Jump to content

- - - - -

Why Not To Judge Performance By Score


58 replies to this topic

#41 Alaskan Nobody

    Member

  • PipPipPipPipPipPipPipPipPipPipPip
  • The Determined
  • The Determined
  • 10,358 posts
  • LocationAlaska!

Posted 20 November 2013 - 12:13 PM

View PostBront, on 20 November 2013 - 08:45 AM, said:

Sure, stats rarely tell the whole story (for example, in my mech with my worst K/D ratio, I have my best winning percentage, so clearly I'm better off to my team dead when I play in that mech :) )


Playing Devils Advocate here:
Not necessarily: I have a lousy K/D ratio on my Locust (obviously) but a decent W/L - because I spend a lot of time capping.
I am Much more valuable to my team alive in that mech than dead - even though it is a Locust. :D

That is more or less the big point of the thread.
Stats only tell part of the story - there are lots of ways you can be the biggest contributor to the team, and still score the lowest.

#42 Victor Morson

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 6,370 posts
  • LocationAnder's Moon

Posted 20 November 2013 - 02:25 PM

View PostDavers, on 19 November 2013 - 09:45 AM, said:

Elo is a measurement of how often you complete your objective.


In practice right now, ELO is a measurement of how moody Match Maker was being that week.

#43 Davers

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 9,886 posts
  • Facebook: Link
  • LocationCanada

Posted 20 November 2013 - 03:58 PM

View PostVictor Morson, on 20 November 2013 - 02:25 PM, said:


In practice right now, ELO is a measurement of how moody Match Maker was being that week.

If that was true then there would be no 'high elo' or 'low elo' just lucky and unlucky. But I have been in many matches where the outcome was uncertain until the final shots were fired. Could the MM be tuned further? Of course. I would rather every variant had it's own Elo for instance. But in general I think win/loss is the best way to do matchmaking.

#44 Katkon

    Member

  • PipPip
  • 20 posts

Posted 20 November 2013 - 06:06 PM

Well, I can only go by one data point: me. I tend to get low scores / inflict relatively low damage with my Heavy Metal (300s is as high as I got once I figured out the key was to ditch the XL engine), and I KNOW I'm cr@p :) Luckily, I play for fun and couldn't care less. Frankly, I enjoy the fact that this game - and its players - generally seem less focused on stats than many other shooters. Based on in-game comments, players generally seem to encourage each other - and the crowd is generally polite. Perhaps it's due to an older, more mature crowd - I can't be sure. But I like it!

#45 Katus

    Member

  • PipPipPip
  • 82 posts
  • LocationTucson AZ

Posted 21 November 2013 - 03:48 PM

I have to agree that damage and mission score are not good ways to judge a pilot. If it were then I would be considered the next Kai Allard-Liao rather than the goof that I am.. ;)

Posted Image

#46 Victor Morson

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 6,370 posts
  • LocationAnder's Moon

Posted 02 December 2013 - 06:09 AM

View PostDavers, on 20 November 2013 - 03:58 PM, said:

If that was true then there would be no 'high elo' or 'low elo' just lucky and unlucky.


Right now? Pretty much what it is. The only reason some people have higher ELO than others (in particular competitive players) is that they run with 4 mans a lot, which CAN influence games a lot, unlike solo drops. Thus they/we get padded stats.

View PostDavers, on 20 November 2013 - 03:58 PM, said:

But I have been in many matches where the outcome was uncertain until the final shots were fired.


Statistically with the coin-flip style ELO matching they're doing right now, this is bound to happen sometimes through sheer luck.

#47 ManaValkyrie

    Member

  • PipPipPipPipPipPipPip
  • Mercenary
  • Mercenary
  • 507 posts
  • LocationUK

Posted 02 December 2013 - 09:09 AM

The major issue with matchmaker is it wants to drive your win/loss ration to 1:1 i.e 50%. If you are having a good set of wins, the match maker will stack the deck against you. At least that is how it's coming across presently, over the course of a weekend i tracked every match.

I recorded wins and losses, match type, my own performance and damage as well as screenshot the final board to see the weight ratios and groupings. I have noticed from this a tendancy for if you have had several wins that to bring you back towards losses it will seemingly heavily weight imbalance the other team. For example an Assault match where one team has 200-300 tons in weight advantage or in a conquest match you end up with no lights or fast mediums and the other team has 4 plus lights and a spread of tonnage in the other weight categories.

The problem is that ELO and matchmaker are actually seemingly at odds with each other right now. ELO is supposed to determine your skill set against other players and matchmaker should be putting games towards a point where the match should literally come down to which team has the better combination of luck, skill and teamwork on a given round rather than a forcible stack the odds.

