7 replies to this topic

[M T]

Ok, Im personally not at all happy on how matchmaking works. I also believe in fact doesnt work properly in the slightest. The constant result of stompmatches where its more then obvious that one team is full of newbies and the other side full of more experienced players. Resulting in games that end up in 12-1's etc.

I have little clue as how the ELO system works, but as far as I understand ELO is based on your Wins/losses (Match). Now unless I'm a total dork, basing a system on these numbers really make zero sense at all. Why? Because the end result of a match says nothing about a player's skill or experience at all. So why base the match making/elo on this?

Trying to balance a match based on classes says very little either. Put up an experienced light pilot versus a newbie player in an assault, Id bet my money on the light any day. Im not saying this isnt at all viable, because in the total 12v12 picture it does make a little bit of sense ofcourse. But this should always be a second priority consideration, not a main one.

What about balancing with using Kills as a number? Thats not going to work either. Why? Because the kills score says nothing about skill or experience either. It only gives you the point if it was your weapons that delivered the final blow.

I think the only way to get a rough result on how players should be classified in the ELO should be based on either: (Assuming PGI only wants to work with the numbers they already implemented in statistics)

- Weapon Accuracy
- Total damage done

Let me elaborate on the method implemention:

Weapon Accuracy
---------------------

Simply compare everyone's accuracy for each weapon type to determine skill. The more people used different weapons, the better. About everyone mustve used lasers in his life, so compare laser accuracy with that of others in a match for the same weapon type. The more data is mined, the better the match making result. If someone in the match is new then yes it will be a bit unbalanced possibly, but still. Also if someone never used a type of weapon before, just determine the result by picking another weapon he did use before.

Total Damage done
----------------------

This method simply uses the amount of damage done in a match, or possibly damage amount including assists (though im not sure how assists are calculated in this game, i assume its x amount of damage done to a mech before it dies)

Now you might be worried about lights etc not being able to do any damage at all. first of all thats incorrect. Sure they cannnot pump out as much damage in the short term, but they can easily get 1000-1500 damage in the long term. If this is really an issue this balancing methid can still take different Classes into account to compensate for this way of balancing.

At the end its all about killing your opponent, and damage equals kills.(Or deaths lol)
Of course another factor that would be hard to get a grasp on is the 'efficiency kills' like headshotting people etc, which wouldnt inflate your numbers as much.

I personally prefer the accuracy method. Tell me what you think PGI, the current system will most likely frustrate people too much, it does frustrate me I wont lie about it.

[OneEyed Jack]

Personal skill/performance might not decide the result of a given match, but that doesn't mean it's not represented. A better/worse player makes his/her team better/worse by some amount, and that will cause a trend for that player to win/lose more often, thereby pushing his/her Elo in the proper direction. No, it's not instant gratification. Get over it.

Weighting by personal performance discourages teamwork and encourages Rambo wannabes. Basing it entirely of damage is grossly misguided and suggests a complete lack of understanding about both the game and tactics in general.

[M T]

OneEyed Jack, on 11 April 2014 - 12:26 PM, said:

Personal skill/performance might not decide the result of a given match, but that doesn't mean it's not represented. A better/worse player makes his/her team better/worse by some amount, and that will cause a trend for that player to win/lose more often, thereby pushing his/her Elo in the proper direction. No, it's not instant gratification. Get over it.

Weighting by personal performance discourages teamwork and encourages Rambo wannabes. Basing it entirely of damage is grossly misguided and suggests a complete lack of understanding about both the game and tactics in general.

And thats the whole reason why they do not tell us exactly how the ELO system works. Other then some general aspects. No matter what, you cant deny the current system doesnt work well at all.

[Mazzyplz]

average out individual average stats per match (accuracy maybe or damage? tough call. accuracy would be difficult to judge since lasers have higher hit percentage than projectiles like ppc.)
with team performance.

works to weed out the noobs being carried. > simple enough to implement?

so something like


{[('dmg per match' times KDR)] + Winloss} / 2 = player's battle value?

compare between players

Edited by Mazzyplz, 11 April 2014 - 10:58 PM.

[Charons Little Helper]

W/L works in the long term.

Law of large numbers should average out your 11 buddies & 12 opponents.

Using any other stats starts to get subjective. (Example: Is that long ranged sniper who pokes all game for 3 kills really more valuble than the brawler who breaks up their formation and dies in a blaze of glory but only gets 150 damage?)

[MischiefSC]

The moment you place value in the matchmaker on something other than winning then you promote a migration to a behavior other than doing what it takes to win. This will then skew the results and make them invalid. If you attempt to include something other than win/loss into creating a score meant to represent how good a player is at winning then you will make an inherently unreliable and inaccurate system.

You can however drill down on it and say that in mechs of X chassis with Y loadout and playing with a team of Z composition you are weighted at W value to impact your teams odds of winning against Q enemy team composition.

At the moment though with current estimated population densities that would be a lot of work for little if any change - Currently the most logical and worthwhile system would be inherently general in estimating player skill across all mechs in a given weight/performance class. Even more important is the switch both to matching in a range (0-1000, 1001 to 1500, 1501 to 2800) and 3/3/3/3 you create broader and more reliably and accurately filled buckets in each of the three tiers for filling matches. It means you'll play with teams on both sides of a more homogeneous skill range, no more than 1 premade on either team and if 1 has one so does the other, also having a set weight class composition of teams broadens the viability of otherwise niche builds (backstabbers will always have 3 assaults and 3 heavies to choose from, light-hunters will always have at least 3 lights if not some fast mediums, etc).

The matchmaker with the incoming changes is going to give about the best possible matchmaking environment for the currently available population, mech/equipment/map balances and game modes.

You will always have stomps. Good games and bad games. If you don't want that, play single player. Otherwise accept the reality that how good or bad you play does matter in the long run. It in fact is what determines your rank in the game; how good you are at helping your team win. That doesn't mean getting kills (kill-stealing, using your team as bait, etc) or doing damage (a damaged enemy can still kill your teammates) or getting assists (spraying lasers /= driving a win) is the same as doing well. Getting a high match score doesn't mean doing well in the context of being a winner.

Being a winner means sometimes you die early and with little damage by being the vanguard of the push that wins the match. It means being a distraction sometimes, or stepping into the line of fire to help a teammate get safely into cover so he can then continue to do damage later in the match, helping your teams odds. Being useful in chat, having good suggestions on tactics, keeping an eye on your team in the map and supporting them by shooting people shooting at your teammates.

There are tons of tactics that lead to winning - not just damage/kills/assists. That's not a metric you can gauge absolutely but we don't have to. In the long run that behavior leads to more wins.

The only viable means of inferring how good someone is at helping their team win is..... looking at how often they successfully help their team win. We all play with the same group of people and teams. That's a constant, as random for everyone as it is for you. Your behavior in matches is the only thing that is consistent between all your matches and, thus, is the only reliable way to gauge your value in the matches that you've played in and predict your value in matches you will play.

[Whoops]

MischiefSC, on 12 April 2014 - 03:54 PM, said:

You will always have stomps. Good games and bad games. If you don't want that, play single player.

Oh, don't I wish this were possible...if for no other reason than to practice my aim against moving targets without dragging down my team.

[Mazzyplz]

i was just afk for a while and my team won.

i started walking when the enemy team had mostly died.
a win.

i still don't see how this is not a problem,
the conducive behavior to winning in this case was.... to keep shutdown basically for the whole match