Do you think Life Support should be Vanilla?

[The_Rocketeer]

12 minutes ago, Alshain said:

I said the poll was inaccurate and that is why I went with the 17% instead of the 10%.

Sorry, I thought your final figure was 10%. On re-reading I see that that was a separate point. The 10% figure is based on a significant skew, but actually your statistical methodology is pretty sound.

That doesn't necessarily mean that @Pthigrivi is wrong though. There isn't enough data to have any confidence in your 17%, however sound the method. How individuals respond to a proposal and how they respond to a delivered product often do not correlate.

[Pthigrivi]

 

9 minutes ago, The_Rocketeer said:

Sorry, I thought your final figure was 10%. On re-reading I see that that was a separate point. The 10% figure is based on a significant skew, but actually your statistical methodology is pretty sound.

Ignoring all data and just arbitrarily giving even weight to all possibilities is not actually a methodology at all. Its usually better to listen to what people have to say on a subject than to assume you know what they're thinking.

Edited September 7, 2016 by Pthigrivi

[Alshain]

10 minutes ago, The_Rocketeer said:

Sorry, I thought your final figure was 10%. On re-reading I see that that was a separate point. The 10% figure is based on a significant skew, but actually your statistical methodology is pretty sound.

That doesn't necessarily mean that @Pthigrivi is wrong though. There isn't enough data to have any confidence in your 17%, however sound the method. How individuals respond to a proposal and how they respond to a delivered product often do not correlate.

It has the benefit that it potentially skews in all directions for all groups.  We don't actually know who it would be giving more of an advantage to because there is no way to know what those groups would really look like. It's intended to give a general idea, not a perfect allocation.  It's the sort of cost/benefit analysis that a company would have at their disposal.  In the end only they can truly decide what is right.

@PthigriviThere is no real data and can't be unless you can somehow poll every player.  Not just on the forums but everywhere.

Edited September 7, 2016 by Alshain

[Pthigrivi]

1 minute ago, Alshain said:

@PthigriviThere is no real data and can't be unless you can somehow poll every player.  Not just on the forums but everywhere.

No poll in the world works like that.

[The_Rocketeer]

Quote

Ignoring all data and just arbitrarily giving even weight to all possibilities is not actually a methodology at all. Its usually better to listen to what people have to say on a subject than to assume you know what they're thinking.

@Pthigrivi that may be true in the presence of reliable and representative data, but this poll, thread or even entire community do not meet that specification. What Alshain has done is dismissed inadequate data to find a benchmark on the basis of all options having equal statistical weight. That is not evidence of anything, as I pointed out, but it is useful for weighing the significance of qualitative, non-metric or poorly sampled evidence.

Essentially Alshain has drawn a line-of-best-fit based on neutral data, which basically just lets us know how far-from-normal or extraordinary any other evidence may be. It's important to realise that being 'normal' is not the same thing as being 'right' - reality is usually far from normal.

Edited September 7, 2016 by The_Rocketeer

[Superfluous J]

2 minutes ago, Pthigrivi said:

No poll in the world works like that.

You don't need to poll every player, but you do have to have at least some reasonable expectation that the people you poll are a representative cross-section of the relevant population.

A poll on a forum will never, ever be that because at the very least it discriminates against people who aren't on the forum.

[Alshain]

1 minute ago, 5thHorseman said:

You don't need to poll every player, but you do have to have at least some reasonable expectation that the people you poll are a representative cross-section of the relevant population.

A poll on a forum will never, ever be that because at the very least it discriminates against people who aren't on the forum.

That was a really good way to explain it.  Just wanted to say that, I'm not that good at explaining that type of thing

[tater]

Many would try ls who would not dl a

mod if it were stock. The cost is effectively zero, don't like it, don't turn it on.

[Superfluous J]

Just now, Alshain said:

That was a really good way to explain it.  Just wanted to say that, I'm not that good at explaining that type of thing

I've taken more than a few statistics classes in my day

[Pthigrivi]

If Alshain is accepting no data then he should assume nothing, including an equal distribution. If he's paying attention to pages upon pages of respondents comments then he would have at least something to go on. Certainly its not a professionally weighted poll, but given the overwhelming answers to questions like which model would make the best fit and would LS be okay if it had a toggle I don't think its remotely necessary for it to be.

