Spell failure annoys me when it is a 'must have' spell . It is part of the fun to obtain a rare scroll, such as cloudkill or ghost armor , and we feel that it was taken from us when failing to learn it.
Odds of a 5/7 outcome assuming 50%? 19% chance Odds of at least 5 or fewer successes assuming 50%? 39% chance
Definitely not the product of something taking place prior to the game save. I'm not sure what the exact odds are. I would say these are consistent with a pretty wide range of underlying probability, including the 50% that I've previously seen.
Really? I thought I read that it was based on a charisma check. That'll teach me to trust what I read on the...wait a minute, these forums are on it too...
(As a side note, I have noticed strings of bad rolls in attack and damage rolls, not just spell writing. Which suggests that it's more widespread, and just only noticeable in the instances where it has irrevocable consequences.)
Now that you've said that, I have to agree. The bad results in to hit rolls are also strange, when you start getting bad results in the to hit roll it's rarely a single one. They seem to for a sequence of bad rolls, just as when the spell learning roll starts to fail.
Next time something like this happens, I'll quit the game and restart to force a new seed. It won't prove anything but may be a clue to indicate a problem in the random number generator.
(As a side note, I have noticed strings of bad rolls in attack and damage rolls, not just spell writing. Which suggests that it's more widespread, and just only noticeable in the instances where it has irrevocable consequences.)
Now that you've said that, I have to agree. The bad results in to hit rolls are also strange, when you start getting bad results in the to hit roll it's rarely a single one. They seem to for a sequence of bad rolls, just as when the spell learning roll starts to fail.
Next time something like this happens, I'll quit the game and restart to force a new seed. It won't prove anything but may be a clue to indicate a problem in the random number generator.
I would be wary on this without collecting more serious data. If you have a 75% chance to hit, for example, you will miss several times in a row not infrequently. 3 misses in a row should occur every 64 attacks, for example. Given the thousands of times you attack someone, these patterns are going to show up.
I would take a methodical approach and document the rate of occurrence over a significant sample size (143 times is statistically significant for a 95% confidence level).
(As a side note, I have noticed strings of bad rolls in attack and damage rolls, not just spell writing. Which suggests that it's more widespread, and just only noticeable in the instances where it has irrevocable consequences.)
Now that you've said that, I have to agree. The bad results in to hit rolls are also strange, when you start getting bad results in the to hit roll it's rarely a single one. They seem to for a sequence of bad rolls, just as when the spell learning roll starts to fail.
Next time something like this happens, I'll quit the game and restart to force a new seed. It won't prove anything but may be a clue to indicate a problem in the random number generator.
I would be wary on this without collecting more serious data. If you have a 75% chance to hit, for example, you will miss several times in a row not infrequently. 3 misses in a row should occur every 64 attacks, for example. Given the thousands of times you attack someone, these patterns are going to show up.
I would take a methodical approach and document the rate of occurrence over a significant sample size (143 times is statistically significant for a 95% confidence level).
I agree that a significant statistical analysis or access to the source code would be required to prove anything. But I also find strange that when you start failing rolls in the game it's never a single occurrence, in my experience. The worst case I had was the 5 failures in a row with a 95% chance of success...
Re: Melicamp and charisma check - I believe that is related to whether or not he will talk to you a second time if you run him off originally.
Similar to the charisma check when picking up Viconia - fail the check and she will not join (or even offer to join) even after you have saved her. I can't recall if it worked this way in vanilla BG or not as I rarely used her but it occured twice already in BG:EE. (with 8 charisma)
I once watched for five minutes while my character and a monster exchanged miss after miss.
And yeah, while this sucks? It is not impossible from a statistics perspective. Not saying that there isn't anything wrong with the engine, there might be. But statistically speaking it is POSSIBLE for this to happen and nothing be wrong.
Way back when I was young, I played D&D with this group. One time this new guy wanted to join. This was back in the day when Stats were rolled (like in BG) using 3D6. He presented a character whom he claimed was legitimately rolled up. The Character in question had all 18 stats and an 18/00 strength. When we questioned him about it, he provided the program and the work sheet that had rolled up some unreasonable number of stats and eventually spit out the character in question. It was Legitimate (after a fashion), if very, VERY improbable.
