Sunday, November 23, 2014

Riffing on 2d6: Reaction Rolls

I've started to take a real liking to 2d6 as a dice mechanic. For the longest time I didn't actually care for it, I was a fan of rolling a d20 for almost everything. But I've spent a lot of time thinking about the classic reaction roll in D&D (inspired in part by this post) and the longer it sat around in my brain, so more useful the probabilities created by 2d6 seemed to get. Here's how that reaction roll works abstractly, at least in most D&D-variants (B/X D&D, page B24 for example):

2really bad2.78%
12really good2.78%

This has some nice properties. Obviously the "neutral" outcome is most likely which is probably just as well. After all, this is meant for a confrontation with a random pack of monsters who just popped around a corner. If both sides indeed "try to chat" that should probably lead to a temporary stalemate more often than not. Then there's a pretty decent chance for making "progress" in one direction or the other. I think that fits the "random negotiation" model again where both parties can usually feel that things are starting to slip before any actual stabbing occurs. Finally there's a pretty slim chance that things lead to immediate hostilities or immediate friendship. Neither of these should be very likely when you turn a corner and suddenly face a dozen kobolds: your side may be worried about being outnumbered, but their side should be worried about your big swords. So to me, this is just about perfect for monster reactions.

It's strange that Gary Gygax decided to mess with this for AD&D where reaction rolls are suddenly percentile-based. Here's what he (unenthusiastically in my opinion) proposes in the Dungeon Master's Guide (page 63), again abstracted a bit:

1-5really bad5%
26-45neutral, negative bias20%
56-75neutral, positive bias20%
96-100really good5%

What exactly are three versions of "neutral" good for except to complicate the table and to force the poor referee to be prepared for seven different outcomes to portray? Five different outcomes already enough of a workload! Also we get slightly increased chances for "really bad" and "really good" at the price of upping the total chance for "not sure yet" to 50% instead of 44.45%. That really only seems to lengthen the average time spent on the mechanic (negotiations take longer to conclude). As far as I am concerned, this is one of several cases where Gary was trying too hard to be "advanced" as it were.

If the point was to make the extreme outcomes more likely, he could have achieved a similar effect just by using different bins:

2-3really bad8.34%
11-12really good8.34%

I am not sure if I am the first person to propose this? Maybe Gary even tried it but decided that 8.34% is too much? Who knows with these things...

Interestingly the reaction roll comes in two variants in OD&D. The "monster reaction" variant was actually an even simpler 2d6 roll with only three possible outcomes (Vol 3, page 12):


This variant is also used in the Confusion spell (Vol 1, page 27). The variant with 5 outcomes (as above from B/X) was used only when recruiting monsters (Vol 1, page 12), not when encountering them.

Using the simpler OD&D table with three outcomes, there's yet another variant we could construct by attaching "special meaning" to rolling doubles. For each of the three outcomes there are two ways of rolling doubles: 1-1 and 2-2 for "really bad", 3-3 and 4-4 for "really neutral", 5-5 and 6-6 for "really good" as it were. It's easy to interpret "really bad" as "monsters attack" and "really good" as "monsters help" but what the heck does "really neutral" mean? Maybe that negotiations end right there and the monsters just decide to leave? Seems reasonable I guess. Opinions?

In terms of probabilities we now get something a bit more complicated to figure out, but it's not impossible. It is, however, a pain to put into a table because the "special outcomes" are not a contiguous range:

2-5really bad5.56%
6-8really neutral5.56%
9-12really good5.56%

So for the "non-neutral" outcomes, that's also close to what Gary did with his AD&D reaction roll. On top of that it's pretty cute to make doubles meaningful, at least I find it fun.

I don't know if this turns anybody else on to 2d6 but I sure had fun writing it all up. There's more in my brain regarding 2d6, at least the "turn undead" mechanic and using 2d6 for skill checks. We'll see when I have the time to write them up. In the meantime, I'd love to hear your take on 2d6 in general and reaction rolls in particular.

Update 2014/12/11: Who would have thunk it, there are still more variants on the reaction roll! I noticed this earlier today when reading this post over at DHBoggs' fascinating blog. The version he gives comes from the "New Easy-to-Master Dungeons & Dragons Game" but it's also in the Rules Cyclopedia (page 93). Here's my reproduction with percentages inserted as I've done for the other versions above:

2-3really bad8.34%
4-6neutral, negative bias33.33%
10-11neutral, positive bias13.89%
12really good2.78%

Given the historical precedent discussed above, those asymmetric ranges are certainly strange: Monsters are now more likely to attack than to be cooperative! (Maybe that's just as well, after all the later versions of the game seem to encourage combat more anyway.) The AD&D idea of "bias" is back again, so there's a good chance (a 47.22% chance in fact!) that you'll roll on this table at least twice before getting a result (the rules recommend stopping after three such rolls). Note, however, that there's no longer a "good" and "bad" category, only the extreme outcomes are actually possible.

