Towards better tactical combat

[fizzix]

In another thread Derakon asked that I outline what I think combat should look like. A justified request since I was saying some provocative things, some of which may have been poorly thought out. Anyway, I've thought about it a few days and tried to organize my thoughts.

What Angband does well

I find Angband to be the most interesting from many areas from dungeon level 5 to about 40. It general starts when the player find some secondary means of attack besides the weapon (or magic missile for mages), and it ends, well we'll get to that later. During this range, players tend to have several different viable approaches to attack. These include both damage wands with limited charges, valuable ammo, and status effect wands and staves. Furthermore, with the exception of out of depth monsters, and super low hp classes, there are few one-shot threats around.

I actually find Angband at this range to be one of the most engaging and rewarding roguelikes. It has interesting tactics and a very nice interplay between tactical and strategic considerations. More on that now.

Tactics versus strategy

I'm going to throw these terms around, so it's probably a good idea to define them. I think my definitions are standard, but I'll define them anyway. Tactics regards to player decisions in the shrot term, usually this refers to a given encounter or objective, like clearing a vault. Strategy refers to what decisions you should make to win the game in total. For Angband, which doesn't have character builds and skill trees, the strategic elements lie entirely in the decision of inventory decisions and use of valuable consumables.

Therefore, a given battle, although tactically focused, involves strategic decisions. Using a wand charge in one battle, means you do not get to use it later. The decision usually starts before the battle begins. Angband forces players to estimate what resources they'll need to use to defeat an enemy, and then decide whether its better to engage or not. This is one of the key decisions that Angband throws at the player, and it's one that's mostly unique to *bands. Other games expect players to clear all enemies on a level before progressing to the next one. Angband absolutely and emphatically does not expect this. The decision of whether to engage is neither strategic or tactical, but a mix of the two. Because it is something unique to Angband, it probably behooves us to ensure that it remains in the game.

Tactically dull situations

There are two types of situations that commonly crop up that are tactically dull. The first situation occurs if the player has a zero-cost combat option that straight up outclasses all others. Usually this is a powerful weapon, but it can also be a powerful launcher that can do loads of damage even with unenchanted ammo. Spells sometimes also fill this role, there is a significant range of the game where Priest Orb of Destruction is the best attack available by far, limited only by spell points. These tactically dull situations occur in every roguelike I've played. There's really no avoiding it. The random nature of drops means that at some point, in some games, you will find a weapon that is just dominant. However, we could consider trying to change up tactically dull situations for spell-casting classes if they're too prominent.

However, it's really the second dull tactical situation that bugs me more. It's what I'd prefer to focus on. The situation arises when the player is put in a situation where they need to take X or Y action immediately, or they have a chance to die before they get another move. Due to the asymmetric nature of roguelike combat, even 1% chances of death are unacceptable, and must be avoided. The player only has the full set of tactical choices when they are not in risk of dying next turn. Angband produces way too many situations where the player is at immediate risk of dying. These situations are often unavoidable. Morgoth (or Vecna even) summoning a single black reaver in LoS is enough to kill all but the most hardy of players on the next turn if both decide to cast Mana Storm.

So I said there were two situations that I wanted to talk about, but there's also a third situation. When monsters can outright kill you in one turn, the decision of whether to engage or not is removed. You absolutely cannot engage these monsters. A very common example are monsters that hit to paralyse. If you don't resist paralysis, you must flee. There is no choice.

Why the game gets less fun after level 40

In my opinion, after about level 40, the game starts throwing many situations at the player in which they go from a perfectly ok situation, to one where they are in immediate danger. This is when you start getting summoning monsters, and monsters with highly deadly attacks, both of which can cause forced move situations. Tactical combat then becomes, at most, a two turn affair. Thinking more than two turns ahead is pointless in many situations.

Because of the proliferation of these situations, Angband also gives the player the ability to handle them. There are many "get out of jail free" cards in Angband. A super-powerful one, the staff of teleportation, is available in town. It's not foolproof, but it's incredibly powerful. Things only get more powerful from there. Stuff like Destruction, teleport level, teleport other, and banishment are incredibly useful to get the player out of tricky situations. Even the lowly phase door would be a fairly powerful escape option in some games. Based on the way Angband works, these escapes are necessary. Because you are often put in a situation where not having a powerful escape could kill you next turn, you must make sure the player has access to these escapes.

