Nethack variants and popularity

[Janis Papanagnou]

The following data is taken from a post from "CSS Dixieland" (that
appeared in a recent thread "nethack gpl"), here reformatted for
Usenet (#columns < 80), and I added a column 'Popularity'.


Junethack 2023 - Played Games

Variant Ascensions Played Ascensions[%] Popularity[%]

NetHack-3.4.3 23 575 4.00 3.67
NotDNetHack 8 211 3.79 1.35
GnollHack 3 92 3.26 0.59
DNetHack 7 248 2.82 1.58
SlashTHEM 2 99 2.02 0.63
AceHack 3 174 1.72 1.11
NetHack4 9 588 1.53 3.75
NetHack-3.6.7 51 3936 1.30 25.11
NetHack-Fourk 1 84 1.19 0.54
NetHack-3.7.0 25 3333 0.75 21.26
XNetHack 3 501 0.60 3.20
EvilHack 4 742 0.54 4.73
Slash'EM 1 221 0.45 1.41
DynaHack 1 358 0.28 2.28
HackEM 2 837 0.24 5.34
GruntHack 1 470 0.21 3.00
UnNetHack 2 1245 0.16 7.94
NetHack-1.3d 0 61 0.00 0.39
SporkHack 0 77 0.00 0.49
FIQHack 0 196 0.00 1.25
SpliceHack 0 1630 0.00 10.40


I think it's interesting to observe that the Vanilla line (represented
by three entries, 343, 367, 370) sums up to 50.0% of the played games!

In other posting contexts I had mentioned that (for my taste) there's
already (more than?) enough variants existing. Given the popularity of
the various variants it may be worth a reconsideration.

About two(?) decades ago (when not that many variants existed) I seem
to recall to have heard that Slash'EM was a "testbed" for new features;
that those features that proved worthwhile made their way into Vanilla.
I think this is an interesting concept. It helped to maintain a standard
("Vanilla") line with consolidated features added (in a balanced way).

Meanwhile I played quite some variants; one problem I had was that many
variants had features that were overlapping, partly or fully, partly
implemented (slightly or completely) differently; in other words, that
there was no overall consolidated concept. That's fine if a variant has
a community that constantly plays just this variant or popularity.

I will not draw any (general) conclusions in this post. (The data gives
hints but not the statistical significance or all the necessary data.)

I think interesting data would also be the number of different players
for each variant, as an addition or refinement to the data column of
the number of played games. Then we'd certainly have a more accurate
definition of "popularity".

Personally, I think, I'll try out SpliceHack (10.4% "popularity") next.

Janis

[nabru]

Post by Janis Papanagnou
Personally, I think, I'll try out SpliceHack (10.4% "popularity") next.

FWIW, SpliceHack has a bunch of bugs and isn't in active development
(the main dev quit working on it).

[Janis Papanagnou]

Post by nabru

Post by Janis Papanagnou
Personally, I think, I'll try out SpliceHack (10.4% "popularity") next.

FWIW, SpliceHack has a bunch of bugs and isn't in active development
(the main dev quit working on it).

Good to know. Thanks!

It's a problem if development is relying on one (or too few) people;
chances are high that at some point in time it won't be continued.

It would be good if the interesting parts would be borrowed from
variants to make their way into a maintained main development line.

Janis

[CSS Dixieland]

== Ascensions per version or variety in Junethack 2023 ==

This text is a comment about the results of Junethack, not a notification related to NetHack Wiki.

The Junethack competition of 2023 has finished. This is the proportion of ascensions for each of the versions or varieties played. Eight servers have participated: one of Nethack Alt Org, one of Nethack 4, three of Gnollhack, and three of Hardfought. Hardfought servers also have few other versions available, such as NetHack version 1.3d, but they have not been included in Junethack. The list is ordered by the proportion of ascensions obtained.

