Role difficulty and number of ascensions

Ohle Claussen <ohle.claus... RemoveThis @ds.mpg.de> wrote:
> cdi wrote:
> > Ohle Claussen <ohle.claus... RemoveThis @ds.mpg.de> wrote:
> >> Janis wrote:
> >> > cdi <cdinchau... RemoveThis @hotmail.com> wrote:
>
> >> >> I wonder, has anyone looked at 78291's NAO ascension rates by class?
> >> >> He has over 200 ascensions within the past year and, by himself, might
> >> >> provide a good read on the real degree of difficulty for each class.
>
> >> So this is a single person? In that case, generalizing the results with
> >> respect to difficulty wouldn't be very objective.
>
> > I think the conventional view that Valks and Wizards are "the easiest
> > classes" might be a self-reinforcing fallacy.  Beginners are often
> > told to play these "easy" classes, and so they do.
>
> That may well be the case.

Except that beginners view things differently than serial ascenders.
If a class has a higher chance of surviving sokoban and the mines
even with a strategy worse than what a serial ascender does, then
that class is easier to a beginner.

I think valk or wiz are "easier" not for that reason but because they
are in some way purer as combat wombat and as spell caster
respectively. Valk can get to expert in long sword and dagger,
samurai to long sword and bow. To me that means valk is better
at learning missile combat early on. Learn valk and the rest of
the combat wombat classes become easy. Learn wiz and the
rest of the spell casting classes become easy.

For the early game I think "easy" means simpler to learn
strategy not likelier to survive.

> > One way to eliminate this bias of perception involves a sample for
> > which it does not exist or is minimized.  Take a group of players with
> > a) high ascension rates, b) high ascension counts, and c) no overly
> > preferred class.  
>
> Such a sample will be hard to find, especially since criterion c) isn't very
> commonly met.
>
> >                   78291 provides such sample (statistically
> > significant to boot!) by himself.
>
> I disagree. There is some spread in his/her choice of classes (total
> number of games varies from 16 to 40). And any single player will have his
> own characteristic preferrences and skills that others don't, so a sample
> of one is always bad. For all we know, he/she might play all their valk
> games as a particularly hard conduct attempt for some personal reason.
>
> Going from 78291s statistics, if he/she really is a perfect sample,
> healers should be the easiest class. Do you really think they are?

For most frequent ascenders? It would take comparison with other
frequent ascenders but it wouldn't be surprising.
 >> Stay informed about: Role difficulty and number of ascensions 

On Feb 8, 11:03 am, Ohle Claussen <ohle.claus....DeleteThis@ds.mpg.de> wrote:
> On 2008-02-08, cdi wrote:
> > n Feb 8, 7:54 am, Ohle Claussen <ohle.claus....DeleteThis@ds.mpg.de> wrote:
> >> On 2008-02-08, Janis wrote:
> >> > On 8 Feb., 04:33, cdi <cdinchau....DeleteThis@hotmail.com> wrote:
>
> > One way to eliminate this bias of perception involves a sample for
> > which it does not exist or is minimized. Take a group of players with
> > a) high ascension rates, b) high ascension counts, and c) no overly
> > preferred class.
>
> Such a sample will be hard to find, especially since criterion c) isn't very
> commonly met.

An "expert, well rounded player" sample could consist of those players
who meet 6 of the following:

- have at least 13 ascensions
- have ascended each class
- have no more than 2/13 of their games in any given class
- have no more than 2/13 of their ascensions in any given class
- have no less than 1/26 of their games in any given class
- have no less than 1/26 of their ascensions in any given class
- have an ascension rate of at least 5.00%

There is a pool of 100 players on alt.org who meet the first
requirement. I wonder how many meet another 5 requirements.

> > 78291 provides such sample (statistically
> > significant to boot!) by himself.
>
> I disagree. There is some spread in his/her choice of classes (total
> number of games varies from 16 to 40). And any single player will have his
> own characteristic preferrences and skills that others don't, so a sample
> of one is always bad. For all we know, he/she might play all their valk
> games as a particularly hard conduct attempt for some personal reason.
>
> Going from 78291s statistics, if he/she really is a perfect sample,
> healers should be the easiest class. Do you really think they are?

There is no such thing as a "perfect sample." So, using more than one
player will give better results. 78291 has enough ascensions and
games played, however, that looking at his results alone will give
meaningful, if imperfect, results.

To answer your question, based on my own healer ascension rate (60%)
and my overall ascension rate (43%), I would argue that Healers are an
easier than average class.

-cdi
 >> Stay informed about: Role difficulty and number of ascensions 

On 2008-02-07, Doug Freyburger <dfreybur.RemoveThis@yahoo.com> wrote:
> If I read the table correctly there are columns of how many ascensions
> are a part of a run of sequential ascensions. It takes very few
> regular serial ascenders to bias such numbers greatly.

