Multi-Table Tournament Strategy

One tournament spread across multiple tables

A multi-table tournament (MTT) is one event whose entrants initially play across several tables. In an ordinary elimination MTT, players are moved as tables are balanced or broken, eventually reaching a final table. This is different from multi-tabling—one person playing several separate games at once. This lesson concerns the tournament, not increasing the number of games on your screen.

We focus on conventional no-limit Hold'em elimination events. A structure sheet tells you the starting chips, blinds and antes, level duration, registration window, breaks, entry rules and prizes. Do not assume a tournament's name establishes those details. Shootouts, satellites, bounties and timed events can have different objectives or procedures.

Tournament chips normally cannot be redeemed at face value. Winning 10,000 chips does not mean earning $10,000. Nor does surviving guarantee a profit: entry fees and other costs remain, even if you reach a paid place. Start with the tournament and cash-game comparison if that distinction is unfamiliar.

Read the tournament stage—not a fixed aggression script

These stages do not arrive at the same blind depth in every event. A short stack can exist in level one; a final table can still contain deep stacks. There is no universal instruction to tighten early, attack the middle, then fold everything on the bubble. Antes may be present from the start, introduced later or absent.

Raising, calling and checking can each belong in a strategy. Naming a stage does not justify “three-bet or fold,” guarantee a blind steal or reveal that another player is afraid. Use position, weighted ranges and the actual price before comparing actions.

The same chip stack can be a very different number of big blinds

Stack depth is stack ÷ current big blind. The following isolates that denominator: hold a hypothetical stack constant at 20,000 and change only the big blind. It is not a record of someone folding through levels; real blinds, antes and hands would change the stack.

Scroll within the table if needed. Keyboard users can focus the table area and use the arrow keys.

Three doublings of the big blind reduce 200 BB to 25 BB, not 15 BB, if the chip count stays fixed. In a real hand, also identify the effective stack: how much the relevant opponents can contest against one another. A player with fewer chips cannot win more from you than the amount you can match in their pot, but losing that amount can still matter greatly.

For chip-pot accounting, the effective stack is useful. For a prize model, it is not enough: a short stack elsewhere can affect the value of avoiding elimination. “Deep,” “medium” and “short” are descriptions, not mathematically universal boundaries. Short-stack play can retain smaller raises, calls or limps in some configurations; a low BB count does not make those actions illegal.

ICM estimates prizes, not a withdrawable balance

The Independent Chip Model estimates expected prize money from the remaining chip stacks and payout list. Its basic rule assigns a player's chance of first place in proportion to their chip share. For lower places, it considers each possible higher finisher, removes that player's stack and repeats the proportional calculation among those left.

George T. Gilbert's 2009 mathematical paper, section 3 describes this recursion. It is a model of finishing probabilities—not proof of the actual players' chances. Basic ICM does not encode skill differences, positions, future blind timing or all the possibilities of subsequent play. Bounty rewards and unusual award structures need further treatment.

Showdown equity is a hand's expected share of a pot under specified cards, ranges and runouts. ICM prize equity is a modelled share of tournament prizes. A 55% chance to win a hand is neither a 55% chance to win the event nor automatically 55% of its money. The ICM glossary entry provides a shorter definition.

A finite three-player example

Three players remain with 10,000 chips each. Fixed prizes are $500, $300 and $200, totalling $1,000. Assume ordinary finishing-place awards, no bounties, deals, re-entry, additional fees or prizes, and use basic ICM without a skill adjustment. These are invented teaching inputs, not a current event or observed player data.

Initially each player is equally likely to occupy each place within the model: ($500 + $300 + $200) ÷ 3 = $333.33, rounded. Now consider a chip exchange in which A either wins all of B's 10,000 chips or loses all 10,000 to B; C's stack stays unchanged. Exactly one player is eliminated in either branch and receives the fixed third-place $200. This is not a rule for simultaneous eliminations.

Scroll within the table if needed. Keyboard users can focus the table area and use the arrow keys. Each row is a separate state, and money is rounded to cents.

