Sprites

[Seravy]

As Catwalk doesn't seem to care anymore, I've used my own calculator to have an estimate of the treasure a player can harvest using Sprites.

Any lair containing only the following monsters, but any number of them, is counted as valid, others as invalid. This is a rough estimate - some lairs containing only valid monsters but too many might be impossible without sacrificing a large amount of sprites, while others with only a single invalid monster might be doable anyway with fewer losses.

Code:

Hell Hounds
Fire Elemental
Chaos Spawn
Skeleton
Werewolves
Zombies
Unicorns
Guardian Spirit
War Bears
Sprites
Cockatrices
Great Wyrm
Phantom Beast
Phantom Warrior

Only lairs on the same plane, no further than 15 tiles in either direction are counted - it's safe to assume the rest won't be discovered early enough to provide early game advantage, if it's even still there by the time the player finds it. Nodes are counted as being worth 500 extra for the power income. Items are counted at the value you get if you turn them into mana.

Out of 8 save files per plane, the following total treasure is counted on average per game file :
5900 on Arcanus and 2900 on Myrror.

Based on these numbers, we need to decide if this is "balanced" or "overpowered".

Each sprites unit costs 80 to summon and an additional 2 to maintain per turn - assuming a 30 turn lifespan, that's a total cost of 140. To successfully exploit all the lairs in such a large distance in the actual early game, including the hard ones, at least 15-18 sprites are necessary (1 big stack and 2-3 smaller ones) -  which costs a total of 2100-2520 mana plus a lot of casting skill and time.

I'd say it's outright not worth it on Myrror in the early game - unless very good artifacts, heroes, or books are found, it barely earns back the investment and leaves the player with a summoned army that cannot be used effectively against wizards or to protect the nodes gained.

On Arcanus it works much better - there profit is at least double the investment. However I don't think this is enough to be considered gamebreaking, especially as having only sprites makes the player quite vulnerable in the early game and conflict on Arcanus is much more likely. If anything, this is good because it rewards the more experienced player - you have to know exactly what is safe to attack and what is not, and have to be very careful - failing to scout a secondary monster can easily waste a stack of 9 sprites.

These numbers also show how this got affected by the monster pool changes  - while high difficulty games had an additional monster pool multiplier, lairs on Arcanus followed numbers closer to what is on Myrror now, while Myrran monsters were even stronger. It's safe to assume stronger monster pools result in a lower effectiveness of the Sprites strategy.

Overall, this much profit, or more if playing some retorts that reduce Sprites costs won't be enough to match the AI bonus on high difficulty, so I don't think there is a problem. Conjurer might be an exception even with all the changes done to cost reduction stacking, but I don't want to nerf an entire unit that is fair for a single retort. If anything, removal of "25% less maintenance" from Conjurer might be the way to go - sprites profit way too much from that.

The calculator is in the tools.zip already - sprites.pas and exe - feel free to test other game files. The expected input is 8 game files, numbered SP1 to SP8.Gam.

[RefSteel]

It doesn't surprise me that it's difficult to profit from casting 15-18 early Sprites; my understanding was that the problem is the return on one casting of Sprites. You don't send out 3-4 smaller stacks plus one big stack; your very first Sprites unit (and then the first pair etc. where needed) is already taking the easy targets before the big group can be assembled. The more Sprites you already have on the board - especially once you have enough to take out your best/hardest target, whether that's nine Sprites or six or one - the smaller the return for the next one. It might be that you can cast nine Sprites and still get a much better return for the investment than anything else you could do with the mana, but even if that weren't true, I wouldn't discount the possibility that early Sprites are overpowered.

[Seravy]

That's certainly true but if we go with that assumption, all targets that require more than one-two sprites has to be removed from the list.

