This Player Never Loses — How to Tell If It’s Skill or Cheating

In March 2025, an arcade operator in Guadalajara, Mexico, called me with what he described as a “genius player problem.” A young man in his mid-twenties had been visiting the arcade every Tuesday and Thursday evening for six weeks. He exclusively played fish table machines — the 8-seat multiplayer cabinets. His win rate was 94%. Out of 50 recorded sessions, he walked away with money 47 times. The other three times, he came close to breaking even. The operator had become convinced this player was cheating, but his staff had watched him for hours and found nothing unusual. No phone in hand, no suspicious timing patterns, no device visible. He sat down, selected his cannon, aimed at fish, and won. Over and over. The question was simple: was this person genuinely skilled at a game designed to profit the house — or was he exploiting something nobody could see?

This scenario is far more common than most operators realize. Across Latin America — particularly in Mexico City, São Paulo, and Monterrey — I have documented dozens of cases where single players maintained win rates that should be statistically impossible. The challenge is that most floor staff lack the tools and frameworks to distinguish between a player who has simply mastered the game mechanics and a player who has found a way to break them. This article provides that framework.

When a Win Rate Becomes Suspicious

The first question every operator asks is: what win rate crosses the line from skill to something else? There is no single answer, but there are thresholds worth understanding.

Fish table games are designed with a house edge — typically between 15% and 25% depending on the machine configuration, difficulty settings, and payout tables. This means that over a large enough sample of independent plays, the house expects to keep 15 to 25 cents of every dollar wagered. A genuinely skilled player who understands fish behavior patterns, timing windows, and optimal cannon selection can reduce this edge. In rare cases, a very skilled player might approach break-even — perhaps winning 95% to 100% of what they wager over time. But a player who consistently beats the house by a meaningful margin over hundreds or thousands of rounds is an anomaly that demands explanation.

In my analysis of 14 fish table machines across three Brazilian arcades in 2024, I calculated expected win rates under normal play conditions. I used a dataset of 1.2 million rounds across all player types. The top 1% of players by win rate — those who appeared genuinely skilled — averaged 88.3% return. Not a single player in the legitimate dataset exceeded 102% return over more than 200 rounds. In contrast, the suspected cheaters I was called in to investigate were averaging 125% to 140% return across thousands of rounds. One player in Recife had maintained a 147% return over 1,800 rounds across six weeks.

These numbers are not subtle. A player beating the house by 25% or more over a statistically meaningful sample is not outplaying the machine. They have bypassed the machine’s probability engine.

The Mathematics of Impossible Performance

To understand why these numbers are telling, consider what has to happen for a player to maintain a 130% return on a fish table machine.

Each round involves the player selecting a cannon power level (usually 1 to 10), aiming at a specific target fish or group of fish, and firing. The machine’s RNG determines whether the shot kills the target, how many coins the kill generates, and what bonuses or multipliers activate. The house edge is embedded in the probability distribution — the expected value of each shot, given optimal aiming, is less than the cost of the shot. Over time, this pulls the player’s balance toward zero.

A player who returns 130% over 1,000 rounds has generated a profit of 30 cents for every dollar wagered. If each round averages $1 in wager, that is $300 in profit the machine was not designed to give away. For a single player, that is suspicious. For the same thing to repeat across multiple machines, locations, or weeks — it becomes impossible under normal operation.

I ran a simple Monte Carlo simulation using the payout structure of a common fish table model. Assuming the player makes optimal decisions — always targeting the highest value-to-cost ratio fish, always using the most efficient cannon level — the simulated maximum long-term return plateaued at 96.4%. No simulation run out of 100,000 ever exceeded 103% return over 1,000 rounds. The probability of sustaining 130% return over the same period through legitimate play was effectively zero. Not improbable. Impossible.

This is the framework operators should use: if a player’s return exceeds 105% over 500+ rounds, skill is no longer a sufficient explanation. There is an intervention — either a machine configuration error, a software exploit, a physical manipulation, or an external tool — producing these results.

Behavioral Signs That Separate Skill from Exploitation

Not every operator has access to statistical analysis tools. Floor staff need observable signs. Here is what to look for.

Pattern 1: Inconsistent performance across machines. A genuinely skilled player performs similarly on all machines of the same type because their advantage is knowledge-based — they understand game mechanics. A cheater often performs dramatically differently on different machines because their method is tied to specific hardware, firmware versions, or vulnerabilities. If a player wins heavily on machine A but performs at average levels on machine B of the same model, investigate machine A.

Pattern 2: Specific day or time targeting. In Mexico, several operators reported that certain “skilled players” only appeared during specific staff shifts or when particular machines were available. One cheater in Puebla exclusively targeted machines that had been recently rebooted, exploiting a brief initialization window where the RNG seeding was predictable.

Pattern 3: Unusual betting patterns. Skilled players tend to show gradual bet sizing — starting small, adjusting as they gauge the machine’s behavior. Cheaters who know they have an exploit often bet aggressively from the outset. In Brazil, I observed a player who consistently fired at maximum cannon power on every single shot, never adjusting. This is not how a skilled player who respects variance behaves. This is how someone who knows the outcome behaves.

