How the AI HUD Reads Your Opponents in Real Time

August 14, 2026

Every time you sit down at a WePoker table, you are surrounded by opponents whose playing styles range from wildly aggressive to painfully passive, from disciplined regulars to complete beginners splashing chips. Your ability to identify these tendencies quickly and adjust your strategy accordingly is one of the biggest determinants of your long-term win rate. This is exactly what the WePoker AI HUD is designed to do — except it does it with a depth and speed that no human player could match.

The AI HUD overlay tracks 255 or more parameters per opponent, updating in real time as each hand plays out. It processes this data through statistical models, identifies patterns and tendencies, and feeds those findings directly into the recommendation engine. The result is an AI assistant that does not just tell you the theoretically correct play — it tells you the most profitable play against the specific opponents sitting at your table right now.

This guide takes you through every layer of the system: the core statistics and what they mean, the advanced metrics that separate casual HUD users from serious grinders, the contextual filters that capture nuanced behaviour, and practical advice for reading and applying the HUD data in your WePoker sessions.

The foundation: core preflop statistics

Before diving into the deep end of 255 parameters, it is essential to understand the handful of statistics that form the foundation of every opponent profile. These core metrics give you the broadest and most immediately actionable read on any player.

VPIP — Voluntarily Put Money in Pot

VPIP measures the percentage of hands in which a player voluntarily puts money into the pot preflop — that is, any action other than posting a blind and folding. It is the single most important statistic for categorising an opponent. A VPIP of 15 or below indicates an extremely tight player who enters pots only with premium hands. A VPIP of 20 to 28 is typical for a solid regular player. A VPIP of 30 to 40 suggests a loose player, and anything above 40 signals a recreational player who plays far too many hands.

On the WePoker platform, the average VPIP tends to be higher than on established online poker sites, especially in the lower-stakes clubs. This is because many WePoker players are recreational and play for entertainment rather than profit. The population baseline in the AI engine accounts for this, but individual reads always override the baseline once enough data accumulates.

VPIP converges relatively quickly — after 30 to 50 hands, you have a reasonably accurate estimate. The AI engine begins incorporating VPIP-based adjustments almost immediately: against a very high VPIP player, it recommends wider value ranges and fewer bluffs; against a very low VPIP player, it tightens your calling range and increases respect for their raises.

PFR — Pre-Flop Raise percentage

PFR measures how often a player raises preflop. The gap between VPIP and PFR tells you about a player's passivity. A player with a VPIP of 30 and a PFR of 8 is a loose-passive player — they enter many pots but rarely take the initiative. This is one of the most exploitable profiles in poker: they call too much and rarely build pots with their strong hands, making them predictable and easy to play against.

Conversely, a player with a VPIP of 22 and a PFR of 20 is tight-aggressive — they play a narrow range but raise with most of it. This is closer to an optimal strategy and harder to exploit, so the AI engine stays closer to GTO against such players.

The VPIP-PFR gap is one of the first things the AI engine evaluates when building an opponent model. A large gap (more than 10 percentage points) is a strong signal of a passive, exploitable player. A small gap (2 to 5 points) indicates an aggressive player whose range is largely defined by their raises.

3-Bet percentage

The 3-bet statistic measures how often a player re-raises a preflop open raise. The typical 3-bet percentage for a solid player is around 6 to 9 percent. Below 4 percent signals that a player 3-bets only with premium hands (typically aces, kings and sometimes queens and ace-king). Above 10 percent suggests a player who uses the 3-bet as an aggressive weapon with a wide range including bluffs.

This stat is critical for the AI engine's preflop strategy. Against a low 3-bet percentage, the engine narrows the range it gives credit to when that player does 3-bet — if they only 3-bet 3 percent of hands, you can confidently put them on a very narrow, strong range and fold marginal hands. Against a high 3-bet percentage, the engine recommends wider 4-bet ranges and more frequent calls, because the opponent's 3-betting range includes many hands that are weak against continued aggression.

