Experimental version merge?

[Seravy]

I think it's about time we make a decision on which of the experimental features we want and make experimental the new "main" version.

So without further ado, we need to consider :

1. Stability.
I think this is fine - I'm not getting any more crashes and weird bugs than in the normal version, and enough time have elapsed so the major problems are most likely all fixed (such as when the AI used Spell Binding without knowing it)


2. AI Doomstack.
Having played against it, I feel the AI's doomstack feature is a great success. Those stacks keep a lot of my troops busy, and if unexpected, even destroy cities. I need significantly more troops to defend myself and this makes attacking the AI a lot more difficult - something the same units left in garrisons never achieved.

3. AI More cautious attacks
Although it's easier to "scare" an AI away from attacking my cities, this didn't lower difficulty, in fact increased it. The troops not daring to attack my heavily defended places will head towards a place with thinner garrisons, where they'll cause trouble. Maintaining a full 9 garrisson of strong unit everywhere is a major strain on economy, but leaving weak points is too risky. Also, the doomstacks just don't care - they are powerful enough to attack anything, and win the battle most of the time. Oh, and the units not finding a city to attack keep staying around and will either steal my lairs and nodes, or cause general inconvenience by preventing use of engineers, shamans, and organizing my troops. In the end I have to attack them anyway, giving them the first turn to cast and shoot.
So I feel this change is also a success.

4. AI Overland Spellcasting
Can't say I played enough to be 100% sure each detail in this is perfect yet, but having easy to modify source code is a big improvement, as we can address problems much easier now. At the very least, being able to use "naval" spells, disjunction priority depending on the intended target, time stop no longer being a curse, "cheap buff stacking" being an actual spell group and tactic, and so on should be massive improvements. Out of these, I especially do not have enough experience facing Disjunction - I usually didn't get a chance to depend on a large number of global enchantments in my games to test this.

5. New banishment rules
This one is where I have most doubts. At the very least I can say I feel less tempted to hit capitals first, which is a step in the right direction. Doing so is still a good strategy and is well rewarded, but not worth it if the capital is very well defended and beating it would cost most or all of my forces - this hasn't been the case before, as banishment was so severe, it was worth it even if it meant losing half my armies - while the wizard couldn't use combat spells, I was able to conquer anything with just a few swordmen tier troops and combat spell spam.

6. AI difficulty
This is the hardest to adjust but I think adding a 6th difficulty level (see the other thread) should make the game playable no matter what, as it gives more fine-grained options for the player to pick the difficulty they can comfortably play.

7. Suppress Magic is strength 500 counter effect.
Not sure about this one, while this reduction is needed to retain original functionality (most very rares are unaffected but commons, uncommons and rares suffer), it might make it took weak. (a 200 cost spell will still have a 40% chance to work which is not that low) I haven't faced this spell in any game and when I had it, I didn't cast it as I was already winning the game by far.

8. Cheaper summoning costs for rares and very rares.
Also having doubts - it certainly plays better and the AI is likely to also have them and even build doomstacks of them so it can't be said it's a "no brainer" tactic as I originally was worried about - yes, I might have that 9 archangel stack but getting it is way slower than having 3 normal stacks each carried by one archangel, and this latter has 3 times the map coverage and mobility. In a situation when I have to race against an enemy in summoning creatures, waiting for my 9 doomstack just buys enough time for the AI to produce two similar doomstacks of their own. Still, in the scenario of having these creatures vs opponents who don't have them, summoning is an easy way to win, and this change makes Spell Blast even more crippling then before, as it prevents the use of these units until Disjunction is researched and high casting skill is reached.

9. AI frontier town choice for summoning
Clearly works better, I've had my lost city filled by 2 Gorgons by the time my armies reached it to take it back, while the doomstack already went to attack another one.

10. New racial unrest table, no change of race by Move Fortress
I don't have enough experience with this, but it should be an improvement considering the move fortress abuse was a massive unfair advantage to the human payer

[Nelphine]

My main concern right now is still overland casting skill from enemies. Having 9 volcanoes on your capital before you have a second city is brutal. I think with the doomstack rules, the AI no longer needs nearly as much casting skill to make summons offensively dangerous, and all the non summons still get the boost from overland casting.

Banish rules I think definitely last too long early game, and with casting skill of AI on extreme or impossible, reduce to 1 turn too quickly. (Due to my primary concern of overland spells, banish still makes the game a virtual walk over if its done early, even if you need more than spearmen to win now.)

Doomstacks are amazing.

Reduced summon cost: I think this working well. Even with my concern of overland casting skill, I'm not being overwhelmed by summons.

I havent had a chance to face natures erarh or suppress magic, but what you consider s problem, I consider a good thing. Before, suppress magic basically reduced the game to globaks and summons. Now, top end curses and buffs have a chance.

The other changes you mention I consider minor, and working well.

One other thing - due to doomstacks, I think AI bonuses across the board (production, gold, power specifically) can be reduced - the doomstacks ensure dangerous attacks will happen regardless of garrisons, which means the AI doesn't need to have the sheer quantitative advantage they used in order to be a threat.

One thing I've been noticing recently - the AI doesn't use any combat spells in certain battles till turn 9 or 10 or so - and then randomly uses reaper slash every few turns. I've never seen that behavior before.

[Seravy]

Volcanoes?
That's weird. The AI shouldn't be able to turn hostile before turn...25 I think in the current version. And by then I would expect people to have other cities (either outposts or neutrals). I can only imagine you attacked the wizard yourself. Even then volcano is unrealistic, it's an uncommon spell and not high on the AI's research priorities. The AI shouldn't have it until much later (unless it was in treasure).

