D&D 5E Why I Am Starting to Prefer 4d6 Drop the Lowest Over the Default Array.

[FormerlyHemlock]

Lanefan, I agree that low Wis is waaaay more fun than low Int.

 

[TheCosmicKid]

If I don't care about Ratskinner's method enough to read it, what makes you think I'm interested in spending hours reverse-engineering your method from seven samples, like some kind of mathematical puzzle?

I'm not asking you to spend hours doing anything. Quite the opposite, in fact. You've said that if characters use normalized arrays, "they're still the same in fundamental ways that I will quickly notice". If you have to spend hours crunching numbers in order to "notice", I think that's a pretty good indicator that you didn't actually notice -- and that if you'd spent those hours doing something else you might be perfectly happy playing with these characters.

Now, there is a way you might quickly notice in which these arrays are the same, and it's that they all total 74. So it seemed reasonable to me that this was the sort of similarity you were talking about. But when I started to talk about totals directly, you contradicted me and said that arrays with the same total can be very different. This is all I'm asking: for you to clarify whether you think the total is a constraint on this "attribute space" of the sort that bores you, or not. Because right now I'm more than a little confused.

How about instead, YOU spend a few minutes or hours coding up a web page with a generator for your distribution, which might or might not be using a normalized method or 4d6 drop lowest; and if I can guess whether it's generating normalized samples I'll tell you? (And if I get it wrong, congratulations, you've got a normalized method which will have fooled me, and which would presumably work as well for me as a random method.)

Sure, I can do this, but again I'm confused, because I've already given you lots of samples, even some unlabeled ones for you to guess at, and you have been dismissive of them. If you need a script to generate enough samples to guess if a method is normalized, isn't that again a pretty good indicator that you can't "quickly notice" if it is?

 

[FormerlyHemlock]

I'm not asking you to spend hours doing anything. Quite the opposite, in fact. You've said that if characters use normalized arrays, "they're still the same in fundamental ways that I will quickly notice". If you have to spend hours crunching numbers in order to "notice", I think that's a pretty good indicator that you didn't actually notice -- and that if you'd spent those hours doing something else you might be perfectly happy playing with these characters.

Hey, hey, hey. I didn't say "in fundamental ways that I will quickly reverse-engineer." I'm not claiming to be a mathematical prodigy who can instantly reconstruct distributions based on a handful of samples from the distribution--I'm saying that a distribution consisting of normalized stats will feel hinky, especially if I'm actually doing the stat generation by drawing cards or whatnot. I do think I will notice a distribution which is normalized in obvious ways (such as your sum-to-74 thing), even if I'm just looking at the outputs from the distribution, but even then I might not know exactly what the similarity is ("they all sum to 74!")--I just think I'll be able to tell them from the real thing.

Now, there is a way you might quickly notice in which these arrays are the same, and it's that they all total 74. So it seemed reasonable to me that this was the sort of similarity you were talking about.

But when I started to talk about totals directly, you contradicted me and said that arrays with the same total can be very different.

I believe that was on a different topic. That was about someone trying to compare point-buy vs. rolled stats' overall utility/power levels--see recent discussion with Lanefan for another treatment of that topic.

To address your confusion directly: the space of all arrays that sum to 74 is a narrow slice of attribute space, and it would be boring to be confined to that space, but that doesn't mean that all arrays within that space have equivalent utility. Summing attribute totals is, IMO, a bad way to measure utility of the array. 3 + 18 play out very differently from 10 + 11, and have different utilities. 3 + 18 has nearly the same utility as 8 + 18 despite having different totals; but from a creative standpoint, 8 + 18 can sometimes be quite different from 3 + 18 and it would be a pity if it never occurred. E.g. I'll probably roleplay a Str 3 wizard quite differently from a Str 8 wizard--the Str 3 guy is probably ancient, or some sort of cripple--even though in terms of the actual die rolls you'd make during the game they are basically identical. But the hypothetical PCs themselves are quite different.

This is all I'm asking: for you to clarify whether you think the total is a constraint on this "attribute space" of the sort that bores you, or not. Because right now I'm more than a little confused.

A method which gives arrays that always sum to 74 bores me, yes. Not interested.

Sure, I can do this, but again I'm confused, because I've already given you lots of samples, even some unlabeled ones for you to guess at, and you have been dismissive of them. If you need a script to generate enough samples to guess if a method is normalized, isn't that again a pretty good indicator that you can't "quickly notice" if it is?

Well, let's see. I accidentally looked at the source code for your first link, so I have to discount that one, but looking at the samples drawn from your second link, they all seem kind of homogenous, so I'm going to guess that they're not actually 4d6 drop lowest--I think they are normalized.

[a moment later, after looking at source code]

And, yep, they are. So I did quickly notice in this case.

If you were my DM and were just spitting out streams of stats, I don't know how long it would take me to catch on for absolute SURE that you weren't using 4d6 drop lowest, but judging from this experiment it appears I would have definite suspicions. The arrays produced are hinky.

