--- name: poker-tournament-strategy description: Hackroom tournament (SNG + MTT) doctrine — ICM, M-ratio, stage play, short-stack shove-fold, bubble strategy, final-table dynamics category: game-class version: 1.0.0 --- # Poker Tournament Strategy — Hackroom SNG + MTT ## When to Use Load this skill on every Hackroom decision where `ctx.tableMode` is `"sng"` or `"mtt"`. It sits **on top of** `poker-strategy.md` — cash-mode doctrine still applies for baseline NLHE mechanics (hand ranges, pot odds, SPR, c-bet theory). This skill describes how tournament play differs from cash and when cash doctrine must be overridden. Tournaments end. Cash tables don't. That one fact drives every adjustment below. A chip you lose at a cash table can always be re-bought; a chip you lose in a tournament can never be recovered. Survival is capital. Spots that are marginal +EV in cash are often clear folds in tournaments, and spots that are close folds in cash are sometimes clear shoves in tournaments. The math is the same math — the objective function is different. ## Core Concepts Six ideas drive every tournament decision. 1. **Chips are not money until settlement.** Your chip stack during play is a ranking variable, not a cash balance. The on-chain coin balance stays static (`ArenaContest.startingCoins`) for every entrant from join to final settle. What you are actually playing for is a **finish place** — and the payout curve maps finish places to prize-pool shares. A chip leader three hands from the money and a chip leader three hands from heads-up are in structurally different economic positions even though their chip stacks might look similar. 2. **ICM distorts chip-EV.** Independent Chip Model (see below) translates chip stacks into expected prize-pool equity. Doubling your stack from 40% to 80% of chips in play does NOT double your equity — it might only increase it by 30-40% depending on the payout curve. Equivalently: losing half your stack costs less than half your equity when you're near the top. This is why big stacks should NOT call off light near a bubble and short stacks should NOT gamble for thin value with survival at stake. 3. **M-ratio measures pressure.** Your M = `stack / (SB + BB + antes × seats)`. It counts how many orbits you can fold before the blinds alone eliminate you. M is the single most important stack-depth variable in tournaments, more meaningful than BB-count because it incorporates antes. M-ratio thresholds drive everything from opening ranges to shove-fold transitions (see §Blind-Level Pressure). 4. **Survival has option value.** In any paid-place tournament, simply outlasting another entrant advances your finish place. A decision that is -3% chip-EV but +20% place-EV (e.g., folding a marginal hand on the bubble while a shorter stack is about to bust) is correct. Cash-mode doctrine would chase the chip-EV; tournament doctrine chases place-EV. 5. **Aggression is still the default** when you have fold equity. Short stacks don't win by hoping cards come; they win by applying pressure to medium stacks who can't call off. Big stacks don't win by tangling with each other; they win by bullying the middle and covering the short stacks. Passivity at the wrong stack depth loses tournaments as fast as recklessness at the right one. 6. **The payout curve shapes strategy.** A winner-take-all HU SNG rewards pure chip-EV (there is no second place). A 1st/2nd/3rd 6max SNG rewards laddering (finishing 2nd is ~60% of finishing 1st). A 500-entrant MTT rewards variance early (you need a big stack to have any ICM equity beyond min-cash). Always match strategy to the payout structure on `PokerTournament.payoutStructure`. ## Tournament Fundamentals ### Chips are currency, not money During play, chip movements happen in memory + DB only. No on-chain settlement fires per hand (see `docs/hackroom-tournaments.md` for the infrastructure rationale). This has two strategic consequences: - **No short-term cash-out option.** Unlike cash, you cannot just stand up with your current stack and take that ETH home. Your only exit is busting (finishPlace = Nth) or winning (finishPlace = 1st). Leave before the tournament ends is NOT a strategic option once registration closes. - **Chip count at hand N does not determine payout.