Consider the following situation. Two creatures A and B sneak up on a creature X, successfully remaining unnoticed by it. Without signalling this to B in any way, A then throws a dagger at X from its hiding place. This starts the encounter: It is determined that X is surprised and everyone has to roll for initiative: A rolls a 1, B rolls a 10, X rolls a 20. So X goes first, but cannot do anything on its turn, as it is surprised. Next up is B, who goes into melee and attacks X but misses. X can then use its reaction (since its turn has ended) for the Riposte Maneuver by which it kills B.

Isn't this really, really strange from a narrative point of view? The idea is that the encounter is triggered by A throwing the dagger from hiding (assume that any corresponding Stealth check was successful), but as it turns out, B is already killed by the surprised X, before that trigger even occurs?!

Am I right that this is in fact the right way to proceed by RAW? Are there alternative ways of handling such a situation as a DM?

Related, but not really solving my problem: When exactly does combat start and surprise take effect? What happens when initiative allows a player to act before the player that started the combat?


9 Answers 9


Yes, That Description is RAW

That is the way mechanics happens in RAW. How you narrate that to make sense is up to you. Remember the rules have to be turn based to make it run and everything that happens in a round is happening inside 6 seconds and at roughly at the "same time".

Two Common Alternatives

There are two common alternatives I've seen:

  1. Surprise Round - Borrowing from other system, including 3.5 and Pathfinder, some DMs I've seen hold on to a full surprise round. It would mean the surprised creature doesn't get its reaction back until after B has gone.

    The Surprise Round

    If some but not all of the combatants are aware of their opponents, a surprise round happens before regular rounds begin. Any combatants aware of the opponents can act in the surprise round, so they roll for initiative. In initiative order (highest to lowest), combatants who started the battle aware of their opponents each take a standard action during the surprise round. You can also take free actions during the surprise round. If no one or everyone is surprised, no surprise round occurs.

    (DnD 3.5)

  2. The action that starts surprise combat happens before the first round - The dagger is thrown and damage dealt, then initiative is rolled, and the creature has the surprised condition for the first round of combat. This is what I play, and I first saw it in how Matt Mercer of Critical Role DMs.

    The players must have passed a stealth check against the character's perception. Note, that players don't know if they are really hidden, until they attempt to sneak by or attack. If they aren't stealthy as they think, I narrate something like, "You ready your dagger to throw, and X looks at you. Roll initiative."

    I allow monsters to do it very occasionally as well (if players can, the NPCs can, too). It works well with narrative. I make the stealth check behind the screen, check the passive perception, or ask for a roll (roll is better here, as it lets the players feel I didn't do arbitrarily). Then, "You feel a sharp pain as you down you notice an arrow has sprouted from the joint in the armor. You look around to find the source. You are surprised to see X staring at you. Roll initiative." And if it fails to beat their perception check when a monster tries it, my players love it. "You see an X creeping in the forest, it appears to think you don't see him. What would you like to do?"

  3. From comments, it has also been suggested that a common method is to give a initiative bonus to the surprise attacker who starts the combat, or to just let them go first in the initiative order. This works with the narrative and isn't as powerful as 1 or 2, and is less likely to have balance issues.

Game Designers and Initiative

There are a few other options for how initiative is done posted by game designer Mike Mearls to Unearthed Arcana back in 2017. In it he proposes initiative is calculated the other way, higher number goes last. Surprise adds 10 to the initiative and the creature can't take reactions until its turn. This isn't a common way, but it is very interesting.

  • \$\begingroup\$ Comments are not for extended discussion; this conversation has been moved to chat. \$\endgroup\$
    – mxyzplk
    Commented Jun 6, 2019 at 19:18
  • 1
    \$\begingroup\$ It is perhaps worth noting that PF/3.5's "standard action" is analogous to 5e's "action" - that is that, during a surprise round in PF/3.5, creatures are basically restricted to what 5e calls taking their action (eg., they could dash, but they couldn't move and also take the attack action). \$\endgroup\$
    – minnmass
    Commented Sep 21, 2021 at 13:51

The narration of the event seems strange to you because you are confusing the planned order of events with the actual order of events.

The plan was that the encounter begins when A throws their dagger. So if the characters had followed the plan then the order of events for round 1 would have been:

  • X does nothing because they don't know they are in an encounter yet.
  • B has their turn and prepares an action to attack X, specifying that the trigger for their action is A's attack.
  • A throws the dagger.
  • B performs their prepared action.

