Original windows BoE: defend 0

Looks like there was a nerf of shirts between E3 and BoE.

This is one of the areas where I think it's fine to depart from the original BoE and treat it as fixing a bug from the original game.

If only it were easy to come up with an actual list of these differences…

There are several different issues at play here.

1. In the item info screen, Exile III shows different item stats than Blades of Exile and OpenBoE. Defend 1 versus Defend 0, as shown at the beginning of this bug report.

2. On the item pickup screen, Exile III and OpenBoE show different item stats than Blades of Exile. Blocks 1-1 damage versus Blocks 0-0 damage. In this case, OpenBoE is actually more correct than the original game in how it displays item stats on pickup.

Exile III:

OpenBoE:

Blades of Exile:

3. In Blades of Exile, Shirts still block 1 damage despite showing Defend 0 and Blocks 0-0 damage. Just a note that the original game had a discrepancy between reported defense versus calculated defense. I tested this with a 5d1 physical damage special where, with all else being equal, wearing a Shirt protected the wearer from 1 damage (expected 0).

Before taking damage:

After taking damage:

4. In Blades of Exile and OpenBoE (using OpenBoE code here as the example), the reason that an equipped Shirt protects the wearer from 1 damage is because it of type armor, and equipped armor that has Defend (internally referred to as item_level) less than 1 is treated as equivalent to 1 during damage calculation. In boe.party.cpp, within the damage_pc() function, we see this line:

short defense = get_ran(1,1,item.item_level);

And in the mathutil.cpp function, we have the following relevant lines of code for get_ran():

short get_ran (short times,short min,short max, bool use_unique_ran){
        ...
	if(max < min) max = min;
	if(max == min) return times * min;
        ...

For the Shirt, it is essentially calling get_ran(1, 1, 0) where min is 1 and max is 0; since max is less than min, max becomes 1, and 1 is returned for defense. Note that, at least in the base game, the only items that this issue affects are Shirts. Pants are not affected because they are of type pants rather than armor, and no other armor item has a Defend (item_level) less than 1 aside from Shirts. It appears that the protection attribute is where negative modifiers to damage protection are calculated (e.g. for Feldspar Charm).

The implications of this are that a Shirt has the same defensive value as a Leather Baldric, and is better than a Crude Buckler (due to both encumbrance and weight). I believe that changing the code here (to where get_ran is called with min set to 0 rather than 1) would make sense, as it will not affect any other base game item and will align player/creator expectations with what is happening in the game engine.

Are other armor types intended to always block at least 1 damage? If so, then I think it would be better to treat armor level 0 as a special case which doesn't call get_ran(), and the other armors keep 1 as the minimum. And if the UI currently says they block 0-x damage then we should change those to say 1-x damage.

I'm not the right person to be making calls about game balance, though. I have no credentials for RPG combat design or tuning.

No matter how we fix this, we should do it with a feature flag (I'll implement those asap) so that old replays can keep their dice rolls the same.

You're right, my mistake. I think armor is expected to always block at least 1 damage. Adding a conditional check for item_level > 0 would be the right approach here, and this is already done for the item bonus and protection values as well. Would want to make sure that the correct values are displayed on the item pickup screen as well.

I greatly appreciate you looking into to this and doing the hard part (figuring it out)!

A way to fix this and PR without waiting for feature flags, would be to add the conditional and make it return 0 for 0 armor level, but still call get_ran(), just dropping the value.

We have a replay system which relies on keeping seeded randomness the same between versions of the game. If we add or remove a dice-roll, combat in replays diverges greatly.

Assigning myself to fix the shirt defense bug. If we need to go further and find/fix more discrepancies, I don't really know how.

Reading back through this issue, I'm seeing that the core problem is the handling of armor value 0 and displaying how much damage it will block. So I will consider it fixed just by handling those things.

Agreed. The issue isn't specific to Shirts, it's just the most obvious example. Other items affected include the Glue Gauntlets and the Magician's Hat, not to mention items in custom scenarios. Your changes should resolve the issue.