How to Win

A friend who has never played asked whether a D&D turn is like blackjack: learnable rules, so you never freeze when the table looks at you. It is.

Both games compress to one law each. Blackjack: "the dealer has no choices, so play off what their up-card threatens." D&D: "actions are the currency, so spend yours to take away theirs." Here are the two algorithms, with the why built in.

Blackjack (one algorithm)

The dealer has no choices. House rules force them to draw until 17 (at many tables they also hit a soft 17). So their up-card tells you almost everything: a 2 to 6 means they are often stuck drawing to a stiff hand and busting; a 7 to an ace means they usually finish on 17 to 21.

Here is how often the dealer busts, by up-card. Computed with an infinite deck, dealer stands on all 17s, and no dealer blackjack (the dealer has already peeked):

Dealer shows	2	3	4	5	6	7	8	9	10	A
Dealer busts	35%	37%	39%	42%	42%	26%	24%	23%	23%	17%

The cliff between 6 and 7 is the whole strategy. The rules below are basic strategy for a 4 to 8 deck shoe, dealer standing on soft 17 [1]; four mathematicians first published a sound version of it in 1956 [2].

1. Pairs first.
   • Always split aces and 8s. Sixteen is the worst hand in the game; two 8s are two fresh starts.
   • Never split 10s or 5s. Twenty already wins; 10 is a doubling hand, not two bad hands.
2. Can you double? Double 11 against anything but an ace. Double 10 unless the dealer shows 10 or ace. Double 9 against 3 to 6. You are the favourite; put the most money on your best spots.
3. Hard 17+: stand. Busting yourself is the only way to lose from here.
4. Hard 13 to 16 (stiff hands):
   • dealer shows 2 to 6: stand. Let the forced hitter bust; don't bust first.
   • dealer shows 7 to ace: hit. They will likely make 17+, and your 13 loses by standing anyway.
5. Hard 12: the awkward one. Stand only against 4 to 6; hit against everything else, including 2 and 3.
6. Hard 11 or less (when not doubling): hit. You cannot bust.
7. Soft hands (an ace counting as 11): always hit soft 17 or less. Stand on soft 18, except hit it against 9, 10 or ace. Stand on soft 19+. The ace absorbs a bust, so drawing is cheap.

Basic strategy does not beat the house. It makes the house edge as small as the rules allow, which is the most any algorithm can promise.

D&D combat turn (one algorithm)

The whole game is action economy. On your turn you get one action, a move, and usually a bonus action; everything on the other side gets the same [3]. So the winning move is whatever removes enemy actions or protects your side's. Run this cascade top-down and take the first branch that applies:

1. Is a teammate dying (at 0 hp)? Heal or stabilise them. Why: a downed ally is a lost turn every round. A character at 0 hp rolls death saves: 10+ on a d20 is a success, three successes or three failures ends it, a natural 1 counts as two failures, and any damage adds a failure [3]. Left alone, they die about 40% of the time (computed; the rest stabilise or roll a natural 20 and get up). Any healing at all, even 1 hp, puts them back in the fight [3]. Healing Word costs a bonus action from 60 feet away. Nothing you could deal in damage is worth more than a whole player back.
2. Are you about to go down (low hp, enemy in reach)? Retreat, Disengage, drink the potion (a bonus action under the 2024 rules [3]). Why: same law, applied to yourself. Staying up keeps your actions on the board.
3. Can you finish an enemy off this turn? Do it, even a weak one. Why: a dead enemy stops rolling dice at you forever. Damage spread across two live enemies changes nothing; damage that kills one halves the incoming pain.
4. Otherwise: your biggest attack at the target your party is already hitting. Why: focus fire is rule 3 in advance. The fastest way to make something dead is everyone on one thing.
5. Spend the leftovers before you end your turn. Movement (step behind cover or next to an ally), bonus action, your one free object interaction [3]. Why: they don't carry over. An unspent resource is a gift to the other side.

Try it on a dragon

Run the combat cascade on this and see how fast it goes.

A young red dragon has the party pinned in a torchlit cavern, with two kobold warriors yapping at its heels. Anya the paladin has her shield up. Borin the dwarf is in front. Elara the ranger has a longbow and a clear line. Malakai the sorcerer is at the back. Kaelen the tiefling rogue has already gone quiet and sideways, which is never good news for whoever he is looking at.