But the characters did not follow the plan. Instead:

  • X did nothing.
  • B got excited and attacked early.
  • A attacked.

As you can see, the narration is clear for both orders of events so long as you are viewing them in the correct context.

  • \$\begingroup\$ Comments are not for extended discussion; this conversation has been moved to chat. \$\endgroup\$
    – mxyzplk
    Commented Jun 7, 2019 at 5:01

Your interpretation of the rule is correct

According to Chapter 9: Combat of the PHB/Basic Rules on D&D Beyond, under Surprise:

If you're surprised, you can't move or take an action on your first turn of the combat, and you can't take a reaction until that turn ends.

After that turn (the surprised creature's turn) ends, it is then free to take a reaction, as the rule removes any such restrictions at the end of that turn.

RAW does (indirectly) provide a narrative way to parse the sequence of events

From the same chapter, under The Order of Combat, D&D Beyond says:

A typical combat encounter is a clash between two sides, a flurry of weapon swings, feints, parries, footwork, and spellcasting. The game organizes the chaos of combat into a cycle of rounds and turns. A round represents about 6 seconds in the game world.

and then later under Movement and Position:

In combat, characters and monsters are in constant motion...

So, our characters aren't actually just standing around and waiting for their turn to move and fight. They're moving around, striking, parrying, flourishing, and so on, all at the same time, and the game just uses the turn system to sort it all out in a useful way.

In other words, there is a lot of maneuvering going on that just doesn't make it into the mechanical turn system. Everything's basically happening at about the same time, and the game just uses turns and rounds to mechanically sort it out.

How does this help us?

Well, let's take your example.

A triggers the fight with a sneaky stealth attack, but X wins on initiative and goes first, with B going second, and A going last. So the scene plays out like this.

A starts to move in to attack. X, going first, becomes aware of A and B, but since it didn't perceive a threat before this, it is surprised and unable to act on this. Nevertheless, X had particularly good reaction time (represented by the high initiative) and quickly recovers from its surprise. Now it's prepared for the attack.

B then gets their turn. Whether they're jumping the gun, noticed the quick recovery of X, or decide to ready an action according to the plan, they still got higher initiative than A and so are just a little bit faster to respond to unfolding events.

Now it's A's turn. At this point, A might know that their plan has been foiled, and that the surprise is lost. They might also be aware that B was just a little bit quicker to pick up on this and react accordingly. Regardless, all of this happened in the mere moments between A deciding to attack and A actually attacking, since all of this is happening at about the same time.

My own experience with this approach

I've used this approach before, though not always with surprise in play. In fact, one situation was almost exactly like the one you've described in every other way.

Player A initiates combat by attacking. Player B gets the highest on initiative, monster X goes second, and Player A goes last. So, I explained it by saying that, though A begins his attack, monster X is ready for this and acts faster. Player B sees this and manages to act even faster than X, and ends up going first.

I compressed the next few actions narratively into a couple of moments of who can react the fastest (since the round is "about" six seconds), and it made a pretty neat narrative sequence that left both players excited, rather than player A feeling cheated out of a cool moment.

Where surprise is in effect, I've explained it by saying that X hesitates before acting as they have to figure out what's going on, but they just recover and collect themselves faster than the other lower initiative monsters do. If X goes before the party, then I just say that as the party begins to act, X becomes aware of them and, though surprised, quickly recovers and reacts.

This generally has had the effect of making my players' enemies feel competent, rather than leaving the players robbed of their moment.

Obviously you don't have to play this way.

But if you want to use RAW in regard to surprise, and you don't want to steal your players' thunder while you're at it, this is a good way to do it.


The issue I see in this scenario is that one player broadcasted their intent ahead of their turn, and the other did not, so the narrative only seems unusual because the decision of what's happening in the ambush was only partially made outside the context of turn order.

To resummarize my understanding of the narrative, A stated they wished to take a combat action, so the DM dropped into initiative to manage the situation. At this point, A's statement of wanting to throw the dagger at X is basically just a statement of intent, not something that has been 'locked in'/happened yet.

X went first (doing nothing but no longer being surprised), so narratively they may have noticed something was happening at the last split second and are no longer completely unaware (can react, etc.).