Fold-to-3-Bet

This metric captures how often a player folds when facing a 3-bet after they have opened. It is the direct complement to the 3-bet percentage and equally actionable. A fold-to-3-bet of 70 percent or higher indicates a player who gives up too easily — the AI engine exploits this by widening your 3-betting range against them, turning marginal hands into profitable preflop bluffs. A fold-to-3-bet below 50 percent signals a player who defends or 4-bets often, so the engine tightens your 3-betting range to include more value hands and fewer bluffs.

Postflop statistics: where edges multiply

Preflop stats tell you what hands an opponent is willing to play. Postflop stats tell you how they play those hands once community cards are dealt. This is where the depth of the 255-parameter model becomes truly powerful, because postflop play is where the largest and most exploitable errors occur.

C-Bet frequency (flop, turn, river)

The continuation bet (c-bet) is one of the most common postflop actions: the preflop aggressor bets again on the flop. The c-bet frequency measures how often a player makes this play, broken down by street. A typical solid player c-bets the flop around 55 to 65 percent of the time, the turn around 40 to 55 percent and the river around 30 to 45 percent.

The AI engine tracks c-bet frequency separately for each street and in different pot types (single-raised pots, 3-bet pots, multi-way pots). A player who c-bets 85 percent on the flop but only 30 percent on the turn is someone who takes one stab and gives up — the engine exploits this by recommending floats on the flop (calling with the intention of taking the pot on the turn when the opponent checks).

Fold-to-C-Bet

The opposite side of the c-bet coin: how often does a player fold when facing a continuation bet? This stat is tracked per street and is one of the most directly actionable metrics in the entire HUD. A fold-to-flop-cbet above 65 percent means the player gives up most of their range when you bet the flop — the engine recommends c-betting almost your entire range against them. A fold-to-flop-cbet below 40 percent means the player defends wide, so the engine reduces your bluffing frequency and sizes up your value bets.

Aggression factor and aggression frequency

The aggression factor (AF) is the ratio of bets and raises to calls. An AF of 2 or higher indicates an aggressive player who bets and raises significantly more than they call. An AF below 1 indicates a passive player who calls more than they bet. Aggression frequency (AFq) converts this into a percentage of how often a player takes an aggressive action when given the opportunity.

These metrics are critical for understanding how to respond to an opponent's actions. A check from an aggressive player (high AF) is more meaningful — it signals genuine weakness because this player would usually bet with anything decent. A check from a passive player (low AF) carries less information because they check routinely even with medium-strength hands.

Check-raise frequency

Check-raising is one of the most aggressive and revealing actions in poker. The check-raise frequency tells you how often a player check-raises on each street. The AI engine tracks this by street, by pot type and by board texture. A high check-raise frequency (above 12 percent on the flop) indicates a player who uses the check-raise as a weapon, potentially with bluffs. A very low check-raise frequency (below 4 percent) suggests a player who check-raises only with extremely strong hands — when they do check-raise, you should believe them.

Donk-bet frequency

A donk-bet is a bet made out of position by a player who did not take the lead on the previous street — for example, the big blind betting into the preflop raiser on the flop. Donk-betting is rare in strong player populations but common among recreational players on WePoker. The AI engine tracks donk-bet frequency and uses it to identify players who bet with weak or marginal hands out of position, typically as a defensive measure rather than for value.

Advanced metrics: the 255-parameter deep dive

Beyond the core and standard postflop stats, the WePoker AI engine tracks dozens of advanced and contextual metrics. These are the parameters that give the engine its precision edge — the ability to make street-specific, position-specific and board-texture-specific adjustments that no human player could calculate in real time.

Positional breakdown

Every core statistic is tracked separately for each position: under the gun, middle position, hijack, cutoff, button, small blind and big blind. This matters because many players have dramatically different tendencies by position. A player might be tight from early position (VPIP 12) but loose from the button (VPIP 40). The positional breakdown prevents the engine from over- or under-exploiting based on an aggregated average that masks position-specific behaviour.

