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

[NinjaMonkey]

And Avatar FE13 would be literally the worst.

How is the Avatar worse than Arran (who has 10% growths for the majority of his stats)?

That does make me wonder, who in Fire Emblem has the honor of having the absolute worst growths?

Sephiran and the Black Knight have 0% growths in every stat. Also there's Arran, with his 10% growths for most stats (with 3% Res and 0% Luck growths).

Edited May 25, 2014 by NinjaMonkey

[Ϲharlie]

How is the Avatar worse than Arran (who has 10% growths for the majority of his stats)?

I think you missed his point.

He was trying to imply that if 50% was a bad growth, the avatar would be very bad because several of his/her growths are 50.

Sephiran and the Black Knight have 0% growths in every stat. Also there's Arran, with his 10% growths for most stats (with 3% Res and 0% Luck growths).

There are characters with even worse growth than Arran. Take a look at all of Bantu's zeroes.

And I don't think Black Knight or Sephiran should count. Otherwise you'd have to include Athos as well.

Edited May 25, 2014 by Starman

[Rehab]

Back in my day, if it was above 30%, we said "great, we should be able to scum that number green"

[NinjaMonkey]

Take a look at all of Bantu's zeroes.

I take it's FE1 Bantu, because the SD growth page says something completely different.

[Book of Ereshkigal]

I take it's FE1 Bantu, because the SD growth page says something completely different.

It's not like there's a page for it, or anything...

Edited May 25, 2014 by L95

[eclipse]

I think you missed my point. I said earlier that the only time 50% can be considered anything close to bad is if there's a situation that requires exactly X in a stat - no more, no less. That was a highly theoretical corner case that would require exactly X, on a unit that had quite a few levels to go off the beaten track. So. . .what was the point of that?

[Jotari]

So FE1 Bantu had the worst growths for any character that can level up in the series. Yet from what I hear he's still a pretty decent unit well into the end game. Course this was back when not all units could promote too so stats really had a different meaning.

[Ϲharlie]

I'm not sure I would consider FE1 Bantu all that great. He has an argument for being a good unit for a while in FE3, although it's rather grating that he (and this applies to Chiki as well) is perpetually in a state where he can never double and will always be doubled by anything with 3 or more attack speed.

[XeKr]

I think you missed my point. I said earlier that the only time 50% can be considered anything close to bad is if there's a situation that requires exactly X in a stat - no more, no less. That was a highly theoretical corner case that would require exactly X, on a unit that had quite a few levels to go off the beaten track. So. . .what was the point of that?

To more accurately take the conditions of your theoretical case, the chance that Priscilla has exactly 16 Res at level 19 is approximately 12.2% with her 50% growth, and 3.9% if she had 40% growth and 19.8% if she had 60% growth. The same trends hold as before and for other lower/higher growths; 50% is not that special. I was describing the or more case, which is generally more relevant since she wouldnt able to staff bait with 16, 17, 18 etc Res. Specifically, I was clarifying the case where higher growths are not "strictly better".

While I think I know what you're getting at, I don't particularly see how your example is a clean illustration of it.

EDIT: For example, I guess you could say if getting exactly 14 Res at level 19 was bad for Priscilla, then she has a marginally higher chance (5.5%ish more) of having 14 Res at 50% growth (it's the average after all) than other growths, for example 40% and 60% in this case or other lower/higher growths. But at every other Res value, it's the inbetween case.

And 50% is still not that special, as you'd see similar trends if comparing for 15 Res at level 19 for 50%, 55%, 60% growths (55% being "bad" here) or other things along the same lines.

Edited May 25, 2014 by XeKr

[Byte2222]

Can I have a go?

First: I'm not trying to say that 50% is a bad growth, but it does have an unusual downside. How much it matters depends on how you're playing.

Now, lets take eclipse's example: we need a unit with less than 16 res to bait a status staff. Now consider 3 units:

Unit A has a res base of 11 and a growth of 30%

Unit B has a res base of 8 and a growth of 50%

Unit C has a res base of 5 and a growth of 70%

Lets assume all 3 of these units grow 15 levels and, as such, they all have an average of 15.5 res. However, what's the probability that they have 'too much' res for the required task?

The probability of unit A having over 16 or more res after 15 levels is 0.278

The probability of unit B having over 16 or more res after 15 levels is 0.303

The probability of unit C having over 16 or more res after 15 levels is 0.297

(this is the same for 'too little' of a stat too)

In this highly contrived example, unit A has the advantage over unit B (assuming my calculations are correct). In most cases, it is better to have the higher growth but if you really need to know if a unit has more or less than a certain value for a stat after a known number of levels, 50% makes it most difficult to know in advance.

[Jave]

Every character in the series should have FE6 Karel's growths.

[RJWalker]

Every character in the series should have FE6 Karel's growths.

