MARVEL SNAP: When to Retreat, Stay or Snap

Retreating can be the correct move even with a playable hand. Staying can also be correct when you expect to lose more often than you win. The difference is the number of cubes you give up by leaving now compared with the number you can win or lose by finishing.

This guide covers cube decisions in ordinary Ranked matches. It does not apply the same arithmetic to Conquest health, special event scoring or Infinite leaderboard points. The examples use hypothetical outcomes to explain the decisions.

Read two amounts before deciding

Separate the current retreat cost from the final stakes. Each player can Snap once; a Snap raises the stakes, and finishing the final turn also doubles them. Completed matches therefore normally settle for 2 cubes without a Snap, 4 with one player's Snap or 8 with both players' Snaps. A retreat can end the match for less.

The cube display distinguishes current stakes from the next increase. An opponent's Snap is not a reason to mentally replace your available retreat cost with the eventual final loss. Decide whether to continue through that increase while the retreat option is available, rather than letting another turn resolve and assuming you can still leave for the earlier amount.

Do not assume that “the last turn” always means turn six. Use the match's actual remaining turns. More importantly, do not price a final result using the current retreat amount alone: losing 1 now and losing 4 at the end are different choices.

Winning more games does not necessarily gain more cubes

Consider two hypothetical ten-match records.

The first record wins more matches and still loses cubes. The second loses or retreats more often and gains cubes. That does not mean you should force every win toward 4 cubes: opponents can retreat, and raising the stakes can make your losses more expensive too.

It means that protecting a small loss is part of progression. A 1-cube retreat is not the same result as staying for an 8-cube defeat just to see the opponent's final card. Your next decision should be based on the present position, not on recovering cubes lost in the previous match.

The final decision: compare staying with retreating, not with zero

Suppose you can retreat for 1 cube or finish for 2. For the example only, assume a 40% chance of winning, a 60% chance of losing and no ties. Imagine that identical decision repeated 100 times:

• Staying would produce 40 wins worth 80 cubes and 60 losses costing 120: a net loss of 40 cubes.
• Retreating every time would lose 100 cubes.
• Staying averages −0.4 cubes per decision, compared with −1 for retreating.

Neither option gains cubes on average. Staying simply loses less under those assumptions. “I am probably losing” is therefore not enough to settle the decision. You need to compare the possible loss with the smaller loss you can already accept by retreating.

Now change the final stakes to 4 while the available retreat still costs 1. At the same hypothetical 40% win chance, staying averages 0.40 × 4 − 0.60 × 4 = −0.8 cubes. That is still better than −1, but only slightly. At a 30% win chance, it becomes 0.30 × 4 − 0.70 × 4 = −1.6, making the 1-cube retreat better.

The important change was the cost of continuing, not a rule that every uncertain game should be played out. If your opponent raises the stakes, reassess the choice at the new final amount.

What the 25% and 37.5% thresholds actually mean

For a final decision with no ties, let R be the cubes lost by retreating and S the cubes won or lost if the game finishes. The break-even win chance is (S − R) ÷ (2 × S). Above that point, staying has a better average cube outcome than retreating; below it, retreating does.

These are terminal comparisons with retreat still available and no further stake changes. They are not deck win-rate targets, early-turn Snap rules or a promise that a position “looks 40%.” At exact equality the mathematical cube averages are the same, not a command to stay.

Earlier in the game, you may draw another card, reveal a location, receive new information or face another Snap. Those future decisions are missing from this small model. Use it to understand the final price of continuing, not to justify accepting every early Snap because you might eventually win one game in four.

Ties have their own outcome

A complete tie changes neither player's cubes. In a calculation that allows ties, compare retreating for −R with S × (chance of winning − chance of losing). The tie cases contribute zero.

For example, suppose the final result is worth 4 cubes, retreat costs 1, and the hypothetical outcomes are 30% wins, 20% ties and 50% losses. Staying averages 4 × (0.30 − 0.50) = −0.8 cubes, which is better than −1. Treating every non-win as a loss would incorrectly turn those ties into defeats.

Do not count a tied location as a tied match. A match can still have a winner when both players win one location and the third is tied; the overall power tiebreak matters. The calculation needs the outcome of the whole match, not how many individual lanes look even.

Turn uncertainty into a board plan, not a made-up percentage

You will rarely know a precise win chance during a match. Instead, identify what has to happen for your final play to win. Start with your strongest legal sequence and the two locations it is intended to secure. Then consider the opponent's plausible replies and which of those actually changes the match result.

A useful review has three parts:

1. Name your route to two locations. A large total at one location is not a complete win condition.
2. Name the replies that defeat that route: more power in the contested lane, a disruption effect, or a way to change where power ends up.
3. Look for a different allocation of your own cards that covers more of those replies without abandoning your second location.

Do not give every imagined reply equal weight. Listing four possible finishes does not make each one a 25% outcome; they may require different cards, draws or earlier setups. The point of the list is to find a robust line and expose a missing answer, not to manufacture a probability.

If every plausible reply you can account for defeats your only route, preserving the smaller cube loss is a sensible decision. If you have a credible route against several likely replies, compare it with the actual cost of staying rather than retreating simply because one scary card could exist.

Snapping is a separate decision from staying

A position worth continuing does not automatically justify raising the stakes. Staying compares two available outcomes: a small certain loss or the current final contest. Snapping changes what your opponent is offered and how they may respond.

Consider a final turn where your opponent would have stayed for a 2-cube result. If you Snap and they retreat for 1, you collect 1 instead of playing that possible 2-cube finish. If they stay and lose, you can collect the larger result; if their answer beats yours, your own loss is larger. The opponent's response is part of the decision.

This is why waiting until your advantage is obvious is not automatically the best time to Snap. Earlier confidence can offer a different choice before your strongest play is revealed, but it also commits you with less information. Neither “always Snap early” nor “never Snap on the final turn” follows from the cube arithmetic.

Before you Snap, answer two questions: what new information makes your position stronger, and what could an opponent who accepts the Snap be representing? A hand you like is not enough if the locations or revealed opponent plan have already undermined its route to winning.

Do not treat a Snap as a reversible test. When retreat is otherwise allowed, your own Snap locks you into that turn unless the opponent Snaps back. Decide whether you are willing to continue before pressing it, not after trying to provoke a response.

Retreat Now versus Retreat Later

Retreat Now ends your participation immediately and loses the displayed current cubes. Retreat Later also commits you to leaving, but waits for the opponent's choice: if they retreat too, the match is a tie; if they play normally, you retreat for the displayed amount.

“Later” does not mean “show me their cards and then let me reconsider.” It is useful when you have already decided to leave but want to allow the mutual-retreat outcome. It does not turn a losing final board into an extra chance to play.

A practical sequence for the last decision

• Identify the retreat loss and the final stakes separately.
• Find your best route to winning the match, not merely adding the most total power.
• Test that route against plausible opposing replies without pretending all replies are equally likely.
• Decide whether to retreat or continue at the existing terms.
• Consider a Snap separately, including the opponent's ability to leave and your own commitment.

The goal is not to avoid every loss. It is to stop turning a small, reasonable loss into a large one without a credible reason—and to recognize when continuing an uncertain match is better than surrendering cubes automatically.