I consider early game banishment a non-issue. If your capital, which is your largest city and probably contributes like half your economy and/or troop production, falls, you've already lost the game.
Midgame (turn 60-120) is a pain, but by then you are supposed to have enough troops to protect the capital, and the AI isn't going to have unstoppable threats yet (or if they do, you lose anyway)

There was this first klackon game where I lost my capital to my own mistake, and if returning was faster I could have still played it but that's about the only scenario I can imagine where this matters - having a very good start, not defending the capital properly, and then returning. And honestly, even if it's not 40 turns just 20 or 15, not being able to cast overland spells that long in the early game (no bears, no nothing) means losing.
There is another side of the coin - if it's still the early game, why bother continuing from a horrible board position? Starting a new game is faster.

Late game banishement, 1 turn is unrealistic. If a wizard can do that, they have already wiped everyone from the map, including you. That much skill can summon 4 Sky Drakes each turn on Extreme, 5 on Impossible. A sorcery wizard with all the skill boosting effects and a large empire might manage 2 turns but if they are still alive at that level, you have greater problems to worry about than their capital. I've never won a game where a Sorcery wizard already reached 500+ skill, I try to take them out earlier. An average wizard will take 3-6 turn to return (assuming we are considering 1415-20 the late game - yes, if the game drags out until 1430 on extreme then they might return in 2 turns. One turn is still unrealistic even then).

Note that AI casting bonus does not apply to Spell of Return.

I'm convinced the new Suppress Magic is going to be better, now that you mention that.

About AI not casting spells - they don't have enough mana crystals to afford blowing it on "unimportant" battles. This is a rare situation as in this case the AI will be producing more mana, and will eventually have enough to be able to cast. In a certain range, the AI will only cast spells that have a priority over a certain amount - the priority of that Reaper Slash might shift enough during battle so that in some cases they can, in others they cannot use it. (priorities also have a random component btw)

[Nelphine]

I'm talking early game AI banishment due to a barbarian AI slaughtering the defending ai sprites. Still happens about 1 in 5 games, maybe 1 in 6. Usually happens in 1403 or 1404.

Late game, on impossible: I have an AI with one city left. In 1407, they have ~240 casting skill. By 1410, they (if I didn't go to war) would have ~500. On extreme it might take till 1413. Put that AI on myrror, and you (almost) can't kill them before then.

I do have to assume you are correct and the game I was thinking of was 2 turns not 1 turn.

But even still, 2 turns in 1415 compared to 25 turns in 1403 is too much difference.

I'd like to see 10 turns at most and 3 turns at least.

[MrBiscuits]

The troops not daring to attack my heavily defended places will head towards a place with thinner garrisons, where they'll cause trouble.

I've noticed this and my problem with it is that the AI seems to have perfect knowledge of my least guarded cities and heads straight for them regardless of whether the human would be able to see their cities if the situation is reversed. I know the AI probably needs that knowledge to work properly, but it seems unfun when they exploit it in such a direct way.

Also with things like disjunction, curses and corruption/volcanoes happening more often, the extra power and casting spell the AI has just makes it feel like they are cheating casting these all the time. My cities get hammered by corruption very early on and if you are playing a race that doesn't have shamans then you are in real trouble with nothing much you can do about it. 

I wouldn't mind so much if the curses meant they had less summons when you get to their cities, but they still have loads of them as well. It makes the game harder but also less fun in my opinion.

[Arnuz]

To fix these points we'd need to have multiplayer... Unless we get some google deepmind machine to play as the AI we're never going to be able to have a challenging game within the frame of the possibilities given to Seravy by MoM - CoM is already a fantastic result from those starting conditions!

[Seravy]

I'd like to see 10 turns at most and 3 turns at least.

Me too but casting skill does not scale that way and unless we hardcode an exception, it does depend on casting skill.
The problem with the exception is, players will then ask "Why can't I cast this faster if I have higher casting skill when it's a spell?".

About corruption and volcanoes, I believe main of the main reasons for the new AI overland casting system was to gain better control over this, and this latest update should have reduced the preference of using them quite a lot.
However, keep in mind that these are 30-50 mana spells - even without casting advantage, it's easy to spam them once or twice a turn early, or many times later, so the only real limiting factor we have is through adjusting AI preferences.

[Nelphine]

But we can limit the overall overland casting skill so that at least if the AI uses spells they don't get so many other spells. So even a maniacal death wizard using lots of curses... Would have noticeably less summons. And since summons are cheaper, and doomstacks work effectively, reducing overall casting skill shouldn't reduce the AI ability to be a threat due to not getting enough summons, which was the primary driving factor for such high overland casting skill in the first place.

I do think we should hard code an exception for spell of return. Then remove the casting cost in the description and call it special. Its not a spell most players ever read the description on anyway - my guess is most aren't even aware it has a casting cost.

However, I'd also be willing to make the had coded exception strictly apply to AI if you felt it would be a big problem for players - player banishment is a much less common occurrence without the player simply starting a new game.

[Domon]

well, i like almost all changes, i'd only suggest trying with slightly highrer summoning costs (75% of original value?) and adding maybe a secondary effect to suppress magic?

i would also like as much "tactical tricks" as possible at the earliest difficulty level.

if after all these changes some assets end up being too obnoxious (vulcanoes?) i'd say change the asset, not the general rules.