Anouleth Posted April 9, 2010 Share Posted April 9, 2010 This is probably going to make little sense to most people, but using Blossom seems to end up, in general, with a prisoners dilemma type situation: Your levels are never increased by more than double, and that's only for exceptionally low growths, and without Blossom you level twice as quickly. Therefore, regardless of your level, it's in your best interest to not use Blossom. However, if you use Blossom non-stop on one person, said person will eventually be so far behind levels-wise that they level at the same speed as everyone else in the party, but gaining about 50% more stats every level, clearly giving you an advantage once you catch up the lost levels (by my estimate, you'd tend towards something like 10 levels behind, after which you gain exp at about the same rate as everyone else). If you want to, as you get towards the end of the game, you can take it off and take advantage of double-speed levelling for a while, too. So taking Blossom off at any time (akin to giving evidence in PD) is advantageous in every way, but actually keeping it on (akin to staying silent in PD) would be better in the long run. I think it's good for Titania and Volug. Their EXP gain is like molasses anyway, yet they have speed issues and only mediocre speed growth (50% and 40% respectively). So you could BEXP them to .99, slap Blossom on them and they can gain the remainder naturally. It's the difference between Titania having a 75% chance of 24AS and a 50% chance of 24AS. Which is insignificant. Of course, Volug can't even take it in Part 1, so it's a moot point. Maybe Haar could use it. Or Royals, since they could do with a few more points here and there. Quote Link to comment Share on other sites More sharing options...
Narga_Rocks Posted April 9, 2010 Share Posted April 9, 2010 This is true, but when I train Mist, I do it without any pretense of efficiency, since she is not an optimal character and I am only doing it because I love Mist more than jellybeans. In other words, I bring her to 3-8 with Blossom and a bazillion staves, and just have her chain-heal people who get hit by lava rocks, until she naturally hits promotion. It takes a while, but she's completely badass as a Valkyrie. You know, I never thought of that idea. I've thought of using laguz in 1-4 for Laura because of infinite weapon enemies, but the lava in 3-8 is pretty cool. Get two units injured, use Reyson on Mist (he doesn't even need to be transformed) and two heals per turn. I think there is at least one enemy that won't ever move unless you get in its range. Reyson might even get to level 40 this way while actually accomplishing something other than raising Reyson. Too bad cards can't double and he'd only ever reach 26 mag max. Also sothe is probably one of the few characters that would rather use random mode than fixed mode for transfers (assuming you don't reset repeatedly in random mode with bexp levels to cap things like Mia's def). I think he'll easily get over most of his RD bases. Do we assume blossom has no effect on bexp levels in RD? Or is there a way to find out if it uses the effective blossom growth rates rather than the non-blossom growth rates? Quote Link to comment Share on other sites More sharing options...
Fei Mao Posted April 10, 2010 Share Posted April 10, 2010 You know, I never thought of that idea. I've thought of using laguz in 1-4 for Laura because of infinite weapon enemies, but the lava in 3-8 is pretty cool. Get two units injured, use Reyson on Mist (he doesn't even need to be transformed) and two heals per turn. I think there is at least one enemy that won't ever move unless you get in its range. Reyson might even get to level 40 this way while actually accomplishing something other than raising Reyson. Too bad cards can't double and he'd only ever reach 26 mag max. Also sothe is probably one of the few characters that would rather use random mode than fixed mode for transfers (assuming you don't reset repeatedly in random mode with bexp levels to cap things like Mia's def). I think he'll easily get over most of his RD bases. Do we assume blossom has no effect on bexp levels in RD? Or is there a way to find out if it uses the effective blossom growth rates rather than the non-blossom growth rates? I think blossom has no effect on bexp levels, either I was either unlucky or it doesn't effect bexp levels but I gave Leonardo a blossom and bexp but he only gains stat ups in the stats he has the highest in growth prior to giving him a blossom (skill,luck,hp,and res) I'll post back later on what characters to test. Quote Link to comment Share on other sites More sharing options...
