Push-Fold Charts Misfire 22% Past 14 Big Blinds
The headline numbers are stark: when tournament stacks drift past 14 big blinds, the standard push-fold charts that most recreational and mid-stakes players rely on are wrong more than one in five times. A new analysis of 40,000 simulated endgame scenarios, run against solver-based equilibrium strategies, found that the charts’ recommended actions misfire 22% of the time at the 15–20 big blind range, with the error rate climbing to 31% at exactly 20 big blinds. The misfires aren’t minor — they swing tournament equity by an average of 1.8% of the prize pool per decision, which over a typical Sunday major is the difference between a min-cash and a final-table run.
Where the Charts Break Down
The core problem isn’t that the charts are lazy. It’s that they’re static. A push-fold chart is a snapshot of a game state that assumes a narrow set of opponent behaviors: that the big blind will call with a prescribed range, that the small blind will fold or shove with predictable frequency, and that no one else at the table is paying attention to your tendencies. Those assumptions hold reasonably well under 10 big blinds, where the threat of being blinded out forces everyone into a binary decision tree. But past 14 big blinds, the geometry of the hand changes. You’re no longer in pure desperation mode. You have fold equity, you have room to min-raise and fold to a shove, and you have opponents who are also not yet desperate.
The 22% figure comes from a dataset of 40,000 simulated hands run through a custom solver built by a quantitative poker coach who asked to remain anonymous because he still plays live tournaments under his real name. The simulation pitted a range of standard push-fold chart outputs — the kind you’d find on any free poker training site or in a popular strategy book — against a GTO+ solver’s equilibrium response. The solver wasn’t playing a fixed opponent. It was playing against the chart’s assumptions, meaning it exploited the chart’s predictable leaks the way a strong human would. The result: at 14 big blinds, the chart’s recommended shove or fold was correct about 88% of the time. At 15 big blinds, that dropped to 84%. By 20 big blinds, it fell to 69%.
The specific leaks are predictable once you see them. The charts over-shove from the small blind with hands like A-5 offsuit and K-9 suited, hands that have decent equity when called but that also have enough playability post-flop that open-raising to 2.2 big blinds is strictly better. Conversely, the charts under-shove from the button with middle pairs and suited connectors in the 14–18 big blind range, where a min-raise forces the blinds to defend with a wider, weaker range that you can out-play post-flop. The charts also fail to account for the ante structure. In a standard tournament with a 12.5% ante (the typical BB ante), the pot at the start of the hand is 2.3 big blinds, not 1.5. That extra 0.8 big blinds of dead money changes the math on marginal shoves significantly — a hand that’s a fold at 1.5 big blinds of dead money becomes a clear shove at 2.3, and vice versa for hands that are only marginally profitable.
The Ante Blind Spot
This is the single largest source of error in the 15–20 big blind range. Most push-fold charts were created in the pre-BB-ante era, or they were updated carelessly by simply adding a constant to the pot size without re-solving the game tree. The difference matters most with hands that have high card removal value but low showdown equity — think A-2 through A-9 offsuit, or K-8 through K-T offsuit. With a 12.5% ante, shoving these from the cutoff with 18 big blinds is profitable if the blinds are folding more than 62% of the time combined. But the chart-based solver showed that most recreational players are folding too much in the blinds in this exact spot, because they’re playing their own chart that says “call a shove with A-10 or better, 88 or better.” That creates a feedback loop of error: the chart tells the shover to fold, the chart tells the caller to fold too much, and both players lose equity to the mathematically optimal line, which is a min-raise with a mixed range that includes both value hands and bluffs.
The simulation found that the biggest single error was with small blind play at 17–19 big blinds. The chart’s recommendation to shove any two cards when the action folds to you in the small blind is correct only when the big blind is folding more than 55% of the time. But against a big blind who is playing a solver-based defense, that fold rate drops to 41%. The result is that shoving 7-2 offsuit or 9-4 suited from the small blind at 18 big blinds is a direct equity donation of about 2.4% of your stack on average. Over the course of a tournament where you’re in the small blind roughly once every six hands, that leak compounds quickly. By the time you’ve played 100 hands at that stack depth, you’ve given up the equivalent of 0.8 big blinds per orbit — which is the entire ante plus the small blind.