Chaos Spawn, Great Wyrm, Werewolves, Unicorns and cockatrices all need more to kill, and in case of the most rewarding Great Wyrms, at least 5-6 are necessary to even have a chance of killing one per combat. Even a War Bears or Nagas usually needs 2 Sprites to kill. Units that can shoot back or reach sprites such as other Sprites or Hell Hounds are also impossible.
So while what you say is true, in that case the treasure obtained will be a fraction of this, especially as the game runs out of "early game" period which we are trying to measure if it takes too long to kill the monsters.
While I admit a single Sprites unit (or more realistically two, one fails to kill almost everything 3 out of 4 tries) for 1000-1500 treasure is a really good deal, those weak monsters can be killed easily with any other tactic, such as a group of bowmen or 1-2 early wolf riders/berserkers/etc. I don't think being able to kill easy monsters only is risk to the game balance and you certainly need larger stacks of sprites for the strong ones - Phantom Beast is probably the only "strong" monster that can be killed by a low number of Sprites but that dies to anything, even the aforementioned bowmen.

[RefSteel]

Makes sense; it may even be that the rewards per Sprite even go up per unit (depending on targets available) until you get 5, 6, or even 9 before you reach a point of diminishing returns; I don't know what the optimal number is, though I'd imagine it's closer to 9 than to 15! Even with just a handful though:

1) You can get your first (and second and third...) Sprites unit out significantly sooner than you could get early wolf riders/berserkers/etc or even bowmen, right? And thus start reaping rewards (and get the snowball rolling) sooner. (Including getting those very units faster than you could without Sprites finding treasure to fund purchases.)

2) Sprites can natively cross water, so they'll be able to attack more lairs sooner than other strategies.

3) For many encounters, Sprites are at zero risk of taking any damage at all. The other aspects contribute, but this is the reason the Sprites opening is overpowered! Bowmen (for example) might be able to kill a Phantom Beast before it can reach them, but then they'll get eaten by its attendant Nagas (or even Phantom Warriors if you didn't bring enough Bowmen) because they can't kill them all in time and can't fly, so can't just time them out. If Sprites fail to take a lair/node/neutral on the first try, they've still killed one or more of the defenders (or a mid-sized group has had their x% shot at the one Great Wyrm) and they're still feeling fine, ready for round 2, and eventually to just move on to the next target once they've inevitably won.

[Seravy]

1. Yes, but the difference is not significant now that there is a starting sawmill. You can have a Wolf Rider done on, like, turn 6-7? Bowmen, even on turn 2 but you need more than one of those. Some units like Berserkers are slower but usually pack a lot more punch than sprites in exchange.

2. I'd say that's a huge benefit, except with 20 mana for Water Walking, or ~200 gold for a ship building and boat to carry any amount of units, not really. If Sprites were not chosen, Water Walking can be instead so comparing with the assumption of having the spell is fair. Faster travel certainly helps which is why we can consider such a large radius, otherwise 30x30 would be impossible to cover. However, large distance is not all happiness - if you manage to take a node (or neutral city) but it's closer to an enemy wizard, you either need to have a massive amount of sprites there to defend it (too expensive so not really an option), or they'll eventually steal it. Sprites in small numbers don't do well against the AI because they will be bombarded with direct damage or webs -they'll stop a few stack (especially if the node counters spells for you) but eventually they get nuked.
Finally, exploring a large area means faster contact with other wizards - while this can be beneficial in trading, in many cases it increases the risk of getting attacked. Not as badly as before the "militarist" effect got delayed, but still it's there.