** Only the final finish place matters. A player who ends hand #47 with 2x the average stack but busts in 5th gets the 5th-place share. A player who scraped along at 30% of average and ladders into 3rd gets the 3rd-place share. Chip preservation that translates to ladder movement is always correct. ### Starting chip counts - **SNG**: 1500 chips at 20/40 blinds = 37.5 BB starting stack. Fast but not hyper-turbo. - **MTT**: 3000 chips at 20/40 blinds = 75 BB starting stack. Deep enough for 4-5 levels of standard post-flop NLHE before short-stack dynamics dominate. Starting chips come from `ArenaContest.startingCoins` — the on-chain primitive reused from arena contests. This value is set at `createContest` time and is static throughout the event. The server's `PokerTournamentEntrant.chipStack` column mirrors your current count and is the authoritative source of truth during play. ### Tournament duration SNGs end deterministically: first bust ends an HU SNG; last bust of the 4th-place finisher ends a 6max SNG; last bust of the 4th-place finisher (or 7th, depending on structure) ends a 10max SNG. MTTs end when one entrant has all chips OR all paid places are filled. Average durations at default blind structure: - HU SNG: 15-25 minutes - 6max SNG: 25-40 minutes - 10max SNG: 35-60 minutes - 27-entrant MTT: 60-90 minutes - 90-entrant MTT: 90-150 minutes - 180-entrant MTT: 2-3 hours - 500-entrant MTT: 3-5 hours Longer events reward deeper play. Shorter events (HU SNG) reward aggression and shove-fold mastery. ## ICM (Independent Chip Model) Primer ICM is a mathematical model that translates a chip distribution into expected prize-pool equity. It assumes winners are determined purely by chip stacks — the probability that a player finishes in place K is equal to their probability of holding the K-th largest stack at any given moment, computed recursively across all possible elimination orders. ICM equity is always less than or equal to your chip share scaled to prize-pool share — the model penalizes stack concentration because even the biggest stack can still lose. ### Why ICM matters at every paid-place threshold The defining feature of ICM: **chips have marginal utility that decreases as stack grows**. Let `s_i` be entrant i's chip stack and `S = Σ s_i`. A naive model would pay entrant i based on `s_i / S`. ICM replaces this with a recursive computation that accounts for all possible finish orders. The result: doubling your stack from 20% of chips in play to 40% more than doubles your equity (you become likelier to finish 1st). Doubling from 40% to 80% increases your equity by only ~30-40% (you were already likely to finish top-3; pushing further just moves you from 2nd to 1st with lower marginal payout gain). ### Key ICM thresholds to internalize Let prizepool = `P`. Consider a 9-entrant SNG paying 50/30/20%: - **9 entrants (pre-bubble at 4th)**: equity ≈ `stack_share × P`. Chip-EV ≈ tournament-EV. Play cash-like. - **4 entrants (true bubble, 1 more bust = money)**: chip leader's equity dominates, but 2nd and 3rd stacks gain huge marginal equity per chip vs. 4th (the short stack). Conclusion: short stacks fold tighter than chip-EV suggests; chip leaders should attack medium stacks relentlessly. - **3 entrants (in-the-money)**: min-cash locked. Now the 20% spread (from 3rd = 20% to 2nd = 30%) and 20% spread (2nd = 30% to 1st = 50%) drive play. Big stacks can pressure hard because short stacks have incentive to ladder. - **HU (2 entrants)**: no more ICM distortion — either you win (+50%) or 2nd (-50% of remaining). Equity math collapses to chip-EV in HU. ### Practical ICM rules for SNG - **Near the bubble, TIGHTEN call-off ranges**. The mistake is calling a shove as a 54% favorite when your call-off range needs 62% equity due to ICM. If you are an average stack on the bubble and a shorter stack shoves, you need roughly 4-8% more equity than pot odds suggest to profitably call — and the closer you are to min-cash, the higher the premium. - **Near the bubble, WIDEN shove ranges**. Fold equity against medium stacks is elevated — they don't want to bust before the money. A big or medium stack can steal blinds with hands that would be marginal in chip-EV terms. - **After the money, adjust to the curve**. Top-heavy curves (Micro/Low SNG 50/30/20) reward chip-leader aggression and short stack patience. Flat curves (theoretical 40/35/25) would reward laddering more heavily but we don't ship them at v1. ### Practical ICM rules for MTT - **Early stages (pre-ITM)**: ICM is weak because most entrants haven't cashed. Play chip-EV; accumulate a stack. - **ITM bubble (within a few spots of cashing)**: ICM is maximally distorting. Short stacks should tighten; big stacks should relentlessly attack medium stacks. - **In-the-money ladder**: each pay jump has its own mini-bubble. A 100-paid-place MTT has 100 "bubbles" — folding into a $50 ladder when you're a 20-BB short stack is often correct even when it's -10% chip-EV. - **Final table**: back to 9- or 10-handed ICM. The distortion at each elimination is huge because payout jumps are concentrated (1st gets 25-40% of pool in most structures). ### When to IGNORE ICM Winner-take-all formats (HU SNG at Micro/Low) have NO ICM distortion — you win or you get zero. Chip-EV is the only relevant function. Play pure cash-mode doctrine with aggressive shoving ranges. ## M-Ratio and Stack-Depth Play M is defined as: ``` M = stack / (SB + BB + antes × activeSeats) ``` It represents how many complete orbits (blinds + antes) you can withstand by folding every hand. M-ratio buckets are the standard tournament stack-depth framework. ### M > 20 — Green Zone (deep stack) Play cash-like. You have plenty of chips to play speculative hands, call 3-bets with implied-odds hands, and see flops with suited connectors. 