Nope. Those numbers are how many ascensions a given player has, period --
sequential isn't considered. The label "serial ascenders" is reserved for
those who have at least 13 ascensions, and more implies that they've
reached a point in the game where they tend to win, as opposed to vice
versa.

>> > Maybe combine the two numbers by multiplying?  Then sort highest
>> > to lowest as a difficulty estimate.
>>
>> Which two numbers are you thinking of?
>
> (First ascensions by type) / (Number of games by type)

I think there's a column in there that does exactly that, isn't there?

I'm going to have to go back and look at it more closely when I get a
chance, but things are kind of busy ATM. (Again, Google probably has
my explanations of this from long ago

> Given how popular valk and wiz are I would not be surprised if their
> first asension ratio would be very low. And that really does mean
> they are "harder" in some nebulous sense,

Wizards are far more popular than any other class, bar none. That by
itself creates a problem and skews things; if you don't apply any sort
of calculations to filter out that garbage, you end up thinking Wizards
are the hardest class, when they're the easiest.

The distinction here is that only a small amount of difficulty in
Nethack is role-based to begin with -- the vast majority of obstacles to
be overcome are nearly identical for every class, and so are the
solutions. (Conflict for Astral, for example, is "get a ring" for every
class but the Caveman.)

> Nethack in general is so hard that measuring difficulty by class
> is like using mathematics to compare different ways to generate
> infinity.

And yet, Nethack is actually very easy for a player who is willing to
use all available resources.

This seems odd, but consider: the majority of Nethack's perceived
"difficulty" falls into the category of "things you have to learn"
(don't move with a footrice corpse wielded) as opposed to "problems you
have to solve" (tactical combat, etc.) Once you commit yourself to
working with complete information, the tactical problems that remain are
trivial.

--
Derek

Game info and change log: http://sporkhack.com
Beta Server: telnet://sporkhack.com
IRC: irc.freenode.net, #sporkhack
 >> Stay informed about: Role difficulty and number of ascensions 

On 2008-02-08, cdi <cdinchauste DeleteThis @hotmail.com> wrote:
> On Feb 7, 5:41 pm, Doug Freyburger <dfrey... DeleteThis @yahoo.com> wrote:
>> Derek Ray <de... DeleteThis @moot.its.only.a.spamtrap.org> wrote:
>> > "Percentage of serial ascensions among ascenders" doesn't make sense as
>> > written; run that by again, rephrased?
>
> I wonder, has anyone looked at 78291's NAO ascension rates by class?
> He has over 200 ascensions within the past year and, by himself, might
> provide a good read on the real degree of difficulty for each class.

He would be the absolute worst subject, actually, for two reasons:

1) He doesn't play for ratio (win every game), he just plays extremely fast
and restarts if it ends in death.

2) His skill level renders all classes equally 'easy' and thus creates
far more difficulty distinguishing between classes than if we take a
spectrum including those players who are happy to have ascended once.

--
Derek

Game info and change log: http://sporkhack.com
Beta Server: telnet://sporkhack.com
IRC: irc.freenode.net, #sporkhack
 >> Stay informed about: Role difficulty and number of ascensions 

On Feb 8, 9:23 am, Derek Ray <de....DeleteThis@moot.its.only.a.spamtrap.org>
wrote:

> The label "serial ascenders" is reserved for
> those who have at least 13 ascensions, and more implies that they've
> reached a point in the game where they tend to win, as opposed to vice
> versa.

I was wondering what your definition of "serial ascender" was. I would
have thought it had at least something to do with ascension
percentage, because even though I had 17 ascensions last year, I had
only around 15% winning percentage, and didn't consider that all that
great an accomplishment...

- funcrunch
 >> Stay informed about: Role difficulty and number of ascensions 

On Feb 8, 9:39 am, cdi <cdinchau....TakeThisOut@hotmail.com> wrote:

> An "expert, well rounded player" sample could consist of those players
> who meet 6 of the following:
>
> - have at least 13 ascensions
> - have ascended each class
> - have no more than 2/13 of their games in any given class
> - have no more than 2/13 of their ascensions in any given class
> - have no less than 1/26 of their games in any given class
> - have no less than 1/26 of their ascensions in any given class
> - have an ascension rate of at least 5.00%
>
> There is a pool of 100 players on alt.org who meet the first
> requirement. I wonder how many meet another 5 requirements.

Hey, I think I meet four of those criteria (1, 2, 6, and 7). Not
all six though, and I hardly hold myself out to be either a typical or
an "expert, well rounded player"...

http://alt.org/nethack/player-stats.php?player=funcrunch

Trying for vegetarian conduct hurt my Tourist ascension percentage
some amount - conducts and startscumming further muddle stats
comparisons for both skilled and unskilled players I think...

