Great Weapon Master — at what AC does -5/+10 stop being worth it?

The setup

Great Weapon Master, in 5e (2014): when you swing a heavy weapon, you can choose to take −5 on the attack roll. If the attack lands, deal +10 damage. Plus a bonus-action attack whenever you crit or drop a creature to 0 HP — the bonus haft section below works that part out.

Note: the 2024 PHB reworked the feat — there's no −5/+10 trade in 2024 GWM, just an always-on +PB damage on a successful Attack-action hit with a Heavy weapon. This pillar is the 2014 math; if you're playing 2024 rules the trade question doesn't apply, but the engine still models the +PB rider — write 1d12+5 @ AC15 +9 rider <PB> (e.g. rider 4 at level 9+) and the rider adds flat damage on hit. The 2024 bonus-action-on-crit-or-kill clause is unchanged and lives in the on-kill pillar.

The question is the same one every Fighter, Barbarian, and Paladin asks at character creation: does the trade pay off, or am I just missing more often for nothing?

The math

Two quantities matter:

• Hit chance. Against AC A with attack bonus +B, you hit on a roll of A − B or higher on the d20 — so the base hit chance is (21 − A + B) / 20, clamped to the [5%, 95%] band that natural-1 / natural-20 enforces.
• Expected damage on hit. The mean of your weapon damage roll. For a Greataxe with +3 STR mod, that's E[1d12+3] = 9.5. For a Greatsword with +3, E[2d6+3] = 10.

Damage-per-round (DPR) without GWM:

DPR_without = hit_chance(A, B) · E[base_damage]

With GWM, hit chance drops by 5 (because the −5 penalty is equivalent to facing AC A + 5), and damage gains a flat 10:

DPR_with = hit_chance(A + 5, B) · (E[base_damage] + 10)

Break-even is where these are equal — solve for AC and you get a clean rule.

The break-even table

Numbers below assume a Greataxe (1d12+STR) for the base damage. The break-even AC shifts by ~1 if you use a Greatsword (2d6+STR, mean 10) or a Halberd / Pike (1d10+STR, mean 8.5) — the headline pattern is identical.

So the rule of thumb is break-even AC ≈ attack-bonus + 8 for a Greataxe-class weapon. With a magic weapon at higher tiers, GWM is always on. At low levels with a mediocre attack bonus it's only worth the trade against lightly-armoured targets.

The damage rolls themselves

The +10 isn't free. It shifts the entire damage distribution right by 10 points. Click into either panel to see the per-outcome rationals; tap a bar to read off P(damage = N) exactly.

Note the shape doesn't change. Variance stays at (12² − 1)/12 = 143/12 for either roll since the flat +10 is a constant offset. What changes is the mean, and where the threshold-crossing probabilities land.

The per-attempt distribution — including misses and crits

The damage rolls above are conditional on a hit. The real per-attempt distribution folds in the d20 attack roll: a probability mass at 0 (miss), the standard damage on a non-crit hit, and a doubled-dice tail on a natural-20 crit. Same engine, one expression. 1d12+5 @ AC15 +9 on the left, 1d12+15 @ AC15 +4 on the right (Greataxe Fighter with +5 STR + magic weapon, +9 base attack bonus, GWM costs −5 to-hit).

The 0-bar is the miss probability. Without GWM (left), nat 1 misses and nat 2–5 missed against AC 15 with +9 to-hit: P(miss) = 5/20 = 25%. With GWM (right), the to-hit drops to +4 so misses extend through nat 10: P(miss) = 10/20 = 50%. The crit-on-20 branch is the same 5% in both panels but contributes more damage on the GWM side because the doubled dice are 2d12 against a +15 modifier rather than +5. Drag the HP slider on the panel headers to read off P(damage ≥ HP) against the AC 15 enemy.

Mean per attempt is the headline number to compare:

• Without GWM: (14/20)·11.5 + (1/20)·18 = 179/20 = 8.95 damage per swing.
• With GWM: (9/20)·21.5 + (1/20)·28 = 443/40 ≈ 11.08 damage per swing.