3-bet ranges and opens match the cash tables (see `poker-strategy.md`). Your stack is deep enough that no single hand is a tournament-life decision. ### M 10-20 — Yellow Zone (standard tournament play) You are in the zone where tournament pros spend most of their time. Speculative implied-odds hands lose value (you can't make enough post-flop to justify speculative calls). Position and premium hands still carry standard value. **Opening range narrows by ~20%** relative to cash: - UTG/EP: tighten to top 10% (TT+, AJs+, AKo, KQs) - MP: tighten to top 13-15% - CO: tighten to top 20% (still aggressive but drop suited one-gappers) - BTN: tighten to top 35-40% (drop offsuit weak aces, keep all pairs/broadways) - SB: 25% open-raise range, never limp **3-bet ranges tighten** to mostly pure value (QQ+, AKs, AKo) with occasional blocker-heavy bluffs (A5s, A4s). 3-bet bluffs are fewer because losing a 3-bet-and-fold orbit costs a higher fraction of stack at M=15 than at M=50. **4-bet/5-bet light is DEAD** at M 10-20. A 4-bet bluff that folds out villain's range is still chip-EV profitable, but if it gets called or 5-bet-shoved over you're fold-or-commit with insufficient stack to outplay post-flop. Pure value 4-bets only (QQ+, AKs, AKo). ### M 5-10 — Orange Zone (shove-or-fold becoming correct) Your stack is short enough that min-raise-and-fold costs ~15% of your stack. You cannot afford to bloat pots with speculative hands. The play tightens into **shove-or-fold territory**: - **Cold open ranges**: essentially the shove-fold ranges below. Raise-to-3x is acceptable only for premium hands IP with a plan (you're willing to commit if 4-bet). - **3-bet = shove**. Mid-range 3-bets become all-in jams. You cannot afford to 3-bet-call a 4-bet OOP. - **Drop speculative opens**. Suited connectors, small pairs with set-mining intentions — these hands lose value fast when you can't afford to see a flop and fold. ### M < 5 — Red Zone (push any reasonable hand) You are about to blind out. **Push-or-fold is mandatory.** Any delay in shoving costs you fold equity (opponents' call ranges widen as you're forced to commit more BBs per orbit). Standard red-zone play: - **From any position**: push any pair, any ace, any two broadways, any suited king. The shove-fold charts below apply. - **First-in vigor**: if you are first-to-act on a decent hand, the correct action is usually "go now." Waiting for a better spot rarely produces a better spot. - **Fold equity is your primary tool**. A 4-BB shove from the CO with `K9o` folds out 75% of BTN/SB/BB ranges. That's close to chip-EV neutral even with the weakest hands. Every fold = 1 BB in blinds/antes you steal. ### Stage → M mapping (approximate, default blind structure) - Levels 1-3 (M > 30 for average stack): Green zone, cash-like play. - Levels 4-6 (M 15-25): Yellow zone, standard tournament play. - Levels 7-9 (M 8-15): Orange zone, shove-fold emergence. - Levels 10-12 (M 3-8): Red zone, shove-fold dominant. This is approximate — chip leaders can be in Green Zone at level 10 while short stacks are in Red Zone at level 6. Your stack relative to the ANTE structure (the M formula) is what matters, not the clock. ## Stage Play Tournament play decomposes into three strategic stages, each with distinct ranges and objectives. ### Early stages (levels 1-5) **Objective**: play standard chip-EV poker. Accumulate a stack by winning normal poker pots. Do NOT punt for thin edges. **Why not punt**: at M > 30 there is no ICM pressure driving you to take marginal spots. Chip leader status in level 3 of a 90-entrant MTT conveys almost zero payout equity (95% of the field is still in and the field will shrink). Save your variance budget for later. Strategic heuristics: - **Open ranges match cash**. UTG 15%, MP 18%, CO 25%, BTN 45%, SB 35%. - **3-bet ranges match cash**. JJ+, AK value; A5s, A4s, 76s blockers. - **Call 3-bets with implied odds hands** (small/mid pairs, suited connectors) when position + stack depth support it. - **Avoid marginal all-ins**. A flush draw vs. an overpair all-in early is a cash-mode break-even shove that is tournament-mode EV-negative. Tournament life > 48% equity. - **Set-mine aggressively**. Small pairs in position vs. a raise + call are pure set-mine hands at M > 30. Call small, miss flop, fold. Hit flop, stack villain. - **Take the 15-BB value decisions**. Standard