298 replies to this topic

[Joseph Mallan]

FrDrake, on 18 November 2013 - 07:52 AM, said:

I play conquest, so maybe assault is different, but there are plenty of times the game is won on the back of the guy with the least amount of damage on the team.

A fair counter argument! Yet you did provide the most effort to win the match by the primary condition.

[Sug]

Ghogiel, on 18 November 2013 - 12:55 AM, said:

Start AFKing and after 1000 matches come tell me how you maintain a 50/50 W/L

I will do 100 afk matches on a fresh account.

[Odin]

Sug, on 18 November 2013 - 08:58 AM, said:

I will do 100 afk matches on a fresh account.

Tell me about it here!
Good Luck!

[FerretGR]

Odin, on 18 November 2013 - 07:15 AM, said:

Standing back?

Read my postings. Don't add your own schemes.

I'm not sure what "standing back" means or what it has to do with my post.

I did "read your postings." The line of thinking I'm trying to address comes directly from one of your "postings:"

Odin, on 18 November 2013 - 07:15 AM, said:

But no matter what, ending up as one of the top pilots of the drop, more often then not, doesn't seem to have any effect overall.
You either get shot out early, score a number of kills or help with lots of damage done, win or loss is not in your hand alone.
I think individually, your to small a part of it.

You're arguing that the individual's impact on the match outcome is so small that W/L is meaningless when calculating personal skill. I'm trying to explain to you that your performance factors into your Elo rating, by way of Ghogiel's example of non-performance, something that you accept. If we don't participate, ie. if we AFK, we negatively impact our team. Logically, the opposite must be true. Therefore, W/L does indeed reflect personal skill. I'm not "adding my own schemes."

Joseph Mallan, on 18 November 2013 - 07:18 AM, said:

How is MY performance based on a stat that is only 8% of my doing? If my 8%, kills 4 and delivers 600+ damage and loses v my doing 26 damage and wins Why should my Elo suffer from a team's loss or victory? If I win cause the team picked up my slack Is it right I get bumped up in Elo? If I lose and put in a spectacular game, should I get lowered because the team failed to pull their weight? W/L is not the correct way to judge individual performance

Yes, that's an issue. Sometimes, we're going to have spectacular games, and our Elo doesn't leap to reflect that. Sometimes, we're going to have crappy games, and despite this, our team wins and we get the Elo boost anyway. Nobody is arguing that Elo is perfect. But on average, these things should work themselves out. If you consistently have games like the one you describe, you're going to be an asset to your team, and you'll help them to win, which will increase your Elo rating. Will it happen every time you get 4 kills? Maybe not. Will it happen often enough that those 4 kill games will usually lead to an increased Elo? Definitely.

On average, yes, there are situations where our personal contribution goes unnoticed, and our team loses despite our Herculean efforts. But I think, Joe Mallan, that if you charted your 4-kill games, you end up winning FAR more of them than you lose, and so that performance is indeed reflected in your Elo.

Like I've been saying, though, it's not perfect, but neither would be basing your Elo on kills/damage/etc, as FrDrake points out with the capping example! We have to go for the one statistic that kills, caps, and everything else becomes a part of, and that's whether the team wins or loses. And while you might only play a small role in that, you most certainly play a role. The fact that you're role is small (ie the 8% we're talking about) means that it will take longer for Elo to give us a true representation of skill, but it also means, from my perspective, at least, that given enough games and enough time, Elo is as good a tool as any other for measuring player skill.

Whether the matchmaker uses it properly or whether the matchmaker needs tweaking, those are another issue

Edited by FerretGR, 18 November 2013 - 10:00 AM.

[Joseph Mallan]

FerretGR, on 18 November 2013 - 09:50 AM, said:

You're arguing that the individual's impact on the match outcome is so small that W/L is meaningless when calculating personal skill. I'm trying to explain to you that your performance factors into your Elo rating, by way of Ghogiel's example of non-performance, something that you accept. If we don't participate, ie. if we AFK, we negatively impact our team. Logically, the opposite must be true. Therefore, W/L does indeed reflect personal skill. I'm not "adding my own schemes."