3. This is true, but cannot be objectively measured. In many cases your other units also don't take damage (enough bowmen will kill the beasts and/or nagas) or are cheap enough to be disposable (You can get 3 Skeletons for the cost of a single unit of Sprites). Also, time is a factor - if the Spites do clear the node but spend 9 turns on it (one monster at a time), then do that for 2 more lairs, we are already out of the early game we are trying to balance. Compared to that, the bowmen might lose 1-2 units but finish the job in a quarter of the time. This is why I assume you need 15 sprites - sure, 9 are enough to clean out everything but to hit all nodes in a radius of 15, meaning a 30x30 map area, you need more stacks. It would take the stack 15 turns just to travel from one side to the other in one direction and to clear everything you'd need to reach all 4 corners and you start from the center. Even with only a single turn per lair (which is unrealistic if any of the "big" targets are included), you'd need far more turns to finish than what can be called the early game with major snowballing potential. (and you have to add the summoning time - yes, the first few units are done right away but to have the 6 stack to be able to hit stronger destinations you need at least a dozen turns, more without retorts)
With the latest retort changes, 1 sprites/3 turns is about the best rate one can reach, so that's turn 18 for even starting on hitting anything powerful and after that you have to still cover the entire 30x30 area to actually get the 5900 calculated. There is no way to do it unless you have more stacks for the easier targets.

So I'd say the ideal number is between 6 and 9 for your main stack - but you need more than one stack to cover the whole area the tool is considering in the timeframe we are considering - too slow and either the treasure will not be relevant enough, or other players steal the targets. Furthermore, I didn't include the additional cost of ~400 mana and skill on casting Earth Lore - without doing that, scouting that whole area is unrealistic (especially using only one stack) and the risk of of running into a nasty secondary monster is huge - Before Earth Lore reveled them, I didn't dare to attack a lot of targets without sending at least a spearmen inside and that's not an option for the entire area.

[RefSteel]

I hear that, but based on Catwalk's actual-game results - achieving an early winning position with Sprites N times out of N (for significant N) on a difficulty where such a perfect result would normally be (pardon the pun) impossible - I think you're describing a limitation of your tool, not of the Sprites: 30x30 is completely arbitrary, as is the "end date" for the "early game," and the number of Sprites needed to grab everything in that arbitrary area in that arbitrary timeframe is irrelevant. The question is how to maximize the amount you can get and minimize the time it takes to get each part, not how to meet a specific benchmark created for convenience of calculation.

I'd like to see a report of an expert playthrough (it's sad that Catwalk decided to stop participating before creating his alternate tool or offering such a report) to demonstrate some of the tactics used to best pull this strategy off, but as I can't supply one myself, not being an actual expert, I don't think there's anything else useful I can add on this question.

[EDIT: Well, apparently I can't stop spamming, so one more thing: I think the alternatives you're proposing are re-using resources. Casting Sprites does not prevent you from building Bowmen (or whatever) - quite the opposite. The bowmen can still capture the nodes they could handle themselves, reducing the number of Sprites you need even for the 15-tile benchmark. Or you can build Settlers or whatever instead while casting Sprites, which you can't do at the same time you build units. And then casting Water Walking on enough regular units to capture overseas lairs seems way less efficient even just intuitively than casting Sprites to get the overseas lairs and using the regular units at home.

Relatedly: Sprites don't suddenly disappear orr stop being useful after the early game, even though enemy wizards can wreck them. The value of exploring lairs may continue to slowly erode, but it doesn't disappear entirely. Also, a Sprites strategy doesn't have to spend all its resources on Sprites; maybe you accept one less Sprite unit to add buffs or use combat spells to existing ones that allow you to take a target that would have stopped basic Sprites. I don't think Sprites are the One Right Thing That Destroys All Else, but they might still be at the heart of an unbeatable strategy.]

[Seravy]

achieving an early winning position with Sprites N times out of N (for significant N) on a difficulty where such a perfect result would normally be (pardon the pun) impossible

He used retort stacking in ALL of those games.

I don't think he ever played a single game where he had to pay full (or at least near full - 10 books do give you a discount) price for the unit without a boost to income and skill from retorts. If your mana/skill budget is doubled and your costs are halved, no matter what you're testing, will be 4 times as effective as normal.
To prove that I even played a game using the same retorts and War Bears and eliminated two wizards and claimed their territory doing that. And that again was too subjective and ended up being wasted time on my part - I think territory taken from wizards is more relevant and he thinks nodes and gold are.

