Failing to learn Identify

moopy

This happen to anyone else? Maybe it is just me, but it seems I fail attempting to learn identify more than anything. I failed it 3 times in a row, and I've failed it at least once every other game. Always with 18-19int.

JuliusBorisov

Try learning Know Alignment. I bet you'll succeed the very first time every other game! Or Infravision))))

Mykra

I use BGTweaks and have the 100% learning rate on. A bit of oddness on my part - I like randomizing HP, but I hate gambling on spell scrolls.

Calmar

CAn't you also lower the game difficulty in order to avoid failure?

Wowo

I failed to learn it 4 times in 1 game. I think it's just something that we pay attention to.

lunar

I copied it to my book on the first try with 17 INT. It's just luck, I guess. Failed to copy melf's acid arrow, and blindness scrolls dropped by ogre encounters. I won't try scribing very important stuff (glitterdust, fireball, stoneskin, emotion) without genius potions.

Silverstar

Well I've been playing Sorcerer starting with EE and I always pick it as one of the first two spells... but I can't remember having trouble with any spell more than another in the past. I do feel that 85% is a fake number though as I've reloaded up to 10 times on occasion, but it's been for several different spells. [Q] and [L] are very good friends at times.

griever0483

I think learning spell % is buggy. I'm pretty sure of it

[Deleted User]

The user and all related content has been deleted.

mlnevese

Someone made a statistical run with 3000 attempts to learn spells and proved that in the long run it does fall within the % to learn spell. I just can't find the link to that thread right now.

Wilbur

I think it's just variance.

SirK8

I failed to learn identify 3x with Dynaheir and a potion of genius, it was very frustrating.

Fighting_Ferret

While not in the % to learn spell from scrolls, I had Khalid with a -2 AC (Plate, Medium Shield, Ring of Princes and the Elves Bane Belt) Which gives him a -5 AC vs missiles took 3 simultaneous crits from the kobolds outside of Molahey's area in the mines killing him instantly (he was only level 2).

Is it just me or do kobolds with bows seem like they hit, even well armored targets, a bit too often.

Agricola

1/20^3 is 1 in 8000. since you kill a thousand+ critters in the game somebody is probably going to roll a natural 3x in a row once or thrice a game.

Since bows attack faster than meelee and kobolds are in groups you'll see more natural rolls from them.

reedmilfam

The number of kobold arrows made me use Sword & Shield proficiency, actually. I know that the effect isn't quite as good as some would like, but any help against arrows is welcome for me! Also, in this case, the character is a paladin, so there's no need to push proficiency toward grand mastery. As a Cavalier, weapon choices are also limited (another reason I'm willing to use points for S&S).

Xavaine

Reduce the game difficulty to easiest any time you are learning a new spell. A bit of cheese here and there is better than constant frustration.

Dee

I'm still on the fence about whether this is a bug or just a string of bad luck. If it's a bug, it's likely caused by the number generator using in-game variables (which wouldn't change after a reload, causing less random numbers) rather than system variables (which would change regardless of the saved game).

BelgarathMTH

18 INT only gets you up to 85 percent chance to learn a spell. With the number of scrolls you'll be scribing, that's still a pretty high failure rate. With some of the most important spells only having one or two copies in the whole game, there's a very good chance you can wind up with no access to emotion:hopeless, cloudkill, lower resistance (bg2), spell immunity (bg2), or even web or fireball.

I've seen critically rare spells fail to scribe even with a potion of genius and 25 INT.

I think there's pretty wide agreement that the INT system for mages is badly implemented in the game, which is why so many people either tweak it out or turn the slider down for spell-scribing. It would have been better to use the "spell level plus ten" system for INT and spells, but oh well, it is what it is.

Pantalion

I failed three times in a row to learn "Skull Trap" once.

This was at the point in the game where there are, at most, three scrolls of skull trap. In the same game I also failed to scribe Flame Arrows, the singular Fireball scroll, and every Lightning Bolt scroll I found.