The perceived conflict at this point is due to B taking an action they did not state before combat began. There's nothing inherently wrong with that- maybe they simply didn't get a chance to, as A was leading the narrative at the time or something along those lines. However, if their intent was to act after A threw their dagger, B should have readied their action with that event as the trigger.

I would suggest that the DM should assist in preventing these conflicts in one of two ways:

  1. Highlight the fact that A's proposed course of action caused initiative order to start, but has not happened yet: "Ok, A wants to throw a dagger at X, so let's roll initiative to manage this ambush. Ok, B, you're up before A- as you're closing on in X, you see A looking like they're pulling back to throw a dagger, what do you do?"

  2. Make sure everyone states what their proposed course of action for the ambush is before entering initiative, so there's a more full image of the plan before people's turns happen and that conflict between planned and actual narrative doesn't exist: "Ok, A wants to throw a dagger at X. B, you're also closing in on X, what are you trying to do? Are you waiting to see what happens with A's throw, or are you also attacking now? [...] Ok then, let's roll initiative."


Summary: X winning the initiative roll doesn't mean they see or know about the attacker before the attack. It merely means that mechanically the Surprised condition isn't still in effect when an attack happens.

(Unsolved problem with this: what if A and B decide not to attack at all after rolling initiative.)

The mechanical consequence of this rule is that to actually surprise-attack someone, you also have to beat them on an initiative roll. As @CTWind says in an answer, declaring intent to attack puts the game into round/turn-based simulation mode to figure out the details of what actually happens. If B chooses to attack before A instead of Readying an action, that's what actually happens. (This sucks for B if they lose out on move+attack or Extra Attack because they happened rolled higher initiative than A: they "failed to coordinate their attack as well as they'd intended".)

I think that solves the problem of B acting before A, but if we leave out B then we still have to explain what it means for X to take their turn before A and lose the Surprised condition. (At which moment does the 'Surprised' state disappear? concludes that the only sensible interpretation of RAW is that a creature is no longer Surprised after they've had their first turn, during which they couldn't act.)

If the ambusher wins the Initiative roll, things go as they anticipated, gaining the initiative over the target. You get 2 turns of attacks before the first turn in which they can actually act, or can Attack and move away without taking an AoA.

If you (the ambusher) don't beat them, the only benefit you got was being able to attack in a round where they didn't get to do anything on their turn. So even though you lost the initiative, you've still had 1 more real turn than they have, or equal, at any point in the rest of combat.

In narration, this can be them recovering right away from the attack, quickly enough to take a reaction like Hellish Rebuke or Riposte. Or to take an AoA at a melee attacker that moved in to attack and then tried to move away.

Even though they weren't expecting it and didn't see it coming, their immediate reaction proved to be useful, and the game models that as being able to take a Reaction to anything they can perceive after their first turn that clears the surprise round.

If nothing happens until A's turn, then there's nothing for X to react to.

This is important for preserving causality in the narration: X can't actually do anything until after something is done to them. (Or they see something, e.g. if B moved past them without attacking, X could make an AoA.)

X's first turn, where they clear their Surprised status, doesn't involve them reacting to anything at all. They don't know that they're "not Surprised" because the thing that might have surprised them still hasn't happened. Just that if/when something they notice does happen, they won't still be Surprised.

One possible problem with this interpretation is that RAW, A could decide not to actually attack once it's their turn. The DM might disallow that as meta-gaming after declaring intent to attack, or might narrate "steadying yourself for the shot, you feel your timing is off; you take a breath and clear your head". I don't think you get to try again right away; you're still in initiative order and the non-surprised target is still doing whatever they were doing (e.g. walking or talking). Maybe if you wait for the target to focus on something else, you might get another chance to surprise them? But these aren't ideas I've playtested, I'm not recommending them; I'm pointing them out as problems with letting A change their mind after getting into initiative order. But in general they need to have some flexibility; what if other party members act first and their intended target is dead, or Banished.

Another possible interpretation: The turn-based system is a mechanism for simulating simultaneous events, simplified/modeled as a sequentially-consistent series of turns. There aren't real gaps of time between turns. X's losing the Surprised condition narratively happens as they're being attacked, and the surprise-attack wasn't perfectly successful so didn't get the real benefit of Surprise. (e.g. Assassin (Rogue)).