To prevent interference from such game mechanics, (and to save time - playing dozens of games takes too long) we decided to use the tool instead.

I don't think the results can be reproduced without those retorts, but anyone is welcome to try - rules are simple, you're allowed up to 10 Nature books and all other picks on anything unrelated (recommended 2 Chaos books), nothing more. You have to win a significant portion of the games.

Casting Sprites does not prevent you from building Bowmen (or whatever) - quite the opposite.

It does quite a bit if you use Alchemy to convert your gold into mana. Assuming you boosted your skill with retorts, you'll still need additional income to be able to pay for it. When I went for this strategy, I could barely afford a settler. (Mana Focusing used to help with that but it's gone now)
This is even more true as income drops faster than skill when retorts are picked instead of books.

Of course if we go for the "base" situation of 12 books there is no such problem, but in that case you can only summon a single unit of sprites every 4 turns using ALL your mana and skill. Not exactly a gamebreaking rate, is it?

Sprites don't suddenly disappear orr stop being useful after the early game, even though enemy wizards can wreck them. The value of exploring lairs may continue to slowly erode, but it doesn't disappear entirely.

True, but it's not a slow erosion.

The gold or other treasure is no longer relevant enough to put you in a snowballing position. It'll still help, maybe even a lot, but there is a difference between buying 10 new settlers (or wolf riders) on turn 12 out of your gold or doing that on turn 78.

And this assumes they actually live to do it, which in my experience is rare. They have low AI combat value so the AI will attack them with their nearby stacks and kill them off one at a time with direct damage, or worse, annihilate the whole stack with Fire Storm directly on the overland map.

I agree the 30x30 area is arbitrary - I can change it to a lower number, calculate the result and compare it to the cost of fewer sprites if you want. If we reduce the number below 9 we have to restrict the types of monsters though - stuff like "many cockatrices" is not doable with fewer. 9 Sprites cost 1260 - a sizeable investment still.

[Nelphine]

While I understand that you are focused on sprites specifically, I think it's a mistake to not consider it with retorts.

For instance, if you take conjuror and spellweaver, that still leaves you with 9 nature books. Yes other summons are strong, but sprites do all of it by themselves - you don't have to summon warbears and cast water walking, you don't have to use web to hit fliers, etc.

I think conjuror + spellweaver should be your baseline for the experiment as that isn't really excessive on retorts.



The other thing is that using earth lore allows you to hit the best lairs nodes immediately. That means your very first sprite is out taking 2-5 lairs/nodes by himself, while your second 2nd to 4th sprites start out taking a weak lair the opposite direction and then start grouping up, while your 5th-9th go in a 3rd direction, with the idea that those two stacks will meet up around the time they run out of weak lairs, but also perfectly positioned to instantly start hitting strong lairs.

This means you've cleared 7-10 lairs or nodes (and even neutral cities) by the time you've summoned your 10th sprite.

And with only a tiny bit of luck, you get mana from at least 1 of the first 2 lairs (call it 3 to be safe), which means after, oh, your 3rd or 4th sprite, literally 100% of your power income is on skill, so you can summon more sprites faster.


So let's combine these two comments:
You have 9 books plus spellweaver. That's.. Uh.. I don't know. 25ish casting skill on turn 1?

You have conjuror. So 1/sprite per 3 turns. So on turn 10 or so, suddenly all your power is now in skill. With spellweaver, that means you boost up to 1 sprite/2 turns very quickly. Then, it keeps going, so call it turn.. 20 or 25 at the latest and you can summon 1/sprite per turn.

By turn 30 you have multiple stacks of 9 sprites.


Catwalk then uses this until turn.. 50-75? I'm not sure of exact details. But, what he does with it, is he clears EVERYTHING that's remotely killable. He was killing stone giants! In nodes! With sprites!

