Master advanced Golf Solitaire strategy. Endgame planning, probability-based decisions and minimum score techniques explained for experienced players.
Golf Solitaire rewards two distinct types of player: the one who wins hands outright — clears all 35 tableau cards before the stock runs out — and the one who loses hands well, finishing each unwinnable deal with the lowest possible penalty score. Most experienced Golf players develop reasonable instincts for the first type and almost none for the second. Yet in cumulative five- or ten-hand scoring, which is how Golf is almost universally played, the ability to score three or four strokes on a l
The endgame in Golf Solitaire begins roughly when seven or fewer stock cards remain and four or fewer tableau columns still have cards. At this point the game transitions from chain-extension optimisation — finding the longest runs — to endgame sequencing: arranging the order of remaining plays to maximise the number of cards removed given the exact stock cards known or likely to remain. Endgame planning is a different skill from mid-game chain building, and players who do not make this transition consciously waste the resources they built up across the earlier game.
Audit the remaining tableau at the endgame threshold. When seven or fewer stock cards remain, pause and take a complete inventory: how many cards are left in each column, what is the top card of each remaining column, and which ranks are still represented among the accessible column tops. This audit transforms the endgame from a reactive chain-extension exercise into a planning problem with known inputs. Without the audit, the final seven stock draws are used as reactively as the first seventeen; with it, each draw can be evaluated against a known remaining-card landscape.
Identify the minimum chain needed to clear and work backward. If five tableau cards remain and seven stock cards remain, clearing the hand requires a chain of at least five cards from those five tableau tops — possibly using one or two intermediate stock draws to bridge chain gaps. Identify the longest possible chain that removes the most remaining tableau cards, trace it backward from the last tableau card to the current chain top, and execute it as a planned sequence rather than as a series of opportunistic individual plays. This backward-chain technique routinely recovers one or two additional cards in the endgame compared to forward-reactive play.
Sequence accessible column tops by rank adjacency before the final draws. In the last four to five stock draws, the chain top after each draw determines which column tops are available for the next extension. Before each final draw, map the rank-adjacency relationships between all remaining accessible column tops: which tops connect to which other tops, and which connection sequence produces the longest chain? A top-card map takes fifteen seconds to construct and routinely reveals chain paths that forward-reactive play misses — specifically, chains that require playing column tops in a non-obvious order to bridge from one rank cluster to another.
Prioritise removing cards from columns that block no other card. In the endgame, some column tops cover no further cards — they are the last card in a one-card column. These are free removals: clearing them costs nothing structurally and reduces the penalty count by one. Prioritise same-chain removal of single-card column tops over removing tops from deeper columns when chain adjacency is equal, because single-card column removal has no opportunity cost while deeper column removal always defers the face-up card beneath it to the next chain segment.
When clearing is impossible, plan the penalty-minimising endgame sequence. If the audit reveals that clearing all remaining tableau cards is mathematically impossible given the remaining stock, shift immediately to minimum score play (see below). Do not exhaust the remaining stock pursuing the impossible clear — each draw in an unclearable endgame should be evaluated for penalty reduction value, not chain length.
Golf Solitaire deals 35 of 52 cards to the tableau and holds 17 in the stock. From move one, the deck composition is known: four cards of each rank, 52 total. As the hand progresses and cards are played — both from the tableau and from the stock — the composition of the remaining unplayed cards shifts in trackable ways. Probability-based decision making uses this shifting composition to make higher-expected-value choices when multiple options look locally equivalent.
Track rank depletion as the hand progresses. Each rank starts with four cards in the combined tableau and stock. As cards are played, the remaining count of each rank decreases. A rank with three cards already played has only one remaining across the entire undealt stock and unplayed tableau; a rank with zero played has all four still in circulation. Tracking rank depletion — even approximately, registering when a rank has gone from four to two to one remaining — gives a probabilistic picture of what the stock is likely to contain in its remaining undrawn portion. This is the foundation of all probability-based decisions in Golf.