Zanarkin Posted April 10, 2010 Share Posted April 10, 2010 (edited) Ok i had a probelm with some that i had chosen to do because they had capped something. Anyways i have done 2 so far. Rolf and Brom. FE10 [spoiler=Rolf]Hp/STR/Def HP/STR/MAG/SPD/LCK/RES HP/STR/SKL/SPD/DEF HP/STR/SKL/SPD/LCK HP/STR/SKL/SPD/DEF/RES HP/STR/MAG/SKL/SPD/DEF/RES HP/STR/MAG/SKL/LCK HP/STR/SKL/SPD/DEF HP/STR/MAG/SPD/LCK/DEF HP/STR/SKL/SPD/LCK/DEF/RES HP/MAG/SPD/DEF/RES STR/SPD/LCK/DEF/RES HP/STR/MAG/SKL/LCK HP/STR/SKL/DEF STR/SKL/SPD/LCK/DEF HP/STR/SKL/LCK HP/STR/SKL/SPD/DEF HP/SKL/DEF HP/STR/SPD/LCK HP/STR/SKL/SPD/RES Total: HP-18/STR-18/MAG-6/SKL-14/SPD-14/LCK-11/DEF-13/RES-6 [spoiler=Brom]STR/SKL/SPD/DEF HP/SKL/SPD/LCK/DEF/RES HP/STR/LCK HP/SKL/SPD/LCK/DEF STR/MAG/SKL/SPD/LCK/DEF HP/STR/SKL/LCK/DEF HP/SPD/LCK/DEF HP/STR/SPD/LCK/DEF HP/SKL/SPD/LCK HP/STR/MAG/LCK/DEF/RES HP/STR/SKL/SPD/LCK/DEF HP/SKL/SPD/LCK/DEF/RES HP/STR/SKL/LCK/DEF HP/STR/LCK/DEF/RES HP/STR/LCK/DEF/RES HP/SKL/LCK/DEF/RES HP/SKL/DEF HP/STR/SKL/SPD/LCK/RES HP/STR/LCK/DEF HP/STR/SPD/LCK/DEF/RES Total: HP-18/STR-13/MAG-2/SKL-11/SPD-11/LCK-18/DEF-17/RES-8 Edited April 11, 2010 by Mordecai Quote Link to comment Share on other sites More sharing options...
Narga_Rocks Posted April 10, 2010 Share Posted April 10, 2010 I think blossom has no effect on bexp levels, either I was either unlucky or it doesn't effect bexp levels but I gave Leonardo a blossom and bexp but he only gains stat ups in the stats he has the highest in growth prior to giving him a blossom (skill,luck,hp,and res) I'll post back later on what characters to test. Well, it's not really surprising that the higher growths would go first. They always do. What I'm wondering is if the proportions change at all. If a unit has 30, 50, 70, 30, 50, 70, 40, 70 as their growths, you can expect to get magic, luck, and res more often than anything else. But how much more often? People have done experiments in the past with various units. Now say that after experiments the experimenter got (in 100 bexp levels) 5, 19, 80, 5, 19, 80, 12, 80 That many points for each stat. Blossom turns those growths into 51, 75, 91, 51, 75, 91, 64, 91 As such, you'd still expect to get those 91s more than anything else, but would they still get: 5, 19, 80, 5, 19, 80, 12, 80 as their proportions, or would they instead get 2, 4, 95, 2, 4, 95, 3, 95 As their proportions because of the high growths being so much higher? Or maybe 11, 25, 70, 11, 25, 70, 18, 70 As a result of the low growths being higher (even though the high growths are also higher.) Basically, would there be any difference whatsoever? If so, then blossom has an effect. If not, then not. Quote Link to comment Share on other sites More sharing options...
Magna Posted April 10, 2010 Share Posted April 10, 2010 I'm nowhere near Blossom in FE10... but I can give you a hand in FE9 if you still want it. I suppose I could hack Blossom into Titania (she complements Sothe's growths nicely) or any character that doesn't have random jumps in the memory address for the skill sets (I haven't got Blossom to work for Soren for example). Quote Link to comment Share on other sites More sharing options...
Silith Posted April 10, 2010 Share Posted April 10, 2010 Did the same thing again with Kurthanaga (only 100 times this time though (It gets boring)) Level 100/100 → 100% <- 100% HP 100/100 → 100% <- 95% Str 69/100 → 69% <- 45% Mag 26/100 → 26% <- 15% Skl 35/100 → 35% <- 20% Spd 59/100 → 59% <- 35% Lck 85/100 → 85% <- 60% Def 45/100 → 45% <- 25% Res 63/100 → 63% <- 40% Stat timesup/timestried -> growth% <- original growth% Quote Link to comment Share on other sites More sharing options...