#48 Buckminster

    Member

  • PipPipPipPipPipPipPipPip
  • Elite Founder
  • 1,577 posts
  • LocationBaltimore, MD

Posted 02 December 2013 - 09:17 AM

Although especially in PUGs, it's really hard to know how well a given team will come together. It's no different than any other team sport - how the group plays as a team is not reflected in individual stats. So sometimes you may get a bunch of statistically good players that just don't work together well, and you end up with a roflstomp.

The one thing I've always wondered is if your ELO takes into account the mech you are using. Do I, as a pilot, have a single ELO number? Or does it vary based on the chassis, or even on the variant? I'd hate to be ranked artificially high because I've been good in other mechs.

#49 TercieI

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 8,161 posts
  • LocationThe Far Country

Posted 02 December 2013 - 09:21 AM

View PostBuckminster, on 02 December 2013 - 09:17 AM, said:

Although especially in PUGs, it's really hard to know how well a given team will come together. It's no different than any other team sport - how the group plays as a team is not reflected in individual stats. So sometimes you may get a bunch of statistically good players that just don't work together well, and you end up with a roflstomp.

The one thing I've always wondered is if your ELO takes into account the mech you are using. Do I, as a pilot, have a single ELO number? Or does it vary based on the chassis, or even on the variant? I'd hate to be ranked artificially high because I've been good in other mechs.


Supposedly it's by weight class.

View PostVictor Morson, on 20 November 2013 - 02:25 PM, said:


In practice right now, ELO is a measurement of how moody Match Maker was being that week.


QFT. I may have a new sig...

#50 Threat Doc

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Bowman
  • The Bowman
  • 3,715 posts
  • LocationO'Shaughnnessy MMW Base, Devon Continent, Rochester, FedCom

Posted 02 December 2013 - 09:45 AM

View PostShar Wolf, on 18 November 2013 - 08:25 PM, said:

A 3 Kill 200 damage match may have been you 'kill stealing!!1!' and the 3 kill 700+ match may have been you 'kill stealing!!1!11111!!' and being a lucky-yet-crappy shot at the same time. :P
I am, perhaps, one of the worst Mech Drivers on the planet... in three weeks (as of Sunday evening), I have played in a total of 247 drops. Of the 69 kills I've had, thus far, I can only confirm four (4) of those as being absolutely clean, two of which happened this past week, one last night. Those kills were 100% mine, no chance for damage by anyone else. What I wish is that PGI would come up with a way to convert kill assists into kills; hell, then I'd have around 100 - 150 kills. I'm an LRM jockey, I love my Catapults, but my highest LRM percentage is my LRM-10s with 24%, which is ridiculously low. Most of my kills have happened with LRMs. Alternately, my highest-use weapon other than LRMs and AMS, which doesn't count, anyway, is my Large Lasers, which I've only actually used in 86 matches, and I barely use them in these other matches, and I have a 72.5 hit percentage. It's utterly ridiculous how bad LRMs are in this game, and my kills would be more if PGI would raise their hit possibilities by about 20-30%.

So, the kill stealing -which I prefer to refer to as good teamwork-, absolutely the truth. If you do the highest percentage of damage, not the final damage, I've noticed, you get the kill... even if you disengaged that target earlier; otherwise, you get a kill assist. I've noticed, more often than not, that I will not get a Kill or Kill Assist until after I've died, in the final stats screen for the drop. If you don't believe me, start watching your chat box, more. I've received kills more than a minute after I disengaged the target, when they went down; same for kill assists and savior kills. I got a huge savior kill last night, thirty seconds after I had moved onto another target, when the guy whose butt I saved, a superior pilot to me, actually finished the 'Mech off AFTER dropping below 35%.

Edited by Kay Wolf, 02 December 2013 - 09:46 AM.


#51 arghmace

    Member

  • PipPipPipPipPipPipPip
  • Elite Founder
  • Elite Founder
  • 845 posts
  • LocationFinland

Posted 02 December 2013 - 09:46 AM

View PostKoniving, on 19 November 2013 - 12:32 AM, said:

While some of my teammates were busy killing disconnected mechs...


And this is why kills per match is also a bad statistic. If you wanna win, stop shooting at mechs that have lost their weaponry. All too often you see people running crazy after weaponless zombies shouting Kill! Kill! Kill! while the battle is being lost to their stupidity.