Oh I'm sure we'll reach that point eventually with all this escalation.

[XeKr]

Can I have a go?

First: I'm not trying to say that 50% is a bad growth, but it does have an unusual downside. How much it matters depends on how you're playing.

Now, lets take eclipse's example: we need a unit with less than 16 res to bait a status staff. Now consider 3 units:

Unit A has a res base of 11 and a growth of 30%

Unit B has a res base of 8 and a growth of 50%

Unit C has a res base of 5 and a growth of 70%

Lets assume all 3 of these units grow 15 levels and, as such, they all have an average of 15.5 res. However, what's the probability that they have 'too much' res for the required task?

The probability of unit A having over 16 or more res after 15 levels is 0.278

The probability of unit B having over 16 or more res after 15 levels is 0.303

The probability of unit C having over 16 or more res after 15 levels is 0.297

(this is the same for 'too little' of a stat too)

In this highly contrived example, unit A has the advantage over unit B (assuming my calculations are correct). In most cases, it is better to have the higher growth but if you really need to know if a unit has more or less than a certain value for a stat after a known number of levels, 50% makes it most difficult to know in advance.

(note I don’t necessarily disagree with the last part of your last statement)

I’m getting (in excel) that:

Note that most cdf calculators deal with a number of successes or less.

15 levels to get 5 procs or more at 30% = 15 levels to get 10 or less misses at 70% (b/c of how cdfs work)

BINOMDIST(10, 15, .7, 1) = 0.484508941

15 levels to get 8 procs or more at 50% = 15 levels to get 7 or less misses at 50%

BINOMDIST(7, 15, .5, 1) = 0.5

15 levels to get 11 procs or more at 70% = 15 levels to get 4 or less misses at 50%

BINOMDIST(4, 15, .3, 1) = 0.515491

I don’t see where 50% is the worst growth in this case.

I believe the values you’re calculating are the chances to get strictly more than 5 procs in 15 levels with a 30% growth, so more than 16 Res for example. Or you’re doing something like 1-[chance of 5 or less procs at 30% growth]. You should be doing 1-[chance of 4 or less procs at 30% growth], since 5 procs is 11+5=16 Res. I have a separate formula to do it but for verification I compared a few numbers by manually summing the pdfs so atm I think this’s what’s happening (could still be mistaken).

Suppose the benchmark was 17 Res (or more) instead. I then get the same numbers as you, where 50% has the highest chance to have 17 or more Res (or more than 16 Res). But what about that 16 number that’s above the average Res? Ignoring it is where you get some of these results, as the chance of getting exactly 5 procs (or 10 misses) in 15 levels is fairly high for 30% growth and the chance of getting 11 procs or 4 misses is fairly high for 70% growth. And it’s being neglected.

Suppose I came up with an equally contrived situation, where we have 10 levels to grow, and an array of bases/growths 7/30%, 5/50%, 3/70%, where the average Res is 10 in all 3 cases. The relevant probabilities for 9 or more Res are 0.849731667, 0.828125, 0.850692, respectively. Here, 50% has a lower chance to have 9 or more Res (or “too much Res”). Why is 50% the “best” growth in this case, in terms of not having “too much” Res?

I’m sure one could tweak numbers however and so on.

(incidentally I think for intuitive purposes we should perhaps shift over to RNG screwage and lower stats being “bad” instead of RNG blessing and higher stats being “bad”. Doesn’t change the actual numbers, but eh?)

I’ll verify mathy stuff later. Good convo tho.

Edited May 25, 2014 by XeKr

[eclipse]

To more accurately take the conditions of your theoretical case, the chance that Priscilla has exactly 16 Res at level 19 is approximately 12.2% with her 50% growth, and 3.9% if she had 40% growth and 19.8% if she had 60% growth. The same trends hold as before and for other lower/higher growths; 50% is not that special. I was describing the or more case, which is generally more relevant since she wouldnt able to staff bait with 16, 17, 18 etc Res. Specifically, I was clarifying the case where higher growths are not "strictly better".

While I think I know what you're getting at, I don't particularly see how your example is a clean illustration of it.

EDIT: For example, I guess you could say if getting exactly 14 Res at level 19 was bad for Priscilla, then she has a marginally higher chance (5.5%ish more) of having 14 Res at 50% growth (it's the average after all) than other growths, for example 40% and 60% in this case or other lower/higher growths. But at every other Res value, it's the inbetween case.

And 50% is still not that special, as you'd see similar trends if comparing for 15 Res at level 19 for 50%, 55%, 60% growths (55% being "bad" here) or other things along the same lines.

If you know what I'm getting at, then why do you keep this up?

[Refa]

Every character in the series should have FE6 Karel's growths.

Too weak, should go Shining Force 1 levels of growths. +10HP in one level, so good.

Edited May 25, 2014 by Refa