129 replies to this topic

[Duke Nedo]

From wikipedia:

Correlation does not imply causation

From Wikipedia, the free encyclopedia

See also: Illusory correlation

Correlation does not imply causation is a phrase used in statistics to emphasize that a correlation between two variables does not necessarily imply that one causes the other.^[1][2] Many statistical tests calculate correlation between variables. A few go further, using correlation as a basis for testing a hypothesis of a true causal relationship; examples are the Granger causality test and convergent cross mapping.^{[clarification needed (hypothesis testing not well explained here)]}
The counter-assumption, that correlation proves causation, is considered a questionable cause logical fallacy in that two events occurring together are taken to have a cause-and-effect relationship. This fallacy is also known as *** hoc ergo propter hoc, Latin for "with this, therefore because of this", and "false cause". A similar fallacy, that an event that follows another was necessarily a consequence of the first event, is sometimes described as post hoc ergo propter hoc (Latin for "after this, therefore because of this").

Just saying.

Edited by Duke Nedo, 21 October 2015 - 06:48 AM.

[0bsidion]

Well, I can certainly understand how frustrating having such long losing streaks must be. Personally, after PSR my KDR and W/L ratio has been slowly going up. Now I rarely lose much more than 2 matches in a row. Way more of my matches are close instead of the complete roflstomps I used to experience with ELO.

[Mudhutwarrior]

Duke Nedo, on 21 October 2015 - 06:46 AM, said:

From wikipedia:

Correlation does not imply causation

From Wikipedia, the free encyclopedia

See also: Illusory correlation

Correlation does not imply causation is a phrase used in statistics to emphasize that a correlation between two variables does not necessarily imply that one causes the other.^[1][2] Many statistical tests calculate correlation between variables. A few go further, using correlation as a basis for testing a hypothesis of a true causal relationship; examples are the Granger causality test and convergent cross mapping.^{[clarification needed (hypothesis testing not well explained here)]}
The counter-assumption, that correlation proves causation, is considered a questionable cause logical fallacy in that two events occurring together are taken to have a cause-and-effect relationship. This fallacy is also known as *** hoc ergo propter hoc, Latin for "with this, therefore because of this", and "false cause". A similar fallacy, that an event that follows another was necessarily a consequence of the first event, is sometimes described as post hoc ergo propter hoc (Latin for "after this, therefore because of this").

Just saying.

How about Empathy leads to understanding? Brick walls of denial do not.

[MeiSooHaityu]

Mudhutwarrior, on 21 October 2015 - 06:58 AM, said:

How about Empathy leads to understanding? Brick walls of denial do not.

Sticking with the Trailer Park Boys theme

[Water Bear]

Duke Nedo, on 21 October 2015 - 06:46 AM, said:

From wikipedia:

Correlation does not imply causation

When do you decide that one thing causes another thing? Our hypothesis is that event A (the OP being in a game) causes event B (loss) to occur more frequently.

In over 80% of games where event A occurred, with every other variable changing, event B is the result. I'd say that's good evidence for cause.

You are right, we are linking cause and effect here when we can't be sure A is causing B. But a more scientific verification of the hypothesis I can't imagine.

It's also possible that he's just on a bad luck streak (see my previous post). But the odds of such a long streak, although positive, are small. I think it's fair to conclude that A is causing B, where I'm wrong with small probability.

Edited by Water Bear, 21 October 2015 - 07:06 AM.

