Skip to content

My bard fails too much when trying to learn spells

2

Comments

  • ZilberZilber Member Posts: 253
    PK2748 said:

    PK2748 said:

    Actually in 1st edition you had to dual Fighter to Thief and Thief to Druid to even become a Bard and in 2nd edition there was no limit to the combinations created by dual class. Bard had prime requisites so you could dual to and from the Bard class. With the complete Bard book you could also make a significant number of multi class Bard combos in 2nd edition.

    And this is why people loathed the 1st edition after 2E came out. It's also why many people (myself included) prefer the intuitive nature of the 5E ruleset.
    What's "intuitive" about somebody who spent 7 years apprenticed to a Mage killing 30 hobgoblins with magic missiles and then suddenly being a rogue?
    Let's not take that discussion here. It started with the question how my Safana ended up a bard, that has been answered (eekeeper). Difference in editions is too big of a discussion to derail this thread with
    jackjackAlonso
  • The user and all related content has been deleted.
  • jackjackjackjack Member Posts: 3,251
    Isn't there a hard cap of 95%? Or is that just at 19 INT?
  • AstroBryGuyAstroBryGuy Member Posts: 3,437

    PK2748 said:

    Actually in 1st edition you had to dual Fighter to Thief and Thief to Druid to even become a Bard and in 2nd edition there was no limit to the combinations created by dual class. Bard had prime requisites so you could dual to and from the Bard class. With the complete Bard book you could also make a significant number of multi class Bard combos in 2nd edition.

    And this is why people loathed the 1st edition after 2E came out. It's also why many people (myself included) prefer the intuitive nature of the 5E ruleset.
    1e's bard is a bit of an aberration. The OD&D bard was a regular class, like the 2e bard.

    http://annarchive.com/files/Strv201.pdf
    jackjack said:

    Actually, I think in 2E you could use your old skills at anytime, but doing so forfeits any experience you gain for that encounter or action. Anytime as you use only the skills available to your second class, you gain experience in that class.
    It's just goofily implemented in the BG series.

    It's a bit worse than that: ...if he uses any of his previous class's abilities during an encounter, he earns no experience for that encounter and only half experience for the adventure. This last part would be hard to implement, since BG awards XP instantly, rather than at the end of the adventure or session, like in PnP.
  • AstroBryGuyAstroBryGuy Member Posts: 3,437
    Alonso said:

    What's your point? That's what we have discussed already (with far less mathematical detail, certainly).

    The last part was chopped for some reason. I suggested testing without the potions to see if the effect was the specialist penalty being applied or the potion INT bonus being ignored.
    Alonso
  • AlonsoAlonso Member Posts: 806
    Buf, I'm getting a headache with so many numbers. I still tried as you suggested: 18 INT, no potions. 100 attempts, 78 successes, so that's 78%. Expected value: 85%. It isn't as bad as other cases, but it still doesn't look good.

    I also tried with 25 INT (two potions). This worked fine. I learned all the spells from a sample of 50.

    Anyway, I say we've made a really nice effort here already, let's let the pros (for programmers) mince the data and point the culprit. Unless you come up with another good idea, of course.
  • joluvjoluv Member Posts: 2,137
    Alonso said:

    100 attempts, 78 successes, so that's 78%. Expected value: 85%.

    If the true probability is 85%, then there's a 3.9% chance of 78 or fewer successes in 100 attempts.
    Alonsobob_veng
  • AlonsoAlonso Member Posts: 806
    I know. That's why I say "it doesn't look good" rather than "it's plainly wrong". But considering how many tests we have performed already around here (you might want to check the bug report and the other similar thread), the evidence has become overwhelming that something is wrong.
    BelgarathMTH
  • joluvjoluv Member Posts: 2,137
    Oh, I wasn't arguing. Just quantifying.
    Alonso
  • DjinnDjinn Member Posts: 76
    edited May 2016
    Just to add my 2 cents, I have this very same problem in IWDEE (v1.4). Also, only with bards kitted/unkitted, not with mages of any kind. Raising INT to 25 eliminates the problem, but anything below that means about a third of the spells go unlearned on anything higher than Easy mode.
    Also, for some reason the problem is worse with some spells more than others. Especially Chromatic Orb and Slow. Bards hate those two spells, I guess, hehe.