- funcrunch
 >> Stay informed about: Role difficulty and number of ascensions 

On 2008-02-08, Janis wrote:
> On 8 Feb., 04:33, cdi <cdinchau... DeleteThis @hotmail.com> wrote:
>>
>> I wonder, has anyone looked at 78291's NAO ascension rates by class?
>> He has over 200 ascensions within the past year and, by himself, might
>> provide a good read on the real degree of difficulty for each class.

So this is a single person? In that case, generalizing the results with
respect to difficulty wouldn't be very objective.

> Good idea!
>
>
> 78291 A N A+N A/(A+N)
>
> Arc 18 17 35 51%
> Bar 19 6 25 76%
> Cav 18 5 23 78%
> Hea 13 3 16 81%
> Kni 15 8 23 65%
> Mon 20 7 27 74%
> Pri 18 22 40 45%
> Ran 28 11 39 72%
> Rog 19 12 31 61%
> Sam 14 7 21 67%
> Tou 16 14 30 53%
> Val 15 7 22 68%
> Wiz 21 12 33 64%

Or sorted:
Hea 81%
Cav 78%
Bar 76%
Mon 74%
Ran 72%
Val 68%
Sam 67%
Kni 65%
Wiz 64%
Rog 61%
Tou 53%
Arc 51%
Pri 45%

Definitely not what I'd expect from a representative player sample.

--
Ohle Claussen | GPG-Key-ID E7149169
----------===========----------
BOFH Excuse #104:
backup tape overwritten with copy of system manager's favourite CD
 >> Stay informed about: Role difficulty and number of ascensions 

On 2008-02-08, funcrunch <web-google2.DeleteThis@funcrunch.org> wrote:
> On Feb 8, 9:23 am, Derek Ray <de....DeleteThis@moot.its.only.a.spamtrap.org>
>> The label "serial ascenders" is reserved for
>> those who have at least 13 ascensions, and more implies that they've
>> reached a point in the game where they tend to win, as opposed to vice
>> versa.
>
> I was wondering what your definition of "serial ascender" was. I would
> have thought it had at least something to do with ascension
> percentage, because even though I had 17 ascensions last year, I had
> only around 15% winning percentage, and didn't consider that all that
> great an accomplishment...

Great compared to zipcode? No. But realistically, once you've done
"one of each", it is a safe bet to say you've mastered Nethack -- and
once you've ascended 13 times _regardless_ of class, it's safe to say
that you're more than familiar with what it actually _takes_ to ascend,
whether or not you've bothered to do it with any given role.

--
Derek

Game info and change log: http://sporkhack.com
Beta Server: telnet://sporkhack.com
IRC: irc.freenode.net, #sporkhack
 >> Stay informed about: Role difficulty and number of ascensions 

On Feb 8, 3:41 pm, Derek Ray <de... DeleteThis @moot.its.only.a.spamtrap.org>
wrote:
> On 2008-02-08, funcrunch <web-goog... DeleteThis @funcrunch.org> wrote:
>
> > On Feb 8, 9:23 am, Derek Ray <de... DeleteThis @moot.its.only.a.spamtrap.org>
> >> The label "serial ascenders" is reserved for
> >> those who have at least 13 ascensions, and more implies that they've
> >> reached a point in the game where they tend to win, as opposed to vice
> >> versa.
>
> > I was wondering what your definition of "serial ascender" was. I would
> > have thought it had at least something to do with ascension
> > percentage, because even though I had 17 ascensions last year, I had
> > only around 15% winning percentage, and didn't consider that all that
> > great an accomplishment...
>
> Great compared to zipcode? No. But realistically, once you've done
> "one of each", it is a safe bet to say you've mastered Nethack -- and
> once you've ascended 13 times _regardless_ of class, it's safe to say
> that you're more than familiar with what it actually _takes_ to ascend,
> whether or not you've bothered to do it with any given role.

Here's my "difficulty for non-experts who know how to play" stat.

Take all the players on NAO who have ascended at least 3 times but no
more than 12 times.
Throw out all the games (ascended or not) up to and including their
2nd ascension.

Those numbers are arbitrary, but it's basically eliminate all the
"just learning to play" games and the "I've totally mastered things"
games.