[lol lol lol lol lol lol lol lol]

Duke Nedo, on 21 October 2015 - 06:46 AM, said:

From wikipedia:

Correlation does not imply causation

From Wikipedia, the free encyclopedia

See also: Illusory correlation

Correlation does not imply causation is a phrase used in statistics to emphasize that a correlation between two variables does not necessarily imply that one causes the other.^[1][2] Many statistical tests calculate correlation between variables. A few go further, using correlation as a basis for testing a hypothesis of a true causal relationship; examples are the Granger causality test and convergent cross mapping.^{[clarification needed (hypothesis testing not well explained here)]}
The counter-assumption, that correlation proves causation, is considered a questionable cause logical fallacy in that two events occurring together are taken to have a cause-and-effect relationship. This fallacy is also known as *** hoc ergo propter hoc, Latin for "with this, therefore because of this", and "false cause". A similar fallacy, that an event that follows another was necessarily a consequence of the first event, is sometimes described as post hoc ergo propter hoc (Latin for "after this, therefore because of this").

Just saying.

You need to forward this to Paul regarding every balance, ever.

Just sayin.

Edited by saKhan Ds00 Kerensky, 21 October 2015 - 07:25 AM.

[Ex Atlas Overlord]

Lugh, on 21 October 2015 - 06:31 AM, said:

True you can't control 11 others,

Actually he can't control the 23 others.

Paigan, on 21 October 2015 - 06:32 AM, said:

Of 100 games of a certain player with more less randomly (or at least unbiased) assigned team mates, the only common element is himself.

Yep out of 100 games.... he is one constant out of 23 other variables... and that's only counting the players.

So over that one hundred games... that's still one constant factor... except now it's 2300 distinct variables.

Paigan, on 21 October 2015 - 06:32 AM, said:

Adding his performance is the only thing that deviates his personal 100 games from the statistical average.

This right here is just moronic.

So he (being one of variables) is capable of upsetting the statistical average...but the other 23 variables can't.

You really need to study actual mathematics.

Paigan, on 21 October 2015 - 06:32 AM, said:

while all others (which are evenly distributed in the long run, no matter the selection algorithm) are supposedly the ones that are significant for his performance.

An algorithm can't "distribute" player behavior.

The fact they other players exist implies nothing at all about their chances of winning or losing.

P.s. - an "algorithm" has literally nothing at all to do with anything we're talking about. Aka - insert X players. Done. The things AFTER the algorithm are what matters.

Paigan, on 21 October 2015 - 06:32 AM, said:

If you want to say something smart, make sure it's really smart and not based on rhetorical ambiguity to make a kind of senseless pseudo-point.

I'm sorry you've only taken Statisics 101?

P.s - I find it amusing you think differential equations and indefinite mathematics to be "rhetorical ambiguity".

Aside: This important thing that everyone overlooks.

A statistical average applied to a system only applies to THAT system.

So any given team may have a 50/50 chance of winning....If every player on both teams stayed the same.... and THEN he didn't even out to a 50/50 average.... THEN he would be the cause....

BUT the second you switch out the variables... you've created a NEW system.

Any given player may be considered a system....but when you add 11 variables... or 23 variables... you've created a new system.

You can't (well you can if you want to be wrong) simply say (x+y+z) and (w+r+v) both have 3 variables so screw it let's pretend they describe the same thing.

They don't.

Edited by The Atlas Overlord, 21 October 2015 - 07:45 AM.

[Duke Nedo]

Water Bear, on 21 October 2015 - 07:04 AM, said:

When do you decide that one thing causes another thing? Our hypothesis is that event A (the OP being in a game) causes event B (loss) to occur more frequently.

In over 80% of games where event A occurred, with every other variable changing, event B is the result. I'd say that's good evidence for cause.

You are right, we are linking cause and effect here when we can't be sure A is causing B. But a more scientific verification of the hypothesis I can't imagine.

It's also possible that he's just on a bad luck streak (see my previous post). But the odds of such a long streak, although positive, are small. I think it's fair to conclude that A is causing B, where I'm wrong with small probability.

He could be the cause, but more likely it's a small sample plus random, spiced up by psychology. If there is a cause, It's more likely that a loss% like that is related to how the MM builds the teams in combination with the OPs tier, mech class, mech choice, habits, syncdrops or whatever. Perhaps the OP perceives 83% losses, but actually lost only 65%. Who knows... I just have to say something when the You Are The Only Constant Factor gets going....