But the sequential-consistency of the turn model still applies: if the target is KOed by the attack, they don't get to take a Reaction because the attack happened first. Or during the flow of combat, if you're KOed before your turn, that means that you don't get to act; your character was probably in the middle of doing something, but it didn't end up being completed or having an effect before the knockout blow.

RAW, to fully successfully ambush someone, you have to beat them on an initiative roll. i.e. they get a roll to reduce the impact of the surprise.

This really matters a lot for an Assassin (Rogue), where a hit on a Surprised creature is automatically a crit, and they get advantage against targets that haven't taken a turn yet in this combat. (The can't-act turn for a Surprised creature counts as a turn).

Thus an Assassin (Rogue) needs to beat their victim on an initiative roll to use their class features as part of an attack from stealth, according to RAW. As CTWind and I discussed in comments, this is sort of like the victim getting a "saving throw" against being assassinated. (Mechanically, Initiative rolls are Dex ability checks, not actual saving throws.)

At which moment does the 'Surprised' state disappear? has some examples about Assassin Rogues, and this is the RAW conclusion.

RAW, this still applies even if the Assassin is far away, in cover, and has been holding a crossbow aimed at the target for minutes. Their only movement is to pull the trigger. So physically there's nothing for the target to react to.

Narration like "saw the assassin move while they were winding up to throw a knife" doesn't work for a hidden crossbow from behind. (Assume magical Silence or whatever if you think they might hear the bolt whoosh through the air.)

But D&D doesn't run purely on physics. Luck is a real thing. Rolling high on initiative could be (if no other explanation works) a kind of "danger sense" (aka spidey sense) tipping you off to change your stride and make the bolt miss a vital area, only doing normal damage spoiling the crit Sneak Attack. High-level characters with lots of initiative bonuses are hard to assassinate in this way, as well as having lots of hit points. (Which are also a pretty nebulous concept).

D&D is (usually?) a fantasy setting where you wouldn't find a laser weapon that's impossible to react to without FTL senses, so usually there will be a plausible explanation of reacting to a projectile in flight, or for melee of an assassin moving up behind you. But regardless, "danger-sense" probably always works. The world is a dangerous place; part of having class levels is this kind of always-ready ability to make an initiative roll in any circumstance.

People that survive an assassination attempt were lucky and/or happened not to be flat-footed. (If a low-HP target does still die from an attack that turned out not to be a crit Sneak Attack, you could still describe it as a going off mostly the way the assassin planned it.)

  • 2
    \$\begingroup\$ "Unsolved problem with this: what if A and B decide not to attack at all after rolling initiative." This is a very important point. Keeping the narrative coherent for any outcome of initiative rolls makes it more or less necessary that A and B have to decide at the beginning of the round what they will do on their turn. This, on the other hand, might severely reduce the benefits of an ambush. \$\endgroup\$ Commented Jun 7, 2019 at 12:30

By strict RAW of your question, your description is correct. You have already rolled initiative, so you have to act in turn order. However, that isn't the only way your DM could have played this scenario.

Combat begins when the DM says so

There isn't any explicit guidance as to when combat encounters technically begin. We know what a combat encounter is, an encounter where we are playing in turns, fighting monsters. But the Basic Rules, the PHB, and the DMG, do not actually explain when the encounter begins. In your example, the DM has ruled that the combat encounter begins as soon as you want to attack, but that isn't the only way.

The rules actually have two example situations for initiating surprise combat:

A band of adventurers sneaks up on a bandit camp, springing from the trees to attack them. A gelatinous cube glides down a dungeon passage, unnoticed by the adventurers until the cube engulfs one of them. In these situations, one side of the battle gains surprise over the other.

The second example of the gelatinous cube is extremely relevant. The example has the cube attack before combat begins, after which the party is surprised. Your DM did the opposite, combat started before you could make the attack. They could have done this instead.

Have combat begin after the action that initiates it

Your DM could allow you to make the attack, then roll initiative after that:

  • A: "I throw a dagger at the creature!"
  • DM: "Roll the attack"
  • A: rolls, hits the creature, deals some damage
  • DM: "The creature grunts in surprise and pain! Roll initiative" rolls 20 for the creature
  • A: rolls 1
  • B: rolls 10
  • DM: "The creature is initially surprised, but recovers quickly, whirling around to glare at you! B's turn"
  • B: "I swing at the creature!"
  • DM: "roll it"
  • B: rolls 1, missing
  • DM: "The creature easily parries your attack, then ripostes" rolls 20, rolls a billion damage "the riposte brutally cleaves through your torso, instantly killing you in a shower of blood. Your turn A"
  • A: "I flee"
  • combat continues

Initiative is not "who starts first"; it is "who finishes first"

Strange issues like this make sense when you realise that the initiative roll is tracking the order of resolution of actions, not the order of the start of actions.