    Edit: Haven't tried with Bards in BGEe/BG2EE/SOD
    Alonso
  • StevevdlStevevdl Member Posts: 73
    @Alonso Am I reading it correctly that you are complaining about missing 1-3 spell learning at 98%.
    You do realise that 98% is not 100% so is not a guaranteed learn. It is quite possible that the ones you missed rolled 99 or 100, which makes a fail.
    Now if you only learned 1-3 out of the 14 then I would suspect a bug or unlikely very bad rolls
  • StevevdlStevevdl Member Posts: 73
    edited May 2016
    Alonso said:

    Buf, I'm getting a headache with so many numbers. I still tried as you suggested: 18 INT, no potions. 100 attempts, 78 successes, so that's 78%. Expected value: 85%. It isn't as bad as other cases, but it still doesn't look good.

    I also tried with 25 INT (two potions). This worked fine. I learned all the spells from a sample of 50.

    Anyway, I say we've made a really nice effort here already, let's let the pros (for programmers) mince the data and point the culprit. Unless you come up with another good idea, of course.

    The problem with your maths is that you have no idea what was rolled on the pass fail, just because you have an 85% chance to do something, and your test gives you 78% pass rate does not make the maths wrong, you are wrong for not taking into consideration what numbers you rolled to make your judgement. All your maths proves that at an 85% chance to learn you rolled 85% or lower 78% of the time.
    When you rolled to learn, did you take note of what the computer rolled for your chance to learn and was it over or under 85%. If any of the rolls showed you rolled 85 or less and was still a fail then it's a bug.

    With an int of 25 and 100% chance of learning spells of course you are not going to fail


  • joluvjoluv Member Posts: 2,137
    @Stevevdl: The randomly generated numbers aren't visible in-game, so there isn't an easy way to check that information. Everyone here understands that it's possible to have only 78 successful rolls out of 100 when the success probability is 85%. We're just saying that it's relatively unlikely. Specifically, as I noted above, there would be a 3.9% chance of that happening.
    Alonso
  • alceryesalceryes Member Posts: 380
    joluv said:

    @Stevevdl: The randomly generated numbers aren't visible in-game, so there isn't an easy way to check that information. Everyone here understands that it's possible to have only 78 successful rolls out of 100 when the success probability is 85%. We're just saying that it's relatively unlikely. Specifically, as I noted above, there would be a 3.9% chance of that happening.

    I guess the question is, how big of a sample size is needed to determine an average?
    Is 50 enough? How about 150? How about 500? The larger your sample size is the closer you 'should' be to the listed percentage chance (taking into account all applicable modifiers).
  • kjeronkjeron Member Posts: 2,367
    How about this:
    Changed the Value in INTMOD for all intelligence values to 100% chance to learn.

    L7 Jester repeatedly learning/erasing L1 Identify - failed 6 of 50.
    Observed: 88% success rate.
    Expected: 100% success rate.

    L1 Jester repeatedly learning/erasing L9 Bigby's Crushing Hand - failed 8 of 50.
    Observed: 84% success rate.
    Expected: 10% success rate.
    Alonso
  • StevevdlStevevdl Member Posts: 73
    joluv said:

    @Stevevdl: The randomly generated numbers aren't visible in-game, so there isn't an easy way to check that information. Everyone here understands that it's possible to have only 78 successful rolls out of 100 when the success probability is 85%. We're just saying that it's relatively unlikely. Specifically, as I noted above, there would be a 3.9% chance of that happening.

    alceryes said:

    joluv said:

    @Stevevdl: The randomly generated numbers aren't visible in-game, so there isn't an easy way to check that information. Everyone here understands that it's possible to have only 78 successful rolls out of 100 when the success probability is 85%. We're just saying that it's relatively unlikely. Specifically, as I noted above, there would be a 3.9% chance of that happening.

    I guess the question is, how big of a sample size is needed to determine an average?
    Is 50 enough? How about 150? How about 500? The larger your sample size is the closer you 'should' be to the listed percentage chance (taking into account all applicable modifiers).
    1. You are dealing with random numbers, you say you understand random numbers, then are looking for an average, for a bunch of numbers that can change every single time you test it.
    What are random numbers?

    Random numbers are sets of digits (i.e., 0, 1, 2, 3, 4, 5, 6, 7, 8, 9) arranged in random order. Because they are randomly ordered, no individual digit can be predicted from knowledge of any other digit or group of digits.