 

[Wulffolk]

What we did ages ago was take the percentile stat increment idea first seen in Cavaliers in the 1e Unearthed Arcana and apply it to all classes.

The write-up on how this works in my game is here: http://www.friendsofgravity.com/gam...blue-book-in-html/decbluebook3.html#increment

You can easily tweak the average rate of stat advancement by changing what dice get rolled each level.

Hope this helps!

Lanefan

Thanks, Lanefan.

I had forgotten about that aspect of the original Cavalier. This could easily be adapted to 5e.

 

[cbwjm]

Well apparently any thread I make is good for at least a dozen pages.

It a gift. The gift of Zard.

Do one asking why Greyhawk is "loved". I'd do it myself but I don't think I have quite the touch to get it over that dozen or so pages.

 

[S'mon]

How many players are going to enjoy playing with one arm already tied behind their back, especially if it's a long-running campaign? One-shots, deadly 'old school' campaigns, etc - IMO these are where most people are more comfortable with random generation of stats.

I don't think my old school Wilderlands game is particularly deadly, but I've come to like the idea of rolling for PC stats in such sandbox games. I think it needs to be combined with the old school assumption that players roll up multiple characters - at level 1/starting level - to adventure in the same campaign world, and
PCs can hire on PC-statted retainers/henchmen who can become PCs over time, especially if the
player likes the retainer better. This isn't doable in 4e, and barely in 3e, but I think in 5e it can be done, especially if the optional feat & multiclass rules aren't used.

I have an idea to run 'Dwimmermount' in 5e and I think this approach might be ideal there - roll best 3 of 4d6 in order for starting PC, allow hiring of retainers (max retainers 2+CHA bonus) who use the same stat gen system, but no feats or multiclassing unless they become full PCs.

 

[Zardnaar]

Do one asking why Greyhawk is "loved". I'd do it myself but I don't think I have quite the touch to get it over that dozen or so pages.

I would not like to make that argument. I have the 83 GH boxed set but I missed all of the 90's meta material and don't know much about the splats and world events or living Greyhawk.

I could dedicate it to 1E modules and the 83 boxed set though.

 

[Swordsman503]

I have been playing Dungeons and Dragons almost since its first release way back when. My groups have varied the stat rolls with the type of campaign we were playing, but the standard was 4d6 drop the lowest. But when it comes down to it, it was not the numbers that mattered, it was what the player did with the numbers and how the player "played" the numbers.

For example: In a first edition campaign, we had a barbarian with an 18 STR, but a 4 INT. The player played these numbers, calling his character "Ugh" and only answering in character "Ugh." The most comical moment of the campaign was when Ugh thought a statue was real and tried to mate with it.

This is what is really at the core of the game, it is not the number on the sheet, they are just aids. It is how you play those numbers. The most important rule of Dungeons and Dragons is and always will be.....Have Fun.

 

[Miladoon]

Just two coppers. I prefer 4d6 drop lowest, but, put simply, I don't fully really positively agree with Zard's put simplies.

My recent Cleric: A human non-variant resulting in:ST9 DX16 CN13 IN18 WS13 CH16 - Neither highest nor lowest rolls at the table.

People are thinking, capped gnome 1st level wizard? Well no.

Ultimately, high scores can give you more options, and lower scores give you less options. Likewise, If everyone is samey, samey ability stats will give you samey options, all the samey time. Samey characters are predictable, Unsamey characters are less so.

Also, make sure you check your DMs house rules before deciding your class.

IME, it is really hard to break a character in 5E. Unless you develop a mindset that your character is broke, like suicidal style.

 

[Oofta]

As long as the characters are roughly equal, it's all good. The DM can adjust up or down as needed. I have never seen a campaign with unbalanced characters work out very well. The trick is to figure out a way to level the playing field. I have seen some very good ideas in here about how to do that.

The problem with rolling (using the rules from the book with no home rule adjustment) is that it does nothing to guarantee your goal that characters are roughly equal.

Truly random rolling effectively guarantees that some characters will be above average while others will be below. Will 1 character at the table have 1 or more 16s? Probably. Will 1 character at the table have a high score of 14 or less? Quite likely.

I suspect many people associate rolling with more powerful characters for two main reasons. First is optimism bias. They assume their character will be the guy with 1 or more 16s or higher. In addition, in many cases the DM will allow rerolls or have other house rules to assure decent characters.

IMHO the best way to assure even footing is to use point buy. Don't like the point buy from the book? Think it's too restrictive? Give people more points. Allow a greater spread of numbers. Use the 3.5 point buy system which goes up to 20. Make your own point buy system to allow some other variation.

To me it doesn't matter if the high stat a character can have is 17 or 20. As long as people have the same options. In my experience the math of 5E seems to work better if you don't have characters with super high stats at first level.