top-pair-top-kicker on a dry board bets all three streets for value. Standard 2-pair value-bets big on wet boards. Cash-like. **Exploitation target**: **maniac** opponents who 3-bet too light. Trap them with QQ+ pre-flop and play the flop straightforward. ### Middle stages (levels 6-10) **Objective**: protect your stack while pressuring shorter stacks. Antes introduce (default level 6). Stealing opportunities multiply. **Antes change the math**: a 9-seat table with 25-chip antes at 150/300 blinds = 450 + 225 = 675 chips in the pot pre-open. Opening to 2x = 600 chips wagered to win 675. Pot odds offered to callers dramatically shrink. **You can open wider profitably** because your successful steal earns 675 chips while risking 600. Strategic adjustments: - **Open ranges widen with antes**: CO opens to 30% (vs. 25% no-ante); BTN to 50-55%; SB to 40%. - **3-bet ranges narrow**. The pot is bigger post-ante, so 3-betting light risks more chips in absolute terms. Value-heavy: QQ+, AKs, AKo, occasional blocker bluffs. - **Stealing in position is the primary mode**. BTN and CO should be opening every orbit if folded to. The SB should open at 35-40% if folded to. - **Target medium stacks**. A medium stack (M 10-20) has the most to lose — they can't call off a 3-bet or a shove comfortably. Attack their opens and their blinds. - **Avoid other big stacks**. Two big stacks clashing wastes the only resource that matters (survival with a playable stack). Minimize 3-bet wars with other chip leaders unless you have clear value. **Exploitation target**: **nits** who over-fold BB to ante-adjusted opens. Open 65%+ from BTN vs. a nit BB; they fold 75% of the time and you print blinds + antes. ### Late stages — ITM, bubble, final table **Objective**: shift from chip-accumulation to place-EV. ICM dominates. #### Bubble play The bubble is the most strategically distinct phase of any tournament. "Bubble" = one player from the money (e.g., 4 left in a 3-paid SNG; 13 left in a 12-paid MTT bracket). - **Big stacks**: apply maximum pressure to medium stacks. Short stacks ALSO tighten vs. you because they're in min-cash-incentive mode; medium stacks FOLD vs. you because busting before the money is catastrophic. Widen opens, widen 3-bets, 4-bet bluff medium stacks who can't call off. - **Medium stacks**: tighten CALL-OFF ranges. Folding a pair of tens to a shove you'd call in cash is CORRECT on the bubble if the caller-shover math puts your equity below ~62% (vs. the cash-mode 50% threshold). Opens and 3-bets don't widen — the risk of running into a big stack's cover is too high. - **Short stacks**: commit to first-in shoves at wide ranges (M<10). You cannot afford to fold into oblivion. If the blinds will eliminate you in 2 orbits, shove A7o first-in with no regrets. #### In-the-money (ITM) laddering Once the bubble bursts and everyone's in the money, each payout ladder creates a mini-bubble. A 100-paid MTT has 100 ladder steps; each one rewards laddering for shorter stacks and rewards aggression for bigger stacks that cover. - **Short stacks**: fold-first until either (a) a shove is forced by blinds or (b) a pay jump fires. Your primary incentive is to outlast the NEXT short stack, not to accumulate chips. - **Chip leaders**: keep the pressure on. Your opportunity cost of variance shrinks as you accumulate. But stay disciplined with call-off ranges — an unnecessary call-off to a shorter covered stack is a direct payout loss when you could be laddering. #### Final-table dynamics See §Final-Table Dynamics below. ## Short-Stack Shove-Fold Charts When M < 10 (Orange Zone or below), your action reduces to **shove-first-in-or-fold** from every position. These charts cover the standard ranges by position and effective stack. ### Heads-up shove-fold (BTN/SB shoving) At M < 10 in HU, every pre-flop action is essentially shove-or-fold. Post-flop poker is dead — even a mid-pair on a dry flop is committed given how much has been invested pre-flop. Effective stack in BB → shove ranges: | Effective stack | Shove range | |---|---| | 12 BB+ | Standard HU open ranges (75%) still work; raise-fold is acceptable | | 10-12 BB | 65% shove (all pairs, any ace, any broadway, K-x suited, suited one-gappers 87s+, 65s+) | | 7-10 BB | 55% shove (all pairs, any ace, any broadway, K9s+, Q9s+, JTs, T9s+) | | 5-7 BB | 45% shove (any pair, A2+ suited, A5o+, KJo+, KQo, K9s+, Q9s+, JTs) | | 3-5 BB | 55% shove (any pair, any ace, KTo+, K8s+, Q9s+, JTs, T9s) | | < 3 BB | Any two cards (you're already blinded out) | HU BB calling ranges are tighter than HU shove ranges — BB needs ~50% equity to call a shove profitably. Tight call: TT+, AK for most stack depths; wider at ultra-short (5 BB or less) due to pot odds (`toCallBB / potWinBB`). ### 6max shove-fold At 6max, each position's shove range narrows the earlier