VincentASM Posted April 10, 2010 Author Share Posted April 10, 2010 (edited) I'm nowhere near Blossom in FE10... but I can give you a hand in FE9 if you still want it. I suppose I could hack Blossom into Titania (she complements Sothe's growths nicely) or any character that doesn't have random jumps in the memory address for the skill sets (I haven't got Blossom to work for Soren for example). Sure, sounds great ^^ I'll add in the extra FE10 numbers soon. EDIT Thanks again for all the numbers ^^ Original growth -> Effective growth with Blossom (# of tests) Theoretical growth with 2 RNs 0 -> 0% (25) 0% 5 -> 10% (20) 9.75% 10 -> 26.37% (91) 19% 15 -> 26.97% (471) 27.75% 20 -> 35.76% (165) 36% 25 -> 43.86% (570) 43.75% 30 -> 54.32% (81) 51% 35 -> 55.61% (241) 57.75% 40 -> 59.36% (246) 64% 45 -> 70.62% (160) 69.75% 50 -> 76.45% (395) 75% 55 -> 76.19% (21) 79.75% 60 -> 84.23% (1281) 84% 65 -> 87.38% (761) 87.75% 70 -> 91.67% (60) 91% 75 -> 85% (40) 93.75% 80 -> 90% (20) 96% 85 -> 90% (20) 97.75% 90 -> --- (0) 99% 95 -> 100 (100) 99.75% Edited April 10, 2010 by VincentASM Quote Link to comment Share on other sites More sharing options...
Magna Posted April 10, 2010 Share Posted April 10, 2010 (edited) Growths for FE9 Titania with Blossom, it follows the trend of the 2RN. Original growth -> Effective growth with Blossom (# of tests) Theoretical growth with 2 RNs 0 -> --- (0) 0% 5 -> --- (0) 9.75% 10 -> --- (0) 19% 15 -> --- (0) 27.75% 20 -> --- (0) 36% 25 -> 44% (200) 43.75% 30 -> --- (0) 51% 35 -> --- (0) 57.75% 40 -> 66% (200) 64% 45 -> 71.33% (600) 69.75% 50 -> 84% (200) 75% 55 -> --- (0) 79.75% 60 -> 79.5% (200) 84% 65 -> --- (0) 87.75% 70 -> --- (0) 91% 75 -> --- (0) 93.75% 80 -> 95.5% (200) 96% 85 -> --- (0) 97.75% 90 -> --- (0) 99% 95 -> --- (0) 99.75% Edited April 10, 2010 by Magna Quote Link to comment Share on other sites More sharing options...
Zanarkin Posted April 11, 2010 Share Posted April 11, 2010 (edited) Illyana's and Nephenee's. FE10 [spoiler=Illyana]HP/STR/MAG/SKL/DEF/RES HP/STR/MAG/SKL/SPD/DEF/RES Everything HP/SKL/RES HP/STR/MAG/SKL/RES HP/STR/MAG/SKL/SPD/LCK/RES HP/STR/SKL/SPD/LCK Everything HP/STR/MAG/SKL/SPD/LCK/RES MAG/SKL/LCK/DEF/RES HP/STR/MAG/SKL/SPD/LCK/RES HP/STR/MAG/LCK/DEF/RES HP/MAG/SPD/RES HP/STR/MAG/SPD/LCK/DEF/RES HP/STR/SKL/SPD/LCK/RES HP/STR/MAG/SKL/SPD/LCK/DEF HP/STR/MAG/SKL/DEF/RES HP/STR/MAG//SKL/SPD/LCK/RES HP/MAG/SPD/RES HP/MAG/SKL/DEF Total: HP-18/STR-15/MAG-17/SKL-16/SPD-13/LCK-12/DEF-10/RES-17 [spoiler=Neph's] HP/STR/SKL/SPD/LCK/DEF/RES MAG/SPD/LCK/DEF/RES HP/STR/SKL/SPD/LCK/DEF/RES STR/SPD/LCK/DEF/RES MAG/SKL/SPD/LCK/DEF/RES SKL/SPD/LCK/DEF/RES HP/SKL/SPD/DEF/RES HP/STR/SKL/LCK/DEF/RES HP/STR/MAG/SKL/SPD/LCK/DEF HP/STR/SKL/SPD/LCK/DEF/RES STR/SPD/RES HP/MAG/SKL/SPD/DEF/RES STR/SPD/LCK/DEF/RES HP/MAG/SKL/SPD/DEF HP/SKL/SPD/LCK/DEF/RES HP/STR/MAG/SKL/SPD/DEF/RES MAG/SKL/SPD/LCK/RES HP/STR/SKL/LCK SKL/SPD/LCK/DEF HP/SKL/SPD/LCK/RES Total: HP-12/STR-10/MAG-7/SKL-16/SPD-18/LCK-15/DEF-16/RES-16 Edited April 11, 2010 by Mordecai Quote Link to comment Share on other sites More sharing options...