The Math Behind the Misfire
Let’s get specific about one hand that illustrates the problem. You’re on the button with 16 big blinds, the blinds are 500/1,000 with a 1,000 BB ante, so the starting pot is 2,500. The small blind has 14 big blinds, the big blind has 18. The chart says shove with A-5 offsuit. The solver says min-raise to 2,000. The difference in expected value is not trivial: the chart’s shove wins the pot outright about 68% of the time, but when called, you’re facing a range of 88+, A-10+, and K-Q, against which A-5 offsuit has only 31% equity. The min-raise, by contrast, keeps the pot smaller when you miss the flop, allows you to fold to a three-bet shove, and lets you apply maximum pressure on flops where you have an ace or a wheel draw.
The solver’s EV for the shove in this spot is +0.52 big blinds. The EV for the min-raise is +1.14 big blinds. That’s a 0.62 big blind difference on a single decision — a 3.9% edge on your stack. Now consider that the chart’s error rate of 22% means roughly one in five decisions at this stack depth is costing you at least half a big blind. If you’re playing a 5,000-entry tournament with a $100 buy-in and the average stack at the money bubble is 40 big blinds, those errors are the difference between surviving to the cash and shipping your stack to a player who min-raised instead of shoved.
The numerical anchor for this entire problem is the 14 big blind threshold. Below that, the charts are actually quite good. The simulation showed a 94% accuracy rate at 10 big blinds and under, and an 88% accuracy rate at 12 big blinds. The drop-off between 12 and 15 is steep, and it’s not linear — it’s a cliff. The reason is that at 14 big blinds, you cross the threshold where a min-raise plus a fold to a three-bet shove costs you only 2.2 big blinds, which leaves you with 11.8 big blinds and 6.5 orbits of playable stack. That’s not a disaster. At 10 big blinds, a min-raise and fold leaves you with 7.8 big blinds and only 4 orbits — which is close to dead. So the chart’s binary “shove or fold” logic is calibrated for the desperation zone, and it simply doesn’t adapt to the new reality of having a workable stack.
The Human Factor
There’s also a psychological component that the charts ignore, and it’s one that live tournament players in the United States feel acutely. When you’re at 16 big blinds in a live event with a $1,500 buy-in, the blinds are going up in 40 minutes, and you’ve been card-dead for two hours, the chart says shove with K-9 suited from the hijack. The solver says min-raise. The difference in EV is about 0.4 big blinds, but the difference in emotional impact is enormous. If you shove and get called by A-10, you’re out of the tournament. If you min-raise and get three-bet, you can fold and still have 13 big blinds — you’re not happy, but you’re alive. The chart doesn’t account for the fact that preserving a playable stack has tournament-life value beyond the raw chip EV. That’s not something a solver can quantify, but it’s something every experienced tournament player intuitively knows.
The simulation tried to account for this by running a secondary model that included an “ICM pressure” variable, which weights the cost of busting higher than the raw EV gain of a double-up. Under that model, the error rate of the charts got worse, not better. At 16 big blinds, the charts misfired 27% of the time when busting was assigned a penalty of 1.5 times the raw chip loss. That’s because the charts recommend shoves that are marginally +EV in raw chips but that expose you to elimination when a min-raise accomplishes the same goal with less risk. In a live tournament where the average field is softer than an online field, the min-raise is even more powerful because your post-flop edge is larger.
What the New Data Actually Recommends
The coach who ran the simulation distilled the findings into a set of heuristics that are simple enough to remember at the table without a chart in front of you. The first is a hard rule: below 14 big blinds, keep using the charts — they’re accurate enough that the error rate doesn’t justify the cognitive load of improvising. At 14 big blinds and above, switch to a “raise-first” strategy from the button and cutoff. The solver’s equilibrium in that range is to open-raise to 2.2 big blinds with about 70% of your range, shove with the top 8% of that range when facing a three-bet, and fold the rest. The remaining 30% of your range — mostly weak aces and low suited connectors — is a mix of shoves and folds depending on the exact stack sizes of the blinds.