Then the dragon breathes: a 30-foot cone, DC 17 Dexterity save, 16d6 fire, 56 on average [3]. Borin fails and drops to 0. Malakai makes the save and is left on 6 hp with a kobold next to him.

Cascade, top down, for each of them:

• Anya: branch 1. Borin is dying. Lay On Hands is a bonus action in the 2024 rules [3]; she touches him, he's up. Her action still swings at the dragon.
• Malakai: branch 2. He is one kobold dagger from the floor. He moves behind a pillar and casts from there. The breath recharges on a 5 or 6 each round [3], a one-in-three chance, so bunching up behind the same pillar as Borin is not the plan.
• Elara: branch 3. A kobold has 7 hp [3]. One arrow ends it, and with it the Pack Tactics the pair were using to get Advantage on Malakai.
• Kaelen: branch 4. Borin is up and adjacent to the dragon, so Kaelen gets Sneak Attack on the thing everyone is already hitting.
• Borin, back up on a handful of hit points: branch 4, then 5. He swings at the dragon, then spends the leftover movement staying beside it instead of chasing the last kobold, which keeps Kaelen's Sneak Attack alive for next round.

Notice nobody needed to be clever. Five people, five branches, one round, and nobody froze. That is the point of an algorithm: it does not need you to be brilliant at the moment you are least able to be.

Then she asked for the same for life

Life runs on the same law as D&D. Actions are the currency, but the round never resets, so there is one extra rule: prefer moves that pay every future turn, not just this one.

One algorithm. Run the cascade top-down, take the first branch that applies:

1. Is someone on your side going down, including you? Stabilise first: sleep, food, safety, treatment, the friend in freefall. Why: a body at 0 hp takes no actions. Every plan assumes a functioning you, so capacity is the stat all other stats secretly depend on.
2. Is something bleeding you every single round? Kill it permanently, even if it costs a whole turn now. Why: a recurring drain rolls dice against you forever; one turn spent ending it buys back every round after. This is why "solve it once" loses to "solve it forever".
3. Can something you do today keep working while you sleep? Build that before anything that must be redone. Why: a system, an asset, a skill, a written-down fact compounds; effort that evaporates at sunset never does. One thing that runs itself beats ten things you push.
4. Otherwise: biggest effort at the one target already moving. Focus fire. Why: spread across ten fronts, nothing dies and nothing ships; stacked on one front, things fall over. Momentum is evidence. Feed the thing that's responding.
5. Spend the leftovers before the day ends. The two-minute ask, the thank-you, the note to your future self, resting on purpose. Why: turns don't carry over, and the cheap moves (asking, recording, appreciating) have the best odds-to-cost ratio in the whole game.

Rule zero underneath all five: you don't control the dice, only the bets. Judge yourself on whether the branch was right, never on how the roll came out.

This post is also a free booklet: download the PDF.

Table Talk

James: So the whole post says stop being clever and pick up the guy on the floor?

Rohan: It says the numbers. Left alone, a dying player dies about 40% of the time. You don't gamble a whole player.

James: But the dragon is right there. Big damage is the fun part.

Rohan: Damage spread across two live enemies changes nothing. Kill one. Then have fun.

James: Fine. Healing Word is a bonus action, so I heal him and still swing at the dragon with my action.

Rohan: And probably miss.

James: Gloriously. Now blackjack: I hit 16 against a dealer 6 because it felt lucky.

Rohan: A dealer showing 6 busts 42% of the time. Stand. Let them bust first.

James: So in both games, let the other side make the mistakes.

Rohan: In both games, judge the branch, not the roll. You'll still roll a 1.

References

1. Shackleford, M. (n.d.). 4-Deck to 8-Deck Blackjack Strategy. Wizard of Odds. https://wizardofodds.com/games/blackjack/strategy/4-decks/
2. Baldwin, R. R., Cantey, W. E., Maisel, H., & McDermott, J. P. (1956). The optimum strategy in blackjack. Journal of the American Statistical Association, 51(275), 429–439. https://doi.org/10.1080/01621459.1956.10501334
3. Wizards of the Coast. (2025). System Reference Document 5.2 (CC-BY-4.0): "Playing the Game" (Actions, Bonus Actions, Time-Limited Object Interactions, Dropping to 0 Hit Points, Death Saving Throws), "Equipment" (Potion of Healing), "Classes" (Paladin: Lay On Hands), "Spells" (Healing Word), "Monsters" (Young Red Dragon, Kobold Warrior; Recharge X–Y). https://www.dndbeyond.com/srd