GWM wins by about 2.1 mean DPR per swing at AC 15 with +9 to-hit, consistent with the break-even table above (which says GWM is on at AC 17 or below for this build). At level 5+ with Extra Attack the per-round mean roughly doubles: 17.9 without GWM vs 22.15 with, a 4.25 DPR lift before we even count the bonus haft attack on a crit.

The bonus haft attack — what most build guides skip

GWM has a second clause every level-1 player forgets: when you score a critical hit (or drop a creature to 0 HP), you can take another attack with a melee weapon as a bonus action. Most break-even tables ignore this because modelling it requires chain-aware probability. "Fires on first crit in the round, at most once per turn" isn't expressible as an additive per-swing term.

The engine's oncrit postfix encodes exactly this semantic. For the level-5+ Greataxe Fighter (Extra Attack ×2), writing the FULL GWM round:

1d12+15 @ AC15 +4 attacks 2 oncrit 1d12+15 @ AC15 +4

Per-round mean climbs to 23.23 — about 1.08 DPR over the chain-only number above. Modest because the bonus only fires when at least one main attack crits, and at standard 1/20 crit chance per attack the probability of any crit in a 2-attack chain is 1 − (19/20)² = 39/400 ≈ 9.75%. Multiply that by the bonus's per-attempt mean (~11.08) and the bonus contribution per round is ~1.08.

Where the bonus-on-crit math really pays is on a Champion crit- fish build. Champion 3 widens the crit range to 19–20, doubling crit chance per attack to 2/20. Per-round mean for 1d12+15 @ AC15 +4 c19 attacks 2 oncrit 1d12+15 @ AC15 +4 c19 is 24.97 — the bonus contribution roughly doubles to ~2.17. That's why GWM and Champion are a canonical multiclass pair: each multiplies the value of the other.

PAM as the alternative — when does GWM lose to Polearm Master?

Polearm Master grants an unconditional bonus-action attack with the haft of a glaive, halberd, quarterstaff, or pike (1d4 + STR mod). Unlike GWM's bonus-on-crit, the PAM haft fires every single turn. Reliable instead of explosive. The engine's bonus postfix encodes this:

1d10+5 @ AC15 +9 attacks 2 bonus 1d4+5 @ AC15 +9

Glaive (1d10 + STR), no GWM, +9 to-hit vs AC 15: per-round mean 22.05. The third attack, even at the smaller d4 die, adds ~5.75 DPR over the 2-attack baseline. Compare against pure GWM (Greataxe, no PAM) at the same build:

Pure PAM (Glaive, no GWM, 22.05) and full GWM (Greataxe with bonus-on-crit, 23.23) are within 1.18 DPR of each other at this AC band, closer than common wisdom suggests. PAM lifts reliably, every turn. GWM frontloads the mean with +10 on every hit and adds a small crit-gated upside. So the choice between the two is about playstyle and build context more than raw DPR. PAM wins on consistency: the third attack always fires, and kill-probability against threshold-HP targets is more predictable. GWM wins on burst: higher per-swing variance plus the crit-doubled bonus give it a fatter right tail, overshooting the kill threshold more often.

The combined PAM + GWM feat is the highest-DPR melee setup regardless. Adding GWM to a PAM build is a substantial upgrade. At the canonical no-rage, no-advantage build PAM+GWM hits 30.09 DPR vs PAM-alone's 22.05, a +8.04 lift from the second feat. With Reckless Attack on a Half-Orc Barb (rage +2, advantage on attacks) the gap widens to PAM+GWM 50.16 vs PAM-alone 33.66, a +16.50 lift. The folk wisdom about it being the optimal pair holds.

