hyperme
Level: Smitemaster

Rank Points: 1173
Registered: 06-23-2006
IP: Logged

55.25%

It's getting closer. Slowly, but surely!

____________________________
[Insert witty comment here]
Qzvlkx?

navithmastero
Level: Smitemaster

Rank Points: 621
Registered: 01-03-2012
IP: Logged

55.5!

-nm

____________________________
Member of the Snake Appreciation Society

One of your local HAs.

My stuff:

Godot23
Level: Master Delver
Rank Points: 114
Registered: 02-28-2013
IP: Logged

Ok, I am also curious about something else. How do the developers judge how complete the game actually is? Is there some kind of math formula or is it just estimated?

____________________________
Roses are red
Violets are red
Everything is red
My house is on fire

mrimer
Level: Legendary Smitemaster

Rank Points: 5141
Registered: 02-04-2003
IP: Logged

Godot23 wrote:
Ok, I am also curious about something else. How do the developers judge how complete the game actually is? Is there some kind of math formula or is it just estimated?

Spreadsheets. Spreadsheets laid out on 3x3 desks. With monthly planners that have 37 days on them. In Erik's basement, a mile underground. Where Erik is being held hostage. By sentient robots, in his own basement.

____________________________
Gandalf? Yes... That's what they used to call me.
Gandalf the Grey. That was my name.
I am Gandalf the White.
And I come back to you now at the turn of the tide.

RabidChild
Level: Smitemaster
Rank Points: 720
Registered: 01-15-2006
IP: Logged

56!

rman
Level: Smiter

Rank Points: 371
Registered: 09-24-2007
IP: Logged

Method 1: 288 days (August 3, 2012 to May 18, 2013) ÷ 42.5% (13.5% to 56%) = 6.776 days/% (previously 6.843)
⇒ end in 298 days, i. e. on March 12, 2014 (previously March 18, 2014).

Method 2: 388 days (April 25, 2012 to May 18, 2013) ÷ 56% (0% to 56%) = 6.929 days/% (previously 6.982)
⇒ end in 305 days, i. e. on March 19, 2014 (previously March 24, 2014).

Method 3: 179 days (November 20, 2012 to May 18, 2013) ÷ 40% (16% to 56%) = exactly 4.475 days/% (previously 4.487)
⇒ end in 197 days, i. e. on December 1, 2013 (previously December 2, 2013).

____________________________

Best quote ever (TSS ending spoiler):

Ezlo
Level: Smitemaster

Rank Points: 1227
Registered: 01-08-2006
IP: Logged

56.25% Who wants to place bets on what the latest .25% represents? I'm guessing it's related to tacos, or some other type of mexican food. No reason, I'm just really craving a fajita right now.

____________________________
Call me Citrus.

hyperme
Level: Smitemaster

Rank Points: 1173
Registered: 06-23-2006
IP: Logged

Ezlo wrote:
56.25% Who wants to place bets on what the latest .25% represents? I'm guessing it's related to tacos, or some other type of mexican food. No reason, I'm just really craving a fajita right now.

A puzzle where Beethro goes back in time to ruin DROD canon, then fix it again.

____________________________
[Insert witty comment here]
Qzvlkx?

mrimer
Level: Legendary Smitemaster

Rank Points: 5141
Registered: 02-04-2003
IP: Logged

Good guesses, all. I'm going to go talk about what the latest 0.25% is on the General board.

____________________________
Gandalf? Yes... That's what they used to call me.
Gandalf the Grey. That was my name.
I am Gandalf the White.
And I come back to you now at the turn of the tide.

Godot23
Level: Master Delver
Rank Points: 114
Registered: 02-28-2013
IP: Logged

mrimer wrote:

Godot23 wrote:
Ok, I am also curious about something else. How do the developers judge how complete the game actually is? Is there some kind of math formula or is it just estimated?

Spreadsheets. Spreadsheets laid out on 3x3 desks. With monthly planners that have 37 days on them. In Erik's basement, a mile underground. Where Erik is being held hostage. By sentient robots, in his own basement.

Ah, yes. That has happened to me before. A lot of my friends too... quite frequently actually... Maybe I should move...

And also, thanks! Just wondering!

____________________________
Roses are red
Violets are red
Everything is red
My house is on fire

---

[Last edited by Godot23 at 05-20-2013 01:52 PM]

navithmastero
Level: Smitemaster

Rank Points: 621
Registered: 01-03-2012
IP: Logged

56.5%

-nm

____________________________
Member of the Snake Appreciation Society

One of your local HAs.