Edited September 7, 2016 by Pthigrivi

[Alshain]

But assuming an equal distribution gives the same potential inaccuracies to all groups.  With the lack of the any good data, it's all there is.

Edited September 7, 2016 by Alshain

[The_Rocketeer]

2 minutes ago, Pthigrivi said:

If Alshain is accepting no data than he should assume nothing, including an equal distribution.

Yes, exactly. The evidence here on which you base your argument does lean in a particular direction. The only problem is, other evidence about other players - players for example who don't bother with forums of polls, but who still have strong feelings about stock LS - hasn't been accounted for. There could be a lot more players who aren't represented by this evidence than those who are, which would overturn your argument.

So Alshain's maths isn't based on this evidence, but it isn't based on any other evidence either. It just tells us what proportion the laws of probability suggest will benefit from stock LS. However, the laws of probability being what they are, the probable answer is still probably wrong.

[Pthigrivi]

8 minutes ago, The_Rocketeer said:

Yes, exactly. The evidence here on which you base your argument does lean in a particular direction. The only problem is, other evidence about other players - players for example who don't bother with forums of polls, but who still have strong feelings about stock LS - hasn't been accounted for. There could be a lot more players who aren't represented by this evidence than those who are, which would overturn your argument.

Player opinions aren't dice rolls. If you've decided to ignore all information you can at best ascribe blank variables, not an equal distribution. Its certainly true that a bigger sample size could give us a clearer picture of what people think, but as stated, when 90%+ of respondents are saying "Fine so long as it has a toggle" its hard to see how the vagueries of sample bias would be a deciding factor. 

[Stargate525]

44 minutes ago, tater said:

Many would try ls who would not dl a

mod if it were stock. The cost is effectively zero, don't like it, don't turn it on.

Not true. The cost is in development time, of Squad putting in yet ANOTHER half-baked system in lieu of bugfixing, balancing, and general polish.

[The_Rocketeer]

@Pthigrivi let's just walk thru this.

Let's say a million people play KSP.
The let's say half of them are registered on the forum: 500,000
Then let's say of them, a fifth are active enough to be online while this thread is still active: 100,000
Of them, half browse the suggestions forum: 50,000
Of them, only 4/5ths bother responding to polls or commenting in their threads: 40,000
Of them, 3/4rs see this thread and are interested enough to respond: 30,000
Of them, a third didn't want stock LS, and would never use it: 20,000
Of the rest, a quarter didn't mind having stock LS but probably wouldn't use it: 15,000
Of the rest, a two thirds prefer USILS-style implementation: 10,000
Of them, half actually vote and post to say as much: 5,000

Even if EVERY SINGLE ONE of those people did this, and even if KSP had ONLY sold 1M copies, the sample represented by this thread is still only half of 1 percent. And I have been EXTREMELY generous in my estimates whittling this down. In other words, the result of this thread's anecdotal evidence, no matter how strongly it leans in any direction, is just not nearly conclusive enough to draw conclusions. The probability of this sample being representative of all players is simply smaller than the probability of Alshain's neutral distribution being correct.

I've tried to help you come to understand this over a few posts, and I'm starting to get the feeling that I'm banging my head against a wall, so this is my last post on this. Apart from anything we're not really on topic anymore - talking about talking about the issue is not talking about the issue.

[tater]

3 minutes ago, Stargate525 said:

Not true. The cost is in development time, of Squad putting in yet ANOTHER half-baked system in lieu of bugfixing, balancing, and general polish.

It's already developed. RoverDude is talking about it, and he already wrote it, you can play with it now, just like you could play with his ISRU stuff before that became stock (a more complex version). 

[Red Iron Crown]

4 minutes ago, The_Rocketeer said:

Even if EVERY SINGLE ONE of those people did this, and even if KSP had ONLY sold 1M copies, the sample represented by this thread is still only half of 1 percent.

Just for amusement I ran the numbers on this. If a poll of such a sample gave a result of 50% it would be accurate to +/-1.4% 19 times out of 20. A sample size of just 500 would give +/- 4.4% 19 times out of 20.

Sampling is surprisingly accurate, assuming it is representative.