In the example in the question, A kicked off the combat by starting to do something hostile (throwing a dagger). All the other people were very quick off the mark and were able to resolve their actions before A could resolve theirs.

  1. A starts to throw the dagger.

  2. X is quick off the mark and would normally have time to react before the dagger hits, but were surprised so can't do much more than say "wha?".

  3. B saw A start to throw the dagger (perhaps they saw the movement of the arm or the drawing in of breath) and reacts by attacking X. They have quicker reactions than A so their attack gets resolved before A's action. As a result, X gets to react (since they were really on the ball, with an initiative of 20).

  4. Finally, A gets the dagger in the air, so we resolve their attack. At this point, the situation has changed, so A's player could change their mind. The character A stands there, a dagger ready to throw, but then goes, "No, wait" and does something else.

In short, the combat wasn't kicked off by A throwing a dagger; it was kicked off by A starting to throw a dagger.

What if you allow A to go first?

Then things can get really nasty. Allowing characters to act outside initiative is very powerful, especially for some classes (Assassin Rogues, UA Deep Stalker Rangers).

For example, an Assassin Rogue sneaks up on the PC camp. They line up on the PC on watch with a crossbow… they fire (attack roll at advantage)… a hit! A critical hit (from the Assassinate feature)! The GM rolls weapon and sneak attack damage twice and tells the player, 'Your character takes 50 damage." The player probably comes back with, "Huh? Don't I get an initiative roll?"

  • 1
    \$\begingroup\$ However, I think the last issue is more a matter of the DM fudging things in order to "beat" the players than anything else. Not at the very least having the players make a perception roll so the player knows something is coming, is probably more "bad DMing" than anything else. If a PC assassinates and NPC, the DM know what they are doing. It doesn't change the results, but none of the players are surprised, even if their characters are. Of course, if they followed RAW, then they would need to call for an initiative roll first instead. Either way, it's not just "you suddenly lose hit point." \$\endgroup\$
    – zeel
    Commented Oct 18, 2019 at 19:53

Are there alternative ways of handling such a situation as a DM?

Part of the problem in your scenario is X's reaction (Riposte) coming before the action that started the initiative roll.

PHB p.189 says about Surprise:

If you're surprised, you can't move or take an action on your first turn of the combat, and you can’t take a reaction until that turn ends.

However, if you tweak that rule by saying "you can't take a reaction until the round ends," it would resolve that trouble.


A Small Tweak to Rules Interpretation that is a Good Compromise

The OP describes RAW, to answer the question. Personally, I rule that a Reaction that is initiated during the PC's Action phase because they took the Ready Action and is executed at the time/trigger specified can't be done purely as a Reaction when no Ready Action was taken.

Therefore, if a PC (or NPC) cannot take an Action, they cannot, by definition, initiate a Readied Reaction for later. This results in something closer to RAW without going full-on "Surprise Round" override.

This might have complications with other Reaction scenarios (involving special abilites for it and such); I haven't puzzled out every possible combination of factors. But in actual play, I haven't yet encountered an issue with this ruling.

  • 4
    \$\begingroup\$ Could you give an example order of events for your house rule? I'm having trouble conceptualizing what it looks like in play. \$\endgroup\$ Commented Sep 21, 2021 at 13:39
  • \$\begingroup\$ Does this just mean that a surprised character doesn't get a reaction during that round? \$\endgroup\$
    – Tiger Guy
    Commented Sep 21, 2021 at 14:01
  • \$\begingroup\$ Edited. Hope this clarifies. \$\endgroup\$
    – David N.
    Commented Sep 21, 2021 at 22:30
  • \$\begingroup\$ Tiger Guy, I think the best way to say it would be that a surprised character doesn't get to take a regular action as a reaction during that round. Special abilities that are executed during the Reaction phase are unaffected by this -- they may be taken, because they are not tied to a prior Ready action having been taken. \$\endgroup\$
    – David N.
    Commented Sep 21, 2021 at 22:39

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .