Torchlight 2 mechanics

I'm quite surprised there are no guide about torchlight 2 mechanics. This forced me to dig runic games forum and do various tests myself. Since finding info really takes a lot of time, I've decided to write this guide to help new players.

Stats

There are four stats in torchlight 2 - strength, dexterity, focus and vitality.

Strength:

One point of strength increases:

• weapon damage by 0.5%. Doesn't seems like there is a realistic damage bonus cap.
  
• critical damage bonus by 0.4%. Capped at 399.6%, which is achieved at 999 strength.

Dexterity:

Increases:

• dodge. Subject to diminishing returns, the generally accepted formulae is %dodge = 0.002*dex*(1-0.001*dex). But since dodge mechanics is subject to accelerating returns by its nature, then you can technically build a plot of your survivability against your dexterity. Read about that in detail[forums.runicgames.com].
  
• critical chance. Subject to diminishing returns, the formulae is essentially the same as for dodge, %crit_chance = 0.002*dex*(1-0.001*dex).
  
• fumble recovery. Subject to diminishing returns, the formulae is %fumble_recovery = 0.003*dex*(1-0.001*dex) + 0.25.

Focus:

Increases:

• 'magic' damage by 0.5% per point. Doesn't seems like there is a realistic damage bonus cap.
  
• execute chance. Subject to diminishing returns. The formulae is %execute = 0.2*foc*(1-0.001*foc) + 9.8, but this bonus is hardcapped at 60%, which is reached once you have 500 focus.
  
• mana pool by 0.5 per point.

Vitality:

Increases:

• block chance. Subject to diminishing returns. The formulae is %block = 0.2*Vit*(1-0.001*Vit). Bonus to block chance from vitality is hardcapped at 50%, which is reached once you have 500 vit.
  
• health pool by 3.6 per point
  
• armor by 0.25% per point

Notes:

1. Strength increases weapon damage, not physical damage. This means that strength improve wand and stave damage on par with focus.
   
2. 'Magic' damage improved by focus is fire, frost, lightning, poison and all kinds of DoT, even one that is stated as 'physical'.

• Since both strength and dexterity improve your damage in a quite complex way, an optimal ratio of str : dex exists.

Defence

There are numerous ways to increase your defence, and a common sense tells you that there is a certain order in which different defence mechanics are applied.

This order was tested long ago, and results were posted here[forums.runicgames.com] and here[forums.runicgames.com]

The order in question is:

1. Kiting
   
2. Missile reflection
   
3. Dodge
   
4. Armor
   
5. Damage reduction
   
6. Damage absorbtion
   
7. Block
   
8. Health

Kiting.

Obviously, if projectile or end-point of attack animation doesn't connect with your character's hit box, you take no damage.

Missile reflection.

If the missile is reflected, you take no damage from it.
Missile reflect chance stacks multiplicatively, so the formulae for your total reflect chance is

%total = 100 * (1 - ( 1 - %source_1 / 100) * (1 - %source_2 / 100) * ... etc for every source of missle reflect ).

The practical view of it is that you lose some portion of %reflect compared to
plain additive stacking the more %reflect you stacked already. A huge single chunk of %reflect is very good though.

One should be very careful, not everything that appears to be a missile could be reflected. I've tried to find a complete list, but no luck there.

Additional info:

• Damage of the reflected projectile is based on your weapon damage, not projectile's original damage, and work the same way as %weapon dps skills. Except that, unlike skills, reflected projectiles can fumble.
  
• If character have multiple sources of missile reflection, the %damage is an average of sources, weighted with corresponding %reflect chance of the source.
  
• The reflected projectile behaviour is somewhat ambiguous. Some projectiles are reflected as auto-hits, they can't be physically evaded. Homing or curved projectiles are usually receive a random direction
  
• After being reflected, while projectiles with a straight trajectory is reflected at it's source. While the latter ones could be re-reflected, auto-hit ones could not.
  
• An interesting thing about %reflection stacking is that it doesn't have a hard cap. You can stack %reflect to get insane amount of it, say, 90% and more.

Dodge, block and damage reduction.

Since the mathematics behind them are essentially the same, I've desided to unite them in one section.

 Let's say you have X health and some %damage_reduction. Then your effective hit points (commonly shorted as EHP) - that is - the raw amount of damage you can take before dying - will be:

EHP = X / (1 - %damage_reduction / 100 )

This formulae suggests that there is an accelerating effect on %damage_reduction. Going from 0 to 25% damage reduction will improve your EHP by 33%. Going from 50% to 75% will improve your EHP by 100%. 