My point is that if the OP truly causes 83% losses, he is not merely in the wrong Tier, he is very actively sabotaging every single game, being truly toxic for winning. Not that I have the answer, but I'd assume that if I did 1000 drops and stayed AFK the whole game, my team would win more than 17% of the games. People used to farm that way in 8 vs 8, and now we have 12 vs 12 so even less impact. Too bad one can't do that experiment. PGI could extract it from their data though, would be interesting read!

[Lugh]

Duke Nedo, on 21 October 2015 - 08:27 AM, said:

He could be the cause, but more likely it's a small sample plus random, spiced up by psychology. If there is a cause, It's more likely that a loss% like that is related to how the MM builds the teams in combination with the OPs tier, mech class, mech choice, habits, syncdrops or whatever. Perhaps the OP perceives 83% losses, but actually lost only 65%. Who knows... I just have to say something when the You Are The Only Constant Factor gets going....

My point is that if the OP truly causes 83% losses, he is not merely in the wrong Tier, he is very actively sabotaging every single game, being truly toxic for winning. Not that I have the answer, but I'd assume that if I did 1000 drops and stayed AFK the whole game, my team would win more than 17% of the games. People used to farm that way in 8 vs 8, and now we have 12 vs 12 so even less impact. Too bad one can't do that experiment. PGI could extract it from their data though, would be interesting read!

One of the many variables we are missing is this, Does he normally play in the group queue as part of a good comp team? Or did he in the past and they are now gone?

That would totally inflate his PSR to better than he deserves values and would be a legit cause of his solo queue woes.

Conclusion: The op needs to up his game. We have no data on him other than that he is on a large losing streak. We have no other data. If the OP wishes to claim it's not his fault he needs to prove it. Show me the statistics.

[Phantom Limb]

Paigan, on 21 October 2015 - 12:18 AM, said:

I have one, but you won't like it.

The point is: however good or bad MM is, it applies to BOTH teams.
So if you drop solo and even if MM were just completely random, you should end up with a 50-50 win chance in the long run (not inside of 10 games, but inside of 100 quite well).

The ONLY constant in your 100 solo drops ... is you.

I' tier 2 as well and I, too, do only solo drops (no time and nerve for a unit).
My win rate is something around 1,25 (overall 1.19 since I started including noob times, 1.29 with the mech I play almost exclusively in the last weeks).

My observation is:

Many people simply play tactically bad.
Simple example: I always shoot UAVs the second I see them. So the enemy has almost never any UAVs. We have some sometimes. Hence huge tactical advantage for us. Hence slightly higher win chance.
Or I cover areas with long range, press enemies into cover, giving our (pardon, mostly stupid) brawlers a chance to close in.
I don't focus on getting kills or spraying out the most damage. I focus on contributing to the win.
(and then they flame me for being "a coward", but that's another story ... funny brawlers)
etc
etc.

Result: 100 matches yield 77 wins.

I'm hardly top scorer or have the most kills. But I know what I contributed.


I GUESS for you it's the other way around: you might fight a lot, but contribute little. Hence a lot of losses.
Sorry.

Have to agree with this assessment. If mm is broken, it's broken for both the blue team and the red team. Seems very unlikely that the mm is screwing just the OP. Given that a single player can impact the match significantly (in either direction) it seems likely that the high loss rate of the OP is due to the performance of the OP, not the mm.

[Phantom Limb]

Onyxian, on 20 October 2015 - 07:03 PM, said:

I've been getting slaughtered with regularity since I restarted playing last week... Wish I had kept track.

An upward bias in the PSR rating system, wtf were they thinking? Moronic

It seems that the upwards bias of the PSR was primarily meant as a way to separate the new players from the experienced players. Problem is, if someone is content to just use mechs because they are fun or because they look badass ( or are just not very good at the game) they will eventually end up in a tier where their skills or the mech they choose results in a quick death.