Narga_Rocks Posted April 11, 2010 Share Posted April 11, 2010 Illyana's and Nephenee's. [spoiler=Illyana]HP/STR/MAG/SKL/DEF/RES HP/STR/MAG/SKL/SPD/DEF/RES Everything HP/SKL/RES HP/STR/MAG/SKL/RES HP/STR/MAG/SKL/SPD/LCK/RES HP/STR/SKL/SPD/LCK Everything HP/STR/MAG/SKL/SPD/LCK/RES MAG/SKL/LCK/DEF/RES HP/STR/MAG/SKL/SPD/LCK/RES HP/STR/MAG/LCK/DEF/RES HP/MAG/SPD/RES HP/STR/MAG/SPD/LCK/DEF/RES HP/STR/SKL/SPD/LCK/RES HP/STR/MAG/SKL/SPD/LCK/DEF HP/STR/MAG/SKL/DEF/RES HP/STR/MAG//SKL/SPD/LCK/RES HP/MAG/SPD/RES HP/MAG/SKL/DEF Total: HP-18/STR-15/MAG-17/SKL-16/SPD-13/LCK-12/DEF-10/RES-17 [spoiler=Neph's] HP/STR/SKL/SPD/LCK/DEF/RES MAG/SPD/LCK/DEF/RES HP/STR/SKL/SPD/LCK/DEF/RES STR/SPD/LCK/DEF/RES MAG/SKL/SPD/LCK/DEF/RES SKL/SPD/LCK/DEF/RES HP/SKL/SPD/DEF/RES HP/STR/SKL/LCK/DEF/RES HP/STR/MAG/SKL/SPD/LCK/DEF HP/STR/SKL/SPD/LCK/DEF/RES STR/SPD/RES HP/MAG/SKL/SPD/DEF/RES STR/SPD/LCK/DEF/RES HP/MAG/SKL/SPD/DEF HP/SKL/SPD/LCK/DEF/RES HP/STR/MAG/SKL/SPD/DEF/RES MAG/SKL/SPD/LCK/RES HP/STR/SKL/LCK SKL/SPD/LCK/DEF HP/SKL/SPD/LCK/RES Total: HP-12/STR-10/MAG-7/SKL-16/SPD-18/LCK-15/DEF-16/RES-16 Is this fe10? (It's important to note it since Vincent is asking for both fe9 and fe10) Quote Link to comment Share on other sites More sharing options...
Zanarkin Posted April 11, 2010 Share Posted April 11, 2010 (edited) FE10 sorry i didn't know The Rolf and Brom are also FE10 Edited April 11, 2010 by Mordecai Quote Link to comment Share on other sites More sharing options...
Fayt Zelpher Posted April 11, 2010 Share Posted April 11, 2010 Hmm... Standard Deviation is np(1-p), right? In which case Not quite. That's the variance of a binomial distribution. Standard Deviation is sqrt(n*p*q) Quote Link to comment Share on other sites More sharing options...
Tables Posted April 11, 2010 Share Posted April 11, 2010 Ah, yes, you're right. Been almost a year since I did any statistics :P. Just sqrt those numbers then. Quote Link to comment Share on other sites More sharing options...
Fayt Zelpher Posted April 11, 2010 Share Posted April 11, 2010 Ah, yes, you're right. Been almost a year since I did any statistics :P. Just sqrt those numbers then. Actually, I made another error. Since you're expressing the output as percentages instead of counts, the standard deviation is actually going to be sqrt (p*q / n)... Which means those numbers are REALLY messed up, actually... Egad... Quote Link to comment Share on other sites More sharing options...
