I never liked the idea of an NPC going after a target simply because that target looked like a fighter, or deals the most damage. I think that this model might work for unintelligent NPCs (animals, perhaps – but even an animal might go for the easiest kill, not the hardest) but it fails miserably for modeling intelligent NPCs.

I prefer to look at things from the perspective of what threat systems are actually trying to represent, in order to get inspiration for how to model the system. What exactly is a tank doing, and why does the NPC attack the tank? It's because the tank is either the NPC's target, or in between the NPC and its target. For example, a diamond formation with a spell-caster in the middle makes it difficult to attack the spell-caster without going through the four corners.

So if you want to call the concept "maneuvering into position" by the name "increasing threat", that's ok. But I think here you start considering that a tank can't be everywhere at once; if I'm busy protecting Fred and Joe over here, I can't run off to protect Bob at the same time over there.

For this reason, I've never really been able to mentally separate the idea of threat and some kind of relative position. You don't need full coordinate-based spatial positioning (although it so happens that you do, Runter, IIRC) but you can get away with something like one tank being able to protect at most X people. You could then have the tank intercept any attacks directed at these people, and also give the tank abilities to make itself a target (e.g. with "taunts" as Quix mentioned).

So to sum up quickly as I have to go: I view tanking as two separate concepts: (a) making yourself a target and (b) getting in between an attacker and its target. These aren't modeled the same way.

Static systems are very learnable. That's good, isn't it? Depends on your target audience. The faster I understand how something works, the less interesting it is to me. I think what you want, in order to get the best results for everyone, is a system that is consistent enough to glean information from, but not consistent enough to be easily learnable. I personally would be tempted to have the weights shift semi-randomly on a daily/hourly basis (using some kind of seed, I'm guessing), just to screw with people.

Random adjustments to static weights would be the best solution, in my opinion. (Then randomizing the actual target off of that)

If you're referring to randomly determining it every round and your range isn't limited, though, it would actually be worse than no system at all, since the randomization could easily override your system, making it a waste of time. You could still randomize by maybe up to 25% of the static value, though. Your original example had 4 people, at threats of 50, 20, 20, 10.

50 could be adjusted by +/- 12%, 20 by +/- 5%, 10 by +/- 2%.

This may or may not make much of a difference on a regular basis, and it'd probably be easier to weight the actual attack roll (to select between the %s), but I'd avoid static numbers for features like this where avoidable.

For randomness to introduce unpredictability, it has to be random enough that you actually can't consistently glean information from it. If you can always, consistently get the information you need, it seems to defeat the whole reason you had for making it random. If you can consistently get information, how have you not learned the system?

I think that you might be somewhat overstating the impact that predictability of a small component has on your "interest" in that component. Presumably, an awful lot of things in games you play are awfully predictable, and you use these things to build larger strategies to deal with other unpredictable things like players.

If you don't mind predictability as long as it takes you a while to figure out, that's another story, I suppose, although at this point you're introducing a barrier to entry moreso than unpredictability. (You're also introducing a barrier in the form of knowing how to find the spreadsheets online that explain how the system works, demystifying the complexity.)

When that happens it's up to the Admin to change the formulas and inner workings.

Personally I would change things in such a way that people might think that the old online spreadsheets are correct but in using them would hurt them more than help. Maybe I'm just a jerk. :devil:

I thought of a good question about this. Will the threat level be expressed to the characters numerically in real time, such as with a prompt variable? I'd think if that were the case, it would be less important for the system to be simplistic and easily learned. It could become more of a, "well….I just used attack X on NPC Y and my threat level jumped Z points….I don't think I want to do THAT again…".

You can be sure, though, that people would figure that out pretty quickly. The kind of people who put together these spreadsheets would definitely notice if something changed; they tend to be very analytical.

It's also hard to change the entire game's system without having to redo all the balance; I think it might be easier to just let them know the system. The game isn't about learning the system, after all, it's about doing things in the system.

