Before and Beyond D&D Reaction Rolls

Most treatments of the reaction roll have the referee roll for reaction once at the start of the adventure, to see how the NPC immediately reacts to the presence of the player-characters. This manifests typically as a 2d6 roll with lower scores indicating a negative reaction (including aggression or hostility) and higher scores indicating a positive reaction. Besides injecting uncertainty into every random encounter, with no guarantee of how an NPC will react to the player-characters, there has been some interesting work building upon it; for example, the total score in Errant represents how many exchanges of words will take place before the NPC ends the conversation; in Troika!, monsters have their own type-specific reaction tables.

I don’t want to condescend anyone by acting like y’all don’t know what a reaction roll is or like no one has talked about them before, but I think there is an aspect of random reaction in OD&D which is elided by later editions of D&D and its reception in the OSR. Unlike how many of these rulesets have the referee roll reaction once at the start of the encounter, OD&D seems to suggest that rolls should be made throughout a conversation. This implies a much more active and procedural model of conversation than those which followed, where the reaction score only represents the party’s first impression and the rest of the conversation is up in the air.

Survey of Reaction Rolls, 1974-1981

There is no reaction roll in OD&D. What we find instead is a table of random actions taken by intelligent monsters: “Other than in pursuit situations, the more intelligent monsters will act randomly according to the score rolled on two (six-sided) dice” (U&WA, p. 12). The table resembles its successors but is simpler, since there are only three possible categories of actions rather than the five more specific reactions (rolled at the start of an encounter) found in later versions. As we will discuss, it also implies a different structure to pre- or non-combat encounters than later editions, since the roll is made not just once but potentially many times to determine the actions of intelligent monsters in general.

The reaction roll as we understand it does not appear until Holmes’ Basic in 1977, which serves as the basis of later D&D revisions (I suspect reflecting some rules changes by Gygax between 1974 and 1979). Holmes says of wandering monsters: “Obviously, some of these creatures will not always be hostile. Some may offer aid and assistance” (HB, p. 11). The 2d6 table he offers is prototypical of later reaction roll tables, offering specific outcomes for monsters like attacking or volunteering help; it also implies that the roll is made only initially to determine the monster’s first impression of the party, rather than to guide the monster’s decision-making throughout an encounter, although it says that an uncertain result should be rerolled after some talks.

A similar table appears in Gygax’s Advanced Dungeons & Dragons: Dungeon Master’s Guide. Instead of being a 2d6 table, however, it is a percentile table—very typical of AD&D—in increments of five-percent, perhaps to be rolled with a twenty-sided die (DMG, p. 63). The bonuses proffered by the player-character’s charisma score are also given in those increments (PHB, p. 13).

There are a couple ways to view the relationship between Holmes’ Basic and Gygax’s AD&D, seeing as they share many concepts not found in OD&D or its supplements. First, the understanding of reaction rolls as a first impression may be reflective of the play culture in TSR before or after OD&D’s publication; in the former case, it would mean that the rules are a clarified or corrected version of what appears in U&WA. However, the Basic and Advanced lines contain other rules which are alien to OD&D and which seem to have originated from Gygax with a view of distinguishing these lines from OD&D. I think it is more likely, then, that the reaction table in Holmes’ Basic is a Gygaxian revision meant to bring the rulebook more in line with Advanced, even as it presents itself as a mere edition of OD&D. The extent to which this was a sincere revision or a strategic one, motivated by Gygax’s desire to distance Arneson from the D&D brand, is a more interesting question but also a more difficult one to answer.

Moldvay’s Basic in 1981 would reproduce Holmes’ version of the reaction table without rerolling on 6-8 (MB, p. 24), but Mentzer’s Basic in 1983 would offer a more complex table closer in spirit to OD&D’s action rolls (MzB, p. 22). The 2d6 table was nested, with non-extreme rolls from 3 to 11 asking the referee to wait one turn before rolling on the next table down; there are up to three turns of conversation on the table, with the third turn posing an ultimatum for the NPC (they either attack, leave, or be friendly). This table encodes the sort of procedural conversation that OD&D seems to have suggested, except much more rigid in its structure with specific tables at different non-terminal points of the conversation. Never have I seen anyone miss this table, nor have I seen it reproduced in a modern rulebook, but I thought it was worthwhile to discuss in light of the original book and its previous revisions.

A Conversation Procedure

Let’s draw out what a conversation in OD&D between player-characters and a monster might look like. The party comes across a band of goblins, the kind with instruments (of war). The referee rolls for their first impression, and they turn out confused. The goblins are freaked out by the party and demand an explanation for their intrusion into the Underworld. The party could respond any number of ways, but let’s assume they put their best foot forward and have their spokesperson with +1 charisma bonus talk to the goblin boss. The spokesperson explains that the party got lost on their way to a cave rave, and needs help finding a way out. The referee rolls for the goblins’ actions again, adding +1 to the score. The chance that the goblins will attack the lying spokesperson is less, but there is a good chance that they will demand more explanations (“Where is this cave rave and why don’t we know about it?”). The best case is that the goblins will let the adventurers go.

