What is the argument for 50% being a bad growth?

[Ansem]

Title, I recall seeing the argument somewhere, but I cant find it. Is there a specific thread.

[dondon151]

are you sure you're not confusing your statement with "50% growth rate has the highest variance"

[Zaprong]

Basically 50% means Random, you can't be sure it will gain a stat or not, as there is not an inclination to being good or bad at it. you can't expect anything at all.

Edited May 23, 2014 by Zaprong

[Dandragon]

Basically 50% means Random, you can't be sure it will gain a stat or not, as there is not an inclination to being good or bad at it. you can't expect anything at all.

the same could be said of any growth really

[Zaprong]

the same could be said of any growth really

you can't expect anything at all.

Keyword: Expect

[Dandragon]

expectations differ from person to person

[Book of Ereshkigal]

Yes, I could expect the earth to crash into the sun within the next five minutes, an individual's expectations aren't always reasonable.

[Elieson]

50% of anything leaves the most room for error, because you're equally likely to have something occur, as you are to have it not occur. With units like Eliwood and Lowen, (growths in the 30s) You are well aware that they're not going to be stellar in X Y Z stats. With a unit that has a 50% growth (i.e. Marcus!Skill), you can only really work off of an average, and nothing else.

[HF Makalov Fanboy Kai]

let me explain it like this.

you got three units

A has 30% str growth

B has 50% str growth

C has 70% str growth

A has on average will get one point of str every 2/3 level ups.

C on average will only fail to get str every 3 level ups or so.

unit B, the chance to get or not get is an even 50/50 split, which means he could get a str point every level up, or none at all, or most of the time, or barely any of the time.

this is what we call a coinflip, these growths are not able to be predetermined accurately, so these units are a risk to use because of that.

you can not rely on them to gain what they need to gain to be good. this is the argument for 50% being bad, though for me its not bad but a gamble.

[Ansem]

let me explain it like this.

you got three units

A has 30% str growth

B has 50% str growth

C has 70% str growth

A has on average will get one point of str every 2/3 level ups.

C on average will only fail to get str every 3 level ups or so.

unit B, the chance to get or not get is an even 50/50 split, which means he could get a str point every level up, or none at all, or most of the time, or barely any of the time.

this is what we call a coinflip, these growths are not able to be predetermined accurately, so these units are a risk to use because of that.

you can not rely on them to gain what they need to gain to be good. this is the argument for 50% being bad, though for me its not bad but a gamble.

this is what i was looking for, thank you

[XeKr]

wtf is happening in this topic. 0% growth has 0 variance, clearly that must be a good growth? >_>

The variance for a binomial with probability (growth rate) p and n trials (level ups) is np(1-p). From this we observe that p = 0.5 gives the highest variance. That’s all.

Higher growths are strictly better. The average is better, the distribution is better (even if spread out marginally more around 0.5). 50% is better than all growths below it and worse than all growths above it.

EDIT: Suppose you have 10 level ups and either 40% or 50% growth. Tell me which distribution you prefer. 40% because it has lower variance?

Edited May 23, 2014 by XeKr

[HF Makalov Fanboy Kai]

wtf is happening in this topic. 0% growth has 0 variance, clearly that must be a good growth? >_>

The variance for a binomial with probability (growth rate) p and n trials (level ups) is np(1-p). From this we observe that p = 0.5 gives the highest variance. That’s all.

Higher growths are strictly better. The average is better, the distribution is better (even if spread out marginally more around 0.5). 50% is better than all growths below it and worse than all growths above it.

EDIT: Suppose you have 10 level ups and either 40% or 50% growth. Tell me which distribution you prefer. 40% because it has lower variance?

i'm sad my post got ignored by you, because what i'm saying is a fact, 50% is a coinflip

sure the number 50% is higher then 40%, but a 50% growth has a bigger risk then a 40% one because its less consistant

[XeKr]

>_>

I legitimately can’t tell if you’re being serious.

Answer this, suppose you have a 50% chance ticket to win the 1 billion dollar lottery or could trade it for any % below 50%. Would you take anything below 50% because of less “risk” or because 50% is a "coinflip"?

Your interpretation of probability makes no sense. Please clarify your logic.

Edited May 23, 2014 by XeKr

[Baldrick]

Risk of what? Having higher stats? That is indeed dangerous.

According to XeKr's post, the difference between the probability of a 40% growth being X points below its average and a 50% growth being X points below its average is less than 1%. So practically, 50% is scarcely more of a gamble.

[HF Makalov Fanboy Kai]

how about this.

i'm judging 50% as a risk because all we have to go by on how a character will turn out are the averages

you can average out 40% and 60% to an reliable average lot easier then 50%, which has an even chance of succeeding, and failing.

i can guaranteed that Caeda gonna gain a button of speed and barely any str.

i can guranteed that Gatrie in FE9 is gonna cap his defense.

i can't guranteed that someone like Raven, Erk, or Eliwood, are gonna become gods of death or pussy's who can't fight a wet paper bag.

this is what i am talking about, 50% is not a bad growth, it is a gamble, i said this in my post. its not edited at all, please check it out.