Ascension percentage by class for the remaining games:
13.37 Bar
13.30 Val
10.51 Sam
10.30 Cav
9.20 Kni
6.01 Rog
5.96 Ran
5.08 Tou
4.74 Pri
4.18 Hea
4.00 Mon
3.52 Wiz
2.58 Arc
 >> Stay informed about: Role difficulty and number of ascensions 

On 2008-02-08, cdi wrote:
> n Feb 8, 7:54 am, Ohle Claussen <ohle.claus....DeleteThis@ds.mpg.de> wrote:
>> On 2008-02-08, Janis wrote:
>> > On 8 Feb., 04:33, cdi <cdinchau....DeleteThis@hotmail.com> wrote:
>>
>> >> I wonder, has anyone looked at 78291's NAO ascension rates by class?
>> >> He has over 200 ascensions within the past year and, by himself, might
>> >> provide a good read on the real degree of difficulty for each class.
>>
>> So this is a single person? In that case, generalizing the results with
>> respect to difficulty wouldn't be very objective.
>
> I think the conventional view that Valks and Wizards are "the easiest
> classes" might be a self-reinforcing fallacy. Beginners are often
> told to play these "easy" classes, and so they do.

That may well be the case.

> One way to eliminate this bias of perception involves a sample for
> which it does not exist or is minimized. Take a group of players with
> a) high ascension rates, b) high ascension counts, and c) no overly
> preferred class.

Such a sample will be hard to find, especially since criterion c) isn't very
commonly met.

> 78291 provides such sample (statistically
> significant to boot!) by himself.

I disagree. There is some spread in his/her choice of classes (total
number of games varies from 16 to 40). And any single player will have his
own characteristic preferrences and skills that others don't, so a sample
of one is always bad. For all we know, he/she might play all their valk
games as a particularly hard conduct attempt for some personal reason.

Going from 78291s statistics, if he/she really is a perfect sample,
healers should be the easiest class. Do you really think they are?

--
Ohle Claussen | GPG-Key-ID E7149169
----------===========----------
BOFH Excuse #104:
backup tape overwritten with copy of system manager's favourite CD
 >> Stay informed about: Role difficulty and number of ascensions 

On 2008-02-08, Doug Freyburger wrote:
> Ohle Claussen <ohle.claus....DeleteThis@ds.mpg.de> wrote:
>> Going from 78291s statistics, if he/she really is a perfect sample,
>> healers should be the easiest class. Do you really think they are?
>
> For most frequent ascenders? It would take comparison with other
> frequent ascenders

Exactly.

> but it wouldn't be surprising.

I'm not sure, but healers don't seem to have any specific advatege going for
them that could make them so much better than, say, tourists, which according
to those stats would rank third to last.

--
Ohle Claussen | GPG-Key-ID E7149169
----------===========----------
BOFH Excuse #104:
backup tape overwritten with copy of system manager's favourite CD
 >> Stay informed about: Role difficulty and number of ascensions 

Quoting cdi <cdinchauste.RemoveThis@hotmail.com>:
>One way to eliminate this bias of perception involves a sample for
>which it does not exist or is minimized. Take a group of players with
>a) high ascension rates, b) high ascension counts, and c) no overly
>preferred class.

Of course, we're missing a definition of "easiest". The class which offers
the highest chance of ascension for an extremely skilled player is
probably not the same as the class which requires a novice player to
develop the lowest level of skill to be able to ascend.

Also looking at a particular skilled player is junk unless you know what
conduct objectives they have had. Maybe my Priest batting rate is low
because I am trying to do an Alpine ascension.
--
David Damerell <damerell.RemoveThis@chiark.greenend.org.uk> Distortion Field!
Today is First Tuesday, February.
 >> Stay informed about: Role difficulty and number of ascensions 

Derek Ray wrote:
> On 2008-02-08, cdi <cdinchauste.RemoveThis@hotmail.com> wrote:
>>On Feb 7, 5:41 pm, Doug Freyburger <dfrey....RemoveThis@yahoo.com> wrote:
>>>Derek Ray <de....RemoveThis@moot.its.only.a.spamtrap.org> wrote:
>>>
>>>>"Percentage of serial ascensions among ascenders" doesn't make sense as
>>>>written; run that by again, rephrased?
>>
>>I wonder, has anyone looked at 78291's NAO ascension rates by class?
>>He has over 200 ascensions within the past year and, by himself, might
>>provide a good read on the real degree of difficulty for each class.
>
> He would be the absolute worst subject, actually, for two reasons:
>
> 1) He doesn't play for ratio (win every game), he just plays extremely fast
> and restarts if it ends in death.

Mind, his ascension ratio is _64%_! That's quite different to someone who
doesn't mind sacrificing a sequence of characters just to ascend 1 in 10.

> 2) His skill level renders all classes equally 'easy'

Quite true. I think that is a good base for a non-biased "Which class is
hardest?" question. There are also other views to look at that question;
e.g. Doug's comment upthread:
>>>>
>>>> Except that beginners view things differently than serial ascenders.
>>>> If a class has a higher chance of surviving sokoban and the mines
>>>> even with a strategy worse than what a serial ascender does, then
>>>> that class is easier to a beginner.

> and thus creates
> far more difficulty distinguishing between classes than if we take a
> spectrum including those players who are happy to have ascended once.

Janis
 >> Stay informed about: Role difficulty and number of ascensions