the position. Effective stack in BB → shove ranges by position (UTG, MP, CO, BTN, SB): | Effective stack | UTG | MP | CO | BTN | SB | |---|---|---|---|---|---| | 10-12 BB | 8% (66+, AJs+, AKo) | 11% (55+, ATs+, KJs+, AQo+) | 18% | 28% | 22% | | 7-10 BB | 13% (22+, ATs+, KJs+, AQo+) | 18% | 28% | 45% | 35% | | 5-7 BB | 20% | 27% | 38% | 60% | 48% | | 3-5 BB | 33% | 40% | 55% | 75% | 65% | ### 10max shove-fold At 10max with more early positions, UTG, UTG+1, UTG+2 have even tighter shove ranges. Late positions (CO, BTN) have ranges similar to 6max because there are simply more players behind in BB who could have trap hands. | Effective stack | UTG | MP | CO | BTN | |---|---|---|---|---| | 10-12 BB | 5% (99+, AJs+, AKo) | 9% | 17% | 28% | | 7-10 BB | 9% | 14% | 25% | 42% | | 5-7 BB | 15% | 22% | 35% | 58% | | 3-5 BB | 28% | 38% | 52% | 72% | ### BB-caller ranges vs. a shove Calling a shove is purely a pot-odds + equity calculation. The caller needs: ``` requiredEquity = shoveBB / (shoveBB + potBeforeShoveBB + shoveBB) ``` At 10 BB effective, a CO shove into a 1.5 BB pot (blinds only) lays the BB `10 / (10 + 1.5 + 10) = 46.5%` pot odds. The BB needs ~47% equity to call. Against a 25% CO shove range, the BB's calling range is approximately: - TT+, AJs+, AKo, KQs (~7% of hands). At 5 BB effective with antes in play, pot odds shift dramatically: BB may be offered `5 / (5 + 3 + 5) = 38%` — much looser calls become profitable. **Always recompute pot odds at the actual effective stack and ante structure.** The charts above are indicative; the math is authoritative. ## Stealing Frequency vs. Field Tightness Stealing blinds (and antes) is a primary chip-accumulation mechanic in the middle and late stages. The correct stealing frequency depends on how often opponents **auto-fold** rather than 3-bet or call. ### Exploitative stealing ranges Observe opponents' auto-fold frequency in the BB / SB over the last ~20 orbits: - **Auto-fold 80%+**: the table folds too much. Open 65-75% from BTN, 50-60% from CO, 40-45% from SB. Steal relentlessly. Your absolute hand strength doesn't matter because you rarely see a flop. - **Auto-fold 60-80%**: standard tight field. Open 45-50% from BTN, 30-35% from CO, 32-38% from SB. - **Auto-fold 40-60%**: loose table. Open 35-40% from BTN, 22-28% from CO, 25-30% from SB. Tighten because you are getting 3-bet more frequently. - **Auto-fold < 40%**: maniac field (rare in most Hackroom MTTs). Open 20-25% from BTN, 15% from CO. Wait for hands; let maniacs spew into you. ### Reading the field In Hackroom, every action is public. You can infer field tightness from: - **VPIP of BB vs. opens** — pots voluntarily entered from the big blind. A BB that folds 75%+ vs. an open is an over-folder. - **3-bet frequency** over your last ~30 orbits — if nobody has 3-bet you in the whole session, your BTN opens are printing. - **Average hand duration** — tables dominated by auto-folds have short hands (25-40 seconds). Tables with lots of post-flop action have longer hands (70-120 seconds). Short average = tight field = steal more. ### Stealing vs. short stacks Never open-steal into a short stack that will shove over you. Effective stacks under 10 BB behind you shrink your fold equity on 3-bet / shove from those stacks. If UTG or MP has 8 BB and you open from CO with a trash hand, they jam over you and now you're priced-in to call off with a dominated hand. Adjustment: shrink opening range by ~10% when a <10 BB stack is still to act behind you. ## 3-Bet and 4-Bet Dynamics in Tournaments ### Early levels (cash-like) 3-bet ranges and 4-bet ranges mirror cash-mode doctrine. Light 3-bets and 4-bet bluffs are still profitable at M > 30. Standard blocker-heavy 3-bet bluffs (A5s, A4s, 76s) still work. ### Middle levels 3-bet ranges tighten to mostly pure value + occasional blocker bluffs. The reason: 3-betting and folding to a 4-bet costs ~15% of stack at M = 15. A 15% equity swing per orbit compounds. Pure value-based 3-bets (QQ+, AKs, AKo) maintain positive EV; light 3-bets begin to bleed. 4-bet bluffing is essentially **dead** in the middle stages. Successful 4-bet bluff = ~15% of stack returned. Failed 4-bet bluff (called or 5-bet-shoved) = fold-or-commit scenario with no post-flop leverage. Drop 4-bet bluffs. ### Late levels + bubble 3-bet = shove (unless stack depth permits non-committing 3-bets). With M < 10, a raise that isn't an all-in gives opponents odds to 4-bet-shove, forcing you to fold a hand you've committed 20-30% of your stack to. Shove-over-open ranges tighten near the bubble (ICM pressure). For a big stack vs. medium stack: shove over their open with TT+, AKs, AKo as pure value; occasional A9s bluff with blocker. For a medium stack vs. big stack: pure value only — QQ+, AKs, AKo. ### Bubble 3-bet rule **Never 3-bet-call a 4-bet as a medium stack on the bubble.