A new take on a Bard role, generally themed around leading a party of other adventurers through the dungeon, with some other additional ideas:

• The bard and healer cannot both be focused around petmongering, so one of them would have to change. Most likely the healer, as the bard’s traditional abilities are only good on pets.
• Instead of the standard little dog or kitten, the bard starts with a “party” of 3-4 player monsters, who are randomly chosen from other roles.
  • This would make the primary strategy focused on having the party do the fighting for you, letting them level up, providing them with weapons and equipment, and not letting them die.
  • Healers and archeologists, and possibly tourists, don’t make very good additions to the party, and so should perhaps be ineligible.
  • Party roles could be randomly chosen from distinct sets of “strong melee fighter” (Bar or Cav or Kni or Val), “ranged combat” (Ran or Rog or Sam or possibly Tou), and “spellcaster/cleric” (possibly Hea or Mon or Pri or Wiz). These could also just be hardcoded.
  • The party will not start with the full starting inventory of their respective roles. They will generally have the same weapons and armor, but will probably not have any of the scrolls, rings, wands they normally start with, and starting spellbooks should be either eliminated or unrandomized. In other words, you shouldn’t be incentivized to slaughter your allies for their stuff.
• Healer strategy should shift away from petmongering - the easiest way to do this is change the protection formula in such a way that the protection racket doesn’t exist; so that the player isn’t incentivized to keep XL as low as possible until they can buy protection.
• Difficult for Healers to find a new niche. (Avoidance of combat is an option, but dtsund’s Class Overhaul Proposal calls for Archeologists to fill that niche). Perhaps something that focuses on pacifying monsters by healing them, but this has issues: you shouldn’t be able to ‘‘tame’’ by healing since that turns Healers back in the direction of acquiring lots of pets, and most monsters start at full HP and don’t need any healing.
• Bards (and probably all players) should be able to equip their pets by #looting them, or a similar mechanism. (This would also apply to mounts.) Intelligent monsters will prefer player-given items above all others and will never willingly replace them for other gear. Possibly make this behavior dependent on options, because some players may want their pets to be pragmatic in their gear choices.
• Player monster pets should perhaps not be able to detect curses.
• Songbooks, + class items which contain a song that the bard can learn.
  • Balancing the songbook generation rate against spellbooks might be tricky, because although bards can still cast spells and other roles can play songs, each is not very likely to have that much use for the other set of books. Hardcoding different probabilities based on whether the character is a bard or not doesn’t seem very clean.
  • Possibly the bard should be able to discover songs by trial and error, although simple brute forcing is definitely not the way to do it. Maybe all spellbooks start identified.
  • The bard could need to practice the song by using the songbook in order to restore “song memory”. Song memory does not work quite like spell memory, it works essentially like direct success rate. There might be some mandatory delay between practices, meaning you can’t take an unlearned song and practice it up to 100% immediately. Or there could be a possible failure of practice, maybe a dexterity penalty (“Dumb move! You strain your fingers.”) to serve to prevent practice-spam. Playing a song would serve to
  • Songbooks do not eventually go blank, and can be used for practice indefinitely.
  • Or to keep it realistic with learning actual songs, you need to have the songbook on hand early on when you want to play it, but as you play it more and gain practice, you can eventually play the song without it anymore.
• The bard should possibly be able to sing some of their songs without requiring an instrument.

See also: Bard implementations.