What's less obvious is that most of GWM's value when stacked on PAM is the −5/+10 trade, not the bonus-on-crit clause. The engine's 1d10+15 @ AC15 +4 attacks 2 bonus 1d4+15 @ AC15 +4 oncrit 1d10+15 @ AC15 +4 encodes the RAW per-turn-bonus-action constraint (PAM haft when no crit, GWM bonus on a crit, one or the other, never both). Of the +8.04 lift in the no-rage case, +7.88 is the −5/+10 trade applied to every attack including the haft (the haft becomes 1d4+15 instead of 1d4+5), and only +0.16 is the GWM bonus-on-crit feature itself. With reckless advantage the bonus-on-crit contribution rises to about +0.47, because the per-attack crit rate roughly doubles, but the trade still dominates. The per-turn-bonus-action constraint caps what bonus-on-crit can contribute when both feats are taken and PAM is your default bonus action. Worth knowing if you're picking feat order: the GWM trade arrives instantly, while the PAM-vs-GWM bonus-on-crit choice only matters at the margin.

When GWM is wrong

The break-even table above misleads in three situations.

You have advantage. Advantage on the attack roll changes the hit-chance math non-linearly (see advantage and disadvantage for the curve). With advantage the −5 penalty hurts much less because you're rolling twice. GWM becomes correct about 2 AC points higher than the non-advantage table suggests.

You're running Bless or some other +1d4 to-hit floater. Each +1 to attack expected from Bless is roughly equivalent to facing 1 lower AC. Add Bless's mean (+2.5) to your effective attack bonus before reading the table.

The target has a kill threshold you're chasing. Mean DPR isn't the question if you need to finish a specific HP this round (see variance and kill probability). The +10 from GWM helps less than you'd expect when the target is already below your single-swing maximum, and helps more than you'd expect when the target is just above.

Try it yourself

↦ /vs/1d12+3/1d12+13 — base vs GWM damage rolls, +3 STR ↦ /vs/1d12+5/1d12+15 — base vs GWM damage rolls, +5 STR ↦ /vs/1d12+5 @ AC15 +9 / 1d12+15 @ AC15 +4 — base vs GWM per-attempt ↦ /strike/1d12+5 @ AC15 +9 — base per-attempt ↦ /strike/1d12+15 @ AC15 +4 — GWM per-attempt ↦ /strike/… attacks 2 oncrit … — full GWM with bonus-on-crit (level 5+) ↦ /strike/… c19 attacks 2 oncrit … c19 — Champion crit-fish + GWM (the optimal stack) ↦ /strike/… attacks 2 bonus … — Glaive + PAM (no GWM), level 5 ↦ /strike/… attacks 2 bonus … oncrit … — Glaive + PAM + GWM (RAW-correct mutex)

Drag the HP slider in any /vs view above. Below the GWM mean, the +10 is dominant; well above, the to-hit penalty starts mattering more — same logic as the kill-probability lesson applied per-attack. The @ AC<n> +<m> forms compare the full per-attempt distribution (including the zero-mass miss spike); the bare-damage forms compare conditional-on-hit distributions only.

Where this matters in practice

The −5/+10 trade is the most-asked min-max question in 5e because it lives in the part of the game where small swings in expected damage compound over an adventuring day. Three places it shows up:

BG3 weapon picks at level 4 vs 8. At level 4 (+5 to-hit, no magic weapon), GWM's break-even is AC 13 — most enemies. At level 8 with a +1 weapon (+9 to-hit), it's AC 17 — almost everything except plate-and-shield. The BG3 weapon table shows where each weapon sits relative to the break-even line.

The Champion / Sharpshooter crit fisher. Crit chance scales the +10 to deal 2× as often. Champion 3's expanded 19+ crit range pushes GWM's break-even AC up by ~1, and pairs cleanly with the crit-chance pillar's "high-HP solo boss" case.

The bonus-action attack on a kill. Against bosses the kill-trigger is dormant — but against minion queues it's the dominant contribution. GWM on kill shows the +60-100% DPR a real cascade adds on top of the per-swing break-even math here.