    You are trying to put order into something that has no order, that is where you are going wrong
    Why is it unlikely to have 78 successful rolls at 85% pass rate? When you are rolling the dice ANY number can come up. Just like the OP complaining he missed a couple of rolls at 98%, and thinks it wrong that he failed any of them even though it is 98%.

    2. Why should you get closer to your answer you want, with a larger sample size. In all this you are forgetting 1 important fact. you are dealing with random numbers. Even if you added more rolls there is nothing stopping the roll from being a ton of fails which will bring your actual success rate lower
    Oh and randomization and maths don't mix, look at you all trying to make sense of something that is naturally chaotic to start with.
    Mr2150
  • AlonsoAlonso Member Posts: 806
    Stevevdl said:

    randomization and maths don't mix

    Looks like you might be a bit mislead here, my friend. You might be interested in finding out a bit more about mathematics and probability. This article may help.

    joluvmf2112
  • AlonsoAlonso Member Posts: 806
    edited May 2016
    kjeron said:

    Changed the Value in INTMOD for all intelligence values to 100% chance to learn.

    Can you explain the meaning of that? I guess I understand it more or less, but I'd like to be sure.
  • joluvjoluv Member Posts: 2,137
    Stevevdl said:

    1. You are dealing with random numbers, you say you understand random numbers, then are looking for an average, for a bunch of numbers that can change every single time you test it.
    What are random numbers?

    Random numbers are sets of digits (i.e., 0, 1, 2, 3, 4, 5, 6, 7, 8, 9) arranged in random order. Because they are randomly ordered, no individual digit can be predicted from knowledge of any other digit or group of digits.

    You are trying to put order into something that has no order, that is where you are going wrong
    Why is it unlikely to have 78 successful rolls at 85% pass rate? When you are rolling the dice ANY number can come up. Just like the OP complaining he missed a couple of rolls at 98%, and thinks it wrong that he failed any of them even though it is 98%.

    2. Why should you get closer to your answer you want, with a larger sample size. In all this you are forgetting 1 important fact. you are dealing with random numbers. Even if you added more rolls there is nothing stopping the roll from being a ton of fails which will bring your actual success rate lower
    Oh and randomization and maths don't mix, look at you all trying to make sense of something that is naturally chaotic to start with.

    My advice is to stay away from casinos.
    mf2112Alonso
  • alceryesalceryes Member Posts: 380
    Stevevdl said:

    joluv said:

    @Stevevdl: The randomly generated numbers aren't visible in-game, so there isn't an easy way to check that information. Everyone here understands that it's possible to have only 78 successful rolls out of 100 when the success probability is 85%. We're just saying that it's relatively unlikely. Specifically, as I noted above, there would be a 3.9% chance of that happening.

    alceryes said:

    joluv said:

    @Stevevdl: The randomly generated numbers aren't visible in-game, so there isn't an easy way to check that information. Everyone here understands that it's possible to have only 78 successful rolls out of 100 when the success probability is 85%. We're just saying that it's relatively unlikely. Specifically, as I noted above, there would be a 3.9% chance of that happening.

    I guess the question is, how big of a sample size is needed to determine an average?
    Is 50 enough? How about 150? How about 500? The larger your sample size is the closer you 'should' be to the listed percentage chance (taking into account all applicable modifiers).
    1. You are dealing with random numbers, you say you understand random numbers, then are looking for an average, for a bunch of numbers that can change every single time you test it.
    What are random numbers?

    Random numbers are sets of digits (i.e., 0, 1, 2, 3, 4, 5, 6, 7, 8, 9) arranged in random order. Because they are randomly ordered, no individual digit can be predicted from knowledge of any other digit or group of digits.

    You are trying to put order into something that has no order, that is where you are going wrong
    Why is it unlikely to have 78 successful rolls at 85% pass rate? When you are rolling the dice ANY number can come up. Just like the OP complaining he missed a couple of rolls at 98%, and thinks it wrong that he failed any of them even though it is 98%.