Now about difference between them.
First of all, while dodge and block could be viewed as a kind of damage reduction, they make your character weak to heavy hits. There is little point in 75% dodge/block if you die in one hit because dodge/block failed. Especially stressed when you play HCE.

Unlike dodge and damage reduction, block doesn't work if you don't wear a shield.

Further difference lies in the scope of things that could be dodged, blocked or reduced.

• dodge: only melee or direct ranged autoattacks could be blocked.
  
• block: most of stuff can be blocked, with notable exception being traps.
  
• damage reduction: all kinds of damage, except traps.

All three are capped at 75%.

Armor.

Armor basically reduces incoming damage by a random integer between 50% and 100% of it's amount.

Damage absorbtion.

Damage absorbtion is a defence mechanics unique to engineer. It basically works as an additional health pool when active. If damage done exceeds remained amount of damage absorbtion, the excess damage is done to your health pool.

Health.

Not much to say there, except that all characters start with 200 health and gain 40 health per level.

Key points and guidelines:

• The most cost-effective defence method is kiting. It doesn't cost you anything and is pretty much necessary even if your character is well protected from damage.
  
• Damage reduction is by far the best (and most important, since it's virtually impossible to play lategame HCE without it) damage mitigation overall. On a bad side, it is ridiculously hard to stack, the easiest way being farming Grell's Eyes for eternity.
  
• Berserkers and engineers have natural 25% damage resistance, so they need to stack only 50% to reach the cap
  
• Probably the easiest way to make your characted tougher is to wear a shield. First of all, shield means shield block. Second, shields are prone to have other defensive modifiers like %reflect. 35% chance to reflect missiles is quite common, even really early on. The downside is that shield block doesn't improve much with level, since high-end shields typically have the same inherent %block, additional modifiers not included. Character with high block usually means quite a few points into vitality.
  
• Block is mostly useless to engineer, since it applies after the damage absorbtion (force field, aegis of fate). You will still see floating text 'blocked' though, which is quite deceptive.
  
• Outlander is notable to be the easiest character to max dodge. Usually 110 dexterity is enough to reach dodge cap at level 100.
  
• Armor is generally considered worthless. It is passable as a main defence for the first 30 levels of elite difficulty, but past that you'll be hit for 2k damage in your face, while your armor
  will hardly surpass 300 at that point. Later in the game you could go to a great lengths to increase your armor up to 2000+, but it is dwarfed by the raw 30k damage you will receive on ng+++.
  
• There is a certain modifier on items - %chance to cast fully heal self when you get hit. Very questionable as a defence mechanic, since it works only if you actually lost health due to attack,
  but definitely could be used as a support to your wall of defence.
  
• There are three kinds of reflection - flat X damage reflected, %damage reflected and %chance to reflect a missile. Only latter actually negates damage, while the former two shoudn't be
  considered as defensive measures. Furthermore, they work only if you actually lost health due to attack.
  
• Traps are dangerous because they go though most of your defence.
  

Damage - Autoattack damage

I won't bother with full formulae here, because it is of little practical use (since your auto-attack weapon dps and %weapon dps used for skills are two different things and you do most damage with skills anyway). So I will focus more on general ideas instead.

1. Most percentage bonuses stack additively. For example, your +%damage on gear and +%damage from strength/focus are first summed up and then applied to your base damage.
   

    Example for a physical weapon
   new_min_dmg = min_dmg * ( 1 + %strength_bonus + %physical_dmg )
   new_max_dmg = max_dmg * ( 1 + %strength_bonus + %physical_dmg )
   All percentage values are considered to be divided by 100 for the purpose of conveniece and to shorten the expression. 
   min_dmg : your weapon's base minumum damage
   max_dmg : your weapon's base maximum damage
   %strength_bonus : weapon damage bonus from strength
   %physical_dmg : physical damage modifiers from gear and skills 
   

   
   
   • One somewhat counter-intuitive and easy to miss concept is that strength increases %weapon damage, not physical damage.
     
   • Yes, this means strength increases wand and staff damage on part with focus.
   
   

    
   Example for a wand. 
   new_min_dmg = min_dmg * ( 1 + %strength_bonus + %focus_bonus )
   Same goes for maximum damage.
   All percentage values are considered to be divided by 100 for the purpose of conveniece and to shorten the expression. 
   min_dmg : your weapon's base minumum damage
   %strength_bonus : weapon damage bonus from strength
   %focus_bonus : magic damage bonus from focus
   

   
   List of bonuses that stack additively:
   
   • +%weapon damage from strength; directly affects AA damage;
     
   • +%magic damage from focus; magic is fire, ice, lightning, poison;
     
   • +%damage type; possibly physical, fire, ice, lightning, poison - add for a corresponding damage type;
     
   • +%all damage; it's basically the same as previous one;
     
   • +%melee damage bonus; only if you do a melee weapon hit;
     
   • +%ranged damage bonus; only if you do a ranged weapon hit;
     
   • +%wand and staff damage bonus; only for wands and staffs; wands also count as ranged, staffs count as melee;
     
   • +%damage bonus when dual-wielding; only if you have a weapon in the offhand slot.
   