I always figured that the healer would look like a more tempting target because the healer ensures the tank and the DPS whore don't die when the mob keeps whacking away at them. I would think that whoever leads to the most damage, which may be DPS in the case of the mage or enhanced survivability for the cleric, should be the one with the highest threat ratio. The ideal tank, then, is the character who can redirect damage away from the squishy party members. The suggestion of formation is a good start, but I think the ideal solution would also include the target the mob wants to kill.

There's something to be said for the idea of a super-high threat rating not being the primary target because it's so scary, but if you're trying to emulate intelligence, there's got to be a point where the mob says, "Screw it. This big ugly guy is the only threat here. I either need to gore him or GTFO."

My issue wouldn't be players knowing how it worked but putting it up on the internet. If someone wants to learn inner workings of a mechanic, I think that should not be easily accessed by "google <insert mechanic>"

Thanks everyone for your posts.



Yes. It's an additional hurdle for me in fact. Consider the problem of having characters attacking a mobile spread out at a distance. The mobile could randomly select characters and ping pong between them. Effectively being kited in dealing little damage. Making that more effective than tanking. There's a few ideas I have to solve it. One would be making characters who only have melee attacks only select characters in melee range for their next swing. While giving practically all mobiles abilities to instantly close distances on cooldown—treating it as one of many ranged attacks they could possibly have. Such as dash for animals, blink for mages, charge for warriors, and dash for ninja-types. In any event, in boss encounters I will likely have just as many deadly ranged attacks so that the problem may be mitigated somewhat.

I didn't want to go into the specifics of my system since it's likely it would be hurdles traditional muds wouldn't have. But yeah, :)

Hmm, interesting problem Runter. One thing you could do is introduce a "cost" in attacking a target. The further you are from somebody, the higher the cost in switching targets. This way, a mob can still change targets even if that target is outside of melee range, however it would only do so if it had a really good reason to switch targets. I guess you could still have the "ping pong" you describe, but it would be harder to set up deliberately. And, it might be hard to come up with good distance costs that give the mob reason to switch eventually but not too often.

I think it's hard to discuss threat/target-selection for your game without knowing anything about its specifics, though :wink: Especially since your game is (from what I recall) rather different from the norm.

I had to leave in the middle of making my last post, so I didn't really get to spend much time on it.

There is a difference between being so random that the underlying system is basically useless and being so static that every observable change leads to a consistent result. What I meant before about gleaning information means that it's ideally consistent enough to be somewhat learnable, but it should also be random enough that it's hard for people to be sure if they have learned it or not. Is X weird happening a result of random happenstance or a misunderstanding in the underlying system? I know for example that raising hitroll raises my chances to hit, but if I can calculate exactly how much hitroll I need to have to fight a certain mob, that's a bit unfortunate. By exact, I mean exactly. It's much easier to infer that you might need, say, >120 hitroll to hit a mob. If you can work out that >120<121 works just as well as >120, then you can equip for maximum effectiveness, and I'm not really a fan of maximum effectiveness.

It's been stated that he's intending to use this system for every day mobs and not for bosses, which makes me much less concerned about these things.

Knowledge about the game is important, as David Haley said, for developing strategies. I'm not a fan of a strategy that is guaranteed to work, however. The best gameplay strategy (in my somewhat less than humble opinion) is one where you have a variety of options and it's hard to be certain which is correct. If you know that wearing a hat of ugliness will end up with you taking 50% of attacks, that's pretty meh from my perspective.

On another note, I'd like to comment on a possible feature of threat systems that I don't ever see used anywhere:

You can make mobs aware of when they're outmatched.

I intended to use a threat system that consisted of working out damage they did to the mob verses damage mob did to them, affected by mob healing and player healing. And have the mob try to work out the best target to hit from there. Except that sometimes the best target to hit isn't really hittable, because, say, he's insubstantial, or he's being guarded by the OMFG INVINCIBLE tank, and so on. Why not have a reasonably intelligent mob run away? Maybe lure them to a more useful terrain, that sort of thing?

