wesnoth-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Wesnoth-dev] About the ulfserker


From: ott
Subject: Re: [Wesnoth-dev] About the ulfserker
Date: Sun, 8 May 2005 20:54:27 GMT

On Thu, May 05, 2005 at 06:36:41PM -0500, David White wrote:
> It seems that in 0.9.1, the Ulfserker is seen as underpowered in the 
> multiplayer community.

>From 0.9.0, we removed berserk-on-defence, which made the unit very
powerful in the eyes of some people (though I am not completely convinced
of this).  We then lowered resistances to compensate, and now the unit
appears underpowered.

The way I understood the comments by active MP players, the main
gripe with Ulfs right now are the lowered resistances. To consider
a commonly cited example, it is claimed this will make an Ulf lose
against a Thief in grassland, when attacking.  As far as I can tell,
the berserk-on-attack-only issue is relatively minor in comparison.
If we raise resistances somewhat, the issue is likely to shift to
berserk-on-defence again.

While composing this message, I realised neither side really had any facts
to back up their arguments, just general comments based on common sense.
I therefore thought about how to calculate the exact probabilities of
units killing each other in a single skirmish, and put these into a perl
script "prkill" so that we would have some facts to inform this debate.
See http://www.wesnoth.org/forum/viewtopic.php?t=5867 for the first
version of this script.  I also attach it here for ease of reference.

In light initial testing, the output of prkill for Ulfserkers attacking
other units like Thieves and Drakes seems to confirm some of the concerns
of people like Dragonking.  In particular, the presence or absence of
specific traits like strong and resilient can dramatically change the
probabilities of which unit dies.  Simply put, berserk amplifies the
effect of traits and can lead to non-intuitive %-to-kill outcomes.

Instead of talking in generalities, I think it would be useful to look
at specific combinations of stats and see how they perform, especially
in the presence of traits.

I would also appreciate critical review of prkill, especially from those
people here who are not intimidated by basic probability theory.

-- address@hidden

Attachment: prkill11.zip
Description: Zip compressed data


reply via email to

[Prev in Thread] Current Thread [Next in Thread]