When choosing between two chain-adjacent column tops, prefer the one whose rank has more remaining cards in circulation. Suppose the current chain top is an 8 and both a 7 of Hearts and a 9 of Clubs are accessible. Playing the 7 makes 6 and 8 the next available ranks; playing the 9 makes 8 and 10. If three 6s have already been played but only one 10 has, the 9 is the higher-probability play — the 10-direction of extension has more remaining cards and is more likely to produce a continuation from either the tableau or the next stock draw. This rank-depletion preference — choosing the play whose resulting chain top has more extension cards remaining — is the core probability habit in Golf and produces a measurable improvement in chain length over random same-value selection.
Calculate implied continuation probability before each stock draw. Before drawing from the stock, count the remaining accessible column tops and their ranks. Each accessible rank has a certain number of cards remaining in the undrawn stock. If four accessible column tops cover ranks 5, 7, 9, and Jack, and the stock has twelve cards remaining, the probability that the next draw produces a rank adjacent to at least one of those four tops is calculable: count the number of remaining 4s, 6s, 8s, 10s, and Queens in the stock, divide by twelve. When this probability is below roughly 30%, the next draw has a high likelihood of producing a dead chain top — and the subsequent draw will also need to be made, burning two stock cards for one chain restart. This low-probability signal means the current position should be re-evaluated: is there a different accessible top that produces a higher continuation probability than the current chain top?
In wrap-around variants, account for K-A and A-K transition probability separately. In implementations where Kings can be followed by Aces and vice versa, the rank-adjacency distribution changes: Kings are now adjacent to both Queens and Aces, and Aces are adjacent to both 2s and Kings. This doubles the continuation value of Kings and Aces as chain tops compared to non-wrapping Golf and changes the rank-depletion preference calculation. In wrap-around Golf, a chain top landing on a King or Ace after a stock draw is not a near-terminal position but a potentially high-continuation position — track remaining Aces and Kings in the undrawn stock with the same attention as mid-rank cards.
Use late-hand rank knowledge to guide final stock draws. By the time seven or fewer stock cards remain, between ten and fifteen tableau-and-stock cards have been played, and the rank distribution of the remaining undrawn stock can be substantially narrowed. If only one stock card of rank 7 has been played, two or three 7s are still in the combined tableau-and-undrawn-stock pool; if all four 7s have been played, no 7 can appear from a stock draw. This late-hand rank knowledge transforms stock draw decisions from pure uncertainty to constrained probability — and in the endgame, constrained probability decisions are meaningfully better than uniformed ones.
In cumulative Golf scoring, the penalty for each hand is the number of tableau cards remaining when the stock is exhausted. A hand that finishes with two cards remaining scores two strokes; a hand that finishes with eight scores eight. Over a ten-hand session, the difference between averaging three strokes on losing hands and averaging seven strokes on the same hands is forty strokes — the equivalent of winning four additional hands outright. Minimum score play is the discipline of treating every card removed in a losing hand as a stroke saved, not as a step toward a clear that is not coming.
Identify the unwinnable hand as early as possible. The earlier an unwinnable hand is identified, the more stock cards remain and the more penalty strokes can be removed. An unwinnable hand typically reveals itself through one of three signals: a circular dependency (two column tops each rank-adjacent only to each other, with no stock card likely to bridge out of the loop); a terminal-rank isolation (a King or Ace column top with no adjacent accessible card and no remaining stock cards of the adjacent rank); or a rank extinction (a rank that appears in multiple column tops but has already had all four of its cards played, meaning those tops can only be cleared by the chain approaching from the other rank direction, which may itself be blocked). Recognise these signals and switch to minimum score play before the stock is nearly exhausted.
In minimum score play, prioritise stock draws that produce the highest-probability chain extensions. Once the hand is identified as unwinnable, the goal of each stock draw is not to restart the longest chain but to remove the maximum number of remaining tableau cards per stock card consumed. Apply the continuation probability calculation above to each potential draw: the draw that produces the highest expected number of subsequent chain extensions is the draw to make. Draws that produce one-card chains followed by another draw are less efficient than draws that produce three- or four-card chains, even if the latter require spending one extra stock card to reach the productive chain top.
Clear Kings and Aces as a batch when they become accessible. In non-wrapping Golf, Kings and Aces are chain terminals — each requires a dedicated stock draw to clear past. In a losing hand, each uncleared King or Ace column top is a guaranteed penalty stroke. When the hand is identified as unwinnable, prioritise any available stock draw that can reach a King or Ace column top, remove it, and immediately count the stroke saved. Multiple King and Ace column tops that are simultaneously accessible via a short chain should be cleared in a single chain segment using the chain order that removes the most of them — trace the adjacency path through K-A-2-3 or Q-K (with A-K wrap if available) and plan the clearing sequence before beginning it.