One of the things that makes fighting players so much more interesting than mobs is that they're smart. Yes, even the really dumb players. If they have enough skill to not get 1 or 2 rounded or spam themselves out, they usually won't just stand there and eat damage from people who are going to kill them. At worst, they'll try to run away, at best, they'll lure you away from your teammates for a fairer* fight.

Regarding David Haley's "target-switching cost", how about something like 3.5 ed. D&D's "attack of opportunity" system.

If you have a ranger standing in the back shooting arrows and a rogue and a warrior fighting your mob, and the mob decides to attack the ranger, he can. He has to move away from the warrior (provokes an attack from the warrior (i.e., warrior gets a free attack)), and the rogue (rogue gets an attack), and moves towards the ranger, who is wielding a ranged weapon (provokes an attack from the ranger). The mob has now eaten 3 attacks before even attacking the ranger. This is assuming a system where each player/mob doesn't get a bajillionty attacks a round, of course.

* a fight more in his favor, which is what "fair" always means in PK


[Edit: Instead of adding a randomization on the percentages, why not have mobs tag people with levels of "dislike" that affects their chances of being hit? You cast fireball on me? Have a +5 dislike. You cast faerie fire on me and gave me a pink aura? +50 dislike. (The pink aura annoys me more than being fireballed, your mileage may vary.)]

Racial feuds would also come into play here. That orc would rather focus his attacks on the elf.

I give each of my mobs ranged tactics. While some will literally sheathe their melee weapons and draw bows, most will just pounce, charge or swoop towards their opponent - this is pretty fast, but it's still just movement, and you can still retreat or run from them. Allowing mobs to instantly close the distance would render ranged tactics somewhat obsolete. It also raises the question of how exactly people will be able to flee from combat, if their opponent won't switch target and can instantly close the distance each time they attack.


If the specifics relate to the feature you're discussing then I think it makes sense to at least mention them, otherwise it can end up being difficult to provide constructive comments.

To go back to the question of tactics, I actually prefer assigning static tactics to each mob, so that players can learn how to fight each opponent (much like arcade fighting games). Knowing how to fight them doesn't ensure victory though, it just gives you a better chance.

Usually flee won't be an option. Against any important encounters there will be no flee option.

Well, I never said each time they attack. Actually used the term "on cooldown." I meant possibly they could instantly close distance every 30 seconds. It would prevent people from absolute kiting. In any event, I do see your point. I don't want to remove ranged combat as an actual component since my system will be based around it.

It'll be interesting to see what sort of social impact that has on exploration and PK.


Could you please give a quick explanation of how "cooldown" works? Are you referring the distance-closing ability itself as having a 30 second cooldown/recharge time, or is it part of an overall balance-style system, or something else?

Well, currently most of my abilities trigger cooldowns for that specific ability. Some of them trigger cooldowns for an entire class of abilities. Using most combat related abilities trigger a global cooldown which is the minimum cooldown for any spell/skill. Currently it's sitting at 1 second.

So it won't be unusual in my game for an ability to have a cooldown that prevents them from using it for any arbitrary amount of time. (Based on the design of the ability.) For example, in this case, the number is going to be high enough where the distance closing won't happen every time. Yet it will happen often enough to fail absolute kiting plans.

In any event, a practical example of my cooldown system could be examined through a certain healer class:

They may have an ability with no associated cooldown but only heals a base amount. And it generates normal aggro + 10%.
Then there may be another ability they have with a 6 second cooldown that heals base amount + 10, costs the same, but also generates normal aggro.
And then there may be yet another ability with a 20 second cooldown that heals the entire party, heals each player base amount - 10%, and generates normal aggro - 25%.

A bit of a contrived example but yes. Cooldowns will be used to balance the strength of abilities as well as give players options for setting up their play style.

To expand on the example, perhaps using the AOE abilities would set a global cooldown off on your stronger single target heals. Perhaps vice versa. In any event, It's nothing I've even began balance testing on. The same stuff will go into play for tank and DPS class abilities.