<pre>
Version or.............Ascensions...............Total games played......Per cent of ascensions
variety................of the hero..............in Junethack 2023.......per version or variety

NetHack 3.4.3...........23..............................575................................4
NotDNetHack..............8..............................211................................3.79
GnollHack................3...............................92................................3.26
DNetHack.................7..............................248................................2.82
SlashTHEM................2...............................99................................2.02
AceHack..................3..............................174................................1.72
NetHack4.................9..............................588................................1.53
NetHack 3.6.7...........51.............................3936................................1.3
NetHack Fourk............1...............................84................................1.19
NetHack 3.7.0...........25.............................3333................................0.75
XNetHack.................3..............................501................................0.6
EvilHack.................4..............................742................................0.54
Slash'EM.................1..............................221................................0.45
DynaHack.................1..............................358................................0.28
HackEM...................2..............................837................................0.24
GruntHack................1..............................470................................0.21
UnNetHack................2.............................1245................................0.16
NetHack 1.3d.............0...............................61................................0
SporkHack................0...............................77................................0
FIQHack..................0..............................196................................0
SpliceHack...............0.............................1630................................0
</pre>

Average: 7 ascensions and 747 games per version or variety, less than 1 per cent ascended. On average, one ascension out of more than a hundred games, the others are not ascended. Dungeon games are amongst the most difficult, but they are also amongst the most addictive. A combination that seems to suggest a certain... 'masochism' amongst dungeon enthusiasts.

The table given above is reasonably well aligned on our screen, but maybe not so well on other screens, or printers, as it depends on the spacing and the character font used by each machine. If the automatic formatting executed by the wiki look gibberish, as I am afraid that it look, please look at the source by activating 'Edit'.

The table shows only the proportion of ascensions in Junethack 2023, not in other competitions, and not in the total number of games played throughout the years. It is also known in Statistical Science that samples with small numbers tend to be less reliable than samples with big numbers, therefore the table should not be taken as an absolute indication of play difficulty, especially for versions or varieties in which the number of games played, or of ascensions, has been small.

Let alone should the table be seen as indication of play 'enjoyability', a very subjective perception that varies greatly from an individual player to another. The style of a player may be rapid, playing our dungeon game almost as he would play a 'point and shoot' arcade game. The style of another player may be slow, playing almost as he would play a classic game of Chess, thinking every action and considering possibilities at every few turns. Or at every turn, in particularly difficult situations.

Turn-based dungeon games are unique in accommodating to both kinds of players, rapid and slow.
One is the 'action' player who goes for muscular Barbarians attacking in melée, at close quarters. The other is the reflexive player who prefers brainy Wizards casting magical spells. In between we have Walkyrie, Ranger, Knight, Samurai, Monk, and the other rôles. Every player has some choice for him.

Of 1630 games played in SpliceHack, not even one has been ascended. FIQHack, SporkHack and NetHack version 1.3d have not been ascended either, but they have been played less than 200 times each (196, 77 and 61 times), not 1630 times. At the opposite extreme we have NetHack version 3.4.3 heading the ranking with NotDNetHack and GnollHack, each with an impressive ascension rate of between three and four per cent. Four per cent exactly, in the case of NetHack version 3.4.3.

Some other factors may account for such disparities in results, from four per cent to virtually zero ascensions. NetHack version 3.4.3 was for many years the only official version available, meaning that plenty of detail in game mechanics is known to a vast number of players. Once the Dev Team decided to take action and release NetHack versions 3.6.x (NetHack version 3.5 never was official), a number of dungeon devotees started to migrate to them, but version 3.4.3 is still very well known.

However, that explanation does not account for NotDNetHack, which is not so well known. Let alone for GnollHack, which is, besides, a recent arrival on the field and known only to a select minority of dungeon enthusiasts. GnollHack can be played off-line in classic or modern mode (permanent death or not), casual or serious commit (preserving older sessions or not), with seven levels of difficulty (damage inflicted to the hero or by the hero to monsters), but in Junethack it is assumed that all the 92 GnollHack participating games have been played in standard configuration. Then, it is difficult to think of what has made GnollHack so successful in the number of ascensions. I have myself played GnollHack a few times and I am playing it at present. It is interesting in the unique features that do not exist in any other variety, but it is not 'excessively' easy. My hero has died a few times already, and in the others I have made him exit the dungeon from first depth, with the remaining pets beside.

It may be that the exclusive features of GnollHack (including even music and speech sounds with the recorded voices of actors when playing off-line) make GnollHack singularly attractive, therefore its players pay more attention and make fewer mistakes. However, that is only a hypothesis. The true cause may be a different one, yet to be discovered.