Stack-depth adjustments

Player behaviour often changes with stack depth. Some players become more conservative as their stack grows (protecting profits), while others become more aggressive with deep stacks (leveraging fold equity). The engine segments statistics by effective stack depth — typically in bands like 20 to 40 big blinds, 40 to 70, 70 to 100 and 100-plus — to capture these tendencies.

Board texture interactions

How a player responds to different board textures is one of the most nuanced aspects of their game. Some players c-bet heavily on dry boards (boards with no draws or connectivity) but check on wet boards (boards with many drawing possibilities). Others do the opposite. The engine tracks c-bet, check-raise and fold frequencies separately for board texture categories: dry, semi-wet, wet, monotone, paired and coordinated. This allows it to predict with remarkable accuracy how an opponent will respond to any given flop.

Multi-way versus heads-up behaviour

Many players dramatically change their strategy in multi-way pots compared to heads-up pots. The presence of additional opponents typically makes players tighter and more cautious, but some recreational players on WePoker maintain their loose-aggressive approach regardless of how many opponents are in the pot. The engine tracks separate statistics for multi-way and heads-up scenarios to make accurate recommendations in both contexts.

Bet sizing tells

One of the most subtle and powerful data points the engine tracks is bet sizing patterns. Some players consistently use different bet sizes for value and bluff — for example, betting 75 percent of the pot with strong hands and 33 percent with bluffs. The engine captures these sizing tendencies and factors them into its hand-reading model. When it detects a consistent sizing tell, the exploit engine adjusts its response: calling small bets (likely bluffs) and folding to large bets (likely value) — or vice versa, depending on the pattern.

Timing tells

While the WePoker AI assistant primarily tracks action-based statistics, the platform also allows for timing analysis. Snap-calls and instant bets often indicate pre-planned actions (strong hands in the case of snap-calls, draws or weak hands in the case of snap-raises). Tanking often indicates a genuine decision point. The engine notes timing patterns as an additional data layer for its opponent model.

How the engine synthesises 255 parameters into a single recommendation

Tracking 255 statistics per opponent is impressive, but the real value lies in how the engine integrates all of this data into a single, actionable recommendation on every decision point. This synthesis happens in several stages.

Stage 1: hand range construction

Based on the opponent's preflop actions and tendencies, the engine constructs a probability-weighted range of hands they could hold. If a player with a VPIP of 35 and a PFR of 12 calls a raise from the cutoff, the engine assigns probabilities to different hand combinations based on that player's historical patterns: how often they flat-call versus 3-bet with strong hands, how often they call with suited connectors versus suited broadway, etc.

Stage 2: range narrowing by street

As the hand progresses through each street, the opponent's actions narrow their range. A flop check from an opponent who c-bets 70 percent of the time eliminates most of their strong holdings. A turn bet from an opponent with low turn aggression suggests a narrower, stronger range. The engine updates the probability distribution after every action, using the 255 parameters as the statistical backbone of these updates.

Stage 3: EV calculation

With a probability-weighted range for the opponent and the current board state, the engine calculates the expected value of each available action: fold, check, call, bet (at various sizes) and raise (at various sizes). The action with the highest expected value is the recommendation. The engine displays the top recommendation along with the EV differential — how much better the recommended action is compared to the next-best alternative.

Stage 4: confidence weighting

The final stage applies confidence weighting based on sample sizes. If the engine's recommendation is driven primarily by a stat with a large, reliable sample (like VPIP over 200 hands), the recommendation carries high confidence. If it depends on a stat with a small sample (like river check-raise frequency over 10 opportunities), the engine tempers its recommendation toward the GTO baseline. This prevents over-exploitation based on unreliable data.

Reading the HUD overlay: a practical walkthrough

The WePoker AI HUD displays a condensed version of the most actionable statistics directly on the table overlay. Understanding how to read this display at a glance is essential for fast decision-making.