Again, not saying there isn't a problem. Just saying that anecdotal observations like this do not prove anything other than a data outlier or anomaly. Because how many "To Hit" rolls does any individual game involve? thousands, probably. Are you sure that there wasn't a run of improbable hits somewhere else that offset this? it kind of sucks, but the only real way to prove anything would be to run maybe ten thousand rolls and compare them. And that would still be on the low end, but at least would give directionally an idea.
Oh, totally, it's definitely possible. And it usually happens in the early stages of the game, which makes it even more likely (since THAC0 is worse). I was just adding it to the list of "things that might or might not be borked in the RNG", because like you said, if it's a problem with the RNG, it's going to be a problem everywhere in the game and not just in the spell writing.
Unless Beamdog updated the RNG which I doubt they did.
I would bet that the RNG isn't super great in terms of being "random" given the year this game was made.
I'm sure over thousands of tests the odds would come out pretty close to being correct, but I would guess you would get pockets of really bad luck or really great luck here and there that aren't truly "random".
Odds of a 5/7 outcome assuming 50%? 19% chance Odds of at least 5 or fewer successes assuming 50%? 39% chance
Definitely not the product of something taking place prior to the game save. I'm not sure what the exact odds are. I would say these are consistent with a pretty wide range of underlying probability, including the 50% that I've previously seen.
Interesting... I've searched through DLG, BCS (the reference to Antichickenator is located in THALANTR.BCS), and SPL files for any mechanism... One thing I found was that the check was entirely reliant on BCS, spell (2989 EFFECT_ONLY, come to think of it, what an appropriate name for it) only providing visual effect. It seems that there are only two outcomes, the computer randomly assigns the outcome in 50 50 chance.
Unless Beamdog updated the RNG which I doubt they did.
I would bet that the RNG isn't super great in terms of being "random" given the year this game was made.
I'm sure over thousands of tests the odds would come out pretty close to being correct, but I would guess you would get pockets of really bad luck or really great luck here and there that aren't truly "random".
This is more or less what I suspect as well. Which is why I was suggesting that people do independent testing to prove their hypothesis. If it turns out true that there is a problem, I think we all would be invested in it being fixed.
However, if the suspicion is based on anecdotal data, not so much.
Next time something like this happens, I'll quit the game and restart to force a new seed. It won't prove anything but may be a clue to indicate a problem in the random number generator.
Getting clumps of similar results is not an indicator of non-random selection. If you flip a coin six times, you are just as likely to get six heads as you are to get heads-tails-heads-tails-heads-tails. In any long list of random numbers, there are guaranteed to be runs of things that look like patterns but aren't. Your brain is optimized for detecting patterns, and tends to see them whether or not there is any underlying cause.
I'm not an expert in random number generation issues, but my understanding is that RNG problems tend to be related to insufficient entropy in the seed which results not in clumping of results, but in predictable patterns. For rolling dice in a computer game, this doesn't matter in the least; it is only when using random numbers do do things like generating cryptographic keys that it matters.
Maybe if someone has gone out and written their own pseudo-random number generating algorithm instead of using a standard system-supplied one, and they did an especially bad job of it, you might get some weird results, but that would be a weird thing for a programmer to do in the first place.
Maybe if someone has gone out and written their own pseudo-random number generating algorithm instead of using a standard system-supplied one, and they did an especially bad job of it, you might get some weird results, but that would be a weird thing for a programmer to do in the first place.
Particularly in this case. I don't imagine that Overhaul games had it in scope to rewrite the RNG. I mean they COULD, but I would be very surprised if they did. And I would expect the net result to be significantly better results rather than worse merely because of the 15 years of advancement between release and now.
I came across something completely at random that I thought pertained to this (possibly defunked) thread. Apparently "Specialty Mages" in BG1/2 suffer a -15% to learning spells that aren't in their chosen school. I kind of wonder if this impacted anyone's perceptions during the pursuit of this thread?
Interesting. I found it on Sorcerer's place and they usually know their stuff. but then it was specifically for BG2? Or maybe they were thinking that it would be implemented? Or maybe someone thought that it was? I know there are a lot of things that were supposed to be there but aren't.
Interesting. I found it on Sorcerer's place and they usually know their stuff. but then it was specifically for BG2? Or maybe they were thinking that it would be implemented? Or maybe someone thought that it was? I know there are a lot of things that were supposed to be there but aren't.