   
2. Out of fumble and critical hit, fumble takes priority. Fumbled attacks cannot crit, and critical attack doesn't prevent fumble. Fumble in it's core is a chance of %fumble_chance to
   decrease your damage by %fumble_damage_penalty.
   

    
   %fumble_damage_penalty = 1 - %fumble_recovery
   fumbled_damage = new_damage * %fumble_recovery
   All percentage values are considered to be divided by 100 for the purpose of conveniece and to shorten the expression. 
   %fumble_recovery : multiplicative modifier to your damage when attack fumbled
   new_damage : either new_min_dmg or new_max_dmg from above
   fumbled_damage : damage done with fumble in effect
   

   • Note that fumble penalty is multiplicative to strenth/focus bonuses, as well as to %damage gear bonuses.
     
   • In case you're clueless - purple floating damage text indicate fumbled attack.
     
   • %fumble_chance is constant and equal to 21%. There are gear pieces that decrease it.
     
   • %fumble_recovery is increased by dexterity and gear
   
   
3. Critical hit is a %crit_chance to increase your damage by %crit_damage.
   

   If the damage wasn't fumbled, then 
   critical_damage = new_max_dmg * ( 1 + %crit_damage )
   All percentage values are considered to be divided by 100 for the purpose of conveniece and to shorten the expression. 
   

   • Important - critical hits always force maximum damage roll.
     
   • Note that %crit_damage is multiplicative to strength/focus bonuses, as well as to %damage gear bonuses.
   
   
4. Exact damage done is random between your minimum and maximum damage, after every bonus and penalty was applied. There is one exception though, and that is critical hit. Critical hit always rolls a maximum damage out of damage range.
   
   
5. On weapons that do more than one damage type. In short, deal with them as if they are separate damage instances that share crits and fumbles.
   
   
6. On +damage bonuses from gear and socketables. It is safe to treat them as separate damage instances. E.g. your weapon does some physical damage, fire damage and you socketed lightning ember on top of it. When you make a hit, first it is checked if the hit is fumbled, and if not, checks if it's critical or not. Then the game checks what kinds of damage were included in this "hit", and it turns out ot be physical, fire and lightning, after which the game just adds every viable modifier, reduces damage if a fumble, increases damage if a crit, and rolls max number from a damage range if a crit. That's pretty much all. Later it's compared with enemy's armor, but that's a different story.
   
   
7. Attack speed is basically a coefficient that converts your damage into damage-per-second.
   

   DPS = avg_dmg * (( 1 + %attack_speed ) / weapon_base_attack_speed)
   All percentage values are considered to be divided by 100 for the purpose of conveniece and to shorten the expression. 
   avg_dmg : average damage you do from hit to hit. 
   

   It's pretty interesting that +%attack_speed can be in fact treated as +%damage_per_second.

This is pretty much everything about your normal autoattacks.

Damage - Splash Damage

Splash damage is an area of effect damage that is done on your auto-attacks. An area of effect has a form of circular sector. Different weapons have different reach and sector angle, as well as different percent of damage done to secondary targets. A secondary target is a term reserved for enemies hit by weapon splash. Main target does not receive additional damage from splash.

Splash effect is not propagated to %weapon DPS skills.

So, this is how weapons are different from each other in terms of reach and splash:

• Claws - range 0.9m, 0 degrees angle and 0% damage;
  
• 1h swords, axes and maces - range 1.1m, 60 degrees angle and 25% damage;
  
• 2h swords, axes and maces - range 1.6m, 120 degrees angle and 50% damage;
  
• Polearms - range 2.6m, 120 degrees angle and 50% damage;
  
• Staves - range 1.6m, 100 degrees angle and 50% damage;
  
• Ranged weapons do not have a splash (see below).

+% secondary target damage is additive to base percent of damage splashed. So, a sword with 50% base splash and +100% to secondary target damage will do a 150% of your base damage as splash damage. Doesn't seems like there's a reasonable cap, I was able to raise splash damage up to 625%.