#52 Tesunie

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The Seeker
  • The Seeker
  • 8,586 posts
  • LocationSeraphim HQ: Asuncion

Posted 02 December 2013 - 10:58 AM

View PostKay Wolf, on 02 December 2013 - 09:45 AM, said:

I am, perhaps, one of the worst Mech Drivers on the planet... in three weeks (as of Sunday evening), I have played in a total of 247 drops. Of the 69 kills I've had, thus far, I can only confirm four (4) of those as being absolutely clean, two of which happened this past week, one last night. Those kills were 100% mine, no chance for damage by anyone else. What I wish is that PGI would come up with a way to convert kill assists into kills; hell, then I'd have around 100 - 150 kills. I'm an LRM jockey, I love my Catapults, but my highest LRM percentage is my LRM-10s with 24%, which is ridiculously low. Most of my kills have happened with LRMs. Alternately, my highest-use weapon other than LRMs and AMS, which doesn't count, anyway, is my Large Lasers, which I've only actually used in 86 matches, and I barely use them in these other matches, and I have a 72.5 hit percentage. It's utterly ridiculous how bad LRMs are in this game, and my kills would be more if PGI would raise their hit possibilities by about 20-30%.

So, the kill stealing -which I prefer to refer to as good teamwork-, absolutely the truth. If you do the highest percentage of damage, not the final damage, I've noticed, you get the kill... even if you disengaged that target earlier; otherwise, you get a kill assist. I've noticed, more often than not, that I will not get a Kill or Kill Assist until after I've died, in the final stats screen for the drop. If you don't believe me, start watching your chat box, more. I've received kills more than a minute after I disengaged the target, when they went down; same for kill assists and savior kills. I got a huge savior kill last night, thirty seconds after I had moved onto another target, when the guy whose butt I saved, a superior pilot to me, actually finished the 'Mech off AFTER dropping below 35%.


Kill assists and Savior Kills are not related to kills. Whomever (as far as I know) get the killing shot gets the kill. Anyone who so much as tapped the target in the match will get a kill assists. If that mech was killed after damaging a friendly to 30% or less, everyone who assisted in that kill should also get the Savior Kill, even though they only assisted with the kill. Savior Kills do not care if it's a kill assist or an actual kill. (Correct me if I'm wrong.)

#53 Amsro

    Member

  • PipPipPipPipPipPipPipPipPip
  • Overlord
  • Overlord
  • 3,436 posts
  • LocationCharging my Gauss Rifle

Posted 02 December 2013 - 11:15 AM

Way too many variables to be able make any accurate value calculations.

The MVP of one match will be the Troll of the next, one misstep can mean disaster. In a pug match all you can do is attempt to work with the team. :P

I think if you can average 200-400 damage per match you're likely an asset to the team. Maybe a bit higher for Assaults and a bit lower for Lights. :D

Its the sub 50 damage matches you want to avoid! :D

#54 Threat Doc

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Bowman
  • The Bowman
  • 3,715 posts
  • LocationO'Shaughnnessy MMW Base, Devon Continent, Rochester, FedCom

Posted 02 December 2013 - 11:25 AM

View PostTesunie, on 02 December 2013 - 10:58 AM, said:

Kill assists and Savior Kills are not related to kills. Whomever (as far as I know) get the killing shot gets the kill.
All I know is what I've observed. I've had Kills come up in my final stats screen, AFTER I was killed, because I did the highest amount of damage, without making the kill shot. I cannot, honestly, say that it wasn't just me not being able to pay attention to my chat box, but I know there have been times I was surprised at the number of kills, assists, saviors, etc., I've had after some games.

Quote

If that mech was killed after damaging a friendly to 30% or less, everyone who assisted in that kill should also get the Savior Kill, even though they only assisted with the kill. Savior Kills do not care if it's a kill assist or an actual kill. (Correct me if I'm wrong.)
I think, in this particular instance, you have it wrong. It increments if I'm not mistaken... a Kill Assist is exactly what it says, and the step up from that is the Kill, and the step up from that is the Savior Kill. A Kill is a Kill, but Savior Kills are pretty much the ultimate expression of getting the Kill.

#55 jper4

    Member

  • PipPipPipPipPipPipPipPip
  • Urban Commando
  • Urban Commando
  • 1,884 posts

Posted 02 December 2013 - 11:28 AM

View PostShar Wolf, on 20 November 2013 - 12:13 PM, said:


Playing Devils Advocate here:
Not necessarily: I have a lousy K/D ratio on my Locust (obviously) but a decent W/L - because I spend a lot of time capping.
I am Much more valuable to my team alive in that mech than dead - even though it is a Locust. :D

That is more or less the big point of the thread.
Stats only tell part of the story - there are lots of ways you can be the biggest contributor to the team, and still score the lowest.