Either way, thanks for the info.
Well, you can mod the 2da file to be 100% spell writing success for any level of intelligence - and you will still be able to fail! So it probably was implemented. A natural 19 intelligence will earn you 150% chance of success.
Comments
Character: Level 1 Inquisitor - 17 charisma
Predicate Actions: Saved Melicamp, brought him to Thalantyr and agreed to help. Got the skull.
Setting Up Conditions for Test: Saved. Each result is loading the saved game followed by handing the skull to Thalantyr and nothing else.
Results:
#1 - Failure
#2 - Success
#3 - Failure
#4 - Failure
#5 - Success
#6 - Failure
#7 - Failure
#8 - Success
#9 - Failure
#10 - Success
#11 - Failure
#12 - Success
5 out of 12 successful / 7 out of 12 failures
Odds of a 5/7 outcome assuming 50%? 19% chance
Odds of at least 5 or fewer successes assuming 50%? 39% chance
Definitely not the product of something taking place prior to the game save. I'm not sure what the exact odds are. I would say these are consistent with a pretty wide range of underlying probability, including the 50% that I've previously seen.
Next time something like this happens, I'll quit the game and restart to force a new seed. It won't prove anything but may be a clue to indicate a problem in the random number generator.
I would take a methodical approach and document the rate of occurrence over a significant sample size (143 times is statistically significant for a 95% confidence level).
Similar to the charisma check when picking up Viconia - fail the check and she will not join (or even offer to join) even after you have saved her. I can't recall if it worked this way in vanilla BG or not as I rarely used her but it occured twice already in BG:EE. (with 8 charisma)
Way back when I was young, I played D&D with this group. One time this new guy wanted to join. This was back in the day when Stats were rolled (like in BG) using 3D6. He presented a character whom he claimed was legitimately rolled up. The Character in question had all 18 stats and an 18/00 strength. When we questioned him about it, he provided the program and the work sheet that had rolled up some unreasonable number of stats and eventually spit out the character in question. It was Legitimate (after a fashion), if very, VERY improbable.
Again, not saying there isn't a problem. Just saying that anecdotal observations like this do not prove anything other than a data outlier or anomaly. Because how many "To Hit" rolls does any individual game involve? thousands, probably. Are you sure that there wasn't a run of improbable hits somewhere else that offset this? it kind of sucks, but the only real way to prove anything would be to run maybe ten thousand rolls and compare them. And that would still be on the low end, but at least would give directionally an idea.
I would bet that the RNG isn't super great in terms of being "random" given the year this game was made.
I'm sure over thousands of tests the odds would come out pretty close to being correct, but I would guess you would get pockets of really bad luck or really great luck here and there that aren't truly "random".
THALANTR.BCS
------------------------------------------------------------------------
IF
Trigger(1)
THEN
RESPONSE #50
StartCutSceneMode()
Face(1)
SmallWait(1)
ForceSpell("Melicamp",EFFECT_ONLY)
Wait(1)
ActionOverride("Melicamp",Polymorph(MAGE_MALE_HUMAN_LOW))
EndCutSceneMode()
RESPONSE #50
StartCutSceneMode()
Face(1)
SmallWait(1)
ForceSpell("Melicamp",EFFECT_ONLY)
Wait(1)
ActionOverride("Melicamp",Polymorph(MAGE_MALE_HUMAN_LOW))
Kill("Melicamp")
EndCutSceneMode()
END
------------------------------------------------------------------------
So I believe you have toe closest answer. It's random 50-50 chance; it either works or it doesn't.
However, if the suspicion is based on anecdotal data, not so much.
I'm not an expert in random number generation issues, but my understanding is that RNG problems tend to be related to insufficient entropy in the seed which results not in clumping of results, but in predictable patterns. For rolling dice in a computer game, this doesn't matter in the least; it is only when using random numbers do do things like generating cryptographic keys that it matters.
Maybe if someone has gone out and written their own pseudo-random number generating algorithm instead of using a standard system-supplied one, and they did an especially bad job of it, you might get some weird results, but that would be a weird thing for a programmer to do in the first place.
Who knew?
Either way, thanks for the info.
Guess you learn something new every day.
A natural 19 intelligence will earn you 150% chance of success.