Role difficulty and number of ascensions

On 2008-02-13, hcobb wrote:
> On Feb 12, 8:01 am, David Damerell <damer....DeleteThis@chiark.greenend.org.uk>
> wrote:
>> You may not deliberately level teleport except to a level which you would
>> be permitted to move to via stairs/ladder/etc, or to the current level.
>> This includes hitting a known levelport trap.
>
> How about Rogue Quest?

Don't play rogue, then. Or have means of self-poly ready.

--
Ohle Claussen | GPG-Key-ID E7149169
----------===========----------
BOFH Excuse #33:
piezo-electric interference
 >> Stay informed about: Role difficulty and number of ascensions 

There's no good definition of "easy", which makes it pretty tough to
handle. One rigorous, if not particularly helpful, definition could
be: assuming the player uses a completely optimal strategy, what is
the probability of ascending a particular role/race/alignment combo?
This isn't particularly helpful for 2 reasons: first, no one knows the
perfect strategy, and second, if the 20+ streaks on NAO are any
indication, the probability of ascending with an optimal strategy is
pretty close to 1, maybe .9 or .95 or something. So then you have to
start asking about the win probability for sub-optimal strategies.
The trouble with this is that there are many many ways to be bad at
something, and I imagine no two beginners are bad at nethack for the
same reason.
 >> Stay informed about: Role difficulty and number of ascensions 

ThomasHarrisonNel....TakeThisOut@gmail.com wrote:
>
> There's no good definition of "easy", which makes it pretty tough to
> handle.  One rigorous, if not particularly helpful, definition could
> be: assuming the player uses a completely optimal strategy, what is
> the probability of ascending a particular role/race/alignment combo?
> This isn't particularly helpful for 2 reasons: first, no one knows the
> perfect strategy, and second, if the 20+ streaks on NAO are any
> indication, the probability of ascending with an optimal strategy is
> pretty close to 1, maybe .9 or .95 or something.  So then you have to
> start asking about the win probability for sub-optimal strategies.
> The trouble with this is that there are many many ways to be bad at
> something, and I imagine no two beginners are bad at nethack for the
> same reason.

The numbers in this discussion did demonstrate one thing -

Roles have very little effect compared to player skill. Asking about
role difficulty is asking about a secondary effect not a primary
effect.
The more analysis that is poured into the effort, the more the data
shows that roles have so little effect there is no use to calling any
role more or less difficult.

Roles don't matter to highly skilled players in that they all have
ascension rates dominated by individual skill.

Roles do matter to poorly skilled players in that focusing on a
single role can as easily lead to over specialization and become
detrimental as beneficial. Playing different roles exapnds the
skills; playing the same roles focuses specific skills; balance
between the two maters. But which roles are selected, the data
does not indicate that matters much.

This thread gives me better appreciation for using random character
generation.
 >> Stay informed about: Role difficulty and number of ascensions 

On Feb 13, 4:35 pm, Doug Freyburger <dfrey....DeleteThis@yahoo.com> wrote:
> ThomasHarrisonNel....DeleteThis@gmail.com wrote:
>
> > There's no good definition of "easy", which makes it pretty tough to
> > handle. One rigorous, if not particularly helpful, definition could
> > be: assuming the player uses a completely optimal strategy, what is
> > the probability of ascending a particular role/race/alignment combo?
> > This isn't particularly helpful for 2 reasons: first, no one knows the
> > perfect strategy, and second, if the 20+ streaks on NAO are any
> > indication, the probability of ascending with an optimal strategy is
> > pretty close to 1, maybe .9 or .95 or something. So then you have to
> > start asking about the win probability for sub-optimal strategies.
> > The trouble with this is that there are many many ways to be bad at
> > something, and I imagine no two beginners are bad at nethack for the
> > same reason.
>
> The numbers in this discussion did demonstrate one thing -
>
> Roles have very little effect compared to player skill. Asking about
> role difficulty is asking about a secondary effect not a primary
> effect.
> The more analysis that is poured into the effort, the more the data
> shows that roles have so little effect there is no use to calling any
> role more or less difficult.

Hmm. Not sure I agree; skill is, overall, more important than role--
but on the other hand, from the numbers a player with 1 ascension
under his belt is more likely to ascend a valk than a player with 20+
ascensions is to ascend a wizard.

The extremes of the roles matter greatly, and the more I look at the
data the more it becomes clear that valk/sam/bar and to a lesser
extent kni/cav really are _much_ easier than everything else
(especially arc/wiz), and that ran/rog are in the next tier (along
with hea for experienced players only).

> Roles don't matter to highly skilled players in that they all have
> ascension rates dominated by individual skill.

Roles seem to be able to make at least a factor-of-two difference even
for pretty highly skilled players.

Asc% for Valk/Wiz
For players with at least 20 ascensions: 33.19 Val/12.39 Wiz
For players with at least back-to-back ascensions at some point: 33.44
Val/10.91 Wiz
For players with at least a 5-game ascension streak: 43.69 Val/20.51
Wiz
For players with at least a 10-game ascension streak: 83.33 Val/47.37
Wiz

> Roles do matter to poorly skilled players in that focusing on a
> single role can as easily lead to over specialization and become
> detrimental as beneficial. Playing different roles exapnds the
> skills; playing the same roles focuses specific skills; balance
> between the two maters. But which roles are selected, the data
> does not indicate that matters much.

Hmm. That's an interesting theory, I'll try to run the numbers on it.

(Personally, I avoid random role choices because I like to keep my #
of characters ascended in each role about even, rather than my number
of games played in each role--so maybe testing both evenness of games
played vs. evenness of ascended would be worthwile).
 >> Stay informed about: Role difficulty and number of ascensions 

Quoting Janis Papanagnou <Janis_Papanagnou.DeleteThis@hotmail.com>:
>I _think_ that's true for the Mines and maybe Sokoban, and untrue for the
>Quest and Ludios and maybe Vlad, for example.

The Quest is special-cased. If you can't leave the branch, that's fine as
well - indeed it avoids the awkward levelported-out-of-branch cases.
--
David Damerell <damerell.DeleteThis@chiark.greenend.org.uk> Distortion Field!
Today is First Sunday, February - a weekend.
 >> Stay informed about: Role difficulty and number of ascensions 

Quoting Carl <carlhnelson RemoveThis @gmail.com>:
>35.3.2 Never moved away from any staircase (> or <) or portal (^) more
>than once before picking up the Amulet, and once after picking up the
>Amulet.

Levelport traps can hose you, which is harsher than I want.
--
David Damerell <damerell RemoveThis @chiark.greenend.org.uk> Distortion Field!
Today is First Sunday, February - a weekend.
 >> Stay informed about: Role difficulty and number of ascensions 

Quoting hcobb <henry.cobb.RemoveThis@gmail.com>:
>On Feb 12, 8:01 am, David Damerell <damer....RemoveThis@chiark.greenend.org.uk>
>>You may not deliberately level teleport except to a level which you would
>>be permitted to move to via stairs/ladder/etc, or to the current level.
>>This includes hitting a known levelport trap.
>How about Rogue Quest?

Better have the selfpoly resources before you do it, so you can Xorn
around when you arrive.
--
David Damerell <damerell.RemoveThis@chiark.greenend.org.uk> Distortion Field!
Today is First Sunday, February - a weekend.
 >> Stay informed about: Role difficulty and number of ascensions