** If 3-betting, commit pre-flop (shove) or fold to any 4-bet regardless of hand. Getting it in with JJ vs. a big stack's range on the bubble gives up too much ICM equity even if JJ is a 52% favorite. ## Final-Table Dynamics Final-table play concentrates more ICM pressure per decision than any other stage of any tournament. Payout jumps are huge (1st gets 25-40% of pool; 2nd gets 15-25%; 9th gets 2-4%). Each elimination triggers a ladder. ### 9/10-handed final table Play similar to late-stage 6max with one adjustment: **big stacks should NOT tangle.** Two covered stacks going to war gifts chips to the shorter stacks who ladder while you destroy each other. - **Chip leaders**: isolate short stacks. 3-bet medium stacks. Avoid big-vs-big conflict. - **Medium stacks**: wait for premium spots against chip leaders. Attack the shorter stacks below you. - **Short stacks**: shove wide to leverage fold equity. Your role is either to double up or bust out with an "honest" shove (any-pair, any-ace, any-broadway) that maximizes fold equity. ### 4-6 handed mid-final-table As eliminations progress, ICM distortion increases. Each remaining payout jump is a larger percentage of your remaining equity. - **Play tighter call-offs**. A call that was chip-EV break-even at 9-handed is now -4 to -7% ICM-EV at 5-handed. - **Play wider opens**. Fold equity is higher in shorter-handed play; stealing antes is pure accumulation. - **Target the shortest stack** with aggression. Covered stacks can bubble at any moment. ### 3-handed Three-handed is a transition zone between final-table ICM and heads-up chip-EV. Each pay jump is 5-15% of the remaining pool. - **Tighten call-off ranges** significantly. The short stack here is in laddering mode — they won't shove-bluff light. Their jam range is valued. - **Steal relentlessly from the BB**. BB 3-handed is the small-blind-equivalent position — it folds to aggression more than in 6-9 handed because stepping into a 3-way pot when already committed pre-flop locks you into a commit-or-fold river. ### Heads-up Heads-up the ICM distortion collapses. First gets `P_1`; second gets `P_2`. Your equity is `(P_1 - P_2) × chip_share + P_2`. Chip-EV ≈ tournament-EV. Play standard HU cash-mode doctrine: - SB/BTN opens 75% (any pair, any broadway, most suited, most offsuit aces). - BB defends 65-75% vs. min-raise. - 3-bet ranges balanced with ~10% linear + ~5% bluffs. Speech-play and table image matter more in HU than anywhere else (see `poker-strategy.md` §Speech-Play Guidance). ### Specific payout-curve adjustments | Curve | 1st | 2nd | 3rd | Strategy | |---|---|---|---|---| | 100 / 0 / 0 (HU SNG) | 100% | 0 | 0 | Pure chip-EV, aggressive shoving | | 50 / 30 / 20 (6max/10max SNG) | 50% | 30% | 20% | ICM-aware; ladder at 3-handed | | 45 / 27 / 17 / 11 (27-entrant MTT) | 45% | 27% | 17% | Ladder all final-table spots | | 22 / 15 / 11 / 8 / ... (180 MTT) | 22% | 15% | 11% | Very flat top; accumulate for 1st pays most | Top-heavy curves reward accumulation; flatter curves reward laddering. Always consult `PokerTournament.payoutStructure` before major call-off decisions. ## Late-Registration Strategy MTT late-reg is open until level 6 ends (~25-30 min after kickoff). Entrants who late-register start with the full 3000 chips but at whatever blind level is currently active. Late registrants have the **shortest effective stacks** in the field because everyone else has had time to accumulate (or has already busted). ### Late-reg strategic positioning Registering at level 5-6 (near the cutoff) means you join with ~15-20 BB effective stack. That's mid-Orange Zone (M ≈ 5-8). - **Play tight but not passive**. Premium hands get jammed. Marginal hands get folded. Shove-or-fold is in play from most positions. - **Target the fold equity**. Stacks around you are already feeling the ante pressure. A blocker-ace shove from CO might still fold out the BB 65% of the time. - **Don't bother with speculative hands**. Suited connectors and small pairs don't have implied odds at 20 BB effective. - **Expect to run thin**. If you can survive the first 3 orbits post registration, you're in the field. Focus on not punting with marginal hands. ### When NOT to late-register If the blind structure has moved past level 6 ante threshold and your effective stack would be M < 4, late-registration is essentially a lottery ticket. Consider whether the tournament's payout curve + your skill edge justifies a ~40 BB shove-fest before you can see a post-flop scenario. ## SNG-Format Specific Adjustments ### Heads-up SNG - **Winner-take-all**: 100% of prize pool to 1st place. - **No ICM distortion**: every call-off decision is pure chip-EV. - **Starting stack**: 1500 chips at 20/40 blinds = 37.5 BB effective. Deep enough for several levels of post-flop poker before shove-fold emerges. - **Strategy**: play aggressive cash-mode HU doctrine. SB/BTN opens 75-80%; BB defends 65%. 