If Elo does its thing properly, each member of a 12 man team provides 8.333% towards victory or defeat.

[FerretGR]

Joseph Mallan, on 18 November 2013 - 09:58 AM, said:

If Elo does its thing properly, each member of a 12 man team provides 8.333% towards victory or defeat.

Indeed... a small contribution, but a contribution nonetheless! Discussed above in an edit added after you posted Mr. Mallan, sorry for the ninja edit

[Doomstryke]

Either way I'd love to have a stat reset, My win/loss isn't great but according to the matchmaker I need to drop with fresh people for another 1000 matches or so before my win/loss evens out to 50/50. That technically means my team automatically starts with -8% or whatever number you guys have that will be stacked against us + all the other noobs who die in under a minute which essentially means that about 2 minutes into a match the team I'm on already has an additional 25-35% chance of loosing.

if that is what you guys are saying then stats or at least win/loss + elo needs to be reset like every 3 months because that just sucks. It also explains why some people can sync drop together almost every time while others can't (well we can just not on same team)

[St4LkeRxF]

FerretGR, on 18 November 2013 - 09:50 AM, said:

You're arguing that the individual's impact on the match outcome is so small that W/L is meaningless when calculating personal skill. I'm trying to explain to you that your performance factors into your Elo rating, by way of Ghogiel's example of non-performance, something that you accept. If we don't participate, ie. if we AFK, we negatively impact our team. Logically, the opposite must be true. Therefore, W/L does indeed reflect personal skill. I'm not "adding my own schemes."

If we talk about 1on1 fights W/L should reflect your skill, but when you drop with trigger happy gang that core you in base trying to set up weapon groups, or when you drop with banzai gang and so on that will leave you damaged or alone vs whole enemy team that can lead to negative W/L, doesn't mean someone is bad player.

Maybe ELO need to look at players W/L when he/she solo drop, drop as 4 man group and in 12 man group.

[Joseph Mallan]

FerretGR, on 18 November 2013 - 10:02 AM, said:

Indeed... a small contribution, but a contribution nonetheless! Discussed above in an edit added after you posted Mr. Mallan, sorry for the ninja edit

My Jager-A Stats: As Anton
144 Games 81/63 (1.29)

156 Kills/97 Deaths (1.61)

My Atlas-D-DC Stats
240 Games 122/116(1.05)

161 Kills/ 146 deaths(1.10)

My (F)Atlas has a
104 Games 57/47 (1.21)
49 Kills/70 Deaths (0.70)

The D-DC is the most used Mech, It was played mostly as a 4 man Mech As was my (F)Atlas. Funny how my personal worst Mech (lowest K/D) is my 2nd best W/L! Carried?

[FerretGR]

St4LkeRxF, on 18 November 2013 - 10:39 AM, said:

If we talk about 1on1 fights W/L should reflect your skill, but when you drop with trigger happy gang that core you in base trying to set up weapon groups, or when you drop with banzai gang and so on that will leave you damaged or alone vs whole enemy team that can lead to negative W/L, doesn't mean someone is bad player.

If you were seeing that sort of thing in every single match, St4LkeR, I'd say you have a point. But the bottom line is that the matches with shenanigans that affect your W/L should average themselves out.

[FerretGR]

Joseph Mallan, on 18 November 2013 - 10:44 AM, said:

The D-DC is the most used Mech, It was played mostly as a 4 man Mech As was my (F)Atlas. Funny how my personal worst Mech (lowest K/D) is my 2nd best W/L! Carried?

I find stats like that very interesting, to be honest! For me it goes to show that there isn't a precise and exact relationship between KD and WL, and it makes for an interesting discussion. What's the difference between those mechs and their loadouts that might lead to numbers like that? If I was to guess, I'd say you ran your (F)Atlas as an LRM boat, which would lead to significantly less kills while still providing damage and support, whereas your Jager-A is a AC40 ride, which means lots of pinpoint damage and killshots. Am I getting warm?

