42 replies to this topic

[sosegado]

Like the title says.

In a game where you can deathstar a lot of matches and still potentially lose because of win conditions why is Win/Loss the definition of a healthy match maker and not Kill/Death?

Wouldn't a healthy match maker be K/D based because your competition would eventually boil down to who is the more effective mech killer?

I understand it's a team game but don't you have more control over how you fight your own mech than you do whether the team helps you win the game?

Shouldn't the goal be a K/D of 1 not W/L of 1?

What am missing?

Edited by Stab Wound, 17 November 2022 - 11:17 AM.

[Duke Falcon]

It is because they not done that the Clan way...

[Vyx]

I suspect it's because they cannot affect your K/D as simply or as easily as they can affect your W/L.

Consider if your K/D got too high. In order to compensate (i.e., make you specifically die more often), they'd have to ... Increase your incoming damage? Lengthen your laser burn time? Slow your mech down? Increase your UAC Jam chance? Or increase your chances of an ammo explosion? None of that seems very fair ... or fun. Nor does it make much sense.

Whereas if your W/L gets too high, they simply match you up with poorer performing teammates.

Edited by Vyx, 17 November 2022 - 01:23 PM.

[Gagis]

The goal of the game is to win, so measuring your success at winning is the most accurate measure you can have. Kills that contribute to a victory are worth much more than kills that do not.

The game does have a fairly large luck element to it, but thats fine. You will get unlucky exactly as often as other players do and lucky exactly as often as other players do. The win/loss record over multiple matches does away with the luck component and is accurate.

Edited by Gagis, 17 November 2022 - 12:41 PM.

[feeWAIVER]

You can kill someone with 1pt of damage.
A kill is meaningless.

[sosegado]

feeWAIVER, on 17 November 2022 - 01:00 PM, said:

You can kill someone with 1pt of damage.
A kill is meaningless.

But can't you also win without even firing a shot? Or even moving from the drop zone?

What does that say about wins?

[Gagis]

Stab Wound, on 17 November 2022 - 01:17 PM, said:

But can't you also win without even firing a shot? Or even moving from the drop zone?

What does that say about wins?

You can sometimes get lucky. Sometimes you'll get unlucky instead. In the long term the luck cancels out and your track record reflects your actual contribution to matches.

[Nightbird]

For example, you can hide for most of the match, not contributing to your team, and come out only to finish weak mechs.

Doing that, it should be easy to maintain a 2.0 KDR, but your team will constantly be missing 1 player and you'll get a WLR around 0.7.

WLR erases most bad habits people do to farm kills or match score that ruins the fun for teammates, and focuses on what they actually do to help the team win. Tanking damage, contesting cap points, etc that don't contribute to kills will contribute to wins (selfless playing). Efficiently killing the enemy is rewarded with WLR but punished with Match Score.

Edited by Nightbird, 17 November 2022 - 01:24 PM.

[sosegado]

Nightbird, on 17 November 2022 - 01:23 PM, said:

WLR erases most bad habits people do to farm kills or match score that ruins the fun for teammates, and focuses on what they actually do to help the team win. Tanking damage, contesting cap points, etc that don't contribute to kills will contribute to wins (selfless playing). Efficiently killing the enemy is rewarded with WLR but punished with Match Score.

But wouldn't using K/D accomplish the same thing?

You wouldn't be farming kills because everyone on the team would be competing for those kills.

W/L seems very random.

If you remove the outliers that feeWAIVER and I brought up of a 1 point of damage kill and a 0 point of damage win the games should settle more towards the middle ground with the matchmaker using statistics that are possibly more skill based (hunting and killing mechs) then luck based (dropping with a group that may or may not be good)?

Edited by Stab Wound, 17 November 2022 - 04:37 PM.

[LordNothing]

you need to win about a third of the time for a game to be engaging. less than that and players tend to leave. nobody wants to play a game where they are cannon fodder for other players.

this is one of the reasons fp had no real uptake of players, because in some situations you could play all day and never win a game. fp never had a functional match maker (what it has now could hardly be called functional due to lack of players), but it does illustrate the consequences of very low w/l ratios. once you are over the rather steep learning curve a wl=1 in fp was sustainable, even as a pug, but most players would have quit long before they get to that point.

[Flea King Fleappuccino]

Want I really want to know how some players get 10-20 W/L!

[Gagis]

King Samu, on 18 November 2022 - 02:54 AM, said:

Want I really want to know how some players get 10-20 W/L!