Thinking about this, by turn 50, he literally has summoned, oh, 40? Ish? Sprites. By turn 75, he probably has cleared everything on the plane that's on your list, plus a fair amount not on your list, plus there's a good chance he's opened at least one tower, which let's him start the weak ones on myrror.

And he's got so much gold he can easily afford to buy troops in every single neutral city, plus a large number of settlers.


That's my guess anyway.

[Seravy]

The baseline 12 books have a mana and skill effectiveness of 26.666.
Adding those two retorts raises skill effectiveness to 37 but mana effectiveness stays at 26.

There shouldn't be a significant difference on the early game - of course it's stronger later - but this is an early game problem. Still, it's only a 42% increase even after sufficient mana crystals are found to fuel the skill.

The other thing is that using earth lore allows you to hit the best lairs nodes immediately. That means your very first sprite is out taking 2-5 lairs/nodes by himself, while your second 2nd to 4th sprites start out taking a weak lair the opposite direction and then start grouping up, while your 5th-9th go in a 3rd direction, with the idea that those two stacks will meet up around the time they run out of weak lairs, but also perfectly positioned to instantly start hitting strong lairs.

I almost never do it that way and send new sprites to where the "large" stack is, unless a very easy target (read 1 guardian spirit) is nearby. Taking out 5-9 units, even if weak, takes so freaking long with only 1 sprite that I try to make the stack larger as soon as I can. My first stack is usually 5-6 sprites, then I start sending out smaller stacks if there is an area the large stack cannot reach.
"weak" lairs aren't weak enough to do all of them with only 1-2 sprites in a reasonable amount of time.

I guess measuring this accurately using a tool is impossible. On the other hand I don't have the time to play through 10 impossible games using the same spell and strategy and I don't consider dropping the game on turn 40 conclusive evidence. I wanted to play one game with specialist conjurer though so might as well do that now.

[Seravy]

I started a game using Spellweaver+Conjurer.

I had a fairly average, or maybe even good map - Sprites gave me 2 nodes, 2 neutrals and I found an Earth Elemental Spell and a Life book.

And this is what such a good start was enough for :
   

Not only did it fail to put me ahead of enemy wizards, or keep me at their level, but I'm about one third as powerful as the next best wizard, and like 20% of the strongest.

It's an interesting game and I wish there was a point to continue, but from this position, there is no way I can win.

Just a guess, but I believe this start would have left me the weakest wizard even on Extreme, but would have been enough to put me ahead on Hard or Normal. Honestly, if someone knows enough of the game to play this strategy, they deserve to win Normal and Hard.

I'm going to try 1-2 more times.

A few observations :
-You can forget about harvesting everything. The enemy took one of the three weak neutral cities before I could finish summoning my third unit of sprites
-A 30x30 map area is unrealistic. I could only cover roughly a quarter to half of that before the game progressed beyond where Sprites were viable.
-I had to dedicate resources to defend and reclaim my nodes. At the very least 1 unit on each node is a must - lone magic spirits will come and meld the node almost immediately otherwise.
-There isn't enough casting skill to have more than 9 Sprites and cast Earth Lores, even with these retorts.
-Mana wasn't a factor after about turn 25 when I found the first place that gave me mana. Until then however, all I could spend was my starting power base. This could have been different if the good targets were closer and weaker but they weren't.
-Myrran AI tactic for high difficulty works well : the Myrran wizard is far ahead of everyone, despite playing white+blue. They also have early access to Arcanus with Astral Gates, popping out units next to my capital.
-Impossible might live up to its name and be unbeatable now - not only due to AI changes but because we removed everything that had a significant impact on the early game from retorts.

the game was recorded : https://www.youtube.com/watch?v=OIsXSUEE...e=youtu.be , it shouldn't be too long to watch because I tried to play fast this time and concentrate my attention on using sprites.

Will start the next game in the afternoon.