When A wins, only $800 remains for the two survivors because B has received $200. A's 2:1 chip lead gives an ICM first-place probability of two-thirds: (2/3 × $500) + (1/3 × $300) = $433.33. C's corresponding equity is $366.67. Including B's awarded $200, the unrounded total remains $1,000.

A has doubled the chips, but not the $333.33 prize equity. The winning branch adds $100 of modelled equity; the losing branch reduces it by $133.33. These values are finite. The “last chip” is not infinitely valuable, and a chip stack does not have one fixed dollars-per-chip rate in this example.

Positive chip EV can still mean negative prize EV

Compare taking that exchange with leaving all three stacks unchanged. This is a deliberately simplified model comparison—not a complete legal hand or a claim that folding to a real bet leaves stacks unchanged. There are no blinds, antes, dead chips, ties or intermediate contributions in this comparison. The only branches are A winning or losing the full 10,000; no additional wagering cost is applied. Win probabilities below are stipulated, not evaluated from cards.

Scroll within the table if needed. Keyboard users can focus the table area and use the arrow keys. Changes use A's initial 10,000 chips and exact $1,000 ÷ 3 prize equity, before rounding.

At 55%, expected chips are 0.55 × 20,000 = 11,000: a gain of 1,000 chips. Expected prize money is 0.55 × $433.333… + 0.45 × $200 = $328.33, about $5 below the unchanged state. At 60%, the same model instead gains about $6.67 in expected prize money.

The chip break-even chance is 50%. The prize break-even chance is ($333.333… − $200) ÷ ($433.333… − $200) = 4/7, or 57.14%. The difference illustrates a modelled risk premium; it is not a universal tournament calling threshold.

C gains modelled prize equity without winning chips because one competitor is eliminated. That does not imply waiting is always best: actual blinds, changing opportunities and other outcomes are absent here. The example compares expected gross awards; tournament entry costs would need to be included in any overall net-return calculation.

Bubble and final-table decisions need all the inputs

The money bubble is the boundary just before paid places. With 100 players left and 99 paid, the next single elimination would leave the paid field. Simultaneous bust-outs can complicate finishing order. In its own online rules, PokerStars describes hand-for-hand synchronization; that is a named house procedure, not a claim that every event resolves every tie identically.

Relative stacks change consequences. A covering stack may threaten another player's elimination, but can still lose a large amount of its own equity. Two big stacks are not exempt from ICM effects. A short stack still has something to lose; a medium stack is not forced to fold without a premium hand.

For a final-table decision, record every stack, the full remaining payouts, positions, blinds and antes, previous actions, legal sizes and plausible responses. “A-J offsuit under the gun with 25 BB and two shorter stacks” is not enough to prove a fold. Missing the other stacks and response ranges prevents an exact recommendation.

Pay jumps are differences between awards. A move from no payout to a positive payout has a meaningful dollar difference, but dividing that increase by zero does not produce a useful percentage comparison. Neither the bubble nor the final table is automatically the point of maximum pressure for every player.

Heads-up is a useful special case

ICM still has a value heads-up. With first prize $500, second $300 and chip share s, the model gives $300 + s × ($500 − $300). That is linear in chip share; the $300 is already secured. Under this ordinary two-prize model, there is no extra nonlinear ICM risk premium between the two remaining players. It does not mean the payout model has ceased to exist or that every hand should be played aggressively.

Likewise, winner-take-all ICM is proportional to chip share. The nonlinearity in the three-player example is not a universal law for every tournament structure.

A push/fold chart solves its listed game—not every short stack

A shove-or-fold chart removes smaller raises, calls and limps from at least some decisions in its action tree. A mathematical answer inside that restricted tree does not prove those omitted actions are inferior in the full game. Stack size alone is not enough to choose a chart or announce a compulsory cutoff.

