66 replies to this topic

[Duke Nedo]

Tarogato, on 06 March 2015 - 01:06 PM, said:

Elo is divided by weight class, so perhaps you should track your W/L for each weight class individually.

I'll start doing the same and tracking my own data, because I love stuff like this. I'll also track my matchscore rating (1-12). See you in about 500 matches, this will take a while...

Aye, I also record which class every match is in so I have intentions to make that plot. Need more games for them to show any trends still though I guess.

[Duke Nedo]

ThrashInc, on 06 March 2015 - 01:41 PM, said:

I think you would have to also math out matches that occur during events or mech release weeks.

For example, regardless of your "best" weight class, the MM was throwing matches with 8 lights vs 6 assaults during Resistance release, so you might have been the only assault, or one of 8 panthers.

My assault W/L is considerably higher than any other bracket, with the overall W/L being 105 games up.

I agree, I have only loose notes in the margin when some tournaments start and when they end. Qualitatively I didn't see a clear difference now, but for some events in the past, especially when it was solo-queue only and lots of group-Elo-players invaded the solo-queue, it felt like the MM went bananas.... Can't see any clear effect after I started tracking results though (all has been solo+group queues).

[Duke Nedo]

I plotted the graphs by Elo-class. Nothing too significant there, all classes end up on the positive side, but it seems like it may be Lights that was mainly responsible for the steep ascent in the original plot.

Edited by Duke Nedo, 08 March 2015 - 01:00 AM.

[Michael Abt]

While the raw numbers on wins is going up, how is your w/l ratio developing in percent compared to total matches played?

[Duke Nedo]

Michael Abt, on 08 March 2015 - 08:09 AM, said:

While the raw numbers on wins is going up, how is your w/l ratio developing in percent compared to total matches played?

Right now it's almost a perfect match compared to the stats on the profile page, 1.125 on stats and 1.14 in my sample, so in that respect this sample is representative.

I think I have by now proven to myself that there is no obvious systematic cyclic-ism in win/loss-streaks, even though it at times really felt that way.

[Murzao]

Meh pug queue is easymode really most people should be winning 1.3 or higher, comp players 1.5.

I myself since stat reset is 772-683 for a 1.13 ratio, but that is 100% in Centurions and that combines group play where I have the most awesomest time in 3-5 man premades vs 6-12 mans in said pure medium chasis in the top elo bracket. That by itself has easily dropped me under a 1.5 ratio solo pug.

And if you consistently play with a large team your wl ratio should be well over 2-3, otherwise you kinda suck hate to say it lol. Like when I'm in a 4 man premade and I don't see a 4 man premade on the other side, 95% guaranteed win time.

Lots of ways to pad your stats far above your skill level if that's important to you.

[Duke Nedo]

Murzao, on 08 March 2015 - 09:24 AM, said:

Meh pug queue is easymode really most people should be winning 1.3 or higher, comp players 1.5.

I myself since stat reset is 772-683 for a 1.13 ratio, but that is 100% in Centurions and that combines group play where I have the most awesomest time in 3-5 man premades vs 6-12 mans in said pure medium chasis in the top elo bracket. That by itself has easily dropped me under a 1.5 ratio solo pug.

And if you consistently play with a large team your wl ratio should be well over 2-3, otherwise you kinda suck hate to say it lol. Like when I'm in a 4 man premade and I don't see a 4 man premade on the other side, 95% guaranteed win time.

Lots of ways to pad your stats far above your skill level if that's important to you.

This is all from the solo queue, 100%, so most people can by definition not have 1.3 w/l ratio. Most people will have 1.0 w/l ratio. It only takes one equally good player on the other team to cancel you out, so your impact will never be that high unless you are a very good "solo-boss" or so good that there is no player at all queuing that can match your überness... Actually, the latter will usually make you lose more because instead of matching you directly you will be matched by several players that are a bit better than your team average players, and then you're really smoked.