In the same game I succeeded in scribing every single Infravision.

moopy

To clarify, it isn't that I think it is buggy. I have great luck learning other spells given 18int is an 85% rate. I just thought it was funny that I managed to buy out High Hedges identify scroll stock once and still fail to learn it. I could cast fireballs and other higher magic but I couldn't identify magic armor/weapons

griever0483

Quick load solves all

the_spyder

Tiax Rules All.

But seriously, it has been discussed that the rng in the game has a tendency to create artificial runs of rolls. That things with low degrees of failures can sometimes tend to fail repeatedly due to this issue.

With that having been said, I could see someone in the development team adding an additional failure rate for Identify because of it's almost universal value and the fact that it is a great way to drain money. This is all 100% fanciful speculation on my part and based in no part on fact. But it is nice to imagine... Maybe it is true? Who knows?

TJ_Hooker

mlnevese said:

Someone made a statistical run with 3000 attempts to learn spells and proved that in the long run it does fall within the % to learn spell. I just can't find the link to that thread right now.

One caveat to these results is that he didn't reload at all during the trials. I don't pretend to know much about random number generation, but I've heard a bunch of people talking about "seed" playing a big part, and that a new seed is generated everytime you start the game (or maybe it was every time you load a game). Most people won't go nearly that long without without reloading/exiting and restarting the game, so would it be possible that a bad seed would cause the first bunch of random numbers to be skewed, but after a while would level out, allowing for the results seen in the above mentioned test? Like I said, I know next to nothing about RNG, so I could be totally wrong here.

griever0483

I agree with @TJ_Hooker . Random algorithms use "seed" (a starting point from which the algorithm goes on). However I think there are others factors otherwise a failed learnt spell will be like that after every load game. But the question is: do we want to play or study information theory? :P
I like to play... and quick load.

griever0483

However...every time my mages fail in learning a spell I curse Tymora!

mlnevese

Actually I commented about probable seed generation problem leading to possible strings of good/bad luck. People just do not notice the strings of good luck.

moopy

I've seen older RNGs with a seed based off of current time in such a way that a number was more likely to lean on the low end or high end a few times in a row, and not dependent on save/load.

Either way, it is bad when you burn through every id scroll at High Hedge and you are playing no reload.

the_spyder

@mlnevese, yep. I'd bet that is indeed the case.

There was a similar thread not to long ago about the general % chance to learn a spell. Lots of anecdotal experiences of fails (and someone trying to relate it to the chicken scenario).

In the end, I'd personally like to see the outcome of a 3000 roll test and see how it fairs. I'd bet it all works out pretty close to the percentages. But who knows. I don't think it is so broken that the percentages are significantly off. If they were, I'd think a whole lot more people would be posting on the issue.

TJ_Hooker

Here's the thread @mlnevese was talking about. I linked to the second page, at the top of which are the results of the 3000 trial run.
http://forum.baldursgate.com/discussion/12514/chance-to-learn-spell-is-bs/p2

Silverstar

mlnevese said:

Actually I commented about probable seed generation problem leading to possible strings of good/bad luck. People just do not notice the strings of good luck.

True. But with 85% chance to learn a spell you'd think "Huh, I failed?" would be a much more common statement than "Huh, I succeeded?". Not going to claim the game is lying because I cannot prove that; I might just have shitty luck in the game as much as I do everywhere else. But it really really doesn't feel like scroll learning triggers most of the time.

mlnevese

@the_spyder the only problem is that to prove it's a problem with the seed you'd have to make 2 experiments, at least.

Let's keep the 3000 rolls for consistency with the previous experiment.

1) 3000 rolls without restarting the game, check not only if the rolls fall within the expected bell curve, but if you have unusual strings of good/bad luck.

2) 3000 rolls, quit the game between each roll and run again. Compare overall bell curve of rolls as well as unusual strings of good/bad numbers.

If my theory of bad seeding mechanism is right this will happen:

a) in both scenarios you end up falling within the bell curve in the long run.

b) in scenario 2 you get a lower probability of long strings of high/low numbers.

The problem is the first scenario is easy to test, you can make a game script for it that will even store the results for you. But to run experiment 2 you have to write some external program that will interact with BG running it, quitting it, and storing the results. I don't even know if it's possible actually.