Mini Heal

[Derrick]

There's about a 3.3% (one in 30) chance of EB doing < 10 hp against someone with GM resist, from a caster with GM Magery and Eval

Maybe this is what it is supposed to be, but I think there may be a coding error. This seems to happen FAR more than 3.3% of the time, and it also seems to happen consecutively (maybe there is a better randomizer?).
When I was investigating the exact range of miniheal for son/felix, I shot four ebolts at myself and I was still above 50% health. Maybe 10-14 dmg ebolts are far, far more common -- but, I think there could very well be a coding discrepancy.

I'm going to do some trials and genrate some tables from actual code. This may be.
The 3.3% was calculated by Kaivan yesterday based on the formula we're supposed to be emulating, but not from actual game code. Hopefully we'll find a bug.

I understand what you mean about hapening consecutively; this is a feature of the ranomization that is very similar to OSI, you also see this in consecutive fails and successes in spell fizzles, skill gains, etc...

[Hicha]

UO Demo - July 1998 - Batlins Decompiled Demo Project wrote:
Actual Code: Q5MJ = ((getSkillLevel(user, 0x19) / 0x0A) + dice(0x01, 0x06));
Code Translation: ((Magery / 10) + 1-6)]

Derrick, this is a bit off-topic, but any reason why they chose to go with hexidecimal instead of just decimal for calculations?

[Batlin]

Derrick, this is a bit off-topic, but any reason why they chose to go with hexidecimal instead of just decimal for calculations?

It's a design choice I made when writing the decompiler. OSI stored constants as 8-bit, 16-bit or 32-bit integers in their scripts. By showing them in their hexadecimal form you can see how they were originally stored. A 0 is 0, but it can 0x00, 0x0000 or 0x00000000. I wanted to know the original size in case I wanted to modify a script. For the record, all constants are to be zero-extended to 32-bit and then interpreted as signed integers.

[son]

Whats with the tangent discussions? Focus on mini heal please.

Maybe the demo's "range" is wrong? What if it was capped at 13 but could go as low as 8? I find it quite weird that the range of gh is 11 but mini heal is 3.


I find it quite frustrating that so many people remember single digit mini heals but we cant find a single amount of proof for it!!

[Derrick]

Mini heal range is correct. So many people have looked into this, and no one has found anything showing a less than 10 point heal for a gm mage, but plenty of examples of greater than 10 point heals.

However, interesting news on the spell damage. I did my trials as I said I would above and discovered shockingly more "less 10 point damages" (after resisted) than the expected 3.3%.

Tracing back the cause of this it was found to be what looks (to me) like an apparent bug in the demo code itself. I don't doubt that this bug existed on OSI as well, the question I have is was it ever fixed, and when. I say apparent because we don't know for certain that it wasn't intentional.

The result of the OSI damage determination for spells, is that instead of doing for example 6 rolls of an 8 sided dice, they are rolling an 8 sided dice once, and then multiplying by 6. The result of this method is obviously wild damage ranges; still within the same range (5-42 base) for eb, but with an absolutely flat probability curve.

Having taken the OSI code for damage verbatim and implementing it, we did not notice this (possible) error.

It is still possible that this error may have persisted into T2A and beyond, but in my opinion it is doubtful, and further research on damage ranges for 6th circle spells in old newsgroups have revealed that damage was believed to be fairly predictable for these spells in 1999.

Therefore, I have implemented the correction to the (theoretical) bug, which is (as in the above example) to actually roll 6 eight sided dice and add them together, rather than roll one and multiply the result by 6. The (theoretical) correction to the damage determination is now live on test center (test.uosecondage.com port 2593).

Please don't fill this thread with posts about what a big piece of crap the demo was, and how it's not accurate to anything, we know that the code within the demo server was mostly production code, especially at this level of detail. OSI servers were buggy too, and this is not nearly the first weird bug we've found in their code.

[Kraarug]

Thanks for the chance to test the correction.

[Arkosh Kovasz]

Very, very, interesting.

[Fwerp]

Woot.

[Hemperor]

Correct or not, very interesting, and that's what I love about this shard.

[Arkosh Kovasz]

I'm at work(as usual) but really anxious to hear from someone who has been on Test Center. Hope this at least helps a little bit with some of the issues PvPers are having with our current mechanics.

[son]

if mini heal is correct/accurate as implemented, then why was it used so sparingly during t2a?

Think back, it was only used in near emergencies. Why?

If the range is correct, is its delay correct?

[Hemperor]

Mini heal range is correct. So many people have looked into this, and no one has found anything showing a less than 10 point heal for a gm mage, but plenty of examples of greater than 10 point heals.

However, interesting news on the spell damage. I did my trials as I said I would above and discovered shockingly more "less 10 point damages" (after resisted) than the expected 3.3%.

Tracing back the cause of this it was found to be what looks (to me) like an apparent bug in the demo code itself. I don't doubt that this bug existed on OSI as well, the question I have is was it ever fixed, and when. I say apparent because we don't know for certain that it wasn't intentional.

The result of the OSI damage determination for spells, is that instead of doing for example 6 rolls of an 8 sided dice, they are rolling an 8 sided dice once, and then multiplying by 6. The result of this method is obviously wild damage ranges; still within the same range (5-42 base) for eb, but with an absolutely flat probability curve.

Having taken the OSI code for damage verbatim and implementing it, we did not notice this (possible) error.

It is still possible that this error may have persisted into T2A and beyond, but in my opinion it is doubtful, and further research on damage ranges for 6th circle spells in old newsgroups have revealed that damage was believed to be fairly predictable for these spells in 1999.

Therefore, I have implemented the correction to the (theoretical) bug, which is (as in the above example) to actually roll 6 eight sided dice and add them together, rather than roll one and multiply the result by 6. The (theoretical) correction to the damage determination is now live on test center (test.uosecondage.com port 2593).

Please don't fill this thread with posts about what a big piece of crap the demo was, and how it's not accurate to anything, we know that the code within the demo server was mostly production code, especially at this level of detail. OSI servers were buggy too, and this is not nearly the first weird bug we've found in their code.

Haven't played in a long time but these changes are far more enjoyable. Mini heal still does sort of ruin what 1v1 was, however at least the fights are NEARLY as random and if you lose you probably saw it coming for a while, rather than one random ass huge roll.

2v2 3v3 and fielding will improve much, much more with these changes, if they do happen to go in.

[Hemperor]

Could the newgroup postings looked at be made public?

I've found a few and are still looking that suggest spells were tweaked late 98/early 99 and they talk of consistant damages etc.

Will post when I find something definitive.

EDIT: I'm really leaning towards Feb99 being when this change could have possibly happened, the meditation patch. Anyone with any info etc. pm me

[Pro]

The new ebolt damages has no real effect as mini heal is still so high its sort of useless.

[Arkosh Kovasz]

The new ebolt damages has no real effect as mini heal is still so high its sort of useless.

But are the ebolt damages much less random? I recently heard this a complaint on why PvP here was no fun;that ebolts were so random it was making PvP more about luck. If the damage is more consistent, thats one more step toward PvP being better.