There are many tactical avenues that are employed to great effect in other roguelikes that are not present in Angband, much to its detriment. One example is, if you've played Sil or DCSS you know how important it is to make sure you have a path to the stairs. Speed isn't as pervasive or as deadly as in Angband, so if you blunder into something you can't handle, you are usually perfectly free to walk back to the stairs and either reset the level (sil) or attempt a different approach (DCSS). This, of course, provided that you were cognizant of the terrain and made sure you had a safe path back to the stairs. In Angband, worries about paths back to the stairs are pretty much non-existent. Levels are huge, and many fast moving monsters (cats, hydras) will kill you before you ever make it back. Pretty much from an early stage in the game, teleport becomes a necessary escape, and phase door is necessary from the very beginning.

The situation gets much worse after level 40. And the main reason is that way too many monsters have attacks that do over half-hp damage to even the best-kitted character, and outright kill weaker characters. You absolutely cannot engage two or more of these monsters at once.

The fact that situations can go from full health to dire in a single turn really limits what attacks are available to monsters. Status effect attacks must either be mild or resistable. You can't have monsters that can confuse through resistance, for example. And you must give players the ability to resist confusion completely. There is also a requirement to avoid specific status effects. Teleportation resistance from either the monster side (resistant to TO) and the player side (prevent player teleportation) is pretty much out of the question.

Where should we be headed

A rule of thumb I suggested in another thread is to try to move from having lots of situations that look like, "If I don't do X,Y or Z now I may die next turn" into "If I don't do X, Y or Z now I may die in 3-5 turns" At first glance this looks similar, both require immediate action, but there's actually a huge difference here. It allows for situations where planning ahead is critical. If you can survive 3 attacks from a monster, you have a good chance of being able to run to the stairs (if stairs are close enough). You still need to start evasive maneuvers immediately, but it extends the range. Another example, in DCSS, teleports exist, but they work after a delay. They are also valuable consumables. What this means is that you may need to recognize the potential for bad things to happen and start teleportation in advance. This is only possible in DCSS because there are few monsters that can kill you in a small number of turns, and, the truly deadly monsters can often be avoided.

I'll list a few things that would help move the game in this direction, knowing full well that many of these are large scale departures. Most of this is theorycrafting, but a lot is informed from playing other roguelikes and trying to steal mechanics that work well.

1. Reduce sight of both player and monster, and make levels smaller (which in turn makes level layout more important)
   
2. Change speed to be movement speed (all attacks take 1 turn regardless of movement speed).
   
3. Rework fast moving early monsters, Grip and Fang are big issues here. But also stuff like crows can kill an early character if they miss their one blow attempt.
   
4. Phase door becomes a mid game escape option. Teleport is a very rare item. Teleport level is delayed
   
5. Destruction damages monsters but does not remove them (kind of like earthquake)
   
6. Teleport other is resistable (player can resist it as well)
   
7. Highly damaging monster attacks are significant nerfed. Non-base element breaths don't do much damage (except maybe nether), the main issue is the status effect. Status effects however get more severe. No 500 damage disenchantment breaths from the Tarrasque anymore.
   
8. Redesign high damage spells from monsters that are not Sauron and Morgoth
   
9. Change INT to magic and have it govern spell points and spell damage
   
10. Change WIS to something else, and have it govern device skill, mental saving through, and the ability to inflict status effects on monsters.
    
11. Remove all resistances to confusion/blindness/stunning. Instead base everything on saving throw. Saving throw can be increased through items directly.
    
12. Greatly nerf monster summoning. A common way of implementing summoning in other games is to have summoned monsters be timed effects that disappear after some number of time, or when the summoner monster is killed. Summoned monsters also don't give XP or drops. We don't need to do it this way of course.
    
13. Recharging is a rare scroll only. Wands and rods are valuable.
    
14. Lightning attacks drain charges from wands rather than destroy them
    
15. Controlled blink is a valuable escape scroll, as in DCSS
    
16. Allow monsters to wander, and go back to sleep.

I wrote this over the course of a few days, so I apologize if it's not entirely coherent. I'll probably come up with more things I should have said later too. Oh well. Better post this before my browser crashes or something.

[Derakon]

Thanks for the writeup, fizzix! It sounds to me (and feel free to correct me if I'm wrong) that your primary complaint is a lack of tactical variety: that is, the player's hand can be forced towards one of their superpowered escape options, because of the large "swinginess" of lategame combat.