Allowing the dice to guide every step of the conversation makes for interesting turns thereof, instead of relegating reaction dice only to first impressions of the adventuring party. For one, it offers a structure for how conversations might go, not just leaving it up to an improv show live at your table. It accomplishes this by actually expanding on what an ‘uncertain’ reaction represents, as the NPC party tries to interrogate the player-character party for why they should let them go (being, typically, armed bandits). This makes more sense to me, especially with the redetermination of reaction after some words have been exchanged, than a ‘neutral’ result which implies to me that the NPC party is idly watching the player-character party fucking around underground. Even an ‘uncertain’ result by itself does not offer as much guidance as specifying that rolls are made per decision point rather than per encounter.

Still, we could do with a clearer rendition of what OD&D seems to prescribe. I tried to include such recursive reaction rolls in my dungeon-crawling pamphlet TURN, where an uncertain result is rerolled with a -1 or +1 after some talks have taken place (since, with such 'context', the monsters are more likely to be aggressive or friendly towards the party) [2]. Below is the conversation procedure and the attached reaction table.

1. Roll reaction.
2. If certain, act on it.
3. If uncertain, interrogate or counteroffer. Reroll with context (+/-1).

Like in Holmes’ Basic, I do think it’d be a good idea to limit the duration of the talks to prevent them from lasting for too long. Mathematically, and with forced modifiers, there is only an 8% chance that the roll will turn out ‘uncertain’ at three turns of talking. At this point, I think the monsters are likely to shrug and leave rather than keep pushing it; it’s just not their problem.

You can use the above procedure beyond the Underworld, too. For example, it can represent haggling between a merchant and a customer where a negative roll results in a rejected offer and a positive roll in an accepted offer. Meanwhile, an uncertain roll means that the other side will make an offer of their own to try to sweeten the pot for themselves. The offer can be accepted or it can be negotiated further, risking the other side saying no altogether. Back in the Underworld, this same rationale can apply to negotiating with monsters to join your party (which although it sounds like a JRPG thing, it’s also front and center of OD&D, M&M, p. 12) [1]. It applies wherever two parties are wanting to talk something out.

Math of 2d6

Before closing out, I wanted to talk about how to adjust 2d6 rolls to account for various non-average factors without making them too deterministic (I like my rolls with a little bit of swinginess). The unmodified score from 2-12 maps to one of three outcomes: negativity (2-5), uncertainty (6-8), and positivity (9-12). The uncertain outcome has a ~44% chance, whereas the other two outcomes each have a ~28% chance. I represent this below, where each asterisk represents a 1-in-36 chance.

Negative:  **********
Uncertain: ****************
Positive:  **********

A modifier of +1 applies unevenly to the three categories because the dice score does not have a flat distribution. There is now a ~17% (1-in-6) chance of a negative reaction, whereas there is an equal ~41% chance of an uncertain or positive reaction.

Negative:  ******
Uncertain: ***************
Positive:  ***************

A modifier of -1 has the inverse effect:

Negative:  ***************
Uncertain: ***************
Positive:  ******

It’s because of the above distributions that I most prefer modifiers of +1 or -1 on 2d6 rolls, or at least in this context. It means that the most extreme case has a 1-in-6 chance of still happening, which is not too high or too low, whereas it’s a toss-up between the two non-extreme cases (which means a good chance that the conversation will continue, or that the players will get what they want). More extreme modifiers of +2 or -2 have more extreme effects. Below is the positive effect:

Negative:  ***
Uncertain: ************
Positive:  *********************

And below is the negative effect:

Negative:  *********************
Uncertain: ************
Positive:  ***

The outcomes seem way too unfair to me. There’s an ~8% (1-in-12) chance of the worst case, and a ~58% (21-in-36!) chance of the best case. That is too overwhelming for me since I think, being a conversation with strangers, them being so willing to take a stranger’s word with little chance of retaliation is ridiculous. For that reason, I’d like to stick with just little modifiers to make sure the roll still matters.

Conclusion

We’ve talked about the history of the reaction roll in early D&D, the impact that an ongoing reaction procedure can have on in-game conversation, and the math behind the typical 2d6 reaction roll. My hope is that, even if you’re a Moldvay purist, you’ll come out of this with a better understanding of how reaction rolls work and why different editors of D&D Basic revised the rules in the different ways that they did. As for myself, I’m excited to try out recursive reaction rolls (for lack of a better phrase? tabletop artificial intelligence?) the next time I run a game since I think they’ll prompt me to take NPC conversations in more interesting ways.

[1] This negotiation roll may have been the more specific basis for reaction rolls in later editions, at least with respect to outcomes.

[2] B., Marcia. 2021. TURN. See also "TURN: Post Mortem" (2022-04-26).