My theory is that once you get out of the kiddy pool, your individual contribution to a match is much more critical, and by not using the best mechs ( or by not being very skilled) you could experience a lot of quick deaths, which isn't much fun. This Is true for players on both sides of the match, so it shouldn't affect the win/loss rations of anyone over time, but would result in some pretty crappy matches for some people.

Because of the upward bias, you would have to score poorly in a lot of matches before your PSR goes down enough to be put back in the kiddy pool, where you don't die as quickly if you make a tactical error or pilot a crappy mech. With a low player population, the debs may have a difficult time finding a way to match players of equal skill. This may be the best we get for now.

[voltdragon]

Sorry for delay in response long day at the job. It is very plausable and most likely there I may be in a tier above my head. With those statistics and since psr has dropped I have gained tier bar. at least 10 % I am nearly tier 1 with an inch to go which is disturbing considering the recent stomps. This still leaves the whole question why the remainder of my team is getting slaughtered along side me in these stomps. I can face facts if im dieing to my mistakes or my misplaced tier level while my team mates play evening with the enemy team but almost 70% of these matches my team mates appear to be getting stomped also. In a few matches other people mentioned playing 20 games that night not winning one of them and interestingly i was placed on their team in solo queue a few times after that (didnt play much longer but would have been intersting to continue collecting data). I guess I have a bit of a sore hiney and thank everyone for their opinions and outlooks.

[Water Bear]

Duke Nedo, on 21 October 2015 - 08:27 AM, said:

He could be the cause, but more likely it's a small sample plus random, spiced up by psychology. If there is a cause, It's more likely that a loss% like that is related to how the MM builds the teams in combination with the OPs tier, mech class, mech choice, habits, syncdrops or whatever. Perhaps the OP perceives 83% losses, but actually lost only 65%. Who knows... I just have to say something when the You Are The Only Constant Factor gets going....

My point is that if the OP truly causes 83% losses, he is not merely in the wrong Tier, he is very actively sabotaging every single game, being truly toxic for winning. Not that I have the answer, but I'd assume that if I did 1000 drops and stayed AFK the whole game, my team would win more than 17% of the games. People used to farm that way in 8 vs 8, and now we have 12 vs 12 so even less impact. Too bad one can't do that experiment. PGI could extract it from their data though, would be interesting read!

It's possible that he's just having bad luck. Since there aren't too many threads about this kind of losing streak, it really might just be the small percentage of players that are predicted to experience such a thing.

All I was saying is that a causation-correlation fallacy wasn't necessarily being made since we have evidence that indicates there might be a cause tied up in here (again, unless he's just having bad luck).

[WarZ]

voltdragon, on 20 October 2015 - 06:34 PM, said:

over the past week I have been keeping track of my games after having some really bad games.

100 games played, 83 loses
67 what i generally consider a stomp with us not able to kill even half their team
22 complete stomps where we get 1 or less.

Before this PSR change I did not notice such horrible losing streaks.

previously i had over 600 more wins then loses that's about even now
k/d dropped 0.1

While I can review changes in my game play I have no explanation why so many of my games are lost with nothing due to my part (mainly the stomps where i can not directly affect teammates gameplay)

This is PUG queue solo tier 2

anyone have any ideas suggestions constructive comments?

As many have noted - they needed to make the tiers actually skill based, vs just grind based. You can be a pretty low skill player and still run up into t2 and 1, given enough matches. Another side of that is people who dont play a lot of games are sitting in tiers well below thier actual skill level, because they didnt "grind" enough games to hit higher tiers.

So they got rid of a system that in theory based players on skill, with one that essentially matches players by # of games played and W / L.

Add to that the tier matching spread is too wide. To be honest the matching seems to be a bit of a disaster due to it. Is it worse than the old system with its wide open release valves ? I dont know. But based on the theory they described to us regarding how PSR works, its sounds wrong and scary if you value actual skill matching.