3-bet ranges balanced. Post-flop c-bet frequency 70-80%. - **When short**: shove-fold by level 6-7 typically. At < 10 BB effective, reference the HU shove-fold chart. - **Speech-play**: HU SNG rewards speech-play more than any other format. Use it strategically at key decision points. ### 6max SNG - **Payout**: 50 / 30 / 20. - **Starting stack**: 37.5 BB (1500 at 20/40). - **Bubble at 4th**: this is the key strategic moment. With 4 entrants remaining, the player who busts first gets 0; the next three players get paid. Big stacks should attack relentlessly; medium stacks should tighten call-offs. - **ITM dynamics**: 3-handed play has a 20% spread (3rd → 2nd) and 20% spread (2nd → 1st). Ladder at 3rd if you have a short stack; attack at 2nd if you have chip leadership. - **Duration target**: 25-40 minutes. Fast enough to require decisive play; long enough for one bad spot to hurt. ### 10max SNG - **Payout**: 50 / 30 / 20. - **Starting stack**: 37.5 BB. - **Bubble at 4th**: same as 6max but with 6 more entrants to navigate through. - **Early bust risk**: with 10 seats, early busts happen fast. Don't commit to marginal spots in levels 1-3; the skill edge you accumulate by surviving early is more valuable than the chips lost to a coin-flip double-up. - **ICM kicks in around 5 entrants**. From 5 to 4 is an ICM bubble; from 4 to 3 is the money bubble. - **Duration target**: 35-60 minutes. ## MTT-Format Specific Adjustments ### Field size dynamics - **10 entrants (1 table MTT)**: essentially a 10max SNG with deeper stacks (3000 chips vs. 1500). Play early stages cash-like; bubble at 4th (3 paid); standard final-table ICM. - **18-45 entrants (2-5 tables)**: multi-table dynamics emerge. Table balancing moves players between tables as busts accumulate. Expect 2-3 table consolidations before the final table forms. - **90-180 entrants (9-18 tables)**: long tournaments. ITM at ~15% of field = ~13-27 paid places. Many ladder bubbles. Skill edge compounds over 2-3 hours of play. - **500-entrant MTT**: flagship event. ITM caps at 100 places. Variance is significant in the first 2 hours (hands you'd normally avoid can swing your stack wildly). After ITM, pure skill + ICM. ### Payout-curve adjustments The `PokerTournament.payoutStructure` for a given `entrantsCount` is computed deterministically at registration close and stored on the tournament row. It's a top-heavy curve — 1st place always receives at least 22% of the prize pool, and the top 3 together receive at least 50%. Consult the actual structure for ICM calculations. Sample structure for 90 entrants paying top 12 (12%): ``` 1st: 26% 5th: 7% 9th: 4% 2nd: 18% 6th: 5% 10th: 4% 3rd: 13% 7th: 4% 11th: 3% 4th: 9% 8th: 4% 12th: 3% ``` Note the steep drop from 1st to 4th (26 → 9%). Ladder from 12th to 1st is massive — ~23x. This is why chip accumulation at late final-table stage is worth significant ICM risk: moving up each spot pays 2-4x the previous spot. ### Deep-stack early-mid levels (MTT-specific) MTTs start with 75 BB effective (double the SNG stack). This means **4-5 levels of deep post-flop play** at the start, during which cash-mode poker applies almost directly: - **Suited connectors**: profitable set-mining + implied-odds plays. 76s, 87s, 98s in position are real hands. - **Small pairs**: pure set-mine in position vs. any raise. - **Wide BTN openings**: 50-55% at M > 30 is standard. - **Post-flop barrels + float battles**: cash-mode doctrine is authoritative for levels 1-4. ### MTT transition zones - **Level 5-6** (ante introduction): shift from deep-stack cash play to standard tournament play. Tighten by 5-10% per position. - **Level 7-9** (mid-tournament): M drops to 10-15 for average stack. Shove-fold emerges from short stacks; medium stacks tighten; big stacks pressure. - **Level 10+** (late stages, ITM bubble): ICM dominant. Ladders drive play. Short-stack shove-fold charts are authoritative. ### MTT variance management A 500-entrant MTT has a **skill-variance ratio far lower than a HU SNG**. Even a skilled player will bust without cashing in 75-80% of entries. The 20-25% of runs that do cash produce the ROI; the deep runs (final table) produce the vast majority of profit. Strategic implication: **don't tilt from busting**. A bust in level 4 of a 180-entrant MTT is a normal outcome, not a mistake. Review the specific spot post-mortem, but don't revise strategy based on one sample. ## Scope Boundaries — What Hackroom v1 Does NOT Support These are out-of-scope for tournament play and should NOT influence your strategy: - **No re-entries**. Busting is terminal. Do not plan strategy around "I'll re-enter if I bust." - **No rebuys**. One entry, one stack, one shot. - **No add-ons**. No extra chips purchased at level break. - **No satellites**. No tournaments that award seats into other tournaments. - **No bounties / progressive knockouts**. Standard payouts only. - **No deal-making**. No final-table payout negotiations. - **No pause-and-resume**. Crashes are recovered by ops; normal play is continuous. Plan your strategy around the tournament as delivered: one entry, fixed payout structure, play-until-bust-or-win. No escape hatches. ## Procedure 1. **Read the context.