Now if you think this is a bad table for the reaction roll, think again. Think hard! Because I am about to show you an even worse one. If my brain is any indication, we simply repress things we find horrible even if we grow up with them. Case in point, I played a long-ish campaign using the Rules Cyclopedia back when it came out. I thought I knew those rules pretty well. But when I first wrote this post, I couldn't remember this whacky reaction roll table at all. I must have repressed it!

But now open your 1983 Mentzer Basic Set, Dungeon Masters Rulebook, and turn to page 22. Yes, look at that triply-nested monstrosity. I won't even type it up, I'll just use a scan:

I know I certainly repressed that one. There's literally no record of this anywhere in my brain. And yet that's the D&D I grew up with, the D&D I played forever and ever before getting the Rules Cyclopedia. Did you repress it too? In protest, I'll leave the analysis of this table to someone else.


  1. "Really neutral" - I would interpret this as the monsters/NPCs completely ignore the party, treating them as if they didn't exist. The monsters/NPCs will do what it takes to avoid, go around the party.

    1. That's possible, but given that the reaction roll is meant to resolve what happens when the PCs communicate with the monsters, I am not sure it's entirely in the spirit of the mechanic. Evasion is usually handled before. But then again there's no "rule" that all the outcomes really have to be post negotiation. Still thinking about it, thanks for your input!

  2. All this does is it makes me scrap the idea I had of assigning a more concrete outcome to each of the values from 2–12 such as "rob" or "blackmail" or "help but take no risks".

    1. I am not sure whether such a table would be good or bad. It could be good because it might give a more concrete suggestion for the DM, or it could be bad since the DM now has to be able to run with a more concrete suggestion. I guess a compromise would be using the super-simple "bad/neutral/good" table and then provide three OPTIONAL tables with suggested interpretations for each of the three possible outcomes? That way I can either make up what happens or if I feel uninspired I can roll on the corresponding "details" table for a hint.

  3. Actually I quite like the Mentzer reaction chart. If you pause at each "roll again" to describe the disposition of the monsters and give the players a few seconds to react, then you can give a modifier to the next roll based on the PCs actions.

    Say the PCs turn a corner and run into a group of goblins. Roll of 4 shows that the goblins are hostile, so I describe them brandishing weapons while hissing and yelling. One of the party members holds out empty hands and tries to communicate and defuse the situation (+1 to next roll), but the rest of the party pulls weapons and prepares to fight (-1 to next roll). The next roll is a 9 (+1 and -1 = 0 modifier), so it seems like Mr. Nice Guy PC is getting somewhere. Most of the goblins are lowering their weapons a bit and keep glancing back at the goblin in charge as if waiting for instruction. The cleric asks everyone to lower their weapons (which most of the other PCs do, so -0 now), and then he steps between the two groups and asks to speak to the goblin leader (+1 because he is trying to defuse the situation, and +1 for his CHA modifier now that the goblins are paying attention to his words). The next roll is a 3 (+2 = 5), so the goblin leader yells "Kill them!". Roll for initiative. If that cleric isn't fast enough, he's going to be mobbed now since he's standing in front.

    1. I think you could get all of that out of the RC table as well. I'd re-balance it to not favor bad outcomes, but if you're really into extended negotiations you can just do the re-rolling with a -4 or +4 as given on the actual table.

      Note that what I have my real problem with is the complex STRUCTURE of the Mentzer table. Maybe it was written in this way to be clear to kids, but for my adult sensibilities it's just an ugly kludge. Check the innermost table for example: It's always the same. The only "difference" is in the intermediate table, there the chances for various outcomes get moved around a bit. I think that's pretty much what the "re-roll with a modifier" table also achieves, and more concisely.

      I like your example BTW. :-)

  4. In general, I really like 2d6 with a target number. The reaction system seems a good use also. I like that a double-1 roll is a significant critical failure as boxcars are a significant critical success. I use them for revised-RC skill checks and ability checks. A typical target number of 8+ with modifiers roughly +1 or +2 at low levels to +4 for mid-level characters. Simple 1d6 rolls are hard to modify with more than one parameter. 1d20 rolls don't have the bell curve that models real-world expectations well.

    1. I am still fooling around with 2d6 as a skill mechanic, but I cannot convince myself completely regarding modifiers. Case in point, in your example, the target number of 8+ is a 41.7% chance. But a +1 bonus makes it a 58.3% chance, a +2 bonus a 72.2% chance. Just seems a little too steep to me. However, I am glad to hear that it works out for you. Maybe I am just excessively careful with the percentages?