+% secondary target damage only works with auto-attacks. Skills do not benefit from this affix.
Attacking while holding shift won't turn every target into secondary one - there always will be at least one primary target.

The reason I haven't mentioned Shotgonnes and Cannons is because they do not have a "true" auto-attack and a "true" splash damage. When you attack with a Shotgonne/Cannon, you actually cast a skill which does damage. This skill has a rather complex damage shape.

Let's consider shotgonnes. What arcane statistics in-game tell you:
Splash angle 30 degrees, damage 30%.

What it actually is:
In a 10% angle of your targeting line it does full damage.
In a 10% angle to the left of central sector it does 30% damage.
In a 10% angle to the right of central sector it does 30% damage too.
So, it's indeed a 30 degree arc total, but the damage in different parts of it is not the same. This basically means that, unlike weapons with "true" splash, this one can do full damage to multiple targets - e.g. the one you hit and ones standing behind it.

For a cannon the game lists a 55 degree arc and 50% damage, but it's the same principle as with Shotgonne, just a little different angles.

The important result is that +%secondary target damage does nothing for shotgonnes and cannons, since splash doesn't propagate to %weapon damage skills, which shotgonne and cannon attacks actually are. Another interesting result of this is that Shotgonnes and Cannons do not fumble, which makes auto-attacking with those two a fair bit better than with any other weapon.

Damage - %Weapon DPS skills

%Weapon DPS that is taken by skills is calculated much easier than your real auto-attack dps.

For both minimum and maximum weapon damage:

1. First, take your base damage and multiply it with %weapon_dps modifier of skill.
   

    
   Scaled_damage = damage * %weapon_dps
   
   All % values are considered to be already divided by 100.
   

2. Apply every %damage bonus, in an additive manner. Keep in mind that if damage is non-physical, both strength and focus bonuses are applied.
   

    Example for a stave
   modified_damage = scaled_damage * ( 1 + %damage_bonus_from_strength + % damage_bonus_from_focus + %damage_bonus_from_gear )
   
   All % values are considered to be already divided by 100.
   

3. If you have multiple instances of damage, do the 1-2 sequence above for every instance, and then add them up.
   
4. Apply attackspeed modifier.
   

    
   DPS = scaled_damage * ( 1 + %attack_speed_modifier ) / base_attack_speed
   

The result is basically an amount of damage that %weapon_dps skill does.

Sounds simple, right? Well, there is a 'catch'. I will just list weird stuff from different skills, to make sure you understand what are we dealing with.

Embermage got a skill called Icy blast. I quote - it 'inflicts 30% of weapon DPS as ice damage'. What could be wrong here, you ask? Conversion ratio.

• First of all, the damage seems to be converted before step 1. Let's say you have physical damage weapon. It is first converted into ice damage, and while computing next steps, the game assumes that weapon does ice damage instead of physical. This means strength and focus +%damage bonus is added, this also means +15% ice damage stacks additively with str and focus bonuses, and this also means that +15% all damage is applied once, not twice. In this case, the damage is converted with 1:1 ratio.
  
• Next, take a wand or a staff. Everything is the same, except that damage done is slightly lower than it is supposed to be. Oh well, it's not much. Conversion rate is 0.95:1.
  But stuff gets weird when you apply the flat damage bonuses from weapons. If weapon got +physical damage, this converts with ratio about 0.85:1. If the damage is elemental,
  the conversion ratio is 0.75:1. I don't have logical explanation for this one, it is quite big to be just a roundup mistake.
  
• This also holds true for skills that don't convert damage types. For example, embermage's skill Arc beam and Shocking burst also shows this - calculations match the combat log only if you multiply base staff damage by 0.95 and other sources damage bonus by 0.75. Strange enough, this appears only on enchanted, rare and unique weapons. If you take a 'white' wand, it's damage will match the formulae precisely. Still no explanation of this. If someone knows what's going on - please tell me in comments. My bet is this is some hidden balance crutch left.

Another quirk is that neither skill description nor floating numbers don't show how much damage you're actually doing. Say, we got Arc beam 1/15. 10% weapon dps does sounds like a crap, yes? Sure, if you assume that this damage is done per second. But this is not how this skill actually work. Without any %cast_spees modifiers, it does listed damage five times per second. With stacked %cast_speed I was able to do those 10% of my DPS six times per second. So yes, from that you can deduce that:

• %weapon dps skills have their inner base cast speed, unrelated to attack speed of the weapon
  
• +%cast_speed shortens this value. I don't tested out formulae for this one, sorry. Also, I'm not sure it works that way for every skill.