Sources:

https://junethack.net/activity

https://junethack.net/post_tournament_statistics

A more readable presentation of the statistical table given above can be read at the GnollHack Forum:

https://community.soundmindgames.com/threads/ascensions-per-version-or-variety-in-junethack-2023.75/#post-366

[[User:Cssdixieland|Cssdixieland]] ([[User talk:Cssdixieland|talk]]) 02:34, 1 July 2023 (UTC)

[Bozo User]

Post by CSS Dixieland
== Ascensions per version or variety in Junethack 2023 ==
This text is a comment about the results of Junethack, not a notification related to NetHack Wiki.
The Junethack competition of 2023 has finished. This is the proportion of ascensions for each of the versions or varieties played. Eight servers have participated: one of Nethack Alt Org, one of Nethack 4, three of Gnollhack, and three of Hardfought. Hardfought servers also have few other versions available, such as NetHack version 1.3d, but they have not been included in Junethack. The list is ordered by the proportion of ascensions obtained.
<pre>
Version or.............Ascensions...............Total games played......Per cent of ascensions
variety................of the hero..............in Junethack 2023.......per version or variety
NetHack 3.4.3...........23..............................575................................4
NotDNetHack..............8..............................211................................3.79
GnollHack................3...............................92................................3.26
DNetHack.................7..............................248................................2.82
SlashTHEM................2...............................99................................2.02
AceHack..................3..............................174................................1.72
NetHack4.................9..............................588................................1.53
NetHack 3.6.7...........51.............................3936................................1.3
NetHack Fourk............1...............................84................................1.19
NetHack 3.7.0...........25.............................3333................................0.75
XNetHack.................3..............................501................................0.6
EvilHack.................4..............................742................................0.54
Slash'EM.................1..............................221................................0.45
DynaHack.................1..............................358................................0.28
HackEM...................2..............................837................................0.24
GruntHack................1..............................470................................0.21
UnNetHack................2.............................1245................................0.16
NetHack 1.3d.............0...............................61................................0
SporkHack................0...............................77................................0
FIQHack..................0..............................196................................0
SpliceHack...............0.............................1630................................0
</pre>
Average: 7 ascensions and 747 games per version or variety, less than 1 per cent ascended. On average, one ascension out of more than a hundred games, the others are not ascended. Dungeon games are amongst the most difficult, but they are also amongst the most addictive. A combination that seems to suggest a certain... 'masochism' amongst dungeon enthusiasts.
The table given above is reasonably well aligned on our screen, but maybe not so well on other screens, or printers, as it depends on the spacing and the character font used by each machine. If the automatic formatting executed by the wiki look gibberish, as I am afraid that it look, please look at the source by activating 'Edit'.
The table shows only the proportion of ascensions in Junethack 2023, not in other competitions, and not in the total number of games played throughout the years. It is also known in Statistical Science that samples with small numbers tend to be less reliable than samples with big numbers, therefore the table should not be taken as an absolute indication of play difficulty, especially for versions or varieties in which the number of games played, or of ascensions, has been small.
Let alone should the table be seen as indication of play 'enjoyability', a very subjective perception that varies greatly from an individual player to another. The style of a player may be rapid, playing our dungeon game almost as he would play a 'point and shoot' arcade game. The style of another player may be slow, playing almost as he would play a classic game of Chess, thinking every action and considering possibilities at every few turns. Or at every turn, in particularly difficult situations.
Turn-based dungeon games are unique in accommodating to both kinds of players, rapid and slow.
One is the 'action' player who goes for muscular Barbarians attacking in melée, at close quarters. The other is the reflexive player who prefers brainy Wizards casting magical spells. In between we have Walkyrie, Ranger, Knight, Samurai, Monk, and the other rôles. Every player has some choice for him.
Of 1630 games played in SpliceHack, not even one has been ascended. FIQHack, SporkHack and NetHack version 1.3d have not been ascended either, but they have been played less than 200 times each (196, 77 and 61 times), not 1630 times. At the opposite extreme we have NetHack version 3.4.3 heading the ranking with NotDNetHack and GnollHack, each with an impressive ascension rate of between three and four per cent. Four per cent exactly, in the case of NetHack version 3.4.3.
Some other factors may account for such disparities in results, from four per cent to virtually zero ascensions. NetHack version 3.4.3 was for many years the only official version available, meaning that plenty of detail in game mechanics is known to a vast number of players. Once the Dev Team decided to take action and release NetHack versions 3.6.x (NetHack version 3.5 never was official), a number of dungeon devotees started to migrate to them, but version 3.4.3 is still very well known.
However, that explanation does not account for NotDNetHack, which is not so well known. Let alone for GnollHack, which is, besides, a recent arrival on the field and known only to a select minority of dungeon enthusiasts. GnollHack can be played off-line in classic or modern mode (permanent death or not), casual or serious commit (preserving older sessions or not), with seven levels of difficulty (damage inflicted to the hero or by the hero to monsters), but in Junethack it is assumed that all the 92 GnollHack participating games have been played in standard configuration. Then, it is difficult to think of what has made GnollHack so successful in the number of ascensions. I have myself played GnollHack a few times and I am playing it at present. It is interesting in the unique features that do not exist in any other variety, but it is not 'excessively' easy. My hero has died a few times already, and in the others I have made him exit the dungeon from first depth, with the remaining pets beside.
It may be that the exclusive features of GnollHack (including even music and speech sounds with the recorded voices of actors when playing off-line) make GnollHack singularly attractive, therefore its players pay more attention and make fewer mistakes. However, that is only a hypothesis. The true cause may be a different one, yet to be discovered.
https://junethack.net/activity
https://junethack.net/post_tournament_statistics
https://community.soundmindgames.com/threads/ascensions-per-version-or-variety-in-junethack-2023.75/#post-366
[[User:Cssdixieland|Cssdixieland]] ([[User talk:Cssdixieland|talk]]) 02:34, 1 July 2023 (UTC)

