Weapon ranking

[heckhoff]

Edit - action plan:

• Setup database and calculate avg. damage (Aufbau ChatGPT)
• Eigenschaften? SR?
• Score based on avg. damage, consistency (less SD = better) and weighted (ChatGPT) properties
• Vis in table format
• Interactive vis in Shiny with similar sliders
• Avg. enemy stats (without tiers, without occurrence probability, just scraped and averaged. Only Bestienmeister or + GRW?)
• Hit probability and player attributes
• Fully interactive vis in Shiny
• More realistic enemy stats (tier-based, assume skewed distribution of encounters)
• Other advanced considerations

Overall "score" as well as usage area based comparison (e.g. highly armored individuals) Score calculation may depend on average enemy calculation as well as their trait distribution (see #4), perhaps a damage matrix for each could provide sufficient input for a ML-based approach. This may rely on an SQLite or hosted database

Can later be extended by personalizing ranking based on character attributes that influence hit probability (consider critical hit/miss outcomes).

Issue of fight back and forth with attrition and speed of death is considered, but unsure about implementation. Fight dynamics like ranged weapon in melee poses problems as well - probably can't be accurately accounted for. Adequate minimal working solution can be simulated back and forth with hit probability, damage (accounting for armor, shield, ...), weapon speed, health pool and attrition from each side. Consider not only average damage, but also standard deviation and its effect

Also considered is the weapon utility and CC capability. Probably custom weighted weapon properties, but again very hard to implement (advanced ML like reinforcement learning fight simulation approach is most likely wishful thinking). Most pragmatic yet boring approach would be a user-defined fixed weighting based on suspected worth of property (e.g. vielseitig > wurffähig). Perhaps allow weight customization.

[heckhoff]

Ranking needs separate one- or two-handed use for qualified weapons

[heckhoff]

Probably overall quite computationally expensive if done properly (avg. player stats, avg. enemy stats, ...), however complex. Can be calculated in advance for specific attribute queries, but then definitely necessary to represent in a database - useful anyways, create issue once needed. How would schemas be organized? GRW vs. beginner's rules? Weapon stats vs. player stats vs. enemy stats, ...?

Does it make sense to rank based on varying player attributes? Probably does due to the large impact of hit probability, but how?

• Idea 1: Calculate all scores for all combinations of attributes, however unlikely. Risk of painting inaccurate picture, but all weapons are treated the same nonetheless
• Idea 2: Assume average attributes. Hard to do without data, could be done roughly according to Heldengrad upper bounds and simple assumptions

How would these ideas compare? Each can be tested out if desired

[heckhoff]

Vis ideas:

https://rpg.stackexchange.com/questions/167980/anydice-how-to-calculate-opposed-roll-probabilities-for-an-unusual-mechanic