My stuff:

Godot23
Level: Master Delver
Rank Points: 114
Registered: 02-28-2013
IP: Logged

navithmastero wrote:
56.5%

-nm

Lets do the 56.5 jive!
I am truly sorry for that... it just kinda happened

____________________________
Roses are red
Violets are red
Everything is red
My house is on fire

---

[Last edited by Godot23 at 05-21-2013 02:37 PM]

rman
Level: Smiter

Rank Points: 371
Registered: 09-24-2007
IP: Logged

Err, is it just me or we missed 57% and the Bar is at 57.5% already?

In that case:

Method 1: 291 days (August 3, 2012 to May 21, 2013) ÷ 44% (13.5% to 57.5%) = 6.614 days/% (previously 6.776)
⇒ end in 281 days, i. e. on February 26, 2014 (previously March 12, 2014).

Method 2: 391 days (April 25, 2012 to May 21, 2013) ÷ 57.5% (0% to 57.5%) = exactly 6.8 days/% (previously 6.929)
⇒ end in 289 days, i. e. on March 6, 2014 (previously March 19, 2014).

Method 3: 182 days (November 20, 2012 to May 21, 2013) ÷ 41.5% (16% to 57.5%) = 4.386 days/% (previously exactly 4.475)
⇒ end in 186 days, i. e. on November 23, 2013 (previously December 1, 2013).

____________________________

Best quote ever (TSS ending spoiler):

---

[Last edited by rman at 05-21-2013 05:28 PM]

stigant
Level: Smitemaster

Rank Points: 1182
Registered: 08-19-2004
IP: Logged

Method 3: 182 days (November 20, 2012 to May 21, 2013) ÷ 41.5% (16% to 57.5%) = 4.386 days/% (previously exactly 4.475)
⇒ end in 186 days, i. e. on November 23, 2013 (previously December 1, 2013).

That would be Thanksgiving weekend (in the US)... I wonder if I would end up being thankful for a new DROD or cursing b/c I have to spend time with family instead of playing a new DROD.

____________________________
Progress Quest Progress

Godot23
Level: Master Delver
Rank Points: 114
Registered: 02-28-2013
IP: Logged

So I just had this amazing idea. I was thinking that I would perform a statistical confidence interval using the various end date predictions on this thread as data. For those of you who don't speak statistics, a confidence interval is basically a method for getting a range of values within which is we are n% confident contains the true average value of the data we are testing (in this case, the true completion date of the game). Anyone object or think that this could be done with a better set of data? I was thinking about using 95% as the confidence level. If you have absolutely no idea what I am talking about, then sorry I can't make it more clear. When it is finished, I will put it into terms that make it clear. I was also thinking about updating it as more predictions came in from you, the forum dwellers.


____________________________
Roses are red
Violets are red
Everything is red
My house is on fire

Kwerulous
Level: Smiter
Rank Points: 340
Registered: 05-15-2008
IP: Logged

Interesting... what are you going to assume the distribution of the date predictions is?

(Well, I guess you'd have to look at the data first, but I'm not sure if it would fit anything obvious. I imagine it'd be quite... erratic).

west.logan
Level: Smitemaster

Rank Points: 608
Registered: 03-09-2011
IP: Logged

File: Bar Movement.xls (66.5 KB)
Downloaded 47 times.
License: Public Domain

I'm really not sure how well that would work as the predictions continually changed and aren't exactly close together but you're welcome to try. I've attached all the data so far.

____________________________
-Logan

Godot23
Level: Master Delver
Rank Points: 114
Registered: 02-28-2013
IP: Logged

I was planning on using a 1 sample T procedure for this. If the number of "trials" is more than 30, the we can assume that the Central Limit Theorem is met and therefore that the T distribution works. If not, I can use a Normal Probability Plot to look at the normality. If not... well then it was a good idea and I just would need to wait for more data to become available .
And most data I have seen is subject to changes. It wouldn't be perfect, but since when has statistics ever been?

____________________________
Roses are red
Violets are red
Everything is red
My house is on fire

---

[Last edited by Godot23 at 05-22-2013 03:17 PM]

Kwerulous
Level: Smiter
Rank Points: 340
Registered: 05-15-2008
IP: Logged

I think the t-distribution is always appropriate (if you think it looks like a bell curve) because you're estimating the standard deviation, but as your sample size increases it approaches a N(0,1).