On nethack 3.4.3, I still play Slash'em because of that. It's like Nethack 3.4.3 but with tons of
nice changes and oddities, such as #technique's and fun roles such as Ice Mages, Vampires
and Doppleganger Monks pretty close to a Dragon Ball gameplay.

Nethack => 3.6.0 it's fine if you like Terry Pratchett and quotes on Discworld.

[Loggers VIII]

I'm late to this, but wanted to comment on some of the reasons for the ascension percentages in different variants.

Post by CSS Dixieland
Of 1630 games played in SpliceHack, not even one has been ascended.

SpliceHack is largely not played anymore; I highly suspect that most, if not all, of these deaths come from post163, as it seems that is one of their favorite variants to die in.

Post by CSS Dixieland
Some other factors may account for such disparities in results, from four per cent to virtually zero ascensions. NetHack version 3.4.3 was for many years the only official version available, meaning that plenty of detail in game mechanics is known to a vast number of players. Once the Dev Team decided to take action and release NetHack versions 3.6.x (NetHack version 3.5 never was official), a number of dungeon devotees started to migrate to them, but version 3.4.3 is still very well known.
However, that explanation does not account for NotDNetHack, which is not so well known. Let alone for GnollHack, which is, besides, a recent arrival on the field and known only to a select minority of dungeon enthusiasts.

DNetHack has a small but extremely dedicated playerbase, who are all quite consistent at ascending; NotDNetHack is often played by only this talented subset of people, and thus has a high ascension rate. As for GnollHack, it looks like the two ascensions were by Malor and Hothraxxa, two very good players with high ascension rates, and the first and second highest z-score respectively. Hothraxxa in particular has ascended every variant on the hardfought server.

[Keith Simpson]

Post by Loggers VIII
I'm late to this, but wanted to comment on some of the reasons for the ascension percentages in different variants.

Post by CSS Dixieland
Of 1630 games played in SpliceHack, not even one has been ascended.

SpliceHack is largely not played anymore; I highly suspect that most, if not all, of these deaths come from post163, as it seems that is one of their favorite variants to die in.

Post by CSS Dixieland
Some other factors may account for such disparities in results, from four per cent to virtually zero ascensions. NetHack version 3.4.3 was for many years the only official version available, meaning that plenty of detail in game mechanics is known to a vast number of players. Once the Dev Team decided to take action and release NetHack versions 3.6.x (NetHack version 3.5 never was official), a number of dungeon devotees started to migrate to them, but version 3.4.3 is still very well known.
However, that explanation does not account for NotDNetHack, which is not so well known. Let alone for GnollHack, which is, besides, a recent arrival on the field and known only to a select minority of dungeon enthusiasts.

DNetHack has a small but extremely dedicated playerbase, who are all quite consistent at ascending; NotDNetHack is often played by only this talented subset of people, and thus has a high ascension rate. As for GnollHack, it looks like the two ascensions were by Malor and Hothraxxa, two very good players with high ascension rates, and the first and second highest z-score respectively. Hothraxxa in particular has ascended every variant on the hardfought server.

I also ascended GnollHack during junethack this year (you can see it if you look at my individual page for K2 or teamsplat's clan page) but for some reason doesn't show up on the page that Loggers referenced.