[Murzao]

I have been known to carry a bunch of (C)s on occasion, in one match me in my souped up Cent-A(C) carried 3 other noob Cs to victory vs a side with mostly tagged enemies with some decently good (above average but not comp team level even if they think they are) game...in which I was able to solo kill 4 people consecutively (at the start of fighting not at the end which is easy killing the last 4 people in a match) in a bunch of tall buildings and had most damage/kills by a wide margin. Just when you get up there you gotta bring your carrying boots especially if you solely use medium mechs. Centurions love a mass of tall buildings!!

Not even that long ago a 5 man premade of us crushed a 12 man NS premade12-2 on caustic (and they think they're pretty good), they tried a full over the hill charge vs a pug team and it ended real bad:) Most of the time though a big premade picks off a few pugs and then easily stomps the rest with snowball effect so if you wanna up your wl ratio just play in a big group....it still won't improve your skill level though.

But if you are used to playing vs top end premades in large groups, you will find the pug queue is easymode even if the pug side is comprised of the same players because of vastly less target calling and less blob effect. Any self respecting comp player that can carry his team should be well above 1.5.........1.3 is easy. Did I mention I ain't a comp player (but have been around since the beginning and at the start of 8 man group play stats before elo I started off 50-2 it's that easy with a big group, doesn't take a lot of skill with big numbers of buddies....was like 250-20 and ended prestart reset with a 2500-1100 wl ratio once elo kicked in, it wasn't hard it didn't take a comp player to do that).

But yea solo pug is where I go to up my wl ratio.....even when I'm trolling/leveling up mechs. If you have a below 1 average you just don't contribute much (and I've seen newbs contribute more to a win than some 'good' players guarding their precious kd ratio often)

Edited by Murzao, 08 March 2015 - 11:52 AM.

[Duke Nedo]

Like I said, by definition, most people cannot have 1.3 or 1.5. In every game 12 players win and 12 players lose. You would need to be exceptionally much better than everyone else, and if you are, that means that you have a deflated Elo rating. If you carry true trial mechs like you imply, then your Elo could not be too high and yes, then you could be much better than both opponents and teammates and then you can carry quite efficiently.

Anyways, that is all off topic. I did not post these stats to brag, it's my little sample of what the MM puts me up to, nothing more nothing less.

[Murzao]

I said souped up champion mech aka a standard modded mech with cbill bonus, not a trial mech......I find carrying 1-2 trials no problem there is usually a few substandard people on the other team to balance out for free kills stat padding. And I only see those in offhours when the fewest people play. In primetime I get both teams loaded with at the very least decent clans to the top.

But totally disagree with everyone being a 1 on average. Just beating up on real bad players that quit the game/not on mics should push you up to 1.1 at the very least even if you are pretty bad. Unless you die early every game and are at the mercy of a 50/50 coin flip......

[Mirumoto Izanami]

Murzao, on 08 March 2015 - 06:02 PM, said:

But totally disagree with everyone being a 1 on average. Just beating up on real bad players that quit the game/not on mics should push you up to 1.1 at the very least even if you are pretty bad. Unless you die early every game and are at the mercy of a 50/50 coin flip......

There is nothing to disagree with. The mathematical average for won/loss is 1.

Edited by Mirumoto Izanami, 08 March 2015 - 06:19 PM.

[Murzao]

YLW Stats (so when I'm actually trying and not trolling):

Matches played 497
Wins: 294
Losses: 203
Ratio 1.45
Kills ratio 1.1 (cuz well I use it on the frontline first man in)

When I pug I get over a 2 wl ratio easy possibly 3.....as those stats include a lot of small premade vs much larger premade games at high elo reducing me back down to a 1.45 average. Pugging is where I go to up my wl ratio average.....it's a myth that there's always players on the other side better than you every game just like it's a myth the big bad 12 man premade touched you. (cuz there's always bad people with inflated stats due to being in said large premades often which the average person will still beat easily in pug land)

How the average player doesn't have a 1.1 average is beyond me (due to power creep).