Also, there is a dual wield. Quite simple here - most skills only take a DPS of main hand weapon into account. The only exception I know is an Arc Beam, which does damage from both hands if
you have two wands, or does two instances of damage instead of one, if you have a wand in a left hand weapon (shield) slot. This pretty much imply that two-handed weapons are better for skills because they have higher DPS.

Damage is not the only thing that scales with %weapon_dps of the skill. Actually, %weapon_dps skill is reported to scale the effect of enemy's armor against the damage of the skill, which is pretty good. This means only 10% of enemy's total armor is applied against your skill that does the 10% of weapon dps, which is quite huge.
On a bad side, your critical chance does the same. I recall there was a fix in changelog where the effect of %weapon_dps on armor soaking and critical chance is limited to 100%. So, 150% weapon dps skills still got your full critical chance, while for a 50% weapon dps skill your critical chance is halved.

Damage - Armor and Damage Soak

Armor works for enemies the same as for you - they reduce damage you inflict on them by a flat number (which is random between X/2 and X, where X - their armor). I'll give a try explaining everything with a set of examples.

Example 1:

Let's say you attempt to inflict 100 physical damage upon your enemy with your autoattack, and enemy has 100 armor. You will actually do from 0 to 50 damage.

Example 2:

Now let's consider a weapon that does 50 physical damage and 50 ice damage, and your enemy has 100 armor and 100 ice resistance. Given the example #1, one could think that you won't do any damage in that case, but the game attempts to be fair here - since your damage is split between physical and ice 50/50, only 50% of armor would be applied against physical part of it, and only 50% of ice resistance would be applied against ice part. So you actually do 0-25 physical damage and 0-25 ice damage.

Example 3:

The same principle works when considering splash damage. If your attack does only 50% splash damage to a secondary target, only 50% of armors/resistances would be applied against it, thus giving it a fair chance.

Things become more complicated for skills. Skills have a hidden stat called Damage Soak Scale, sometimes shortened as either Soak Scale or Damage Soaking. You have no way of checking this parameter in-game, but it exists. Basically, it means to balance low %Weapon DPS skills over high %Weapon DPS skills. One at that point would assume that if a skill does 10% weapon DPS, only 10% of armor/resistances would be applied, and for majority of %weapon DPS skills this is true, but it's not a rule and there are exceptions. For example:

• Embermage - Shocking Burst. Damage Soak is 40% at all levels, %Weapon DPS range from 20% at level 1 to 34% at level 15.
  
• Berserker - Raze. Damage Soak is 100% at all levels, while %Weapon DPS starts at 140% at level 1.
  
• Berserker - Storm Hatchet. Damage Soak is 100% at all levels.
  
• Berserker - Wolfstrike. Damage Soak is 100% at all levels.
  
• Engineer - Blast Cannon. Damage Soak is 100% at all levels.
  
• Engineer - Ember Hammer. Damage Soak is equal to %Weapon DPS up till level 10, and is set to 100% at levels 11-15.
  
• Engineer - Flame Hammer. Actually, a very sophisticated skill. Every part of it is a 100% damage soak, though some projectiles of this skill do only 12% of weapon DPS.
  
• Engineer - Onslaugh. Damage Soak is 100% at all levels.
  
• Outlander - Shadowshot. Damage Soak is 100% for primary projectile and 40% for secondary projectiles (before you rejoice - secondary projectiles only do 40% of the damage that tooltip says they do as well).
  
• Outlander - Poison Burst. Damage Soak is 100% at all levels.

As you see, there are plenty of exceptions. Some of them make the given skill better, some vise versa.

Hidden Damage Soak parameter also exists for plain damage skills, and while you can at least try to guess damage soaking scaling for %Weapon DPS skills, for non-%weapon DPS ones it could be whatever, it even lacks this kind of exception-filled system. Well, not quite. All skills that do flat damage that improves with character level actually have a common damage-versus-level graph. Different skills just multiply this damage by some %X value specific for the skill. For example, Embermage's Brands at level 1 do from 70% to 70% of the base damage, and at level 15 they do from 70% to 210% of the base damage. Damage soaking sometimes equals to either minimum or maximum %, e.g. Brands got 70% damage soak scaling at every level. But since the game doesn't tell you how much damage skill does in relation to base curve, there are no convenient way to deduce the value of damage soaking. And there are much more exceptions than those that abide that rule. Actually, more correct would be to say that for skills that do 100% of base damage of a curve or more, damage soak is set to 100%, which makes those skills even better. And skills that do really minor damage, like Prismatic Bolt, or Arcbeam's fixed electric part of damage, will have a 'fair' damage soak scaling modifier.