    2. Why should you get closer to your answer you want, with a larger sample size. In all this you are forgetting 1 important fact. you are dealing with random numbers. Even if you added more rolls there is nothing stopping the roll from being a ton of fails which will bring your actual success rate lower
    Oh and randomization and maths don't mix, look at you all trying to make sense of something that is naturally chaotic to start with.
    Hahahaha! I understand random numbers, pretty well in fact.
    The 'fact' remains that as your sample size increases, so does the accuracy of the percentage chance of your represented randomization. I was merely asking how large the sample size needs to be to make the conclusion that the listed randomization is, in fact, not being correctly represented.
    I even surmised (in a different post) that the randomization results many are experiencing could be the product of a very small sample size that is extrapolated out instead of individual random functions being called.
  • AstroBryGuyAstroBryGuy Member Posts: 3,437
    kjeron said:

    How about this:
    Changed the Value in INTMOD for all intelligence values to 100% chance to learn.

    L7 Jester repeatedly learning/erasing L1 Identify - failed 6 of 50.
    Observed: 88% success rate.
    Expected: 100% success rate.

    L1 Jester repeatedly learning/erasing L9 Bigby's Crushing Hand - failed 8 of 50.
    Observed: 84% success rate.
    Expected: 10% success rate.

    That is certainly consistent with a specialist -15% penalty and no "too high level penalty".

    Do you get the same results with a vanilla bard? Does setting INTMOD to 115 raise the rate to 100% success?
    Alonso
  • kjeronkjeron Member Posts: 2,367

    Do you get the same results with a vanilla bard? Does setting INTMOD to 115 raise the rate to 100% success?

    Unkitted bard had no failures in 50 attempts at INTMOD: 100%.

    Tried from the other side:

    INTMOD: 15% chance to learn
    L7 Jester repeatedly learning/erasing L1 Identify - failed 100 of 100.
    Observed: 0% success rate.
    Expected: 15% success rate.

    INTMOD: 30% chance to learn
    L7 Jester repeatedly learning/erasing L1 Identify - failed 83 of 100.
    Observed: 17% success rate.
    Expected: 30% success rate.
    Alonso said:

    kjeron said:

    Changed the Value in INTMOD for all intelligence values to 100% chance to learn.

    Can you explain the meaning of that? I guess I understand it more or less, but I'd like to be sure.
    The file "INTMOD.2da" has a column that dictates what % chance to learn a spell each level of Intelligence has. I set it to 100% for all values of Intelligence (0-25).
    AlonsoAstroBryGuy
  • AlonsoAlonso Member Posts: 806
    @kjeron: Nice analysis. That definitely points to kitted bards being treated as specialist mages. You might want to add your findings to the bug report.

    Unfortunately, it looks like the bug report hasn't received attention from the developers yet :(
    AstroBryGuy
  • StevevdlStevevdl Member Posts: 73
    edited May 2016
    Alceryes said

    The 'fact' remains that as your sample size increases, so does the accuracy of the percentage chance of your represented randomization. I was merely asking how large the sample size needs to be to make the conclusion that the listed randomization is, in fact, not being correctly represented


    And the fact remains that if 50 people did a million rolls each, they will not come up with the exact same numbers as one another
    And Statistics show that people will roll dice in this situation to the end of time till they get the answer they want, just like everyone here.

    The OP was crying because he missed 1-3 chances to learn 14 spells at 98%. Please explain to me why this is an impossibility, and they are getting ripped off with their spell learning. Answer this question and I won't join this discussion again, and lets not forget that it is possible that he could of failed all of them

    @Alonso Just because their is a certain probability that something is going to happen does not mean it will
    there is a 1 in 6 chance of rolling a 1 on a die, each and every time you roll the dice, it is also possible that you can roll 100 times the die and 1 will not come up

    Seeing as you are of the opinion that your numbers should come up sooner rather than later, why can I run auto roller for 48 hours on a fighter and not get 100 percentile on strength, there should of been ample opportunity in that time to get 18/00 strength in the sheer number of rolls in that time and the best I got was 92%. I guess at the time I should of said there is something wrong the program rolling percentile wasn't working because I should of got 18/00 in all that time. Did I come here and cry because I didn't get it. No. I accepted that it is not a guaranteed outcome, and I could roll till the end of time and still not get the result I want
    When the weather man says there is 99% of rain and it doesn't rain, does that mean there is something wrong with the weather because it should of rained but didn't?
    Mr2150
  • AlonsoAlonso Member Posts: 806
    Stevevdl said:

    Answer this question and I won't join this discussion again

    Sounds too good to be true. Can you guarantee that?
    joluvmf2112
  • joluvjoluv Member Posts: 2,137
    Stevevdl said:

    there is a 1 in 6 chance of rolling a 1 on a die, each and every time you roll the dice, it is also possible that you can roll 100 times the die and 1 will not come up

    Yes, but the chance of that happening with a fair die is about 1 in a hundred million. If that happens, it's more likely that there's something wrong with the die.
    Stevevdl said:

    why can I run auto roller for 48 hours on a fighter and not get 100 percentile on strength, there should of been ample opportunity in that time to get 18/00 strength in the sheer number of rolls in that time and the best I got was 92%. I guess at the time I should of said there is something wrong the program rolling percentile wasn't working because I should of got 18/00 in all that time.

    If you rolled for 48 hours and never got above 18/92 strength, then there was something wrong with your autoroller.
    Stevevdl said:

    When the weather man says there is 99% of rain and it doesn't rain, does that mean there is something wrong with the weather because it should of rained but didn't?

    No, but there might be something wrong with the weather man.
    sparkleavmf2112
  • mf2112mf2112 Member, Moderator Posts: 1,919
    @Stevevdl No one is saying the results are impossible, just so improbable as to indicate that there is something affecting the outcome beyond sheer randomness.

    It is possible for you to roll a die 100 times and not get a single 1. It is fairly improbable for that to happen however.

    It is exponentially rather less probable that you could roll 100 times and not get a single 1, and then roll 100 more times and not get a single 1.
  • StevevdlStevevdl Member Posts: 73
    edited May 2016
    Alonso said:

    Stevevdl said:

    Answer this question and I won't join this discussion again

    Sounds too good to be true. Can you guarantee that?
    Sure I can, Now tell me why it is impossible for you to roll 99-100% on a 98% to learn 1-3 times out of 14. I don't care how improbable it is, that's not the issue, the issue is why is it impossible?
    Why can't you accept that you could of rolled over 98% , but instead the program is broken because you didn't get all 14 learnt.
    joluv said:

    Stevevdl said:

    there is a 1 in 6 chance of rolling a 1 on a die, each and every time you roll the dice, it is also possible that you can roll 100 times the die and 1 will not come up

    Yes, but the chance of that happening with a fair die is about 1 in a hundred million. If that happens, it's more likely that there's something wrong with the die.
    Stevevdl said:

    why can I run auto roller for 48 hours on a fighter and not get 100 percentile on strength, there should of been ample opportunity in that time to get 18/00 strength in the sheer number of rolls in that time and the best I got was 92%. I guess at the time I should of said there is something wrong the program rolling percentile wasn't working because I should of got 18/00 in all that time.

    If you rolled for 48 hours and never got above 18/92 strength, then there was something wrong with your autoroller.
    Stevevdl said:

    When the weather man says there is 99% of rain and it doesn't rain, does that mean there is something wrong with the weather because it should of rained but didn't?

    No, but there might be something wrong with the weather man.
    So why is it that if you don't get what you want with the die there is automatically something wrong with the die and the same goes for the autoroller, why is there automatically something wrong with the program when you don't get what you want?
    Post edited by Stevevdl on
  • StevevdlStevevdl Member Posts: 73
    edited May 2016
    mf2112 said:

    @Stevevdl No one is saying the results are impossible, just so improbable as to indicate that there is something affecting the outcome beyond sheer randomness.

    It is possible for you to roll a die 100 times and not get a single 1. It is fairly improbable for that to happen however.

    It is exponentially rather less probable that you could roll 100 times and not get a single 1, and then roll 100 more times and not get a single 1.

    But that is the point of randomness, their is no guarantee of what you want to happen to happen, in dealing with random numbers anything is possible no matter how improbable. The way you talk, if there was a 1% chance of learning, the roll is pointless because you have already decided you aren't going to get it and if you do something must be wrong because you shouldn't be able to roll 1%

    You are also forgetting that when dealing with random numbers all previous rolls are immaterial to what you are going to roll next
    Mr2150
  • joluvjoluv Member Posts: 2,137
    brb, got 2 go buy a bunch of lottery tickets

    ANYTHING IS POSSIBLE Y'ALL
    AstroBryGuymf2112
Sign In or Register to comment.