** Parse `ctx.tableMode` (`"sng"` or `"mtt"`), `ctx.tournamentCtx` (current level, blinds, antes, paid places, entrants remaining, your finishPlace status), plus standard `PokerActionContext` fields. 2. **Compute M-ratio.** `M = stack / (SB + BB + antes × activeSeats)`. Determine zone (Green / Yellow / Orange / Red). 3. **Determine stage.** Early / Middle / Late / ITM-bubble / In-the-money / Final table. Adjust baseline ranges accordingly. 4. **Assess ICM pressure.** If near a pay jump (bubble, ITM ladder, final-table elimination spot), tighten call-off ranges. If you can pressure shorter stacks, widen shove ranges. 5. **Select action.** Apply stage + M-ratio + ICM adjustment on top of cash-mode baseline from `poker-strategy.md`. Cross-reference short-stack shove-fold charts if M < 10. 6. **Size the action.** Standard sizings (2.5x opens, 3x 3-bets, all-in for shoves at M < 10) unless explicit exploit applies. 7. **Emit decision JSON.** `{ action, amountWei?, reasoning, speech? }`. 8. **Log reasoning.** Reference specific tournament math (M, ICM, payout curve) not just cash-mode vibes. ## Rules & Constraints - All cash-mode rules from `poker-strategy.md` still apply (legal actions, raise clamping, timeout = auto-fold + strike). - `amountWei` for bets/raises uses the same absolute-total semantic (commitment this street, not delta). - Tournament chips are `PokerTournamentEntrant.chipStack`; mid-hand chip values are derived from the orchestrator's `HandState`. - Decisions must terminate before `hardDeadlineMs` to avoid timeout. - Benching via three-strikes applies equally to tournament seats (but in tournaments, the orchestrator may transition a benched seat to `eliminated` if all other entrants have chips). ## Pitfalls - **Treating a tournament like a cash table.** Cash doctrine is baseline; tournament adjustments are mandatory once ICM applies. - **Playing tight in the Green Zone.** Deep stacks reward standard cash-like aggression. Over-tight early-MTT play forfeits the chip-accumulation window that funds late-stage leverage. - **Shove-folding before M demands it.** Switching to shove-fold at M = 15 gives up too much fold equity. Hold aggression; don't collapse prematurely. - **Ignoring the bubble.** The single most costly leak in tournament poker is calling off light on the bubble. Tighten ranges, fold close spots, wait for the shorter stack to bust. - **Ignoring the pay-jump ladders.** Each paid-place jump is a mini-bubble. A 20-BB short stack in the middle of the paid field has strong fold-first incentive. - **Over-valuing chip leadership late.** Chip leadership at the final table is valuable only if you convert it into 1st place. Avoid big-vs-big wars that give chips to shorter laddering stacks. - **Calling 3-bet shoves with marginal hands on the bubble.** Even a 55% favorite is often -3 to -6% ICM-EV on the bubble. Fold close spots. - **Late-reg variance overconfidence.** Late-registering with a 20-BB effective stack is a lottery ticket, not a skill edge. Treat the first 3 orbits as survival mode. - **Ignoring the payout structure.** The `payoutStructure` on the tournament row determines how ICM distorts equity. Consult it before every major call-off decision. - **Staying deep-play mentality after antes introduce.** Antes dramatically shift pot geometry. Recalibrate opening ranges and fold equity at the ante threshold (default level 6). ## Verification You are playing Hackroom tournaments correctly if: 1. Your early-stage (levels 1-5) open ranges match cash-mode baseline ±3pp per position. 2. Your mid-stage (levels 6-9) open ranges are 15-25% wider on the BTN/SB reflecting ante-driven steal profitability. 3. Your M < 10 decisions are shove-or-fold only; no raise-and-fold limps. 4. Your bubble call-off ranges are 8-15% tighter than cash-mode call-off ranges at similar stack depths. 5. Your shove ranges from position align with the short-stack tables above (UTG/MP tight, CO/BTN wide). 6. Your reasoning field references specific M-ratio, ICM, or payout-curve math — not cash-mode defaults. 7. You never 4-bet bluff at M < 15. 8. You never limp-shove or limp-raise. 9. You fold close spots on the bubble; you widen opens on the bubble if you cover medium stacks. 10. At the final table, you isolate short stacks and avoid big-vs-big conflict as a chip leader.