There's a few players who game the website statistics by playing just a handful of games at opportune times and then stop for the season, at least on that account. Grimmechs rankings use the same data but filter them more for fewer outliers.

[sosegado]

I would be curious to see a deeper dive into the matches that end in 12-4 and 12-0.

For instance:

Both Blue team and Red team have a balanced W/L of 1...but...Blue team has an average K/D of 2 and Red team has an average K/D of .75.

In this scenario I'm going to bet on Blue team for the Win, and possibly a stomp of a win.

So if W/L ratio is the primary factor in the MatchMaker, shouldn't K/D ratio be at least the secondary factor(and heavily weighted)? Is there a secondary factor?

Maybe try changing the weight order for a season or three? Put K/D 1st and W/L 2nd?

I'm at the point in this game where winning a match isn't necessary for me to have fun. I honestly don't remember a time when it was.

I enjoy 11-12 losses way more than I do 12-0 wins.

Edited by Stab Wound, 18 November 2022 - 08:46 AM.

[Gagis]

The matchmaker is mostly based on match score, which is not accurate by every measure. It has a bit of win bonus, a lot of damage, a bit of kill bonus and a bit of random stuff in it.

How well the match score based matchmaker predicts long term win and loss records by how it ranks players, is the measure of how good of a score it is.

Its not all that accurate. Making it weigh kills more than wins would make it less accurate.

Making it weigh damage less might be a good idea, though.

Edited by Gagis, 18 November 2022 - 09:13 AM.

[Ken Harkin]

W/L in an ideal balanced world would be 50/50.

Most wins see 100% death on the L side nd a variable percentage of deaths on the W side. In other words in a perfectly balanced game K/D would be < 50/50

W/L is simpler to track and balance against.

[LordNothing]

King Samu, on 18 November 2022 - 02:54 AM, said:

Want I really want to know how some players get 10-20 W/L!

even with normal distributions you have your ourliers. just remember for every one of those 20 kdr players, you got an outlier at the other end of the bell curve that dies 20 times for each kill and yet for whatever reason (i presume they are masochists) still plays the game.

[crazytimes]

Who said that's the definition of a good match maker?

For this kind of game, my main interest is just some kind of seperation from new players and potatoes from people who know what they're doing. It can't really get much more granular than that because there just isn't the population to support it.

A lot of the 'metrics' that people try and use to justify perceived problems are largely subjective and based on nothing but the persons preference for playstyle.

[Nightbird]

crazytimes, on 18 November 2022 - 03:57 PM, said:

Who said that's the definition of a good match maker?

For this kind of game, my main interest is just some kind of seperation from new players and potatoes from people who know what they're doing. It can't really get much more granular than that because there just isn't the population to support it.

A lot of the 'metrics' that people try and use to justify perceived problems are largely subjective and based on nothing but the persons preference for playstyle.

If a game puts you at a skill tier where playing puts you at 0.5 WLR or 2.0 WLR, is that the right skill tier?

If not, then what should your WLR be when you are placed in the right tier?

[crazytimes]

Nightbird, on 18 November 2022 - 04:55 PM, said:

If a game puts you at a skill tier where playing puts you at 0.5 WLR or 2.0 WLR, is that the right skill tier?

If not, then what should your WLR be when you are placed in the right tier?

So what should we do with all the people with thousands of games at the +4 W/L? Put them in their own MM pool? They have to wait until they're all online at exactly the same time to get a match?

Like I said, there isn't even the population to granulate into 5 tiers- there's a 3 tier spread which means there's only really two MM pool as it stands. For other games that have a concurrent population in the tens to hundreds of thousands- use whatever metric you want. For MWO- as long as we can broadly seperate the potatoes from the rest and not have long queue times, most people will be happy.

[Nightbird]

crazytimes, on 18 November 2022 - 06:00 PM, said:

So what should we do with all the people with thousands of games at the +4 W/L? Put them in their own MM pool? They have to wait until they're all online at exactly the same time to get a match?

Like I said, there isn't even the population to granulate into 5 tiers- there's a 3 tier spread which means there's only really two MM pool as it stands. For other games that have a concurrent population in the tens to hundreds of thousands- use whatever metric you want. For MWO- as long as we can broadly seperate the potatoes from the rest and not have long queue times, most people will be happy.

You just have to ensure there is the same number of those +4 W/L people are on each team. MWO's match maker will happily put 8 of them on one team, 0 on the other, because there are 8 "T1" people with 1.0 W/L to put on the other team lol.

Edited by Nightbird, 18 November 2022 - 06:08 PM.