The Alternating Veto and Snake Draft: How Esports Teams Use the Pick-Ban Phase to Create a Strategic Advantage

Before a single digital ability is cast or a tower is struck, the match is already mathematically tilted. In the high-stakes arenas of professional esports, the mechanical execution of a 5-player roster is heavily constrained by the tools they are allowed to bring onto the server. This constraint is engineered through the pick-ban phase, a zero-sum negotiation that occurs before the game engine even loads the map.[6]

The scale of this strategic puzzle is massive. As of 2026, Riot Games' League of Legends features 168 distinct champions, while Valve's Dota 2 boasts 124 unique heroes. If teams were allowed to freely select any combination, the mathematical permutations would be virtually infinite, leading to stagnant, optimized strategies. To force adaptation, developers implemented the alternating veto and snake draft.[3][6]

The alternating veto, commonly referred to as the ban phase, allows teams to explicitly remove specific characters from the available pool for that specific match. In League of Legends, teams are allotted 10 total bans, divided into two distinct phases. This mechanic serves a dual purpose: it allows teams to remove statistically overpowered characters from the current meta, and it provides a surgical tool to target and eliminate the signature characters of specific opposing players.[1][2]

Following the initial vetoes, the snake draft begins. Unlike a simultaneous blind pick, the snake draft forces teams to alternate their selections in a 1-2-2-2-2-1 pattern. When Team A selects one character, Team B immediately responds by selecting two. This structure ensures that neither team can secure a perfect synergistic combination without exposing their strategy and allowing the opponent to draft direct counters.[3][6]

The complexity peaks in Dota 2's Captains Mode, the standard format for its professional circuit. According to Liquipedia's documentation of the mode, the draft is broken into three distinct ban phases and three pick phases, totaling 14 bans and 10 picks between the two squads. This fragmented structure forces drafters to constantly re-evaluate their win conditions as the available hero pool shrinks in real-time.[3]

To quantify exactly how decisive this pre-game phase is, data scientists have turned to machine learning. Researchers publishing in ResearchGate analyzed feature datasets to predict League of Legends victories based entirely on the picks and bans phase. Their models demonstrate that the structural advantages secured during the draft carry immense statistical weight, often predicting the winning team before the players leave the fountain.[1]

Similar predictive modeling has been applied to the individual selections themselves. Analyst Tim Inzitari built models specifically designed to predict the exact sequence of picks and bans in League of Legends games. By analyzing historical meta data and player tendencies, algorithms can anticipate a drafter's next move, highlighting how rigid and mathematically solvable certain meta-games can become at the highest levels of play.[2]

The academic focus on Dota 2 is equally rigorous. A study published via ScholarSpace utilized a Generative AI approach, specifically employing BERT (Bidirectional Encoder Representations from Transformers) models, to predict bans and picks. By treating the sequence of drafted heroes like a sequence of words in natural language processing, the AI successfully mapped the contextual relationships and synergies that professional captains prioritize.[4]

This computational approach to drafting is not entirely new. As far back as 2016, researchers presented a paper at the AAAI Conference titled "Draft-Analysis of the Ancients: Predicting Draft Picks in DotA 2 using Machine Learning." Even a decade ago, the data indicated that the draft was a highly structured, predictable environment where mathematical optimization could rival human intuition.[5]

What these models collectively reveal is the concept of the "flex pick." A flex pick is a character drafted early in the sequence that can be played in multiple different roles on the map. By securing a flex pick, a team obscures their final strategy, forcing the opponent to guess which lane the character will occupy and potentially baiting them into drafting the wrong counter-matchup.[6]

Despite the wealth of quantitative data, the human element of the draft room remains elusive in the literature. Because the cited academic literature relies entirely on API datasets and quantitative modeling, none of the researchers provide direct qualitative quotations from professional coaches, leaving the psychological warfare of the draft room unrecorded in these studies.[1][4][5]

The models also carry inherent limitations. Machine learning algorithms rely heavily on historical patch data and struggle to account for sudden, mid-tournament meta shifts or highly unconventional "pocket picks" that a team has kept hidden in scrimmages. Furthermore, a mathematically perfect draft still requires flawless human execution under the pressure of a live stadium audience.[2][6]

Ultimately, the draft is the architectural blueprint of the match. The alternating vetoes tear down the opponent's foundation, while the snake draft builds the synergistic framework. The data proves that a superior blueprint provides a massive statistical advantage, but the 5 players on the stage must still successfully build the house.[6]