There is a common belief among the modders that Damage Soak also affect critical hit chance. That's not true, at least not for %Weapon DPS skills - it's %Weapon DPS itself that scales critical hit chance down, if %Weapon DPS is below 100%. If your skill does 20% Weapon DPS but has a 100% Damage Soak, and you have a 50% crit chance, that skill has only 0.5*0.2 = 0.1 = 10% chance to crit. Damage soaking modifier has no effect on that.

Damage - Other

Pretty much the only thing that is left is skills that do damage depending solely on your level. In most cases, this damage is considered to be a 'magic', so is modified by focus. Those skills typically have either a cooldown, or a hidden cast-speed value as a speed-limiting factor. While the former one can't be changed, the latter is modified by your %cast_speed. Another hidden value is a critical chance factor. Basically, your critical chance is scaled (typically down) for skills that do flat damage depending on level quite the same way as those based on %weapon_dps. But unlike them, you have no means on knowing this value in-game. There is other hidden stuff, like, Arc Beam's range is actually improved with skill level. Who would've guessed?

Damage - DoTs and related

Most typical example of DoT is a Convey X damage over Y seconds item modifiers.
This line actually means that X is a total damage done in Y seconds. Actual damage per second is calculated as follows:

Base tick = 0.5 seconds
Number of ticks = Y / 0.5
Damage per tick = X / Number of ticks  -- this is the floating numbers you'll see
Damage per second = Damage per tick * 2

DoT mechanics:

• DoTs from different sources stack, even if they have the same type of damage
  
• DoT from single source stacks when applied multiple times. Consider each instance of DoT as independant from others.
  
• All DoT's magnitude is improved by focus, even if damage type is physical.
  
• DoTs bypass enemy armor.
  
• DoTs are conveyed to %Weapon DPS skills, but the magnitude of the DoT is changed accordingly.
  
• +% to All damage works with DoTs. +% to damage of type should work with a relevant DoT as well.

Status Effects

Four primary status effects are Burn, Freeze, Shock and Poison.

• Burn's only effect is a damage over time. The amount of damage is scaled by your level and increased by focus and +%fire damage. The damage is rather low - around 400 damage per second at level 100 (excluding focus and +%fire modifier).
  
• Freeze reduces the movement, attack and cast speed by 33%.
  
• Shock makes target spread three lightning particles when struck (this includes both the hit that have inflicted shock and all hits while debuff lasts). Lightning particles move chaotically and damage enemy. It's pretty difficult to detemine how much damage they do. I've done a few tests, base damage it does seems to be ok, but it looks like it's affected by enemy's armor and doesn't scale from your focus.
  
• Poison debuff (don't confuse with poison damage type) decreases target's physical armor and damage by 33%.

There are two ways to inflict aforementioned status effects.

• status effects have a chance to be applied when you do a damage of corresponding type to the enemy. The duration of effects seems to be random between 3 and 5 seconds for freeze, shock and poison, and between 4 and 7 seconds for burn. If target already has the effect and you roll a new one, the effect is overwritten (refreshed).
  
• skills that explicitly state that they do that.

There are also a few patterns:

• Type 1 - for attacks and %weapon dps skills that do not have an explicitly stated status effect. Critical hits always place the debuff provided that the (corresponding) damage you've done, after it is reduced by enemy's armor (of corresponding type), is different from zero. Non-critical hits also can inflict status effects, and the chance is proportional to how much damage you do. Just like in the case of critical hit, zero damage means no status effects. In realistic conditions chances for non-critical hit inflicting a status will be around 5-10%. Also, it happens that status effects are priority encoded. Let's say you have multiple elemental damage types on a single hit, and you roll a critical. This will inflict only one effect - the oldest available from the list: burn > freeze > shock > poison.
  
• Type 2 - skills that are not %weapond dps and do not have an explicitly stated status effect. Critical hits don't guarantee a status effect placed. Probably no priorities. Probably chance proportional to damage you do.
  
• For skills that are not %weapond dps and have an explicitly stated status effect. If explicit status effect occurs, it is inflicted instead, even if zero damage was done. If there are more than one type of explicit status effect (e.g. prismatic bolt) priority encoding doesn't occurs - more than one effect can pass. Critical hits doesn't induce an explicit effect proc. Otherwise it's the same as type 2.
  
• For skills that are %weapon dps and have an explicitly stated status effect. Unstable result, unpredictable overwrites may occur in case of multiple hits in short amount of time, which is often the case. In theory should follow the logic "if explicit effect occurs, then it's the one inflicted, otherwise like type 1".
  