[Superfluous J]

11 minutes ago, The_Rocketeer said:

@Pthigrivi

@Pthigrivi

 

Let's say a million people play KSP.
The let's say half of them are registered on the forum: 500,000
Then let's say of them, a fifth are active enough to be online while this thread is still active: 100,000
Of them, half browse the suggestions forum: 50,000
Of them, only 4/5ths bother responding to polls or commenting in their threads: 40,000
Of them, 3/4rs see this thread and are interested enough to respond: 30,000
Of them, a third didn't want stock LS, and would never use it: 20,000
Of the rest, a quarter didn't mind having stock LS but probably wouldn't use it: 15,000
Of the rest, a two thirds prefer USILS-style implementation: 10,000
Of them, half actually vote and post to say as much: 5,000

(Sorry for the double @tags, I can't get the forums to remove them yet somehow got the first one duplicated. At least I can see it, so I know I'll get it)

The key here (As RIC notes) isn't the raw numbers, it's if the reasons for being here (or not) are related to preferring one choice over the other (or not). The best example of this is an online poll that asks "Do you use the Internet?" Assuming everybody tells the truth, the poll should say 100% of people use the Internet.

Not quite so obvious is, did a Life Support mod get a recent update? Maybe its users are busy playing with it. Or maybe it has problems so they are more on the forums now than another life support mod's users. Maybe people who like Life Support mods also tend to gravitate toward another genre of game and a new game in that genre just came out, so they are currently underrepresented.

Those are just some random thoughts, but they all boil down to Forum Polls Are Useless To Determine Anything Useful. About the best you can say is that they accurately describe the feelings of the people who bothered to answer it.

[The_Rocketeer]

10 minutes ago, Red Iron Crown said:

Sampling is surprisingly accurate, assuming it is representative.

Emphasis mine - the point being that at every step in my post above the sample becomes less representative. Not that your findings aren't interesting, Red, but it could seem like an endorsement of this sample, which I don't believe it is supposed to be.

Also, the actual sample that Pthigrivi draws his conclusions from is still way, way smaller than 500 responders.

Edited September 7, 2016 by The_Rocketeer

[Red Iron Crown]

5 minutes ago, The_Rocketeer said:

Emphasis mine - the point being that at every step in my post above the sample becomes less representative. Not that your findings aren't interesting, Red, but it could seem like an endorsement of this sample, which I don't believe it is supposed to be.

It is assuredly not meant as an endorsement. Forum polls are of negligible value for anything but forum issues, IMHO.

[Pthigrivi]

Well, if if selection bias is so strong as to make all commentary worthless why comment here at all?

[Alshain]

6 minutes ago, Pthigrivi said:

Well, if if selection bias is so strong as to make all commentary worthless why comment here at all?

It's not worthless, it's just not enough for statistical analysis.

[tater]

I see it this way, we could have a game where someone can legitimately post, "If you want life support, download a mod!"

Alternately, we could have a game where someone could post, "If you DON'T want LS, download a mod to turn it off!"

Or, we could have "The game comes with LS, turn it on or off as you prefer."

It's kind of stunning that anyone would vote for any option but the last one, as it's the only option not telling people that their way of playing is wrong. 

 

[Alshain]

16 minutes ago, tater said:

I see it this way, we could have a game where someone can legitimately post, "If you want life support, download a mod!"

Alternately, we could have a game where someone could post, "If you DON'T want LS, download a mod to turn it off!"

Or, we could have "The game comes with LS, turn it on or off as you prefer."

It's kind of stunning that anyone would vote for any option but the last one, as it's the only option not telling people that their way of playing is wrong. 

 

Except what you want is not the last option what you want is "The game comes with an LS implementation that a small fraction of players want, if it isn't what you want turn it of and download a mod."

In most cases, that option is almost synonymous with the first option and yet it will undoubtedly take up a lot of development time that could be used elsewhere.

Edited September 7, 2016 by Alshain

[tater]

12 minutes ago, Alshain said:

Except what you want is not the last option what you want is "The game comes with an LS implementation that a small fraction of players want, if it isn't what you want turn it of and download a mod."

In most cases, that option is almost synonymous with the first option and yet it will undoubtedly take up a lot of development time that could be used elsewhere.

Except the most likely candidate was already done by RoverDude, and won't take any dev time to speak of.

I could just as well say, "The game comes with an aerodynamics implementation that a small fraction of players want, if it isn't what you want turn it of and download a mod." So you want the soup back, since we all DL FAR, anyway?