In other words "X happened, therefore I must Y or take an unacceptably large risk of death". Whether that death is immediate or not is largely moot, though; for example, if you get a mortal wound and have no healing options, then you're going to die, just in a few turns rather than immediately. In other words, dead-man-walking is dead.

That in mind, I feel like there isn't a strong differentiation between "this monster showed up, therefore I must teleport immediately lest it breathe" and "this monster showed up, therefore I must flee for the stairs immediately lest it overwhelm me before I can escape." In either case the player must take immediate action or be dead, it's just that in one they die in a single turn while in the other they die over many turns.

It's certainly possible that the latter setup leaves more room for interesting tactical decisions...but if you want to remove all escapes that aren't "walk away" then I have a feeling that won't be very tactically interesting. Whatever approach you want to come up with should be IMO focused on the decisions the player is making each turn moreso than it is on how swingy the combat is or how apparently powerful the player's/monsters' options are.

EDIT: I will also note that modern Vanilla has plenty of "I failed to take an action X turns ago and therefore am now dead" scenarios. They just tend to be things like "I didn't flee as soon as I detected Kavlax/that pack of Plasma Hounds, and when they got into LOS they one-shot me" or "I wasn't carefully detecting, so I didn't notice that Drolem, so it one-shot me" or "I decided to fight that Great Crystal Drake and its breath does more damage than I thought it would". You can always push back the critical decision point to some extent. That doesn't change that when you choose to engage with a summoner or a powerful breather, you may be forced to flee at a moment's notice, though.

[fizzix]

There's a subtlety which didn't come across, probably because I'm bad at explaining these things with words. I'm not really proposing a "mortal wounds" type situation, but rather one where you can calculate the probability of dying some number of turns out (provided you're good enough of a computer).

So imagine the situation where your escape is running to the stairs. This comes with the risk of suffering some ranged attacks along the way, and possibly being intercepted by another monster. Now imagine the situation where you can calculate the probability of reaching the stairs after some number of turns of non-evasive actions. Let's say it looks something like:

Flee now: 100%
Flee in 1 turn: 99%
Flee in 2 turns: 90%
Flee in 3 turns: 70%
Flee in 4 turns: 40%
Flee in 5 turns: 10%
Flee in 6 turns: 1%

This is already a dire situation, even though it looks like the player has a good chance of getting out of it alive. Most players probably will not recognize it as a dire one though. Not fleeing on turn 1, could lead to a situation where a series of bad rolls leads to death. The math obviously gets upgraded after each turn, since some of the uncertainties will have resolved after the first move. The 1% probably is a bad case scenario of another monster appearing (through summons or whatever) along with you failing to do significant neutralization of the threat.

This also assumes the player doesn't have "get out of jail free" cards. In reality you need to give players some of these, because no one can calculate the probabilities for strings of bad rolls. So even the most competent of players are going to make bad estimates, and there need to be ways to survive some number of these. The trick is that these cards need to be rare items, not spammable escapes. You might get something like 10-15 blink scrolls in a game of DCSS, that's it. The probabilities above assume you don't have any of these get out of jail free cards.

Right now angband math usually looks like:

Flee now: 100%
Flee next turn: 90%
Flee in 2 turns: 90%
Flee in 3 turns: 90%
Flee in 4 turns: 90%
....

At any turns you have the chance to escape so that 10% loss is the probability of you dying on the next turn, or your escape failing after you've engaged. One of the differences between this scenario and above, is that this situation is not dire at all. This can be any late game encounter with a summoner for example.

[Derakon]

So if I wait a turn before fleeing, is that not roughly equivalent to (in modern Angband) taking that 1% chance that both monsters in LOS will choose to use their strongest attacks?

Anyway, I don't really want to bicker over whether Angband is equivalent to other games when viewed through the right lens. I'm more interested in what success and failure look like under your proposal. What kinds of decisions will the player have to make to either a) succeed, or failing that, b) escape without dying? Conversely, how would they screw up that would result in them dying?

[fizzix]

Originally Posted by Derakon

Anyway, I don't really want to bicker over whether Angband is equivalent to other games when viewed through the right lens. I'm more interested in what success and failure look like under your proposal. What kinds of decisions will the player have to make to either a) succeed, or failing that, b) escape without dying? Conversely, how would they screw up that would result in them dying?

Sure, I think we can approach this from a plausible situation in early Angband that works sort of appropriately.