• For skills that do both %weapon dps and non-%weapon dps - type 1 for the former part and type 2 for the latter.

Experience

The main source of experience is killing monsters. The rate of experience gain is penaltized if monster level is way too different from your own. See the table below for more info. The rule of thumb is to farm monsters from two levels below your up to 5-6 levels above your. Source[torchlight.ingame.de]

In the table below:
YourLevel - level of your character;
EnemyLevel - level of monster you killed;
LevelDifference = EnemyLevel - YourLevel.

Enemy Level	Exp rate, %
YourLevel+38 and onwards	33
From YourLevel+5 to YourLevel+37	99 - 2*(LevelDifference - 5)
From YourLevel to YourLevel+4	100
YourLevel - 1	97
YourLevel - 2	92
YourLevel - 3	80
YourLevel - 4	53
YourLevel - 5	25
YourLevel - 6	13
YourLevel - 7	8
YourLevel - 8	6
YourLevel - 9	5

Also there was a report on an exp multiplayer bug. Some say it was partially fixed, but may still be an issue. The key idea is that exp in multiplayer isn't split evenly between the players. It depends on whoever gains the kill, and the one who does gains less exp. More info here[forums.runicgames.com].

Magic Find

In Torchlight 2, %magic finding luck improves quality of items found. It has no effect on drop rate, quantity and probably rarity too.

Let's break it down.

• Drop rate. TL2 uses concept of spawn classes, which are basically functions that spawn objects (items, monsters) when called (on monster death, on chest opening, on map creation, etc). For example, here is a spawn class which drops Grell's Eye from the Grell:
  

  [SPAWNCLASS]
  <STRING>NAME:GRELL_TREASURE_EYE_CHANCE
  [OBJECT]
  <STRING>UNITTYPE:NONE
  <INTEGER>WEIGHT:85
  [/object]
  [OBJECT]
  <STRING>UNIT:TL2_EYEOFGRELL
  <INTEGER>WEIGHT:15
  <INTEGER>RARITY_OVERRIDE:1
  <INTEGER>MINCOUNT:1
  <INTEGER>MAXCOUNT:1
  <BOOL>IGNORE_RANGE:true
  [/object]
  [/spawnclass]
  

  So, the chance of drop is regulated by weights. You have a 15/(15+85) = 15% chance to get a TL2_EYEOFGRELL and a 85/(15+85) = 85% chance to get a wholesome nothing. A hunch tells you that magic find should improve your chances to get a valuable eye, but it doesn't happen. Magic finding luck has no impact on weights. This means that chances to get an eye from a boss are forever 15%, chances to get a skull from blue chests are forever 4.5% and chances to get a skull from a mimic are forever 1.5%.
  
  
• Quantity. The amount of item you get is also regulated by spawn classes - check the MINCOUNT and MAXCOUNT in the code example above. %magic finding luck does not improve your chances to get max amount of items in a spawn class. And since spawn classes is the base game mechanic of spawning stuff, %magic finding item doesn't increase a general quantity of items found.
  
  
• Quality. Let's consider a simple thing - a spawn class that drops a sword. Not a certain item, not an unique item, we just know it's a unittype: sword, nothing else (yet). Most of stuff dropped is actually like that - things like skulls, eyes and pre-determined uniques are more like exceptions. So, the game just rolled a sword for you, and now it has to determine what kind of sword it is. And that's where quality system kicks in:
  

  <INTEGER>NORMAL_ITEM_WEIGHT:10000
  <INTEGER>MAGIC_ITEM_WEIGHT:90
  <INTEGER>UNIQUE_ITEM_WEIGHT:7
  
  <FLOAT>MAGICFIND_MAGIC_INFLUENCE:2000
  <FLOAT>MAGICFIND_UNIQUE_INFLUENCE:2000
  <FLOAT>MAGICFIND_RANDOM_INFLUENCE:1000
  

  So, basically, it's the same system as with drop chance, but is global. It suggests that magicfind has a say in whether the item dropped would be of MAGIC and UNIQUE grade. Experimental results confirm that. Based on experimental data and my hunch, I've deduced the way it works:
  

  NORMAL_ITEM_WEIGHT: 10000
  MAGIC_ITEM_WEIGHT: 90 + 90*(%MF/100)*((2000 + Random(0 to 1000))/100)
  UNIQUE_ITEM_WEIGHT: 7 + 7*(%MF/100)*((2000 + Random(0 to 1000))/100)
  

  This basically decreases the relative weight of normal items, since weight of magic and unique items goes up, increasing the chance that some item becomes magic or unique when deciding it's quality.
  
