same base rating. Each time a mech variant is played in a match (possibly separated by queue) the MR of the variant would be adjusted by the outcome of that match. Over time, the qualities of the individual variants - hardpoints, engine size limit, ECM capability etc. - would be reflected by its Mech Rating.

For the actual matchmaking process both the the player's Personal Rating and the Mech Rating of the mech variant the player drops with should be combined into a Drop Rating (DR). The individual players' DRs would congeal into a Team Drop Rating (TDR). Finally the matchmaker would try to match two teams with similar TDR.

I'll leave open how PR and MR are combined into DR for each player and how those can be combined into TDR. For the sake of simplicity they could just be averaged, but using median (for TDR) or weighted averages are options to be considered.

Example
Joe McWarrior (1450 PR) is a slightly below average (1500 PR) MechWarrior. He drops into solo queue in his trusty Hunchback HBK-4SP. The 4SP variant of the Hunchback is one of the most liked - and for good reason. It has established a nice 1870 MR. His Drop Rating works out to 1660 DR.

Mason Chist (1920 PR) is a hot shot player trying to unlock the Elite tier of his Raven piloting skills. So he drops in his craptastic RVN-2X (1350 MR) achieving a 1635 DR.

Being at a similar Drop Rating those two players may now face each other in the upcoming match. If Mason Chist had used his RVN-3L (1790 MR) instead then his 1855 DR would have been far out of Joe McWarrior's league.

Summary
I think that expanding an Elo'esque rating system to mechs and then combining the mech variant's Mech Rating with the pilot's Personal Rating to a Drop Rating would create a matchmaking system that takes the players skill as well as the advantages or disadvantages of the mech he drops in into account. Furthermore, when expanded to a Team Drop Rating, team matchmaking would account for "all Atlas" or "8 ECM" teams.
Once implemented such a system would be very low maintenance. The ratings would adjust naturally and reflect changes in gameplay mechanics dynamically - at least over time.

Do you think this would be a worthwhile addition to the matchmaking process or did I miss huge flaws in my proposed system?
Do you think rating values - especially Mech Ratings - should be publically visible? Or should ratings be hidden to prevent stagnation in the pool of "viable" mechs based on rating-biased perception of chassis or variants?

[tl;dr]
Give mechs an Elo rating just like the players.

Edited by FiveDigits, 18 December 2012 - 03:37 AM.

[Vassago Rain]

No, thanks.

[Antonio]

No, part of what affects a persons performance are the mechs that they typically choose so that variable is already folded in with their Elo rating.

Edited by Antonio, 18 December 2012 - 03:09 AM.

[FiveDigits]

Antonio, on 18 December 2012 - 03:09 AM, said:

No, part of what affects a persons performance are the mechs that they typically choose so that variable is already folded in with their Elo rating.

This may be true for a players individual Drop Rating, but the system would still be valuable to match team loadouts against each other.
Actually, I still do think that a good player in a "crap" mech should be rated differently from the same player in an "OP" mech.

[elbloom]

sound great !

[The Legendary Samurai]

Yeah... no.

[Redshift2k5]

Mech rating needs to be based on an aggregate score of the mech's current equipment, not it's base configuration. XL engines, double heatsinks, and ECM all have huge impacts on a variant/custom build's performance and must count towards it's score.

I'm pretty sure it was already said that phase 3 would include player skill as well as a modified BV-like system for the mech.

[FiveDigits]

Redshift2k5, on 18 December 2012 - 04:44 AM, said:

Mech rating needs to be based on an aggregate score of the mech's current equipment, not it's base configuration. XL engines, double heatsinks, and ECM all have huge impacts on a variant/custom build's performance and must count towards it's score. [...]

I'm not sure about that. This would come down to tracking rating for each mech component. This might become a little too complicated. It would of course be more precise than tracking variants only.
On the other hand, the mech variant decides which equipment a mech can fit - ECM or not, engine size, hard points etc. Other than that every player can effectively always choose the "optimal build" for his chassis. So, I'd consider it fair to match players based on their mech's potential instead of their actual equipment - for the sake of keeping the system simple.