The second heuristic is about the small blind specifically. The charts’ biggest leak is over-shoving from the small blind with hands that have reverse implied odds. The solver’s recommendation at 15–20 big blinds is to min-raise or limp-shove with a polarized range: shove with premium pairs and big aces, min-raise with everything else, and only shove with middle pairs when the big blind has less than 12 big blinds. This is counterintuitive to anyone who learned poker from a chart, because the chart says “shove to deny equity.” But the solver says that at 16 big blinds, the big blind’s defense range is wide enough that you’re better off seeing a flop and using your positional advantage to out-play them, rather than flipping for your tournament life with 88 against A-K.
The third heuristic is about reading your opponent’s chart usage. If you’re at a table where the players in the blinds are snap-calling shoves with A-10 and 99, you can exploit them by simply not shoving — min-raise more, and fold to their three-bets when you have a marginal hand. Conversely, if the blinds are folding too much (which the simulation found is the default for players who memorize charts but don’t adjust), you can shove wider than the chart says, because your fold equity is higher than the chart’s model assumes. The 22% error rate is an average — against a player who is glued to their chart, the error rate on your shoves when you deviate from the chart’s recommendation is closer to 40% in your favor.
The Software Gap
There’s an uncomfortable truth in the poker training industry that this data exposes: the free charts that most players use are outdated by at least five years. The solver revolution in poker happened between 2018 and 2022, and the major commercial solvers (PioSOLVER, GTO+, MonkerSolver) all updated their preflop solutions to account for BB ante structures and mixed strategies. But the charts that are shared on Reddit, in Facebook groups, and on free training sites are still based on older models that assumed no ante or a small ante, and that treated every shove as an all-or-nothing proposition. The result is that the average tournament player is studying a strategy that was already obsolete when they found it.
The coach’s simulation found that even the most popular paid chart sets — the ones sold by well-known tournament pros for $50 to $150 — have the same 22% error rate in the 15–20 big blind range. The issue isn’t the quality of the charts’ authors; it’s the format itself. A static chart cannot represent a mixed strategy, and the solver’s equilibrium at these stack depths is heavily mixed. At 17 big blinds from the cutoff with A-5 offsuit, the solver shoves 34% of the time, min-raises 41% of the time, and folds 25% of the time. A chart that says “shove” is correct only 34% of the time in that exact spot. A chart that says “min-raise” is correct 41% of the time. Neither is “wrong” in a binary sense, but both are wrong most of the time if you treat them as deterministic rules.
The 22% misfire rate isn’t a bug in the charts — it’s a feature of trying to compress a multi-dimensional game tree into a two-dimensional grid. The charts are a heuristic, and heuristics break down at the edges of their applicability. The 14 big blind threshold is the edge, and the data shows that players who rely on charts past that point are leaving meaningful equity on the table every single hand they play.
The Open Question
The deeper question this data raises isn’t about whether you should memorize a new chart — it’s about whether the entire concept of push-fold strategy is obsolete in the modern tournament game. If the solver’s equilibrium at 15–20 big blinds is a mixed strategy that requires you to shove sometimes, min-raise other times, and fold a meaningful percentage of the time with the same hand, then no static chart can ever be accurate. The only way to play correctly at these stack depths is to either internalize the solver’s ranges through thousands of hours of study, or to develop a feel for the specific dynamics of your table and adjust on the fly.
For the recreational player who doesn’t have time to run 40,000 simulations, the practical takeaway is simpler: when your stack is between 14 and 20 big blinds, don’t trust the chart. Trust the situation. If the blinds are tight and folding too much, shove wider. If they’re calling stations, min-raise and play poker. If you’re card-dead and the blinds are about to go up, the chart’s shove with K-9 might be your only move — but understand that you’re making a desperation play, not a +EV one. The chart was never designed to save you from a bad situation; it was designed to help you avoid making it worse.
The 22% number isn’t a call to throw away your charts. It’s a call to know when they’re lying to you. And the uncomfortable answer is that they start lying exactly when the tournament gets interesting — when the bubble is approaching, when the stacks are shallow enough to matter, and when every decision has real money riding on it. The charts got you to the final table, but they won’t win you the final table. The question is whether you’re willing to trust your own judgment when the math says the chart is wrong — and whether you can tell the difference between the two.