[MischiefSC]

Joseph Mallan, on 18 November 2013 - 10:44 AM, said:

My Jager-A Stats: As Anton
144 Games 81/63 (1.29)

156 Kills/97 Deaths (1.61)

My Atlas-D-DC Stats
240 Games 122/116(1.05)

161 Kills/ 146 deaths(1.10)

My (F)Atlas has a
104 Games 57/47 (1.21)
49 Kills/70 Deaths (0.70)

The D-DC is the most used Mech, It was played mostly as a 4 man Mech As was my (F)Atlas. Funny how my personal worst Mech (lowest K/D) is my 2nd best W/L! Carried?

Remember, it's about how your performance plays out in the long run. Same reason you keep your weapon well maintained - even if you cleaned it the other day you clean it again because statistically in the long run those odds play out to keep you alive.

So in a match where you do 600 damage and 4 kills but your team sucked and you still lost it feels like your talent was wasted. That time you do 600 damage and 4 kills but your team clicked, you were the winning factor. A lot goes into winning and losing a match but if you're always bringing extra then you'll skew those odds in your favor. That's why it takes a lot of matches though in 12v12 to seat Elo effectively.

Your stats help prove the point though - your Jag is clearly a mech you do a lot of successful killing in. Consistently in fact, so it skews your wins up. I'm curious to know how you play your D-DC and its loadout vs your Fatlas? I know that my D-DC has a high win/loss and a moderate KDR but that's because of how I play - I tank for my team, soak damage and don't hesitate to keep drawing fire until it kills me if it keeps the enemy exposed to my team.

Kills and damage don't always contribute directly to wins. Especially in an Atlas or a Raven or ECM light. There's a lot you can do in how you play that contributes to wins without directly killing people.

You're a soldier, Joseph. Sometimes not everyone on your team is bringing 110% all the time. Sick, tired, depressed, hurt, hungry, fussy, people struggle to pull their weight sometimes and then someone else has to. Sometimes you've got someone who pulls 110% every-single-day. Between the two, when the opportunity comes up to make a difference, who is likely to get the most out of the opportunity?

That's where the odds play out. It's about opportunities. You don't have the same opportunities in every match but based on skill you'll take advantage of more if you're more skilled, thus turning situations where a lost opportunity would have cost your side the match you turn into a win. Maybe you didn't take the opportunity but instead created it - got enemies to chase you into the open, drew attention off a damaged ally who got back into cover and did a ton more damage before going down. There are countless opportunities created, taken and lost in any given match.

Win/loss is the end result of all 12 people on each team creating, taking or missing those opportunities. The better people are at doing that the more they create, the more they take, the less they miss. From match to match it may not be enough. Maybe the other team was better or luckier this time. But when the ball comes your way if you consistently play it better you'll drive more wins. Simple as that.

[Joseph Mallan]

MischiefSC, on 18 November 2013 - 11:43 AM, said:

Remember, it's about how your performance plays out in the long run. Same reason you keep your weapon well maintained - even if you cleaned it the other day you clean it again because statistically in the long run those odds play out to keep you alive.

So in a match where you do 600 damage and 4 kills but your team sucked and you still lost it feels like your talent was wasted. That time you do 600 damage and 4 kills but your team clicked, you were the winning factor. A lot goes into winning and losing a match but if you're always bringing extra then you'll skew those odds in your favor. That's why it takes a lot of matches though in 12v12 to seat Elo effectively.

Your stats help prove the point though - your Jag is clearly a mech you do a lot of successful killing in. Consistently in fact, so it skews your wins up. I'm curious to know how you play your D-DC and its loadout vs your Fatlas? I know that my D-DC has a high win/loss and a moderate KDR but that's because of how I play - I tank for my team, soak damage and don't hesitate to keep drawing fire until it kills me if it keeps the enemy exposed to my team.