I'm not an expert on statistics, so I'm probably not the best to take advice from - but I would say it would be best to check if the distribution even looks like a normal/t first. It might even be uniform.

mrimer
Level: Legendary Smitemaster

Rank Points: 5141
Registered: 02-04-2003
IP: Logged

Godot23 wrote:
Anyone object

iObject!

Ahh...carry on

____________________________
Gandalf? Yes... That's what they used to call me.
Gandalf the Grey. That was my name.
I am Gandalf the White.
And I come back to you now at the turn of the tide.

stigant
Level: Smitemaster

Rank Points: 1182
Registered: 08-19-2004
IP: Logged

58%

____________________________
Progress Quest Progress

Pearls
Level: Smitemaster

Rank Points: 775
Registered: 12-21-2009
IP: Logged

There's some weird interpretations going on here, mostly tied around a bit, er, liberal understanding of the Central Limit Theorem.

Keep in mind that the CLT describes the distribution of the mean of the samples, which while one-to-one with sample size, is not a perfect map regardless of distribution. That is to say, the more non-normal your data is, the more data you need to have the means behave normally.

Therefore, I would suggest using the Wilcoxon test, to make some determination about centrality, since you don't need to depend on normality. It's even pretty resistant to a lack of symmetry.

If our data are not normal, then the Wilcoxon is generally arbitrarily more powerful a test than a T test, and even if the data are, it's somewhere around 95% as efficient.








____________________________
Hey, w-wait, that's the guy who shamelessly promotes his own...

Godot23
Level: Master Delver
Rank Points: 114
Registered: 02-28-2013
IP: Logged

Pearls wrote:
There's some weird interpretations going on here, mostly tied around a bit, er, liberal understanding of the Central Limit Theorem.

Keep in mind that the CLT describes the distribution of the mean of the samples, which while one-to-one with sample size, is not a perfect map regardless of distribution. That is to say, the more non-normal your data is, the more data you need to have the means behave normally.

Therefore, I would suggest using the Wilcoxon test, to make some determination about centrality, since you don't need to depend on normality. It's even pretty resistant to a lack of symmetry.

If our data are not normal, then the Wilcoxon is generally arbitrarily more powerful a test than a T test, and even if the data are, it's somewhere around 95% as efficient.

That sounds really interesting! Unfortunately, I am speaking from the point of view of someone who just finished his first year statistics college class (AKA, I can do up to about tests for slope of linear regression and that's kinda it... ) , and therefore I honestly have no idea how to perform that test. I will do some research, however! I would be interested in learning more about it!

____________________________
Roses are red
Violets are red
Everything is red
My house is on fire

TripleM
Level: Smitemaster
Rank Points: 1377
Registered: 02-05-2003
IP: Logged

To be honest, I don't really see a confidence interval of the predicted end dates giving much useful information. If you want a confidence interval, it should be of the real end date based on the observed points - which can be done as specified here: http://en.wikipedia.org/wiki/Simple_linear_regression#Confidence_intervals

Godot23
Level: Master Delver
Rank Points: 114
Registered: 02-28-2013
IP: Logged

TripleM wrote:
To be honest, I don't really see a confidence interval of the predicted end dates giving much useful information. If you want a confidence interval, it should be of the real end date based on the observed points - which can be done as specified here: http://en.wikipedia.org/wiki/Simple_linear_regression#Confidence_intervals

That actually does seem more logical... How did I miss that?

____________________________
Roses are red
Violets are red
Everything is red
My house is on fire

navithmastero
Level: Smitemaster

Rank Points: 621
Registered: 01-03-2012
IP: Logged

58.25!

-nm

____________________________
Member of the Snake Appreciation Society

One of your local HAs.

My stuff:

navithmastero
Level: Smitemaster

Rank Points: 621
Registered: 01-03-2012
IP: Logged

58.5%

-nm

____________________________
Member of the Snake Appreciation Society

One of your local HAs.

My stuff:

navithmastero
Level: Smitemaster

Rank Points: 621
Registered: 01-03-2012
IP: Logged

58.75!

-nm

____________________________
Member of the Snake Appreciation Society

One of your local HAs.

My stuff:

navithmastero
Level: Smitemaster

Rank Points: 621
Registered: 01-03-2012
IP: Logged

58.9% What is going on?

-nm

____________________________
Member of the Snake Appreciation Society

One of your local HAs.

My stuff:

The Architest
Level: Master Delver

Rank Points: 255
Registered: 05-25-2008
IP: Logged

I have my own theories readied.