Edited May 23, 2014 by HF Makalov Fanboy Kai

[Florete]

I think you're both saying the same thing but HF Makalov Fanboy Kai is saying it in a very weird way that I'm not sure actually makes sense.

Basically just that 50% has the highest variance.

[Jotari]

Less than fifty is bad. More than fifty is good. Fifty is neural. That's the way I see it. You can't depend on a 50% growth. Anything less than 50% and you're not expecting a stat up in that area because the character probably isn't designed for that stat, when you get one it's a pleasant addition to the level up.

[XeKr]

Yeah I'm going to try to avoid a stats lecture...

The fact that 50% growth has the highest variance is (true but...) both trivial and irrelevant. It’s still statistically better than all lower growths, in every reasonable sense.

Edited May 24, 2014 by XeKr

[Jotari]

Yeah I'm going to try to avoid a stats lecture...

The fact that 50% growth has the highest variance is (true but...) both trivial and irrelevant. It’s still statistically better than all lower growths, in every reasonable sense.

Of course it's statistically better. That;s obvious and no one was saying otherwise. The point is a unit with 50% strength growth is a unit who could be a physical fighter but a unit with 10% strength growth will (almost) never grow to be a good physical fighter because with a 10% strength growth their either a mage or a Jagen. 50% is still a better growth than 10% but you have more certainty with 10% so you know what direction the unit is going. A unit with a 10% growth in one stat will likely have a 60% or 70% in another to balance it out. In essence you don't care about that 10% growth as much as a 50% one because it's unlikely to happen while a unit with all 50% growths who has no guarantee to excel in any particular field isn't a unit you can depend on.

[XeKr]

Of course it's statistically better. That;s obvious and no one was saying otherwise. The point is a unit with 50% strength growth is a unit who could be a physical fighter but a unit with 10% strength growth will (almost) never grow to be a good physical fighter because with a 10% strength growth their either a mage or a Jagen. 50% is still a better growth than 10% but you have more certainty with 10% so you know what direction the unit is going. A unit with a 10% growth in one stat will likely have a 60% or 70% in another to balance it out. In essence you don't care about that 10% growth as much as a 50% one because it's unlikely to happen while a unit with all 50% growths who has no guarantee to excel in any particular field isn't a unit you can depend on.

…both trivial and irrelevant…

(apologizes for any perceived elitism, but I seriously don’t have the time or interest, to go into detail. >_>)

Edited May 24, 2014 by XeKr

[Rewjeo]

If you're depending on growths for that particular unit and it has a 10% growth, you can at least play knowing pretty well what that unit will be able to do with that stat in the long term. A 50% growth leaves everything up in the air and you cannot reasonably depend on that unit reaching whatever value by whatever point. It really shouldn't be a problem to deal with the RNG in Fire Emblem, but there is something better about a 10% growth than a 50% growth, even if it doesn't come close to outweighing the 50% growth's obvious advantage.

[Thor Odinson]

I think the variance issues is in the types of play or estimation of plays where you're using expected values of the unit's projected stats-for-level at the time. The 50% unit will hit their expected value more inconsistently than one with 30%, even though the 50% one will have a better expected value. It's more relevant in discussing what a unit can be expected to do than anything.

50% is definitely better than < 50% as a stat value and there's no denying that, though. The variance just comes up a lot in character ability discussions on reaching certain thresholds (in relation to whatever other stats they have and what they need to do, etc).

It's not a bad growth except in maybe awakening where the average strong stat is around 60-70%, in the other games it seem to be along the higher percentile of growths as far as I can see, it's just a tad inconsistent when we want to make some expectations.

Edited May 24, 2014 by Thor Odinson

[bottlegnomes]

XeKr, how do they have probabilities at 0? It's been a while since I had any stats, but wouldn't n be 0, so the probability would always be 0?

[dondon151]

Your interpretation of probability makes no sense. Please clarify your logic.

i agree with this completely. all of the other interpretations of probability in this thread prior to this post are garbage.

XeKr, how do they have probabilities at 0? It's been a while since I had any stats, but wouldn't n be 0, so the probability would always be 0?

i'm not sure what you mean by this question. XeKr's distribution shows the relative frequency of units who gained x points in a stat given 10 level ups. there would be some units who have gained 0 points in that stat.

when describing variance, n is the sample size. in XeKr's example, the sample size is 10 level ups, so the variance for a 50% growth is 2.5. the variance does not depend on how many points a unit has gained in a stat in one particular case; it is a parameter that describes the probability distribution.

[bottlegnomes]

i'm not sure what you mean by this question. XeKr's distribution shows the relative frequency of units who gained x points in a stat given 10 level ups. there would be some units who have gained 0 points in that stat.

when describing variance, n is the sample size. in XeKr's example, the sample size is 10 level ups, so the variance for a 50% growth is 2.5. the variance does not depend on how many points a unit has gained in a stat in one particular case; it is a parameter that describes the probability distribution.

Ah, thanks. I misread the chart. I was reading it as 0 was the number of levels the unit had gained, not the stat gains.