Thanks for taking the time to compile those random wizard stats. They are interesting. When I was developing my algorithm, the single most important stat I kept in mind was "average number of books (across all realms) per game". It ended up being ~7.8 books out of 11 total picks, or 71% of total picks. If you have time to tweak your stats generator, I'd be interested in seeing what your average books per game is out of 12 picks. I've attached the distribution stats of my algorithm to this post for comparison purposes. Certainly your distribution doesn't have to match mine, but it is interesting to compare the two, and it might give you ideas of how to tweak things to make the generated games more interesting for players.
For Archmange and Mana Focusing, if you want to keep or change the book requirements, I'd tweak the multiplication factor until the distribution more closely matches the other retorts at ~9%. Or you could remove the requirements for those 2 altogether. Either approach would be fine.
One curious thing I noticed about your stats is the big difference between 0 retorts (45%) and 1 retort (12%). That's a big % of games without a single retort. Looking at the code it looks like every iteration of the main loop there is a 1/6 chance to pick retorts, which would explain it.
Also, are you planning to add spells to your random game generator? Btw congrats also on your 2000th post!
(January 4th, 2017, 19:24)Tlaloc Wrote: Thanks for taking the time to compile those random wizard stats. They are interesting. When I was developing my algorithm, the single most important stat I kept in mind was "average number of books (across all realms) per game". It ended up being ~7.8 books out of 11 total picks, or 71% of total picks. If you have time to tweak your stats generator, I'd be interested in seeing what your average books per game is out of 12 picks. I've attached the distribution stats of my algorithm to this post for comparison purposes. Certainly your distribution doesn't have to match mine, but it is interesting to compare the two, and it might give you ideas of how to tweak things to make the generated games more interesting for players.
For Archmange and Mana Focusing, if you want to keep or change the book requirements, I'd tweak the multiplication factor until the distribution more closely matches the other retorts at ~9%. Or you could remove the requirements for those 2 altogether. Either approach would be fine.
One curious thing I noticed about your stats is the big difference between 0 retorts (45%) and 1 retort (12%). That's a big % of games without a single retort. Looking at the code it looks like every iteration of the main loop there is a 1/6 chance to pick retorts, which would explain it.
Also, are you planning to add spells to your random game generator? Btw congrats also on your 2000th post!
Interesting to see how large effect not being able to have Life and Death at the same time has on the numbers.
Spells are intentionally left out. I consider that part of the game itself, kinda. Picking the wrong starting spells can turn most wizards into unplayable garbage. If you want to play a random wizard on extreme (tho with how quickly the AI improves, it might even be true for hard now), having proper spells for the start is essential. I could easily add the option but I'm not sure if it's a good idea.
I was able to improve the retort distribution by adding a few extra lines of code. It's something like 22/22/20/11/4/1 now. I decided to get rid of the restriction on Archmage and Mana Focusing.
Anyway, here is the update :
Quote:3.02
-AI Web spell targeting : If city walls are present, prefer the unit at the gate – this preference is lower than the preference for targeting a flying unit.
-AI Vertigo, Warp Creature, Shatter priority : Do not use if target has 4 or lower health – direct damage is almost always more effective on those!
-AI Healing, Healing Charge priority : +10 is added to the priority if target is enchanted by Invulnerability.
-Fixed bug : Sometimes spell cannot be offered as tribute in diplomacy
-AI combat : The AI can now use the stalling tactic with invisible units if the enemy army has no unit immune to illusions, in addition to using it with flying or fast units. This invisible type stalling is allowed regardless of enemy ranged units.
-AI combat spell priority – Resist Magic : The additional priority if the enemy has Death books is not applied if the target is immune to Death magic.
-New procedure for AI farmer decisions. The AI now considers food gained, production lost and various other factors when selecting which city to use for producing food and how many farmers are needed in each.
-Updated the random game generator. Chance of no retorts or 4+ retorts is lower. Obsolete retort requirements are no longer observed.
-Archmage and Mana Focusing no longer requires having 4 spellbooks in the same realm.
-The AI now recognizes enemy teleporting/merging units as infinitely fast when deciding to do a movement speed based stalling tactic.
-Fixed bug : the human player can sometimes move their teleporting/merging units to a tile that already has a unit. (bug introduced by a previous teleport bug fix)
Yes having mutually exclusive life/death books was probably the biggest hurdle I had to overcome to achieve a good book/retort distribution.
I agree your dramatic AI improvements make starting spells much more important than before and that randomizing them would likely lead to frustration at higher difficulty levels. It would still be nice to have the option to generate them for players that wanted the additional layer of difficulty, but I guess they could generate them on their own if they really wanted to.
I saw the new lines of code for the retorts. Very nice, terse, and efficient. Once your algorithm for the random game generator has stabilized, would you be ok with me implementing it on my site? That would solve the inconvenience Hadriex noted in his video of having to write down the picks in order to enter them in the game. If not, that's fine too.
(January 5th, 2017, 18:08)Tlaloc Wrote: I saw the new lines of code for the retorts. Very nice, terse, and efficient. Once your algorithm for the random game generator has stabilized, would you be ok with me implementing it on my site? That would solve the inconvenience Hadriex noted in his video of having to write down the picks in order to enter them in the game. If not, that's fine too.
I actually have no idea what Hadriex's problem is, I mean, how is alt-tabbing between a DOS/CMD prompt and CoM or a browser and CoM any different? It's not like he is forced to run CoM in the same window as the generator...
Either way if you feel it's necessary sure, you can make one as long as you keep it updated - I'm not planning any more changes to it but we can never know what happens in the future, maybe new options get added or something.
Possible bug, I invaded a neutral Draconian town with a group of beastmen, mostly archers. Towards the end of the battle, the Draconian units stopped attacking... or doing anything really.
Edit: attached a save right before the battle, in case you needed it.
(January 6th, 2017, 22:43)RobertusAmor Wrote: Possible bug, I invaded a neutral Draconian town with a group of beastmen, mostly archers. Towards the end of the battle, the Draconian units stopped attacking... or doing anything really.
Edit: attached a save right before the battle, in case you needed it.
It doesn't seem to happen all of the time.
Bug. The AI isn't supposed to use the flying stall tactic against ranged units unless they are low on ammo.
It was caused by normal and below difficulty skipping the ammo check by accident in addition to the intended parts.
Fixed version is up :
Quote: 3.02a
-Fixed bug introduced in 3.02 : On normal and easy difficulty the AI flying stall tactic skips checking enemy ammo.
Focus Magic increases breath attacks by 3. It's not the best way to use the spell but if there is no better target, it's still better than cancelling the spell and losing the mana.
Focus Magic increases breath attacks by 3. It's not the best way to use the spell but if there is no better target, it's still better than cancelling the spell and losing the mana.
Ah, wasn't aware of that.
It has been a few months since I last played the game and I can say the game got more stable and the AI even better. Hard isn't an easy win for me any more. Very nice!
I am very worried now about what air elementals do in an enemy sorcery node? they're fast, they're flying, they're invisible, there are lots of them, there's no enemy wizard to command them and they don't have to be summoned because they appear as normal enemies.
I haven't started a 3.02 version game yet though to test it, but does this happen and if so is it intended? sorcery nodes are already quite tough without requiring an illusion immune unit.