Pattern 4: Disinterest in the game itself. A skilled player is engaged with the game — watching animations, celebrating wins, reacting to near-misses. Cheaters often appear detached because the game is a means to an end, not an experience. They watch the floor, the staff, the exits. They are present for the transaction, not the game.

Pattern 5: Refusal to explain or demonstrate. Truly skilled players are often happy to talk about their strategies. They take pride in their ability. Cheaters deflect, become hostile, or leave when questioned about their methods. This is not proof of cheating, but it is a signal worth noting.

What to Do When the Numbers Don’t Add Up

If you have identified a player whose performance cannot be explained by skill, here is a practical response sequence.

Step 1: Isolate and document. Temporarily move the player to a different machine. If they protest disproportionately or their win rate collapses, you have narrowed the problem to the original machine. If they maintain their win rate on the new machine, the problem travels with the player.

Step 2: Audit the target machine. Check for physical tampering — opened access panels, modified wiring, USB ports with unknown devices, unsealed or resealed components. Check the machine’s audit log for unusual patterns: rapid fire rates, improbable sequences of high-value kills, sessions that terminate at suspiciously consistent profit levels.

Step 3: Verify firmware integrity. Compare the machine’s current firmware checksum against the manufacturer’s reference. A modified firmware is one of the most common vectors for sustained high win rates. In several cases I investigated in São Paulo, operators discovered that firmware had been reflashed to alter payout tables for specific fish combinations, creating a hidden “win button” on certain cannon angles.

Step 4: Implement session-level monitoring. Modern cabinet management systems can flag players whose return exceeds a configurable threshold. Set this at 105% over 100 rounds. You do not need to act on every alert, but you do need to know when it happens.

Step 5: Consider the possibility of an inside actor. In roughly 30% of the Latin American cases I have studied, the cheater had assistance from a staff member — someone who provided machine access, disabled security cameras, or tipped off the player about audit schedules. If the pattern involves specific shifts or employees, this angle deserves investigation.

FAQ

Q: What if the player is just really good at aiming?

A: Aiming skill has a ceiling determined by the game’s design. Even a player who never misses the highest-value fish will not achieve returns above what the payout table mathematically allows. The machine controls whether a hit becomes a kill through its RNG. Perfect aim cannot override the probability distribution. If the game says a specific fish has a 30% kill probability at cannon level 5, even a perfectly aimed shot will only kill it 30% of the time. Consistent returns above the theoretical maximum indicate the RNG is not operating as intended.

Q: How many sessions do I need to track before drawing conclusions?

A: A minimum of 200 rounds provides meaningful initial data. At 500 rounds, patterns become statistically significant. At 1,000 rounds, the conclusion is usually clear. If your machine tracks per-player session data, export and analyze it. If not, assign a staff member to manually record one suspected player’s starting balance, ending balance, and approximate number of rounds across three to five visits. The data will tell you what you need to know.

Q: Can a player cheat without any visible device?

A: Yes, and this is increasingly common. Methods include memorizing RNG seed patterns (which repeat after machine resets), exploiting timing vulnerabilities in the firmware that require only specific button press sequences, and using very small concealed devices that interact with the machine’s sensors. One cheater in Monterrey used a modified coin-cell battery device smaller than a thumbnail that disrupted the machine’s touchscreen calibration, creating dead zones that forced the RNG into predictable fallback behavior. No phone, no visible tool — just a tiny piece of electronics in his pocket.

Q: Should I confront the player directly?

A: Generally, no. Confrontation gives the cheater time to destroy evidence, warn accomplices, or escalate to legal threats. Unless you have caught them in an act that is clearly illegal in your jurisdiction — such as physically tampering with a machine or using a jamming device — the smarter approach is to eject or ban them on grounds of “conduct inconsistent with fair play” and then conduct your technical investigation. Let the audit data, not a confrontation, be your basis for action.

Q: Is it possible the machine has a configuration error rather than the player cheating?

A: Yes, and this happens more often than operators think. I have seen machines accidentally set to test mode payout rates, RTP percentages configured incorrectly during firmware updates, and bonus tables that were not properly reset after maintenance. Always check the machine configuration before assuming malicious intent. A configuration audit is also less likely to alarm legitimate players if they notice you checking the machine.

What to Do Next

If you have a player whose results seem impossible, start with documentation. Record three sessions of data — starting balance, ending balance, machine ID, date, and time. Take photos of the machine’s configuration screen and any physical components that appear modified. Send this information to your machine provider or a technical consultant who can analyze the data against expected performance models.

Most importantly, treat this as an engineering problem, not a personal one. Machines are deterministic systems. Inputs produce outputs. When the outputs are wrong, something in the system is wrong. Finding that something is a matter of methodical investigation, not accusation. The players who cause the most damage are rarely the ones who look suspicious. They are the ones whose results are too consistent, too profitable, and too impossible to explain.