Magic Missile vs Scorching Ray — when does auto-hit beat higher mean?

The setups

Magic Missile (1st level slot)

Three darts at 1d4+1 force damage each, no attack roll, no save. Total per cast: 3d4+3 guaranteed.

Scorching Ray (2nd level slot)

Three rays, each a separate attack roll vs AC, dealing 2d6 fire damage per hit. Mean per ray (assuming hit) is 7. Spell-attack-mod typically +5 at level 5 caster (proficiency +3, ability +2 to +4 depending on progression).

Each ray is an independent attack roll, so the per-cast distribution is the convolution of three single-ray attack-modulated distributions. The engine handles this via the multi-attack postfix attacks 3 on top of the attack-roll syntax. Compare side-by-side: MM at 3rd-level slot (auto-hit, tight 3d4+3 + 3) vs SR at 3rd-level vs AC 15 with +5 spell-attack mod (three independent rays of 2d6 each):

Notice the SR distribution has visible peaks at multiples of ~7 (the per-ray hit mean): three modes corresponding to "1 hit / 2 hits / 3 hits" rays landed. The 0-bar is P(all rays missed) = (5/20)³ ≈ 1.5%. MM has no 0-bar at all: every cast deals at least 6 damage (three darts of 1d4+1 worst-case).

The per-cast math (closed form)

MM damage: E[3d4+3] = 3·2.5+3 = 10.5 damage. Variance 3·15/12 = 45/12 = 3.75 (small, tight distribution). Force damage; bypasses most resistances.

SR damage against AC A with attack mod +B: per-ray hit chance is P(hit) = clamp((21 + B − A) / 20, 0.05, 0.95), expected rays landed 3 · P(hit), mean damage per ray on hit 2d6 = 7, so total expected damage 3 · P(hit) · 7 = 21 · P(hit).

Crossover at AC ≈ 16 with a +5 spell-attack mod. Above that, the auto-hit on MM beats the SR mean despite SR using a higher slot.

The kill-probability angle (the lesson MM keeps winning on)

Mean DPR isn't the question if you need to finish a target at a specific HP. MM's variance is much smaller (3.75 vs SR's variance, which depends on the binomial of rays-landed times damage roll, substantially higher). Below the SR mean, the higher-variance SR has the advantage; above the mean, MM's reliability wins.

Concretely: if you need to drop a target from 10 HP to 0, MM's P(deal ≥ 10) is ≈ 50% (roughly, since 10.5 is the mean and the distribution is symmetric around it). SR at AC 14 has expected damage 12.6 but the spread is huge: could land 0 rays (12% chance) and deal 0, or land 3 rays for 6-36 damage. MM is more likely to deliver the kill on a single cast.

This is the same lesson as variance and kill probability: when your mean sits above the threshold, lower variance wins; when it sits below, higher variance wins. MM's 10.5 mean is right around the threshold for low-tier enemies; SR's higher mean covers more thresholds but with worse reliability per-cast.

The damage-type angle

MM does force damage. Almost nothing in the Monster Manual resists or is immune to force. SR does fire damage. Many planar / undead / fey creatures resist or are immune to fire. Against a fire-resistant target, SR damage gets halved: per-hit drops from 2d6 to 1d6, expected per-ray to 3.5, total expected at AC 14 to 3 · 0.6 · 3.5 = 6.3. MM stays unchanged at 10.5.

So against fire-resistant targets, MM beats SR even at low AC. This is the case for Mephits, lower-tier devils, Fire Elementals (immune), and a non-trivial fraction of mid-tier monsters. MM's damage type is one of its quiet strengths.

See split damage and resistance for the resistance math worked out in detail.

The slot-economy angle

MM is a 1st-level slot; SR is a 2nd. Upcasting MM to a 2nd-level slot adds one dart (4d4+4 = mean 14, vs SR's 14.7 at AC 12). So upcast MM matches SR at low AC and beats it everywhere higher.

The slot-economy argument: at low levels, your 1st-level slots are abundant and your 2nd-level slots are precious. Saving SR for emergencies and using upcast MM as your default single-target damage is the conservative play.

For boss-HP targets, this is the expected-strikes-to-kill regime: variance washes out, and the spell with higher reliable mean-per-slot wins the war of attrition. Upcast MM stays the conservative pick across the whole HP range past mid-tier.

The crit asymmetry

Magic Missile is auto-hit but cannot crit (no attack roll). Scorching Ray rolls per ray and so eats the full crit upside. A Champion-tier crit on a single ray adds 2d6 (mean 7) at the same AC. The asymmetry shifts SR's effective mean upward by roughly the crit-rate fraction of one ray's bonus damage: ~5% × 7 ≈ 0.35 mean per ray, ×3 rays ≈ +1.05 mean per cast. Not nothing. See when crit chance beats base damage for when this is enough to flip the break-even AC. Short version: against most build configurations, the crit upside doesn't quite cover the auto-hit reliability gap that MM enjoys at AC 16+.

Try it yourself

↦ /strike/3d4+3 — Magic Missile L1 distribution ↦ /kill/3d4+3/10 — MM vs 10 HP target ↦ /strike/2d6 — single Scorching Ray hit (raw, no attack roll) ↦ /strike/2d6 @ AC15 +5 attacks 3 — full per-cast SR distribution ↦ /strike/2d6 @ AC18 +5 attacks 3 — SR vs higher AC (where MM closes the gap) ↦ /vs MM vs SR per-cast — the head-to-head

Where this matters in practice

The auto-hit-vs-roll-to-hit decision shows up everywhere a caster has both options available. The crossover AC moves with modifier and resistance, but the shape stays the same.

BG3 mid-game wizards. By act 2 most casters have +5 spell-attack mod and access to both spells. Against fire-resistant enemies (a chunk of act 2's bestiary), MM's force damage skips the resistance halver entirely — a hidden +50% effective damage you don't see in the tooltip. See the BG3 weapon table for typical AC and modifier ranges.

Reliable-vs-nuke applied to spellcasting. Magic Missile is the textbook reliable build (low variance, auto-hit); Scorching Ray is the nuke (higher mean, can whiff). The reliable-vs-nuke pillar shows the same kill-threshold crossover applies — at low HP, SR's variance helps; at high HP, MM's reliability wins.

The slot-economy question. Both spells scale with slot level, and MM gets +1 dart per upcast vs SR's +1 ray. The divine smite math pillar's "spread vs burst" framing applies: MM spreads, SR bursts.