Combination Prediction Systems: Complete Guide to Trixies, Yankees and Multi-Selection Forecasting

Introduction

System predictions — analytical frameworks built around multiple football match selections combined into structured combinations — represent one of the most sophisticated approaches to football forecasting. The Trixie, Yankee, Heinz, and related system structures allow analysts to express confidence across a portfolio of predictions simultaneously, providing a level of coverage and risk management that single-match analysis cannot offer. Understanding how these systems work mathematically, how to select the most appropriate system structure for a given set of predictions, and how to choose the individual selections that maximise the analytical value of the system is the subject of this comprehensive guide.

System predictions fundamentally change the way analysts think about multiple football match forecasts. Rather than treating each prediction as a standalone event, system frameworks treat multiple predictions as an integrated portfolio where the relationships between selections — their independence, their correlations, and their relative confidence levels — determine the optimal combination structure. A well-constructed Trixie built around three carefully analysed, genuinely high-probability predictions is analytically very different from an accumulator containing the same three selections: the system's combination structure protects against single-selection failure while still delivering returns when two or more predictions prove correct, whereas the accumulator requires all three to succeed. This structural difference has profound implications for how analysts should approach the selection process and the allocation of predictive confidence across multiple events.

Understanding the System Structures: Trixie, Yankee, and Beyond

Before evaluating the analytical merits of different system structures, it is essential to understand precisely what each system contains. These structures are defined by their mathematical combination properties, which determine how many individual combinations are included and what patterns of success produce positive returns.

The Trixie: Three Selections, Four Bets

The Trixie is built around three selections and contains four combinations: three two-fold doubles (Selection 1 + Selection 2, Selection 1 + Selection 3, Selection 2 + Selection 3) and one three-fold treble (Selection 1 + Selection 2 + Selection 3). To produce any positive return from a Trixie, at least two of the three selections must be correct. If all three are correct, the full return combines the three doubles and the treble. The Trixie's value as an analytical framework lies in its tolerance for one failure: an analyst who identifies three strong predictions can be wrong about one and still achieve a meaningful return from the remaining two doubles.

The Patent is the Trixie's companion system, adding three singles to the four Trixie combinations to create a seven-combination structure covering all possible combinations of three selections. The Patent provides a return even if only one selection is correct (the single), making it more forgiving than the Trixie but requiring a larger total cost. For analysts who are highly confident in all three selections but want maximum coverage, the Patent represents the most comprehensive three-selection system structure.

The Yankee: Four Selections, Eleven Bets

The Yankee is built around four selections and contains eleven combinations: six two-fold doubles, four three-fold trebles, and one four-fold accumulator. To produce any return from a Yankee, at least two of the four selections must be correct (producing a minimum of one double). The Yankee is an enormously popular system structure because four selections is a natural number for a weekly football prediction portfolio — an analyst who has identified four confident predictions across a weekend of fixtures can combine them in the Yankee structure to provide coverage for all patterns of success and failure while a full four-fold accumulator would require all four to succeed.

Heinz and Canadian: Larger System Structures

The Lucky 15 extends the Yankee by adding four singles, creating a fifteen-combination system that returns from as little as one correct selection. It represents the most comprehensive four-selection system. The Heinz covers six selections with 57 combinations (15 doubles, 20 trebles, 15 four-folds, 6 five-folds, 1 six-fold accumulator), providing two-selection minimum return coverage across six simultaneous predictions. The Super Heinz covers seven selections with 120 combinations, and the Goliath covers eight selections with 247 combinations — the two most expansive standard system structures, appropriate for analysts who want to build a comprehensive prediction portfolio across a full weekend of football fixtures.

The Mathematics of System Coverage: Why Systems Outperform Accumulators in Specific Contexts

The fundamental mathematical distinction between a system and a straight accumulator is how they handle partial success. An accumulator requires every selection to win — a single failure means the entire combination fails. A system produces multiple sub-combinations, each of which can independently succeed or fail, creating a much more robust return profile when prediction accuracy is imperfect.

How Systems Distribute Risk Across Outcomes

Consider an analyst who identifies four predictions, each with an estimated win probability of 65%. The probability of all four succeeding (required for an accumulator return) is 0.65⁴ = 0.179, or approximately 18%. The expected return from the accumulator depends on the combined return when it does succeed versus the loss when it fails. Now consider the same four predictions in a Yankee system. The probability of exactly three succeeding (producing three trebles, three doubles, and some doubles from the two remaining doubles) is C(4,3) × 0.65³ × 0.35 = 4 × 0.274 × 0.35 = 0.384, approximately 38%. The probability of exactly two succeeding (one double at minimum) is C(4,2) × 0.65² × 0.35² = 6 × 0.4225 × 0.1225 = 0.311, approximately 31%. The probability that at least two succeed (the minimum for a return) is 0.179 + 0.384 + 0.311 = 0.874, approximately 87%. This comparison dramatically illustrates the coverage difference: an accumulator returns roughly 18% of the time from these four selections, while the Yankee system returns from approximately 87% of cases.

Expected Return Calculations for Different Systems

The trade-off is that the Yankee costs eleven times as much (eleven combinations versus one) and returns less per unit when all four succeed, because the return is divided across multiple component combinations. System predictions are not free — they require larger total cost to cover all combinations. The analytical question is whether the coverage and partial-success returns justify the additional cost given the specific probability profile of the selections involved. This is a central question of system prediction strategy, and it is addressed in detail throughout this guide.

Selection Quality: The Foundation of Every System

The most important determinant of system prediction success is not the system structure itself but the quality of the individual selections that populate it. A well-designed Trixie built around three genuinely high-quality predictions will consistently outperform a Trixie built around three mediocre selections, regardless of how carefully the system structure has been chosen. Selection quality is the engine; the system structure is the vehicle — and no vehicle can overcome a weak engine over time.

Minimum Quality Thresholds for System Selections

What constitutes a high-quality selection for inclusion in a system? The foundational requirement is that the selection must reflect genuine analytical edge — your probability estimate for the outcome must be higher than the effective probability implied by the combination structure. If you are combining three selections each with a 60% estimated win probability, your expected return from the doubles is based on implied win probabilities at those estimated levels. Any significant overestimation of your selections' true probabilities — any systematic optimism bias in your assessment — will erode the system's long-term performance.

Identifying Correlation Risks in System Building

The most reliable selections for system inclusion are those grounded in the kind of rigorous analytical framework described throughout our prediction guides. A selection based on form guide analysis showing clear and sustained momentum advantage, combined with expected goals data confirming that underlying performance matches the surface-level results, and team news showing the favoured team at full strength versus an injury-depleted opponent, is a much stronger system component than one based primarily on league table position or recent media narrative. The pre-match analysis checklist provides the systematic analytical process for building the strongest possible individual selections before combining them into a system.

Correlation and Independence: Understanding How Selections Interact

One of the most important and most frequently neglected aspects of system prediction strategy is the relationship between individual selections. Most basic analysis of system predictions assumes that each selection is independent — that the outcome of one match has no bearing on the outcome of any other. In reality, football predictions made across a single weekend of fixtures can have meaningful correlations that affect the system's overall probability distribution.

Positive correlation between selections occurs when conditions that make one selection more likely to succeed also make others more likely to succeed. For example, if your system includes three selections involving top-half teams playing at home against bottom-half opponents, and the broader football context on that weekend favours attacking football (good weather, many teams playing with full squads for the first time after international breaks), then all three selections may be simultaneously elevated by a shared contextual factor. This positive correlation means that success outcomes tend to cluster — all three winning or none winning is more likely than a mixed pattern of two wins and one loss.

Negative correlation occurs when success in one selection makes another less likely. A clear example is selecting both a team to win a match and the match to go over 2.5 goals — these selections are not fully independent because a high-scoring match makes it more likely that both teams scored, which reduces the dominant win probability and the two-selection correlation becomes meaningfully negative. Understanding which combinations of selection types are positively or negatively correlated allows analysts to choose system structures that best exploit their specific correlation profile. The data-driven predictions guide discusses how to think about correlation in prediction portfolios from a statistical perspective.

Choosing the Right System Structure: A Decision Framework

Selecting the appropriate system structure for a given set of predictions requires balancing several competing considerations: the number of selections available, the confidence level in each selection, the correlation structure across selections, and the analytical goals of the system (maximising expected return versus maximising coverage versus achieving a specific return profile).

Matching System Size to Selection Confidence

For three selections, the choice is primarily between the Trixie (four combinations, returns from two wins) and the Patent (seven combinations, returns from one win). The Trixie is appropriate when the analyst is genuinely confident in all three selections and is willing to accept a void outcome if two of the three fail. The Patent is appropriate when the analyst wants insurance against two failures but has more moderate confidence across the three selections. The additional cost of three singles in the Patent (compared to the Trixie) is only worthwhile if the analyst places meaningful probability on exactly one selection winning while the other two fail — which is only sensible if the individual win probabilities are genuinely below 65%.

Budget Considerations in System Selection

For four selections, the Yankee is generally preferred over the Lucky 15 for analysts with high confidence (estimated 65%+ probability) across all four selections, because the singles add disproportionate cost for outcomes (one selection winning, three failing) that are relatively rare at high probability levels. The Lucky 15 becomes more appropriate when confidence across selections is more moderate (55-65% range) and the probability of exactly one win is non-trivial. The guide on how many selections in an accumulator provides relevant methodology for thinking about the optimal number of selections given your confidence levels, which applies equally to system selection decisions.

For larger systems (five to eight selections), the decision between Heinz-type systems and alternative structures should be driven primarily by the analyst's practical capacity to identify and thoroughly analyse that many selections to the required standard. It is far better to have four genuinely well-analysed Yankee selections than eight selections that include weaker analyses to fill the system. The temptation to add selections to populate a larger system at the cost of analytical rigour is one of the most common errors in system prediction, and it systematically undermines the quality of the portfolio. The selection sizing guide provides specific guidance on how many selections can realistically be maintained at high analytical quality, which applies directly to system construction.

The Trixie in Practice: Three-Selection System Strategy

The Trixie is probably the most widely used system structure in football prediction because it balances simplicity (only three selections required), meaningful coverage (three doubles plus one treble), and a rational error tolerance (one wrong selection still allows returns from two doubles). Understanding how to optimise a Trixie — both in selection choice and in structuring the selections within the system — can significantly improve its analytical effectiveness.

Optimal Market Types for Trixie Systems

Sequencing selections within a Trixie does not affect the mathematical return (each selection appears equally often in the combinations), but it affects the psychological and analytical clarity of your evaluation. Organising Trixie selections in order of confidence — from highest to lowest estimated win probability — helps you assess whether the system as a whole meets the analytical standard you are applying. If your three selections have estimated probabilities of 72%, 65%, and 58%, the Trixie's minimum-return scenario (exactly two of three correct) has a probability of: (0.72 × 0.65 × 0.42) + (0.72 × 0.35 × 0.58) + (0.28 × 0.65 × 0.58) = 0.197 + 0.146 + 0.106 = 0.449, approximately 45%. The full-success scenario (all three correct) has probability 0.72 × 0.65 × 0.58 = 0.272, approximately 27%. This means the Trixie returns from exactly two wins approximately 45% of the time and from all three wins approximately 27% of the time — a combined return frequency of about 72%. This kind of explicit probability calculation for the system as a whole — not just for each selection individually — is the analytical foundation of rigorous system construction.

Managing Trixie Performance Over Time

Selection independence is particularly important in Trixie systems because with only three selections, any correlation between selections has a disproportionately large effect on the overall return distribution. The ideal Trixie contains three selections from genuinely different analytical contexts — different leagues, different match situations, different outcome types — that limit the possibility of shared contextual factors creating simultaneous failures. When all three Trixie selections come from the same matchday in the same league, they are all subject to common factors (weather, refereeing trends on a specific day, tactical patterns within the league in a particular week) that create more correlated outcomes than you might expect from treating each selection as independent. Diversifying across leagues and competition types improves the analytical robustness of your Trixie.

The Yankee in Practice: Four-Selection System Strategy

The Yankee's eleven combinations across four selections create a rich return profile that rewards multiple patterns of partial success — three of four correct (seven winning combinations), exactly two of four correct (one to three winning doubles depending on which selections win), and the jackpot full return when all four prove correct. Building an analytically strong Yankee requires slightly different thinking from the Trixie because the greater number of selections creates more complexity in correlation assessment.

Selection Diversity in Four-Leg Systems

The most common analytical error in Yankee construction is conflating confidence with certainty. Selecting four matches where you believe a specific outcome is "almost certain" invites systematic overconfidence bias — the same cognitive trap that makes long accumulators fail much more often than their component probabilities would suggest. A Yankee built around four selections each estimated at 75% win probability has a combined full-success probability of 0.75⁴ = 0.316, only about 32%. Including one overly optimistic selection — estimated at 75% but genuinely 55% — drops this to 0.75³ × 0.55 = 0.232. The marginal impact of one mis-estimated selection is substantial in a system that requires all four to succeed for the maximum return.

Yankee vs Four-Fold Accumulator Comparison

A better approach to Yankee construction is to explicitly identify your four best analytical selections for the week across all available fixtures, estimate their true probability as rigorously as possible using the frameworks in the data-driven predictions guide, and then assess whether the four-selection combination as a portfolio makes analytical sense. If three of the four are strong (estimated 65-70%) and one is weaker (estimated 55-60%), consider whether the weaker selection should be included at all or whether a more thorough analysis might reveal a stronger alternative selection from a different fixture. The discipline to exclude weak selections from a system — accepting a smaller system rather than including a marginal selection to fill a structure — is one of the most valuable analytical habits in system prediction.

Heinz and Larger Systems: When More Selections Make Sense

The Heinz (six selections, 57 combinations), Super Heinz (seven selections, 120 combinations), and Goliath (eight selections, 247 combinations) represent the most expansive end of standard system structures. These larger systems are appropriate in specific contexts — primarily when an analyst has conducted thorough analysis across a full weekend of football fixtures and has identified a genuinely large number of high-quality predictions. However, larger systems require proportionally more total cost to cover all combinations, and the minimum success threshold (typically two or three correct selections from the total) means that even a small number of correct predictions can generate returns.

The analytical challenge with Heinz-type systems is selection quality maintenance at scale. Identifying six independently analysed, genuinely high-quality predictions to the same analytical standard as a Trixie or Yankee requires significantly more analytical effort. Analysts who construct Heinz systems by including four strong selections and filling the remaining two spots with less thoroughly analysed "makeweights" are undermining the analytical integrity of the system. The correlation effects are also more complex with six selections: shared contextual factors across a full weekend of fixtures can create more extensive positive or negative correlations that change the return distribution in ways that basic combination mathematics does not capture.

For most analysts, the practical wisdom is to match the system size to the number of genuinely strong predictions available. A week with three strong predictions calls for a Trixie; a week with four calls for a Yankee; a week where six genuinely strong predictions can be rigorously supported calls for a Heinz. Artificially expanding the system to reach a larger structure by including weaker selections is a systematic error that reduces the expected return of the system over time. The accumulator strategy guide and the over 2.5 goals accumulators guide provide complementary strategic guidance on building high-quality multi-selection portfolios that inform the same selection discipline required for systems.

System Predictions Across Different Markets

System structures can be applied to any consistent prediction market — they are not limited to match outcome (1X2) selections. Some of the most analytically interesting applications of system predictions involve combining selections from different markets within the same system, or constructing systems entirely from a single alternative market where you have identified genuine analytical edge.

Both teams to score (BTTS) selections are particularly amenable to system structures because they provide binary outcomes (BTTS yes or BTTS no) and can be assessed with well-developed analytical frameworks. A Trixie or Yankee built from four BTTS selections across different fixtures combines the coverage of the system structure with the analytical rigour of a specialised market approach. The BTTS predictions guide and the BTTS accumulators guide provide the analytical foundation for identifying strong BTTS selections that can be combined into system structures.

Asian handicap selections offer another strong foundation for system construction. A Yankee of four Asian handicap selections — where the void provision of quarter-goal and half-goal lines provides additional protection against draws and marginal results — creates a system with built-in risk management at both the individual selection level (Asian handicap voids on draws) and the system level (the combination structure protecting against single-selection failure). The Asian handicap guide provides the selection framework, while the quarter-goal handicaps guide explains the specific void mechanics that make Asian handicap selections particularly suitable for system combination.

Over/under goals selections are another popular system component, particularly over 2.5 goals in high-scoring league environments. A Heinz of six over 2.5 goals selections from high-scoring leagues (Bundesliga, Eredivisie, Premier League fixtures between attacking teams) represents a system built around a well-understood statistical pattern with consistent analytical backing. The over/under goals guide provides the full analytical framework for identifying strong over/under selections suitable for system construction.

Expert Insight: Experienced football analysts who use system prediction frameworks consistently emphasise that the system structure is a consequence of the selection quality, not a substitute for it. The most productive analytical mindset approaches system construction from the selections outward — rigorously identifying the best available predictions, determining how many genuinely strong selections are available in a given period, and then selecting the system structure that most efficiently covers that specific portfolio. The opposite approach — choosing a system structure first (say, a Yankee) and then searching for four selections to fill it — almost inevitably produces weaker selections than the available analytical material can support, because the structure drives the selection process rather than the selection quality driving the structural choice. Systems are extremely powerful tools when used with this discipline; they produce systematically inferior results when used as a template to fill rather than a framework for expressing genuine analytical confidence.

Analyst Note: When constructing a system prediction, work through the following analytical process. Begin by conducting thorough pre-match analysis for every available fixture in your target period — not to find selections to fill a system, but to identify genuine analytical edges wherever they exist. After completing this analysis, rank your selections by analytical confidence, assigning estimated win probabilities based on rigorous framework-based assessment. Include only selections with estimated probabilities above your pre-set quality threshold (typically 62-65% minimum). Count how many selections meet this threshold, and select the appropriate system structure for that number (3 = Trixie, 4 = Yankee, etc.). Do not include selections below threshold to reach a larger system. Document the analytical reasoning for each selection, including the key factors driving your confidence estimate. After the system resolves, review each selection outcome and the accuracy of your probability estimates. This retrospective tracking is the most valuable long-term development tool in system prediction — it reveals whether your analytical process is well-calibrated, and it identifies specific types of selections or analysis approaches that consistently over- or under-estimate probability, allowing you to refine your process over time.

Case Studies: System Predictions in Practice

A practical Trixie case study from the Premier League illustrates the system framework in action. An analyst working through the weekend's fixtures identifies three selections with genuine analytical backing. Selection 1: Manchester City to win at home against Southampton. City's recent xG dominance over home opponents (averaging 2.6 xG generated versus 0.7 conceded in their last five home matches), combined with Southampton's away record of one win in eight matches and their key midfielder suspended, supports an estimated 73% win probability. Selection 2: Liverpool to win at home against Crystal Palace. Liverpool's home record (seven wins in eight), Crystal Palace's limited away attacking threat (0.6 average xG per away match), and Liverpool's full squad availability support a 68% estimate. Selection 3: Arsenal to win away at Brentford. Arsenal's superior form (five wins in six), Brentford's recent defensive vulnerabilities (conceding from counter-attacks repeatedly), and head-to-head history showing Arsenal winning four of the last five meetings, supports a 62% estimate. The Trixie combines these three selections. The combined probability of all three winning is 0.73 × 0.68 × 0.62 = 0.307, approximately 31%. The probability of at least two winning (the minimum return threshold) is approximately 69%. This probability distribution — a 69% chance of some return, with a 31% chance of maximum return — reflects a genuine analytical portfolio rather than random selection filling.

A Yankee case study involves a midweek European fixture round. The analyst identifies four selections: Bayern Munich to win in the Champions League group stage (estimated 78%), Atletico Madrid to win in the Europa League (estimated 65%), Celtic to win in the Conference League (estimated 71%), and Real Sociedad to win in their La Liga fixture (estimated 63%). Combined into a Yankee (11 combinations), the probability of all four winning is 0.78 × 0.65 × 0.71 × 0.63 = 0.227, approximately 23%. The probability of at least three winning (four trebles firing plus various doubles) is approximately 55%. The probability of at least two winning (at least one double) is approximately 84%. The Yankee structure expresses this four-way analytical portfolio in a combination structure that returns in 84% of scenarios — a robust analytical expression of confidence across four independently analysed selections. The Conference League analysis guide and form analysis provided the underlying analytical frameworks for assessing these individual selections.