It seems they should create scoring for actual skill levels (which in theory they have some way of valuing skill, instead of just w/l), and you get to play only in your tier level. No level spreads. And for goodness sakes, stop with the release valve bull****. Sure you put in release valves to shorten wait times right ? However they COMPLETELY counter the reasons for having a MM. It negates all you are trying to achieve.

[Mycrus]

The Atlas Overlord, on 21 October 2015 - 06:20 AM, said:

You ever heard of a taking a limit?

If I take the limit of an equation with both variables (kinda like this game where there are 11 variable players, the variable of the map, the variable of if hit reg will work, etc) and a constant (The only one being him).... guess what you can usually do with the constant?

Ignore it, b/c those variables affect the equation to such a extant that the constant isn't even worth noting....

Nevermind, I'll make it simple for you.

Every single car in an accident has tires on it.

Are tires therefore the cause of every single accident? Of course not.

Are tires even a factor in every accident? Or course not.

Does the tires being on every car imply anything about cars getting into accidents? Of course not.


Stop being intentionally stupid.

Stay lurmy.

[Onyxian]

Duke Nedo, on 21 October 2015 - 06:46 AM, said:

From wikipedia:

Correlation does not imply causation

From Wikipedia, the free encyclopedia

See also: Illusory correlation

Correlation does not imply causation is a phrase used in statistics to emphasize that a correlation between two variables does not necessarily imply that one causes the other.^[1][2] Many statistical tests calculate correlation between variables. A few go further, using correlation as a basis for testing a hypothesis of a true causal relationship; examples are the Granger causality test and convergent cross mapping.^{[clarification needed (hypothesis testing not well explained here)]}
The counter-assumption, that correlation proves causation, is considered a questionable cause logical fallacy in that two events occurring together are taken to have a cause-and-effect relationship. This fallacy is also known as *** hoc ergo propter hoc, Latin for "with this, therefore because of this", and "false cause". A similar fallacy, that an event that follows another was necessarily a consequence of the first event, is sometimes described as post hoc ergo propter hoc (Latin for "after this, therefore because of this").

Just saying.

Tobacco companies used this very argument to try and argue smoking didn't cause cancer, heart disease, emphysema.

[PurpleNinja]

This is what I'm talking about.
Playing assault class right now.
About 5 minutes searching for a match.
Maybe I just read the status wrong all the time.

[Duke Nedo]

Onyxian, on 21 October 2015 - 06:37 PM, said:

Tobacco companies used this very argument to try and argue smoking didn't cause cancer, heart disease, emphysema.

...and science and proper statistical treatment or huge data sets proved them wrong. I am just saying one should have an open mind and not jump to conclusions just because one think that there is a correlation among a shetload of other contributing factors...

Here are some nice examples:

Water Bear, on 21 October 2015 - 05:06 PM, said:

It's possible that he's just having bad luck. Since there aren't too many threads about this kind of losing streak, it really might just be the small percentage of players that are predicted to experience such a thing.

All I was saying is that a causation-correlation fallacy wasn't necessarily being made since we have evidence that indicates there might be a cause tied up in here (again, unless he's just having bad luck).

My main point is that one player out of 12 players on the team slightly under-performing (I say slightly because if you got to Tier 2 there are limits how bad you can be and as Tier 2 you can't be expected to carry the whole team in most cases anyway) would not be enough to drag a team down so horrendously that you lose 83% of your matches. The way I see it, it's just a very unlikely explanation, there are many other more probable explanations because I think that 11 players + 1 AFK player will do better than 17% win rate.

[Shismar]

The OP fell victim to the randomness of match outcomes. Does not mean one cannot try to shift the probability towards more wins by playing min/max'ed mechs, but that is something many players will do on both teams. So 83 losses out of 100 matches is just bad luck.

The MM has the fatal flaw that it primarily awards match wins, not individual performance. That way most of the rating is random, depending on the 50/50 chance to be on the winning team.

[LordDante]