[Vosje]

Good suggestion.
I, for one, would like to know how well I perform in my different chassis/model.

It does bring out a brainteaser however: It would require an Elo-score per chassis/model and possibly per role. Take a medium: It can be build as a heavy-scout or light-brawler. When should my Elo-score rise the most? On spotting the enemy? Assisting the Assaults? By Kills? Do I select a role up front? Do I have a detailed breakdown to work with and therefor multiple scores? (Player A is a great scout in his Jenner, but terrible at assisting.) Or do I just look at XP gained?

Note: Elo-scores are set by 1-on-1 engagements (like chess) by comparing scores, making a detailed outcome prediction and changing the score based on outcome. Anybody have any suggestions how to do this with 8 vs. 8? Adding the scores together?

[FiveDigits]

In a group situation the group gets assigned a group rating. This would be an aggregate of the players' Personal Ratings and their Mech Ratings - this creates the Team Drop Rating mentioned in the OP. In the most simple case this could be achieved by taking the average of all those values. (Alternatively, take an aggregated Team Personal Rating and Team Mech Rating to separate the personal and mech rating pools.)

After the match the winner receives a positive rating adjustement (e.g. +15), the loser a negative one of the same magnitude (-15). The rating adjustment magnitude is based on the rating difference between the teams. Winning vs. a higher rated team grants a larger positive adjustment and vice versa. This adjustment is then applied to all ratings involved:

• each player's Personal Rating
  
• each variant's Mech Rating
  
• the team's rating (if tracked seperately and not aggregated)

Edited by FiveDigits, 18 December 2012 - 08:21 AM.

[FiveDigits]

Vassago Rain, on 18 December 2012 - 03:06 AM, said:

No, thanks.

The Legendary Samurai, on 18 December 2012 - 04:08 AM, said:

Yeah... no.

What obvious flaw in the proposed system did I miss to warrant such blanket negative responses? Would you please elaborate?

[Vosje]

FiveDigits, on 18 December 2012 - 08:19 AM, said:

In a group situation the group gets assigned a group rating. This would be an aggregate of the players' Personal Ratings and their Mech Ratings - this creates the Team Drop Rating mentioned in the OP. In the most simple case this could be achieved by taking the average of all those values. (Alternatively, take an aggregated Team Personal Rating and Team Mech Rating to separate the personal and mech rating pools.)

Interesting.
It does make matchmaking for PUG's a bit harder as you first put together a team, calculate the score and then search for others to make a matching opposing team. But if you start out by building multiple pools of players based on Personal PR and MR, it should not be that hard.

I only see one problem at the moment: How do you recalculate the personal numbers?
Suppose you win with a (lone wolf) team of 5 great players (top of the pool), 2 stragglers (just worked themselfs up or having a bad day) and 1 unlucky disconnect. Do they all get the +15 (from your example) ?

[FiveDigits]

Vosje, on 20 December 2012 - 02:27 AM, said:

[...] I only see one problem at the moment: How do you recalculate the personal numbers?
Suppose you win with a (lone wolf) team of 5 great players (top of the pool), 2 stragglers (just worked themselfs up or having a bad day) and 1 unlucky disconnect. Do they all get the +15 (from your example) ?

Giving all players in the team the same adjustment based on both teams' Team Personal Ratings is the simple approach and I'd be fine with it.
Alternatively you could calculate each players rating adjustement based on his Personal Rating vs. the opposing team's Team Personal Rating.

[Brilig]

I suppose the only issue with the system would be if you could figure out a way to "game" it.

For instance, say you have a contract coming up, or are going to make a bid to attack a specific planet. Could you drive the MR down on a particular variant that would normally have a high MR?