Primary display

The default HUD view shows three lines of data per opponent. The first line displays VPIP and PFR as two numbers side by side — this gives you the instant loose-tight and passive-aggressive classification. The second line shows the 3-bet percentage and the fold-to-3-bet — critical for preflop strategy. The third line shows the total number of hands tracked, which tells you how much confidence to place in the displayed numbers.

Expanded display

Tapping on an opponent's HUD display expands it to show a detailed breakdown. This includes postflop c-bet frequencies by street, fold-to-c-bet by street, aggression factor, check-raise frequency, and any detected sizing tells. The expanded display also shows the opponent's hand history — recent hands that illustrate their tendencies in context.

Colour coding

The HUD uses colour coding to highlight extreme values. Stats that deviate significantly from optimal ranges are highlighted in green (indicating an exploitable weakness you can profit from) or red (indicating a dangerous tendency you should respect). For example, a fold-to-c-bet of 72 percent would be highlighted in green because it represents a clear fold-too-much leak. A 3-bet percentage of 2 percent would also be highlighted, because it means their 3-bets are always premium hands.

Practical scenarios: using HUD data to make better decisions

To illustrate how all of this comes together, consider several common WePoker scenarios where HUD data directly shapes the AI's recommendation and your decision.

Scenario: value betting the river against a calling station

You hold top pair with a good kicker on a K-8-3-6-2 board. Your opponent has a VPIP of 52, a fold-to-river-bet of 28 percent and a calling frequency on the river that is nearly double the population average. The AI engine recommends a large value bet — 80 percent of the pot — rather than the smaller, pot-controlling bet that GTO might suggest. The logic: this opponent calls too much on the river, so you want to extract the maximum from their wide calling range. A smaller bet still gets called, but you leave money on the table. A larger bet gets called almost as often due to their tendency, and each call earns you significantly more.

Scenario: thin value versus a nit

You hold second pair on the river against a player with a VPIP of 14, a PFR of 12 and a fold-to-river-bet of 58 percent. The AI engine recommends checking back rather than value betting. Against a tighter player, your second pair is often behind when called — the player's river calling range consists mostly of top pair and better. The HUD data confirms that this player folds marginal holdings on the river, so a bet would only get called by hands that beat you. The correct play is to show down and collect whatever equity your hand has.

Scenario: floating the flop against an auto-c-bettor

The preflop raiser bets the flop, and you are in position with a gutshot and overcards. The opponent's HUD shows a flop c-bet of 82 percent and a turn c-bet of only 35 percent. The AI engine recommends a call (float) on the flop with the intention of taking the pot on the turn when the opponent checks — which they will do roughly 65 percent of the time according to their stats. Your hand has some equity against their range, and the fold equity on the turn makes the float profitable. Without the HUD data, this play would be a guess; with it, it is a calculated, positive-EV decision.

Building your own reads alongside the AI

While the AI HUD handles the quantitative analysis, developing your own qualitative reads is valuable. Pay attention to timing tells, chat patterns and atypical actions that the statistical model may not capture. Does an opponent tank before bluffing? Do they chat more when they have a strong hand? Do they change their strategy after losing a big pot? These human-level observations complement the AI's statistical engine and can give you an edge in specific hands that pure numbers miss.

The best WePoker players use the AI HUD as a foundation and layer their own observations on top. The statistical model tells you what an opponent does on average; your own observations tell you what they might do right now, in this specific moment, given the context of the session.

Privacy and ethics considerations

Using HUD overlays raises legitimate questions about fairness and platform compliance. The WePoker AI assistant runs locally on your Android device and does not share data with other players. However, using real-time assistance tools generally violates the terms of service of WePoker and similar platforms. You should be aware of this and accept full responsibility for your choice to use such tools. Additionally, be aware that platform operators may take action against accounts that are detected using third-party assistance, including account suspension or chip confiscation.