Accept a one-card chain if it removes a terminal-rank column top. In minimum score play, the standard preference for longer chains over shorter chains is overridden when a one-card chain removes a King or Ace column top. A one-card chain that clears a terminal is worth one guaranteed stroke saved at the cost of one stock card — this is almost always a positive expected value exchange in a losing hand, because the alternative (deferring the terminal until a chain reaches the adjacent rank) risks the stock running out before that chain arrives. Take the terminal removal whenever it is accessible, even at the cost of chain momentum.
In the final three stock draws, evaluate each draw for pure stroke expectation. With three stock cards remaining, each draw has a calculable probability of producing a chain extension given the rank-depletion state. Draw one: if the continuation probability is above 50%, take it and play whatever chain follows. If below 50%, consider whether the current chain top already has an accessible extension — if it does, play that extension before drawing, preserving the draw for later when the chain has advanced to a less productive rank neighbourhood. Draw two and three: apply the same probability-first logic, using the now-narrowed rank distribution from all previously played and drawn cards. The final three stock draws in a losing hand are the highest-leverage decisions in minimum score play — making them probabilistically rather than reactively saves between one and two strokes per session on average.
Maintain a running count of played cards by rank. Mentally or on paper, track how many cards of each rank have been removed from the game — played from the tableau or drawn and played from the stock. This count, updated after every card removal, is the direct input to every probability calculation in the game. Players who track rank depletion make better chain direction choices, better stock draw timing decisions, and better minimum score play decisions than players who rely on impression. The tracking habit is cognitively demanding at first but becomes automatic within three to four practice sessions.
Separate chain quality from chain length as evaluation criteria. A long chain is better than a short chain when all other factors are equal — but chain quality, measured as the expected number of future extensions it enables, is more important than raw length. A five-card chain that ends on a King (non-wrapping) is lower quality than a three-card chain that ends on a 7, because the 7-ending chain has continuation options on both sides while the King has none. Prefer quality-ending chains in the mid-game when stock conservation matters most; accept length-over-quality chains only in the endgame when the remaining stock is small enough that future continuation probability is low regardless of the chain top rank.
In multi-hand sessions, adjust per-hand risk tolerance based on running score. In a five-hand session where the running score is already low (fewer than ten total strokes after three hands), the remaining two hands can be played conservatively — preferring stock-conserving moves and minimum score discipline over aggressive chain-length maximisation. When the running score is high (more than twenty total strokes after three hands), the remaining hands require aggressive play — prioritising complete clears over stroke minimisation, accepting draws that might produce dead tops in pursuit of the chain momentum needed for an outright win. Adjusting risk tolerance based on running score is the session-level strategic layer that separates Golf as a cumulative game from Golf as a series of independent hands.
When seven or fewer stock cards remain, stop playing reactively and begin the endgame audit: count remaining tableau cards, identify accessible column top ranks, map rank-adjacency relationships between tops, and trace the longest possible removal chain backward from the last accessible card to the current chain top. Execute this planned sequence rather than extending the chain opportunistically. Single-card column tops are free removals that should be prioritised when chain adjacency is equal. If the audit reveals that a complete clear is impossible, switch immediately to minimum score play and evaluate every remaining draw for stroke-reduction value rather than chain length. Play our free Golf Solitaire game across a five-hand session and apply the endgame audit from the first hand to build the habit before it matters most.
Golf deals from a known 52-card deck, and every card played — from the tableau or the stock — reduces the pool of remaining unplayed cards in a trackable way. The core probability habit is rank depletion tracking: registering how many cards of each rank have been removed and using that count to evaluate the relative continuation value of each chain direction at every decision point. When two column tops are both chain-adjacent to the current top, prefer the one whose resulting chain top has more remaining cards in the undrawn pool — it has a higher probability of producing a continuation from the next stock draw. In the endgame, rank depletion knowledge narrows stock draw outcomes from pure uncertainty to constrained probability, making draw timing decisions meaningfully better than uninformed ones. See our Golf Solitaire strategy guide for the foundational framework this builds upon.