Edited by Murzao, 09 March 2015 - 06:06 AM.

[Mirumoto Izanami]

Murzao, on 09 March 2015 - 06:05 AM, said:

How the average player doesn't have a 1.1 average is beyond me (due to power creep).

Because its mathematically impossible for the average to be 1.1 win loss. Is math beyond you?

[Iskareot]

Meh until we tier up the entire ELO/MM system I say the glass is half empty.

I am still just baffled at the punishment of having to carry people and that idea there of. Make tiers, at least put us with generally the same skill player... why is this so bad?

[Mawai]

Cool data. I wish PGI would publish numbers since they have all the data instead of relying on the great work by individual players.

Also ...

If Elo and MM are working then you should see two effects ...

1) Over time your W/L line should level out ... which it seems to do in your figure. On average, if you are matched up against a team of equally skilled opponents then your chances of winning the match should be 50% ... so on average your W/L line should level out and the W/L ratio will trend toward 1.0. This assumes that the player's skill remains constant. If the player is learning to play better, use a particular mech or mech class better than they did previously ... then the wins and losses will trend upward until your Elo value reflects your new skill level and you are grouped with the appropriate level of people in most games.

2) I would expect your number of mechs lost in wins vs losses graph to converge ... over the very long term ... numbers lost in both wins and losses should look the same. In your figure, your side loses fewer mechs on wins than it does on losses. I would think that this means that on average your losses have been closer games than your wins ... which means your team was usually better .. or you were improving.

Also, the two peaks are interesting. I think that may be due to the initial conditions of the match in which the 3 lances of 4 mechs each start separately. If lances stay together then they are more likely to encounter opposition as a group ... and thus either be eliminated or damaged together unless one group is formed before contact is made (which is fairly infrequent).

In your wins, you seem to catch a couple of opposing lances more frequently than it happens in your losses .. which may indicate that the teams you are on tend to group up more.

Edited by Mawai, 09 March 2015 - 06:33 AM.

[Duke Nedo]

Murzao, on 09 March 2015 - 06:05 AM, said:

YLW Stats (so when I'm actually trying and not trolling):

Matches played 497
Wins: 294
Losses: 203
Ratio 1.45
Kills ratio 1.1 (cuz well I use it on the frontline first man in)

When I pug I get over a 2 wl ratio easy possibly 3.....as those stats include a lot of small premade vs much larger premade games at high elo reducing me back down to a 1.45 average. Pugging is where I go to up my wl ratio average.....it's a myth that there's always players on the other side better than you every game just like it's a myth the big bad 12 man premade touched you. (cuz there's always bad people with inflated stats due to being in said large premades often which the average person will still beat easily in pug land)

How the average player doesn't have a 1.1 average is beyond me (due to power creep).

From what you wrote I think we can conclude that you play quite a bit in the Group queue with a small gang, where you lose a lot versus larger groups and build a low Elo. You then jump into the Solo queue with your low Elo and get matched against new players and win a lot. If you see trial mechs, this is likely.

There are good players in the Solo queue, promise!

[Duke Nedo]

Mawai, on 09 March 2015 - 06:32 AM, said:

1) Over time your W/L line should level out ... which it seems to do in your figure. On average, if you are matched up against a team of equally skilled opponents then your chances of winning the match should be 50% ... so on average your W/L line should level out and the W/L ratio will trend toward 1.0. This assumes that the player's skill remains constant. If the player is learning to play better, use a particular mech or mech class better than they did previously ... then the wins and losses will trend upward until your Elo value reflects your new skill level and you are grouped with the appropriate level of people in most games.

Definately, but I don't think I see the leveling out this quickly. It started to climb a bit again so the plateu is over, which I think is good because it was almost "too" flat.