• Configuration: how many players remain, where they sit, and which stacks cover which others.
• Price: blinds, ante arrangement and any prior raise; an unopened button decision differs from facing an all-in.
• Objective: chip EV, specified prize equity, bounty value or an equal-seat satellite.
• Actions and responses: which raises/calls are allowed and which opponent strategy the result assumes.
• Evidence: the chart's inputs, method and accuracy—not an unexplained “Nash” label or percentage.

No solved ranges are supplied here. Any-ace or suited-connector shove rules, and a quoted percentage for “10 BB on the button,” would be incomplete without those inputs. Study range construction and call prices before interpreting a result. Only use tools in ways the relevant rules permit; offline education is not permission for real-time assistance.

Satellites, bounties and other structures change the question

A satellite awards entry or qualification into another event. In a conventional equal-seat satellite, all qualifying seats have the same stated award, so first place need not be more valuable than the last qualifying place. But leftover cash, different tickets, target-stack rules or registration restrictions can alter that description. Read the actual awards and conditions; a ticket is not automatically transferable or redeemable cash.

A chip lead is not a universal guarantee that folding every hand will qualify. You need the remaining stacks, seats available, blind schedule and event rules. “Eliminate a short stack risk-free” is not a valid default: even a covering stack can lose chips and qualification prospects.

Bounty events add elimination rewards, sometimes changing during play. A finishing-prize-only ICM calculation omits those rewards. A shootout changes table progression. These labels can overlap: a satellite can be multi-table or a sit-and-go, and a bounty event can also use ordinary elimination. Do not treat every format label as a mutually exclusive choice.

Re-entry is a new cost—not a reason to take an unsupported risk

A freezeout provides no new entry after elimination. A re-entry format may permit another paid entry within defined limits; a rebuy or add-on has its own conditions. PokerStars' published type definitions distinguish these mechanisms. They do not establish the rules or availability of an unrelated event.

The possibility of another entry does not make the first one free to lose. Count all fees and entries in a spending boundary established before play; a limit may be one entry or none. No generic “two or three bullets” allowance proves affordability. Money already spent is not a reason to spend again. See budgets and limits.

Deals, absences and stopping

An ICM estimate is not an entitlement to cash out, and a proposed deal is not automatically wise because it exceeds one model value. Check whether deals are permitted, how agreement is recorded and what remains to play for. PokerStars' cited rules include event-specific restrictions; do not arrange a private seat handoff or assume selling a stack is allowed.

You can choose to stop participating. What happens to an absent stack follows the event's rules; leaving does not normally let you redeem tournament chips, and it does not universally erase every possible payout. If circumstances require you to leave, ask the organiser about the procedure rather than treating continuation as an obligation to recover money. Psychology and stopping and independent support are relevant when gambling is causing pressure.

A tournament review should identify a question you can check

1. Record the event structure and all remaining prizes. Separate entry fees and awarded money from tournament chips.
2. Reconstruct the actual stacks, positions, pot and previous commitments. A “25 BB hand” is not a complete situation.
3. State uncertain ranges or win probabilities as assumptions. Do not infer a permanent tendency from the result.
4. Compare legal alternatives under the same objective. Chip EV and expected prizes can rank a decision differently.
5. Review costs and your stopping boundary separately. No model requires another entry or more playing volume to prove itself.

These examples can be studied away from a game. Continue with betting purposes and alternatives, action order or the format comparison—not a promise that tournament play will produce a return.

Sources and scope

Primary sources inspected 24 September 2026:

• Gilbert, The Independent Chip Model and Risk Aversion, version 1 (16 November 2009): the finishing-probability model and its mathematical scope. Our finite numbers are independently derived examples, not a player study.
• PokerStars tournament rules: selected house procedures for table progression, hand-for-hand play and deals. These are platform-specific, not universal tournament law.
• PokerStars tournament types: terminology for MTTs, entries, satellites and other structures. No listed event is presented as available or suitable for the reader.

No solver run, played-hand record, forecast, legal review or financial recommendation is claimed. Basic ICM omits important real-game information. All values are finite, all examples hypothetical, and source-access dates are not claims of newly published research.