Tables Posted April 11, 2010 Share Posted April 11, 2010 Oh, right, yeah, except I was just doing it for the theoretical value, so n = 1, so that doesn't matter :P. But then if I wanted to find the expected number of increases, wouldn't I multiply through by n for the average (np), the find the standard deviation as sqrt(np(1-p)) like normal? Bah, I can't remember this statistics stuff. I need to revise it. Quote Link to comment Share on other sites More sharing options...
Fayt Zelpher Posted April 12, 2010 Share Posted April 12, 2010 Oh, right, yeah, except I was just doing it for the theoretical value, so n = 1, so that doesn't matter :P. But then if I wanted to find the expected number of increases, wouldn't I multiply through by n for the average (np), the find the standard deviation as sqrt(np(1-p)) like normal? Bah, I can't remember this statistics stuff. I need to revise it. Yes and no. You're taking a sample of n to compute the empirical chances of a level up. So the standard deviation of the growth rate is sqrt (p*q/n). I honestly don't know how to determine the standard deviation of the number of level ups in x trials. My intuition would suggest that since variance adds, the standard deviation is sqrt(x*p*q/n), but I'm not sure.... Quote Link to comment Share on other sites More sharing options...
Tables Posted April 12, 2010 Share Posted April 12, 2010 Someone needs to start a Maths thread for this kind of thing... :P. But I'm a bit confused, what's n? I was under the impression n was the number of trials, not x... Quote Link to comment Share on other sites More sharing options...
Fayt Zelpher Posted April 12, 2010 Share Posted April 12, 2010 Someone needs to start a Maths thread for this kind of thing... :P. But I'm a bit confused, what's n? I was under the impression n was the number of trials, not x... Well, yes, that's true. But we've got a problem since we don't know the true mean OR the true standard deviation in this case. So we'd have to use the sample standard deviation to estimate the standard deviation of the sampling distribution. Which runs into issues since now the normal curve doesn't apply and we need to use 't' distributions to estimate the actual distribution of the data... Sorry for the technical jargon, but there's no easy way to explain it. Quote Link to comment Share on other sites More sharing options...
Narga_Rocks Posted April 12, 2010 Share Posted April 12, 2010 (edited) Well, yes, that's true. But we've got a problem since we don't know the true mean OR the true standard deviation in this case. So we'd have to use the sample standard deviation to estimate the standard deviation of the sampling distribution. Which runs into issues since now the normal curve doesn't apply and we need to use 't' distributions to estimate the actual distribution of the data... Sorry for the technical jargon, but there's no easy way to explain it. But what does it matter? t-distributions can be found online. Or you can just ignore it if you have 400+ tests since they are so close anyway. But how do you find a sample standard deviation for a binomial test? Isn't all that other stuff only when you are testing with things that can be any number like average height? If p is the proportion of the sample that proc'd, and n is your total number of trials, can't you just take standard error = sqrt(p * (1-p)/n)? p +/- C*se is your confidence interval. C is based off normal distribution using whatever the heck you want for your desired interval. Like 1.96 or something for 95%, higher for a 99% interval, etc. Trouble is, though, what is your null hypothesis? You can't use the theoretical growth rate. That's what we want to prove, not what we want to reject. I suppose you could pick various things like 1.5x growth and 1.4x growth and just repeatedly reject them, but that's all. In fact I don't think you even can do that. At best what you'll get is take a 99% confidence interval around our point estimate (what we get from testing) and hope that the theoretical growth rate is within the interval. At that point you've only proved that it could be a best of 2RN system for blossom, not that it actually is. Edited April 12, 2010 by Narga_Rocks Quote Link to comment Share on other sites More sharing options...