The scenario is a player vs a uruk or black orc. They are about 30 steps from the stairs. They're inventory consists of a couple potions of cure serious wounds and two scrolls of teleportation. Those scrolls are valuable to the player, so they'd prefer not to use them if possible. Same with the potions. Let's assume no phase door scrolls for some reason, so we're breaking some believability.

Also, to make this work we have to sort of assume that uruk is a bit beefier on the HP front (or the player does less average damage per turn) and simultaneous does less damage, especially with arrows. So imagine that the player has 120 HP, the Uruk has 100 HP. Both have about a 2/3 chance to hit and do about 10 damage per hit. The Uruk can also fire an arrow for about 10 HP damage with about a 50% chance to hit. Let's say if the Uruk is 2 squares or more from the player it will fire an arrow 20% of the time and move towards the player the other 80%.

Let's also assume that the Uruk is one of the more difficult monsters the player is expected to find. A teleport, or use of an up (or down) stairs should on average land them in a safer environment.

Decision 1) Do you engage with the uruk or not. This is the standard risk reward decision. There are generic issues with this in Angband, as Philip points out. But I don't really know how to solve those besides enforcing stuff like forced descent. I'm not proposing really any difference here.

Let's say the player engages with the Uruk and they start exchanging blows. The player should win the battle, based on stats, but with a couple of bad rolls things can start getting a little closer.

Decision 2.1) When should the player abandon the battle in order to reach the stairs without having to use the potion or the scroll? With 30 steps towards the stairs, assuming the Uruk can always see the player, the Uruk will get on average about 6 shots off and hit about 3 of those. So from this point the player can safely flee around 50 HP or so and will usually reach the stairs before dying, but it will be close. They'll also be very wounded when they go up stairs, and it's not guaranteed that they will be safe up (or down) stairs, especially if they are at 20 or 30 HP. 70 HP would be safer to start fleeing.

Decision 2.2) Should the player use a healing item to win the battle? Here it's a question of what is more valuable, the scroll of teleportation or the healing item?

Decision 2) in all. The player is about 70 hp, the Uruk is at about 50 or 60 HP. The player needs to decide whether to continue fighting, in which case they may need to use a healing item or teleportation. Or they could choose to flee to the stairs, where they'll almost definitely make it with a good amount of HP.

Decision 3) occurs if the player continues to fight and things go bad so that they look unlikely to win. Decision 3 is the player is at 40 HP, the Uruk is at 30 HP. They can either attempt to reach the stairs in which case they'll be at deaths door, or they can suck it up and use a heal or escape.

The situation often happens in current Angband where you get into a fight and realize that it was a bad choice. The monster does more damage than you thought, and you are doing less damage than you thought. The only issue is that since escapes are so cheap, you usually just abandon the battle immediately.

[Derakon]

In your example, then, the player basically has the following choices:
* Engage or not
* Whether to use hard-to-replace consumables
* When if at all to disengage

I feel like the main difference between those choices and what's in Vanilla right now is just that whether to disengage becomes harder to calculate, and the act of disengaging is more involved and potentially riskier. But the actual fighting remains basically similar to how it is in current Vanilla -- that is, a sequence of bump-attacking. You're even proposing removing the shoot-n-scoot tactic (that is, Phase Door plus a missile weapon), so arguably the player has fewer options in deciding how to persecute combat.

I guess when I was thinking of a deeper tactical game, I was imagining something where the player has a wide range of options, not a narrow range of options that require complex statistical analyses to choose between. Something like:

* Use my most expensive, highest-damage spell, and hope that it kills the monster before I run out of mana (and that nothing else shows up while I'm resting)
* Ditto, but spend a rare consumable to refresh SP if/when I run out
* Use a cheaper spell which I know will be enough to kill the monster, but I'll take more damage because the fight takes longer
* Use wands to get high damage without spending SP, at the risk of the wands not being available later
* Use a bow and arrows for moderate damage without spending SP, at the cost of losing some arrows and having to lug around the bug and arrows in the first place

I'd like to see basic combat have a similar depth of decision-making as the above, ideally.

[fizzix]

Originally Posted by Derakon

I guess when I was thinking of a deeper tactical game, I was imagining something where the player has a wide range of options, not a narrow range of options that require complex statistical analyses to choose between. Something like:

I was trying to distill a specific tactical idea, that is missing. Obviously other options would be available. I said at the beginning that I like the gameplay between levels 10-40 a lot, and a big reason for that is that there are many different options for dealing with problems, and few monsters are going to insta-kill you.