Kills and damage don't always contribute directly to wins. Especially in an Atlas or a Raven or ECM light. There's a lot you can do in how you play that contributes to wins without directly killing people.

You're a soldier, Joseph. Sometimes not everyone on your team is bringing 110% all the time. Sick, tired, depressed, hurt, hungry, fussy, people struggle to pull their weight sometimes and then someone else has to. Sometimes you've got someone who pulls 110% every-single-day. Between the two, when the opportunity comes up to make a difference, who is likely to get the most out of the opportunity?

That's where the odds play out. It's about opportunities. You don't have the same opportunities in every match but based on skill you'll take advantage of more if you're more skilled, thus turning situations where a lost opportunity would have cost your side the match you turn into a win. Maybe you didn't take the opportunity but instead created it - got enemies to chase you into the open, drew attention off a damaged ally who got back into cover and did a ton more damage before going down. There are countless opportunities created, taken and lost in any given match.

Win/loss is the end result of all 12 people on each team creating, taking or missing those opportunities. The better people are at doing that the more they create, the more they take, the less they miss. From match to match it may not be enough. Maybe the other team was better or luckier this time. But when the ball comes your way if you consistently play it better you'll drive more wins. Simple as that.

For the record. When I do 660+ damage and lose, I still feel great cause I did do good. I played for about 6-8 hours this weekend, I have no idea how I did Win loss wise. I had a good time though i did lose another 0.01 on my K/D record.

[Macbrea]

The more games you play the closer it will get to the 50/50 mark. The matchmaking system tries and finds you games that are competitive for you to play in. In theory, they can use that same matchmaking system to shift the map. Your Elo rating is determined by your win/loss record per category. I will use mine as an example:

Mech'	Matches Played	Wins	Losses	Ratio
Lights	877	404	465	0.87
Mediums	40	16	24	0.67
Heavies	779	367	409	0.90
Assaults	50	23	27	0.85

As you can see, the more matches I play the more it tries to get me toward a 1.0 ratio. If I double the number of matches I have played for Lights or Heavies, I bet those numbers will be almost exactly 1.0. For Medium and Assaults, I will need to play a huge number more games to have very little deviation.

Now, having said that, I wish it would break that up alittle bit more to Chasis vs Entire category. It is very noticable in the heavies if I am in a Dragon vs a Catapult. And my play in a Jester seems like all my other mechs are broken.

Mech'	Matches Played	Wins	Losses	Ratio
JESTER	68	42	25	1.68
FLAME	68	33	35	0.94
DRAGON DRG-1N	70	32	38	0.84
CATAPULT CPLT-C1	60	31	29	1.07
CATAPULT CPLT-A1	96	44	52	0.85
CATAPULT CPLT-K2	219	99	120	0.83
DRAGON DRG-5N	55	27	27	1.00
CATAPULT CPLT-C4	104	39	64	0.61
THUNDERBOLT TDR-5S	14	10	4	2.50
THUNDERBOLT TDR-5SS	14	5	9	0.56
THUNDERBOLT TDR-9SE	10	5	5	1.00
QUICKDRAW QKD-4G	1	0	1	0.00

Edited by Macbrea, 18 November 2013 - 11:53 AM.

[Joseph Mallan]

FerretGR, on 18 November 2013 - 11:37 AM, said:

I find stats like that very interesting, to be honest! For me it goes to show that there isn't a precise and exact relationship between KD and WL, and it makes for an interesting discussion. What's the difference between those mechs and their loadouts that might lead to numbers like that? If I was to guess, I'd say you ran your (F)Atlas as an LRM boat, which would lead to significantly less kills while still providing damage and support, whereas your Jager-A is a AC40 ride, which means lots of pinpoint damage and killshots. Am I getting warm?