Once you have driven it down say by getting a whole bunch of guys from your group to drop into matches and intentionally lose in that variant. Then when it comes time for your contract/land grab you can take a whole bunch of what really is a good variant at their now lower MR and go up against another team that will be in mechs that normally have a lower MR.

Kind of like how wrestlers will try to lose some weight before weigh in so they can compete in a weight class that is normally below theirs.


Edit: Actually this aught to be impossible to do as long as the data isn't cleared often. After a few months of games it would take a huge amount of negative drops to drive a mechs rating down.

Edited by Brilig, 20 December 2012 - 02:52 AM.

[Omigir]

Sounds suspiciously like Battle Value (BV)

Brilig, on 20 December 2012 - 02:49 AM, said:

I suppose the only issue with the system would be if you could figure out a way to "game" it.

For instance, say you have a contract coming up, or are going to make a bid to attack a specific planet. Could you drive the MR down on a particular variant that would normally have a high MR?

Once you have driven it down say by getting a whole bunch of guys from your group to drop into matches and intentionally lose in that variant. Then when it comes time for your contract/land grab you can take a whole bunch of what really is a good variant at their now lower MR and go up against another team that will be in mechs that normally have a lower MR.

Kind of like how wrestlers will try to lose some weight before weigh in so they can compete in a weight class that is normally below theirs.


Edit: Actually this aught to be impossible to do as long as the data isn't cleared often. After a few months of games it would take a huge amount of negative drops to drive a mechs rating down.

The funny thing about this^ is when BOTH sides do it.. they basically compete at the save level in the end. Only ones getting rick rolled are those who do not conform to the 'exploit'.

Edited by Omigir, 20 December 2012 - 02:55 AM.

[FiveDigits]

Scale is the key here. MR is calculated globally. Each and every match contributes to a variant's MR. To influence it in a meaningful way you'd have to manipulate hundreds or thousands of matches (by losing intentionally).
You could theoretically manipulate your PR to get easier opponents, but one's rating is of course a matter of prestige.
I think it might be wise to hide MR from public eyes, not because of fear of manipulation, but to prevent discrimination of players that specialized in an "inferior" chassis.

[CrazyPenguin]

I would like to see a BV system in play, perhaps modified by an ELO, with the caveat that some drops would not affect nor be affected by your ELO. I do not always like ELO systems because sometimes I want to play seriously, and sometimes I want to do something silly just for fun. An always-on ELO system would penalize me for that.

[FiveDigits]

Omigir, on 20 December 2012 - 02:53 AM, said:

Sounds suspiciously like Battle Value (BV)

Yes, it serves the same purpose as BV. The big advantage - at least in my eyes - is that MR evolves naturally. If a mech variant is good (because it's got a good hard point layout or can carry ECM etc.) then its MR will rise to the top. When the game changes (e.g. ECM gets nerfed) then the MR of the affected mechs will adjust.

Battle Value is static. People sit down and assign numbers to mechs/components based on their perceived value in the game at that time. If the game changes (MW:O will) those values become less accurate. They have to be adjusted manually.
It's also an infinite source of contention. People will argue over specific values all the time.

With a an Elo'esque system those values evolve naturally and always reflect the state of the game (admittedly lagging a bit behind). There's no point arguing against the MR value.

Edited by FiveDigits, 20 December 2012 - 03:06 AM.

[Brilig]

I really do like the idea. I have been trying to think of other ways to break it, and I am not coming up with anything.

[bob1234567890]

WOW if that's how they are doing it that is easily exploitable and i look forward to exploiting it in my rather large Merc Group. They dont have the populations to keep breaking people into groups like this. 8v8 almost never finds a match and that's cause of tonnage problems, now they want to make it so pilots affect how to find a match. If they don't allow rating differences of several hundred AT MINIMUM people will have trouble finding matches. Then you'll get massive differences between the teams. so dumb, matchmaking is so easy to do and balance why do we have to go through 50 phases so they just keep going in circles