Mawai, on 09 March 2015 - 06:32 AM, said:

2) I would expect your number of mechs lost in wins vs losses graph to converge ... over the very long term ... numbers lost in both wins and losses should look the same. In your figure, your side loses fewer mechs on wins than it does on losses. I would think that this means that on average your losses have been closer games than your wins ... which means your team was usually better .. or you were improving.

Also, the two peaks are interesting. I think that may be due to the initial conditions of the match in which the 3 lances of 4 mechs each start separately. If lances stay together then they are more likely to encounter opposition as a group ... and thus either be eliminated or damaged together unless one group is formed before contact is made (which is fairly infrequent).

In your wins, you seem to catch a couple of opposing lances more frequently than it happens in your losses .. which may indicate that the teams you are on tend to group up more.

Hmm, never thought about reading out that much from these plots. I usually try to tell people to group up in games where people spread out too much, would be nice if that actually had an effect. Also, I think as long as the w/l is positive, the wins should on average be slightly clearer than the losses, for the same reasons there are more wins than losses.

[Roadkill]

Mirumoto Izanami, on 09 March 2015 - 06:12 AM, said:

Because its mathematically impossible for the average to be 1.1 win loss. Is math beyond you?

Semantics. Is someone who plays once, loses, and never plays again really a player or are they just padding for everyone else's W/L ratio?

Games aren't zero sum for the people who stick around and keep playing. Some people pick the game up, play a few games (sucking horribly the entire time), then drop out and never play again. Those handful of losses by someone who never plays again effectively add extra wins to the player pool for those who do stick around.

So while mathematically what you're saying is true, most people don't count those 1-and-done people as actual players. So yes, and average player probably should have a W/L of 1.1 or so, because their success is being padded by a continuous flow of 1-and-done losers.

Edited by Roadkill, 09 March 2015 - 09:27 AM.

[Murzao]

Duke Nedo, on 09 March 2015 - 06:47 AM, said:

From what you wrote I think we can conclude that you play quite a bit in the Group queue with a small gang, where you lose a lot versus larger groups and build a low Elo. You then jump into the Solo queue with your low Elo and get matched against new players and win a lot. If you see trial mechs, this is likely.

There are good players in the Solo queue, promise!

When I join group queue I end up getting paired with/against the who's who of comp teams...in which a 1wl ratio or anything close is actually impressive most people would get smashed in 30 seconds repeatedly (especially with a small group vs a larger premade group you have to work doubletime to carry). Then I jump in the top end pub queue and do 2 or 3wl to 'average' it at 1.45.

On the other hand if the elo system is based purely off wins (and it is), and both teams are statistically equal elo......if you won one and lost one for your 1wl ratio your elo would not go up at all therefore you are statistically in the 50th percentile of player and are by definition the average player. You actually do need to be above 1 to climb any elo, and at 1.1 that isn't even a fast climb either. You pretty much need a 1.3wl ratio to climb the percentile ladder from 'stat average' into the 'decent', 'good', 'pro', and 'comp' percentile. It is your wl ratio that determines your elo because most games have statistically similar elos (with some people having inflated elos due to large amounts of games in large premades which skews wins and therefore their elos way up even if they are terrible....you know the bottom 4 people in yer group getting free wins without getting large scores in any game, that's where you see 12-0 or 12-2 stomps because your 'high' elo players really got carried by teammates for their games)

Edited by Murzao, 09 March 2015 - 09:47 AM.

[Mirumoto Izanami]

Roadkill, on 09 March 2015 - 09:25 AM, said:

So while mathematically what you're saying is true, most people don't count those 1-and-done people as actual players. So yes, and average player probably should have a W/L of 1.1 or so, because their success is being padded by a continuous flow of 1-and-done losers.

Assumptions.

There are likely to be as many 1 and done losers as there are to be one and done winners.

Edited by Mirumoto Izanami, 09 March 2015 - 09:42 AM.