The big differences in the scenario I laid out is that levels themselves become an important aspect of terrain. You consider where you are and how far from the stairs you are before engaging an enemy. This is a defining feature of Sil for example. It doesn't exist in Angband because levels are too large and teleport is too plentiful. Angband is worse off for it.

The second difference is that there are a lot more turns available for the player to make choices. Things don't go from 10 to 100 immediately, as they often do after a powerful summon.


Edit: I should elaborate on some of the other options that could be available.

Obviously, powerful attack wands could be available to the player. The player could expend a valuable charge instead of melee.

The player could opt for ranged attacks that do less damage on average, but they would take less damage in return.

The player could use valuable branded/magical ammo that they risk losing.

When fleeing, the player could "hold monster" before they round a corner, and then get away without any future attacks. Other status effects could come into play here too. Confusion, sleep, scare etc.

The player could slow the monster (assuming it only affects movement speed here) allowing them to get away easily. Similarly they could haste themselves to escape. These spells are still powerful though, so care is needed. Stuff like "mushrooms of sprinting" and "rings of escaping" are great ideas that just can't fit into current Angband gameplay except in the narrowest of windows.

We used to have spikes and would allow players to jam doors. This was removed because in a world of infinite ?phase this is pointless. This is a great way of introducing terrain management.

[Sky]

I have a (loaded) question for @Nick

How many downloads of 4.1 have you had ?

[Nick]

Originally Posted by Sky

I have a (loaded) question for @Nick

How many downloads of 4.1 have you had ?

No idea - I don't think we keep stats. There are probably a couple of hundred ladder dumps, if that helps.

[fizzix]

I think Pete is close to what I'm thinking. But really what I want is to try to take the elements that I've found interesting from many other roguelikes, and incorporate them into Angband. It's also perhaps not a fair thing to ask for. A good course of action for me might just be to play a bunch of DCSS (or Tome or Brogue or whatever) games until I get disenchanted with its gameplay, at which point Angband will feel fresh and appealing.

Angband has always been a tough, or perhaps impossible, game to balance because you can continually reset the level. You can scum as much or as little as you like. It's probably always been true that the main killer in Angband is complacency and carelessness. I never have had a good solution for how to best balance Angband, the best I came up with was to balance around a single clear of every level, which is how we balanced 3.4 to be roughly in line with 3.0.6. Other roguelikes have it much easier.

I also give my full support to Nick to pursue the game development in the manner he sees fit. To me the most important thing is that Angband is continually developed, even if it's not aligned with my ideas. I'll continue to play and offer feedback (and outlandish ideas) although I may take hiatuses from time to time. Those breaks are good for me, because, I've always found the first game after a break to be fun.

I also agree with Nick that the game should feel like it has increasing danger later in the dungeon. The disconnect for me is that in current Angband, I don't think it's true. I'm more likely to die on dungeon level 1 than I am dungeon level 30 (almost always from Grip/Fang or some other fast moving OoD monster). And I'm way more likely to die on 30 than I am on 80. In fact, usually if I get a character to about dungeon level 40, my win percentage is pretty good, even with challenges like ironman or forced descent. The reason is that I just have so many escape tools by level 40 that I can get myself out of any jam, no matter how dire. I really only get sniped by insta-kill attacks. Getting feedback from players with lots of wins is, and experience in the second half of the dungeon, is necessary, and probably should be actively sought out. My guess is that Grip and Fang have many more kills than Gothmog and Lungorthin.

Also, it's probably a very good idea to get some sort of Monte-Carlo statistics working again. I think there's currently some memory leak which causes the game to crash after generating some 1000 or so levels, but I couldn't find it. Item balance is a bit out of whack now, it seems. But I'm just going from experience over a handful of games, which is always a bad way to balance. Making OP items and artifact rarer and more valuable will probably help the general situation. Right now it's pretty easy to stay ahead of the challenge curve in forced descent, except for a brief period in the 30s and 40s where Ancient Dragons start showing up and you don't have any permanent speed yet.

In the past I have tried to implement some of the changes I've talked about, but I often get stuck from lack of ability. (It's also tough to work on coding for fun after spending 8 hours a day writing physics code, but that's a lame excuse). For example, a long time ago I tried to implement wandering monsters, by loading a bunch of different heat maps into memory and having monsters choose one, but I couldn't get it to work. I think this would be a good change and opens up possibilities.