yeah same here, my Locusts come out to a positive w/l while my kdr in every one of them is under .4 but i usually almost always play them on conquest and probably a third of the kills were discos at base that i pinged away at while capping. like someone else also mentioned i also have a few mechs who help the team win by being dead. current "leader" is my trebbie 5J 1.32 w/l .57 kdr. trebbie 7M is just under a 1.0 w/l with a kdr of .4- so clearly if you see me with a trebbie i'm going to die soon but the rest of you will be getting a win out of it.

i do feel embarrassed though those matches where i'm at 100 or so dmg but have 2-3 kills (usually play a medium so mopping up is one of their better options). other day i had a match in Terra where i went into the caldera where a friendly hawk was harassing an atlas with a quickdraw sniping away a few hundred meters away. i saw the quickdraw- fired one ac2 round at it and it died. turned to the atlas fired one alpha (ML, 2ac2 and a streak or two) at it and it died too. was like "what the hell just happened?" had to apologize in chat cause i felt bad for such a random pair of Kills :P then again that makes up for the srm/lrm cat my streaktaro knocked from 100% to 50% only to have some lrm finish it off in one volley i suppose

#56 Tesunie

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The Seeker
  • The Seeker
  • 8,586 posts
  • LocationSeraphim HQ: Asuncion

Posted 02 December 2013 - 11:45 AM

View PostKay Wolf, on 02 December 2013 - 11:25 AM, said:

All I know is what I've observed. I've had Kills come up in my final stats screen, AFTER I was killed, because I did the highest amount of damage, without making the kill shot. I cannot, honestly, say that it wasn't just me not being able to pay attention to my chat box, but I know there have been times I was surprised at the number of kills, assists, saviors, etc., I've had after some games.
I think, in this particular instance, you have it wrong. It increments if I'm not mistaken... a Kill Assist is exactly what it says, and the step up from that is the Kill, and the step up from that is the Savior Kill. A Kill is a Kill, but Savior Kills are pretty much the ultimate expression of getting the Kill.


I have had matches where I got no kills, but I would see in my chat box "Kill Assists, Savior Kill" from the same mech. As far as my understanding goes, he who gets the kill shot gets the kill, but anyone who scratched the paint gets to claim an assist. I can say I've had targets were I one on one them till they had red CT internals before, just to not get the kill as some stray (not really stray per say) shot hits them before I can hit them again and someone else gets the kill.

Then again, I've had several matches were I kept count of those mechs I hurt, and I wasn't having the correct assists assigned to me. I've had it where I almost killed someone before, someone else got the kill shot, and I didn't even get an assist, which was strange.

I think the rewards might still have some bugs in it from my observations about assists. Maybe you are seeing other bugs with kills? I mentioned this to support, but they wanted me to send video evidence of it happening... and my computer is old so I am worried about compromised game performance if I was to try and record the match in the background... I was not able to prove nor disprove this to support personally...


PS: I do recall a match where I got a kill somehow and did only 1 point of damage... I'm sure I didn't kill anyone that match too... :P

Edited by Tesunie, 02 December 2013 - 11:46 AM.


#57 Davers

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 9,886 posts
  • Facebook: Link
  • LocationCanada

Posted 02 December 2013 - 11:51 AM

View PostVictor Morson, on 02 December 2013 - 06:09 AM, said:


Right now? Pretty much what it is. The only reason some people have higher ELO than others (in particular competitive players) is that they run with 4 mans a lot, which CAN influence games a lot, unlike solo drops. Thus they/we get padded stats.



Statistically with the coin-flip style ELO matching they're doing right now, this is bound to happen sometimes through sheer luck.


Statistically then every outcome is possible. This does not invalidate Elo. The only thing that would prove the matchmaker is completely broken as some people claim would be players who (after a suitably large pool of matches) who won 90+ percent of games or who lose the same. Even then it just might be the fault of the size of the community- after all, a grand master dropping randomly into a local chess club is unlikely to find a good match.

#58 SniperCon

    Member

  • PipPipPipPipPipPip
  • 243 posts

Posted 02 December 2013 - 12:40 PM

I have yet to play with a group. My first 'Mech (183 drops) hasa 0.74 W/L. My latest 'Mech (138 drops) has a 1.67 W/L. Don't tell me that solo drops don't have a measurable influence on match results.

I would agree that the upper limit to W/L dropping solo is lower than the upper limit dropping 4 mans. However, the upper limit for solo drops is greater than 2.0.

#59 Victor Morson

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Elite Founder
  • 6,370 posts
  • LocationAnder's Moon

Posted 02 December 2013 - 02:23 PM

It's a shame there's not a "Components destroyed" and a breakdown of which ones and average damage to each. Now that would be an interesting gauge of a player.





5 user(s) are reading this topic

0 members, 5 guests, 0 anonymous users