The third case study examines a Heinz system across six selections from a full weekend of Premier League and Championship fixtures. The analyst has conducted thorough pre-match analysis for all weekend matches and identified six selections meeting their 63%+ confidence threshold. The selections include: three matches with clear home advantages, two matches involving strong form teams against in-form opponents, and one selection based on a significant tactical mismatch identified through the tactical formations analysis. The Heinz's 57 combinations cover this portfolio comprehensively — returning from any pattern of at least two correct selections out of six. The analyst expects at least four of the six to prove correct based on their probability estimates, which would activate a significant number of the Heinz's trebles, four-folds, and potentially the five-fold. This case illustrates how the Heinz is most powerful as a tool for expressing confidence across a full weekend of well-analysed fixtures, rather than as a structure requiring specific numbers of selections to fill it.

System Predictions and Long-Term Performance Tracking

System predictions, like all football prediction frameworks, require long-term performance tracking to assess whether they are generating genuine analytical value or producing results consistent with random selection. Building a tracking record of your system predictions — including the analytical reasoning behind each selection, the system structure used, and the actual outcomes — is essential for determining whether your approach is working and how it can be improved.

The key performance metrics for system prediction tracking are: the rate at which systems return (how often do you achieve at least the minimum return from your system?), the distribution of outcomes across full success, partial success, and no return, and the calibration of your individual selection probability estimates (are selections you estimate at 65% actually winning approximately 65% of the time?). Well-calibrated probability estimates and consistent return rates above what random selection would produce are the two signals that your analytical framework is genuinely working. The prediction model building guide provides the full methodology for tracking and validating analytical performance over time, which applies directly to system prediction portfolio assessment.

Patterns in which system structures and which selection types produce the best analytical results will emerge over time in your tracking record. Some analysts find that they are particularly effective at identifying BTTS selections (perhaps through deep knowledge of specific leagues and teams) and that BTTS-based Trixies and Yankees outperform their match outcome systems. Others find that their analytical edge is most pronounced in specific competition contexts (home favourites in the Champions League, or away wins in the Championship). Identifying your specific analytical strengths and structuring your systems to maximise exposure to those strengths is the long-term development path for system prediction excellence.

Expert Insight: The most common error in system prediction analysis is treating the system structure as a substitute for selection quality rather than a complement to it. A Trixie built on three mediocre selections does not become analytically sound because of the coverage structure — it simply loses money in a more organised way. Systems reward analysts who already have strong individual selection processes; they cannot compensate for weak analysis at the selection level.

Conclusion

System predictions — Trixies, Yankees, Heinz, and beyond — represent the most structured and mathematically rigorous approach to multi-selection football forecasting. They convert a portfolio of individual predictions into a carefully engineered combination structure that provides partial-success coverage unavailable from straight accumulators, while rewarding the full accuracy of a well-constructed selection portfolio with maximum returns. The mathematical framework is elegant and well-defined; the analytical challenge is entirely in building the selection quality that makes the system perform.

The central lesson of this guide is that system structure follows selection quality, not the reverse. Building three, four, or six genuinely strong analytical selections — grounded in expected goals data, form analysis, tactical context, team news, and all the other dimensions of rigorous football prediction — and then choosing the system structure that best covers that specific portfolio is the correct analytical process. Choosing a Yankee or Heinz structure first and then searching for selections to fill it is the wrong process, and it produces systematically weaker results over time. Applied with analytical discipline and selection rigour, system predictions provide a powerful framework for expressing football forecasting confidence across multiple events simultaneously. For deeper analytical foundations, explore our guides on accumulator strategy, selection sizing, how many selections in an accumulator, data-driven predictions, and the pre-match analysis checklist to build the strongest possible analytical foundations for your system selections.