And also as Loggers said, post163 skewed every single variant and vanilla nethack result. His sole purpose in playing is to die via falling off of their mount. Can't really count those as a serious attempt. So for example, if you look at total splicehack games played (1630) vs the number of games post163 played (1566, under the account 31PwnMoat), only 64 splicehack games were attempted seriously. May even be less if some of those 64 were just players getting their minimum 1000 turns in to get credit for trying a variant. Point being, the total number of games played for every single game that junethack covered for 2023 is wildly inaccurate if you're looking at it from a popularity or serious attempt standpoint.

I've asked bhaak to filter those games out so variant devs could get a more accurate idea of who played their game seriously, but he hasn't yet afaik.

[Erik L]

Post by nabru

Post by Janis Papanagnou
Personally, I think, I'll try out SpliceHack (10.4% "popularity") next.

FWIW, SpliceHack has a bunch of bugs and isn't in active development
(the main dev quit working on it).
--
nabru

You might notice a lot of Splice stuff in Hackem. I really like a lot of stuff in splice, not everything, but since it's run into inactive status I have had a goal to port over most of the content from it. I have also fixed innumerable bugs from said content as I have with slashem and slashthem.

Grenades have had lots of issues, manticore spikes have been a thorn in my side. I've been able to fix up scrolls of cloning and time so they work better, vents have a bit more play to them (you can drop bombs down them, and sometimes bugs crawl out of them now), masks have had a few issues which needed cleaning up. I also borrowed a bunch of skills from splice and turned them into techniques. I don't mind the skill feature, but for hackem it just worked much better to transform them into techniques. Like the sunder and tumble skills, I gave them to the pirate since they didn't have much for techs and they seem kinda piratey.

I still need to port over the dragon rider and cartomancer roles, and the level generation is really nice in splice, like the drawbridge levels, and the occasional tapestry or level name you get. Stuff like that adds a lot of flavor. The major thing I don't like in splice is the leveling-multiclass thing, I'll never port that because I believe a role should have their own set of weaknesses and strengths and that system dilutes that concept. It's also really annoying to deal with a popup after every level.

I actually discussed maybe taking over splice and fixing all the bugs with Kes, but sadly I don't think I'll have time for a while. Maybe someday?

[Erik L]

Post by Janis Papanagnou
Meanwhile I played quite some variants; one problem I had was that many
variants had features that were overlapping, partly or fully, partly
implemented (slightly or completely) differently; in other words, that
there was no overall consolidated concept. That's fine if a variant has
a community that constantly plays just this variant or popularity.

Can you elaborate further on what you mean by "overall consolidated concept"? If I'm trying to do one thing with hackem it's to refine some of the features from other variants that felt unfinished. Can you provide any specific examples?

[Janis Papanagnou]

Post by Erik L

Post by Janis Papanagnou
Meanwhile I played quite some variants; one problem I had was that
many variants had features that were overlapping, partly or fully,
partly implemented (slightly or completely) differently; in other
words, that there was no overall consolidated concept. That's fine
if a variant has a community that constantly plays just this
variant or popularity.

Can you elaborate further on what you mean by "overall consolidated
concept"?

Hi Erik,

I had made two attempts to describe it, and both posts became overly
long, yet incomplete. So I decided to reduce my answer to a minimum
by just quoting a part of a statement of yours from this thread:
"[...] I have had a goal to port over most of the content from it."
(And you mentioned that in context of SpliceHack, SlashEM, SlashThem,
and earlier I seem to recall you also mentioned EvilHack as source.)

My subjective impression is (and YMMV) that a lot of the variants,
also the above listed ones and the derived ones, got overloaded by
features (and in parts not that balanced).

Post by Erik L
If I'm trying to do one thing with hackem it's to refine
some of the features from other variants that felt unfinished.

This is a good thing, please don't get me wrong.

Post by Erik L
Can you provide any specific examples?

My humble personal preference - so please take it as that, and ignore
it otherwise - would be a variant significantly trimmed by abundances
and remaining features adjusted to each other to get a balanced game.
I spoke about concrete imbalances in the past, and by abundances I
mean all the stuff; monsters (types/species and numbers), objects
(specialities and numbers), artifacts, [hard coded] dungeon levels.
Ideally a consolidated version would also streamline user interface
(configuration files, menu handling, sensible defaults), fixed bugs,
or course, documentation (an updated guidebook with all necessary
basics, including new mechanics and other complex or non-obvious
additions), and so on.

This was not meant as criticism of your (or others') variant or work.
Just a brief attempt to answer your question, based on the thoughts
and experiences after having played a couple variants.

It's difficult and heavy-duty if there are many non-converging own
development lines...

Janis