  
• Rarity. Let's say you've rolled an unique sword. Rarity is the weight-based system that determines whether it would be some super-rare one you've always wanted or "oh, not this one again". %magic finding luck doesn't seem to have impact on making rare items more common, but no particular testing was done.

Bottom line: you'll generally see more occasional unique items, but it won't make impact on your purposeful farming. It will improve situation when your chances to get a goodie are low, but will do nothing in cases when your chances are already programmed to be quite high.

Also, there had been multiple reports about weird bug (feature?). People playing with high magic find ratio reported that they indeed had seen more uniques, but there had been a steady decline in average item level. Check that[forums.runicgames.com] if you're interested. I'm not sure if this was fixed or not.

World cache

Torchlight 2 stores the state of locations you've visited (monsters killed, fog of war, chests opened etc), which had raised a question of how to reset them, for farming purposes. The typical advice would be to host a LAN game while checking "reroll world". In reality, that's not necessary, because the game doesn't actually stores the state of everything. Technically speaking, it caches only last few of locations you've visited, hence the name "world cache". This allows resetting stuff without even leaving the game to the main screen.

• Towns - only last two are stored. Resetting a town leads to resetting every vendor in it. Visiting next town will cause the first one of the two towns you've visited before to reset.
  
  How it's used: let's say you're in Imperial camp. The previous town you've visited is Zery-something. Those are the two towns that are currently stored. Go back to Zeryphesh. Zeryphesh is already stored, so nothing resets; but it does become the town you've visited the last. Go to the Enclaves from here. Enclaves are not in the cache, so it's added to it. Since Zeryphesh is the last you've visited, the one before that - Imperial camp - will be reset. This could be used to shopfarm Act3 and Act 4.
  
  
• Hubs are those big locations with monsters, dungeons and other stuff. Only last two are stored. Same rules as for towns.
  
  
• Passes are the locations between a town and a hub. Only last two are stored. Same rules.
  
  
• Dungeons are everything else. Leaving one dungeon and entering another will cause the former to be reset after 10 minutes.

Using a portal scroll in the area will somehow prevent it from resetting, until you move your personal portal someplace else.

Keep in mind that "reroll world" via checkbox in multiplayer and world cache reset is not the same. Rerolling a world may change the layout of map in your game, while resetting will never change the layout. A combination of both is often used to switch the layout to the most wanted one and then farming it constantly for something.

Conclusion

This guide is written for vanilla version of the game.

The guide is supposed to cover game mechanics only. It wasn't intended to give build advice or tips for farming.

If you think my guide doesn't cover something - feel free to suggest. Who knows, maybe I will do some tests and figure it out.

I'm not a native english speaker, and I'm actually sorry for poor quality of it. It's completely fine to point grammar mistakes in the guide to me in the comments.

[WickedRequiem]

Thank you so much.
Information was very well organized and presented in an easy to understand writing.
Great communication skill.

Kudos to you, sir.

[Ukatan Hunter]

The mechanics in this game are awful. Much worse than in Path of Exile (which is the example of the most perfect Diablo-like game). Everything that you see in in-game descriptions is misleading and, in the best case, half-true.

Author, thank you very much for your time and effort you spent to prove this.

[Weiszklee]

Some more observations:

"health stolen on hit" does not work from spells, only auto-attacks.

everything involving "convey" (silence, stun, DOTs ...) does work from spells, but ONLY if it's in your main hand weapon, not off-hand. Same goes for "minus X to All Armor per hit". You can test that by watching the debuffs on enemies/target dummies.

@心之刃 This might just be the result of having a passive "dual wielding" spell equipped, it's easy to overlook.

[心之刃]

It seems that from https://gaming.stackexchange.com/a/85799 , dual-wielding increases the damage of %dps skills by 10%.

[Jean]

Wow, I know my comment is a bit late but that was very informative, thank you.

[Val]

@Captain McPants
Yes, I did. There isn't much mystery about it though.

[Captain McPants]

I believe you've missed CHARGE.

[Val]

@tyoc213 This guide is dedicated to explaining how things work in TL2. It wasn't meant to give a direct advice on how to build a character, or to be a collection of tips and tricks. There are plenty of other guides for that.

[tyoc213]

You could include a nice balance of stats for each class.

[supremebloodorange]

This guide is nice. For damage reduction, you might want to use Skulls of Limoany instead. One skull of Limoany provides 5% damage reduction. Depending on class, the damage reduction cap can be met with 10 to 15 skulls.
Check the following forum discussion for tips for farming them, specifically post #10. https://steamcommunity.com/app/200710/discussions/0/648812304897053461/