Fayt Zelpher Posted April 12, 2010 Share Posted April 12, 2010 But how do you find a sample standard deviation for a binomial test? Isn't all that other stuff only when you are testing with things that can be any number like average height? If p is the proportion of the sample that proc'd, and n is your total number of trials, can't you just take standard error = sqrt(p * (1-p)/n)? p +/- C*se is your confidence interval. C is based off normal distribution using whatever the heck you want for your desired interval. Like 1.96 or something for 95%, higher for a 99% interval, etc. Trouble is, though, what is your null hypothesis? You can't use the theoretical growth rate. That's what we want to prove, not what we want to reject. I suppose you could pick various things like 1.5x growth and 1.4x growth and just repeatedly reject them, but that's all. In fact I don't think you even can do that. At best what you'll get is take a 99% confidence interval around our point estimate (what we get from testing) and hope that the theoretical growth rate is within the interval. At that point you've only proved that it could be a best of 2RN system for blossom, not that it actually is. Well, being in a college statistics class, I can answer some of those questions: Sample standard deviation is sqrt(n*p*(1-p)) for a binomial distribution. For a proportion, the standard deviation is sqrt(p*(1-p)/n). So yeah, you use the standard error to construct the confidence interval. I guess I just was lapsing or something earlier. And then you see the problem with all hypothesis testing - you can't PROVE anything. All you can do is show that a given null hypothesis is plausible (or implausible, as the case may be). To answer the question being asked, you would usually put a strawman theory as your null to attempt to show the null, but in this case, there is no one null that you can use. So I think that the best idea in this case would be to set your null as what your theory is, and show that the theory is plausible (or at least, sufficiently plausible that we can't rule it out, as decided by our confidence level). However, given the nature of the theory and the wide number of implications it has (that is, you can do separate tests for each growth rate - 5, 10, 15, and so on to 95), if all of the results suggest that the theory is plausible, then I imagine that you have enough evidence to, if not prove that the theory is correct, show that it is a good representation of what the game actually does in this situation. And without someone hacking the game or a ROM to determine the exact mechanism, I think that's as good was we'll be able to do here. Quote Link to comment Share on other sites More sharing options...
Narga_Rocks Posted April 12, 2010 Share Posted April 12, 2010 Well, being in a college statistics class, I can answer some of those questions: Sample standard deviation is sqrt(n*p*(1-p)) for a binomial distribution. For a proportion, the standard deviation is sqrt(p*(1-p)/n). So yeah, you use the standard error to construct the confidence interval. I guess I just was lapsing or something earlier. And then you see the problem with all hypothesis testing - you can't PROVE anything. All you can do is show that a given null hypothesis is plausible (or implausible, as the case may be). To answer the question being asked, you would usually put a strawman theory as your null to attempt to show the null, but in this case, there is no one null that you can use. So I think that the best idea in this case would be to set your null as what your theory is, and show that the theory is plausible (or at least, sufficiently plausible that we can't rule it out, as decided by our confidence level). However, given the nature of the theory and the wide number of implications it has (that is, you can do separate tests for each growth rate - 5, 10, 15, and so on to 95), if all of the results suggest that the theory is plausible, then I imagine that you have enough evidence to, if not prove that the theory is correct, show that it is a good representation of what the game actually does in this situation. And without someone hacking the game or a ROM to determine the exact mechanism, I think that's as good was we'll be able to do here. And then there's the fact that it is pseudo rng to throw a wrench in things. stuff like GJ's soren not getting nearly enough mag (I think he got 15 out of 20 when no blossom should result in ~16 out of 20 already) can potentially be explained by supposing that something you are doing or your timing can result in the rng being stuck on something and messing the results. Or it can be explained by random variance, but if that's all it is then his numbers should get overpowered by the hundreds of other tests that went properly. But if the rng can get "stuck" depending on what you do, then it is entirely possible that some of these things could be skewed and the confidence intervals wouldn't actually tell us anything. And so we are left with we can do these confidence intervals, but even if we are then supposed to reject the null hypothesis (best of 2 RNs) then it doesn't actually mean the theory is probably wrong. Quote Link to comment Share on other sites More sharing options...
Fayt Zelpher Posted April 12, 2010 Share Posted April 12, 2010 And then there's the fact that it is pseudo rng to throw a wrench in things. stuff like GJ's soren not getting nearly enough mag (I think he got 15 out of 20 when no blossom should result in ~16 out of 20 already) can potentially be explained by supposing that something you are doing or your timing can result in the rng being stuck on something and messing the results. Or it can be explained by random variance, but if that's all it is then his numbers should get overpowered by the hundreds of other tests that went properly. But if the rng can get "stuck" depending on what you do, then it is entirely possible that some of these things could be skewed and the confidence intervals wouldn't actually tell us anything. And so we are left with we can do these confidence intervals, but even if we are then supposed to reject the null hypothesis (best of 2 RNs) then it doesn't actually mean the theory is probably wrong. I wonder... Would it by any chance be possible to feed the RNG numbers (that is, tell it to pull numbers from a certain string)? If so, there might be a way to definitively show that this theory is correct by testing results (i.e. feeding the RNG certain numbers) that will only succeed if this theory is correct. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.