The LRMAtlas was before I got Murphy's Tower. (A gaming tower. A gift last thanksgiving) I used mostly as a 2 Large 2 Medium AC20 2 SRM6, Now I roll with One Large, 2 Medium, 2 AC5, 2 SRM6. My D-DC was a PPC,Gauss, 3 SRM6, Now it's Almost the same as my (F)Atlas, but no mediums and an extra 6 pack. The Jager is a 40 Jager, But also is used only as a PUG ride. I don't use it when I'm with the Law.

[MischiefSC]

Joseph Mallan, on 18 November 2013 - 11:49 AM, said:

For the record. When I do 660+ damage and lose, I still feel great cause I did do good. I played for about 6-8 hours this weekend, I have no idea how I did Win loss wise. I had a good time though i did lose another 0.01 on my K/D record.

I wish you could teach classes. There's a distinct drive for some people that keeps them from enjoying just doing better - they have to be doing better than *someone else*.

Joseph Mallan, on 18 November 2013 - 11:55 AM, said:

The LRMAtlas was before I got Murphy's Tower. (A gaming tower. A gift last thanksgiving) I used mostly as a 2 Large 2 Medium AC20 2 SRM6, Now I roll with One Large, 2 Medium, 2 AC5, 2 SRM6. My D-DC was a PPC,Gauss, 3 SRM6, Now it's Almost the same as my (F)Atlas, but no mediums and an extra 6 pack. The Jager is a 40 Jager, But also is used only as a PUG ride. I don't use it when I'm with the Law.

Why not run the D-DC as a face-humper? I still do that with mine sometimes and it works terribly well. With the PPC heat nerf snipers are just not the grief they used to be and the AC20/SRM6/LL cycle time works well against dakka mechs - you swing, spread damage, BOOM!!!!! swing, spread damage, BOOM!!!!! Step on their corpse as you advance on their now crying friends.

I love me some D-DC.

[St4LkeRxF]

FerretGR, on 18 November 2013 - 11:32 AM, said:

If you were seeing that sort of thing in every single match, St4LkeR, I'd say you have a point. But the bottom line is that the matches with shenanigans that affect your W/L should average themselves out.

3-4 days ago had match where whole team start killing each other over DC centurion, and today got something similar but i was able to escape before my rear was red.

No luck here but i don't care if i loose or win as long it is fun.

[Sug]

Odin, on 18 November 2013 - 09:06 AM, said:

Tell me about it here!
Good Luck!

Using the trial Blackjack every game so mm uses the same elo.

So far it's 6 wins 7 losses but people are starting to yell at me for afking/botting. I'm running into almost the exact same players every game.

[Nightfire]

I'm actually rather busy to put in a detailed response to what I've read in this thread and advance the argument of primary influenced variables vs secondary influenced variables and the incidence of correlation evidenced over sample size.

I will instead posit this main point for further discussion:
A) It has been stated (paraphrased, don't have the time to search for an exact quote) that the less directly the variable is influenced then the more samples it takes to make a desired outcome observable. In MWO Elo implementation, I think it was stated it takes upwards of 10,000 matches to somewhat accurately determine a players Elo rank. I'm not trying to put words in people's mouths here, working from an imprecise memory is all.

In any random system, given an infinite sample, you will find a sample to match any given criteria. (That part is fact gentlemen, disputing that will not be responded to). The larger the sample size, the more likely you are to match a given criteria.

The method here is to be matched with enough teams, close enough to the desired outcome to place you at your "correct" Elo rank.

My postulation to Elo defenders: At what sample size do you admit that you have as much practical probability of getting that series of matches by pure chance as opposed to by intelligent design?

Edited by Nightfire, 19 November 2013 - 01:15 AM.

[Odin]

Sug, on 18 November 2013 - 05:56 PM, said:

Using the trial Blackjack every game so mm uses the same elo.

So far it's 6 wins 7 losses but people are starting to yell at me for afking/botting. I'm running into almost the exact same players every game.

They have no doubt the same ELO, almost 50/50 then for you.
Not bad I say, for not have done anything useful. IMO one guy missing isn't much of a hindrance to victory...whatsoever.

Thanks for your efforts