Skip to content
Data & Analysis · Coverage 8 November 2026

G2 Offseason Data Guide: Five Stats More Useful Than KDA

Evaluate offseason arguments with resource-adjusted laning, objective setup, participation context, conversion and repeatability instead of a single highlight-friendly ratio.

A G2 player sits in a gaming chair under tournament stage lighting
Player statistics need the context of a role, draft and opponent. Photo via bo3.gg.

Start with the decision the number must inform

Offseason debates often begin with a player comparison and end with a screenshot of kills, deaths and assists. That ratio is easy to recognize, but it does not explain every job a professional player is asked to perform. G2 fans looking ahead to a new season need a better first question: what problem is the proposed player or roster change supposed to solve? A statistic becomes useful when it helps answer that question with clearly defined evidence.

This guide proposes five review measures, not a proprietary G2 ranking. Every calculation below uses hypothetical numbers created for explanation. No player is assigned an invented performance score. The framework is intended for a fan's spreadsheet or editorial notebook, where each figure can be traced to an identified match and a consistent definition. It also explains when a number should remain unfilled. Missing data is less misleading than a confident estimate quietly presented as an observation.

Measure one: resource-adjusted lane state

A lane advantage means more when you know what was invested to obtain it. Start with a fixed checkpoint, such as a stated game timestamp, and record the visible gold or creep-score difference. Beside it, note observable help: jungle visits, support movements, lane swaps and whether either player missed a wave while participating elsewhere. Keep these contextual columns separate from the raw difference. Do not hide a subjective adjustment inside a supposedly objective total.

For a hypothetical example, Player A leads by twelve creeps after receiving two visible jungle visits, while Player B leads by six without a visible visit. That does not prove Player B is better. The matchups, wave positions and purpose of the visits matter. It tells you that a twelve-to-six comparison is incomplete. A responsible review describes A's lead as partly supported by team resources and asks whether that investment produced something valuable, such as access to the next neutral objective or a reliable side-lane threat.

Avoid a fake formula for lane independence

It is tempting to subtract an arbitrary number of creeps for every jungle visit and call the result independent laning value. That shortcut confuses countable events with equivalent events. A brief appearance in river may change almost nothing; a coordinated dive can decide several waves. The opposing jungler's revealed location can also create freedom without a teammate ever entering the lane. Some of the most important resources are informational.

Instead, use categories with written criteria. 'Direct intervention' could mean a teammate damages the lane opponent or forces a summoner spell on screen. 'Pressure only' could mean the teammate appears nearby without committing. If a play is not visible, mark it unknown. Across a sample, those notes let a reader distinguish patterns without pretending the events are identical. When discussing G2's next lineup, the useful conclusion concerns the conditions under which a player creates an advantage, rather than a claim that all assistance has been mathematically removed.

Measure two: objective setup success

Objective totals describe who collected a reward. Setup measures describe whether a team created a favorable chance to contest it. Define a review window before an objective and record lane availability, visible recall timing, nearby numbers and access to safe entrances. You can then classify whether the team arrived prepared, arrived under pressure or chose a deliberate trade elsewhere. This is a manual review measure, so its definitions should be published with the result.

Imagine ten hypothetical objective windows. A team has favorable entry conditions in six, unfavorable conditions in two and deliberately trades the other side of the map in two. Calling the setup rate six out of ten may be acceptable if your denominator includes every window, but it obscures the deliberate trades. Reporting six favorable setups out of eight attempted contests, with two trades shown separately, answers a different question. Neither denominator should be changed after seeing which version makes a favored player or team look better.

Distinguish preparation from the last hit

A well-prepared contest can still be lost, while a poorly prepared contest can succeed through an exceptional mechanical play. If a steal immediately turns an unfavorable setup into a favorable score, the metric becomes a disguised outcome count. Keep the pre-contest assessment and the final result in different fields. That separation allows the review to recognize both preparation and execution without forcing one to erase the other.

Use a hypothetical sequence to make the distinction concrete. A team enters river late, cannot safely check an entrance and has a side wave pushing into its base. Its jungler nevertheless secures the objective during a chaotic fight. The result is valuable. The setup remains fragile. An offseason argument about communication or map planning should not use the successful last hit as proof that no improvement is needed. Equally, the setup classification should not deny the individual player credit for a difficult execution under pressure.

Measure three: opportunity-aware participation

Kill participation can reward proximity without explaining responsibility. A side-lane player may be correctly collecting a wave while teammates take an uncomplicated skirmish. A support may be present for many kills because the composition constantly groups. Before comparing participation figures, identify which encounters a player could realistically join, given location, available movement tools and the team's visible plan. The analysis should not penalize a player simply for performing a different job.

Suppose a hypothetical player participates in eight of twelve team kills. The familiar percentage is about sixty-seven percent. A manual review finds that three of the four missed kills occurred while the player was visibly completing a planned cross-map trade. That observation changes the interpretation; it does not authorize changing the official participation figure to a more flattering number. Publish the original count and the contextual note side by side. Readers can then debate the role assignment rather than arguing over a ratio that silently treats every absence as a mistake.

Measure four: advantage conversion

An advantage is only a starting condition. Conversion asks what happens after a team gains something: a kill, a favorable recall, a numbers advantage or control of a side lane. Choose a limited follow-up window and define the possible outputs before reviewing the games. Depending on the question, those outputs might include a turret, safe objective access, a denied wave or a reset that preserves tempo. Avoid crediting every later success to one earlier play.

Consider five hypothetical early kills. Two lead to a turret or objective opportunity, one creates a clean reset, and two produce no visible lasting gain. A conversion discussion can distinguish those outcomes without assigning an arbitrary gold value to each. More importantly, inspect why conversion failed. Perhaps the winning players were too low on health to remain. Perhaps the next wave was unfavorable. Perhaps the kill itself was strategically low value. The measure becomes useful for G2 analysis when it identifies a repeatable decision that could improve, rather than merely restating that winning teams collect more rewards.

Measure five: repeatability across conditions

A player's best game demonstrates a possible ceiling. Roster planning also requires an estimate of how often useful behavior appears under different conditions. Split a sample by relevant factors: opponent style, champion role, map side, match stage and resource allocation. Use only categories that answer your original question. Subdividing a small sample into twenty groups can create an illusion of sophistication while leaving every group too small to support a conclusion.

For example, an imaginary player records positive lane differences in seven of ten games. If six of those seven occur on one comfortable matchup, the headline needs that qualification. It does not make the performance worthless. It suggests a narrower claim: the player has shown a reliable advantage under a specific condition, while flexibility remains less certain. A useful G2 offseason discussion asks whether the team can consistently create that condition and what it sacrifices elsewhere to do so. Repeatability is about fit as much as individual consistency.

Build an audit trail before a ranking

Each row in a fan dataset should identify the event, opponent, game, patch and source. Manual observations need a timestamp and a brief definition. Riot's developer documentation describes several game data interfaces, but access to a documented interface does not automatically provide a complete professional esports dataset. Do not imply that ranked-game statistics are official stage data or that an unavailable field has been verified by an API.

Check a small portion of your entries twice, ideally without looking at your first classification. If you regularly disagree with yourself about whether a setup was favorable, improve the definition before expanding the sample. Keep corrections in a change log. A transparent dataset with twelve carefully reviewed games can be more useful than a large table with inconsistent categories. The point is not to imitate a professional scouting department. It is to make a public argument that another fan can inspect, reproduce and disagree with on identifiable grounds.

A small hypothetical audit shows why denominators deserve their own column. Suppose an analyst initially classifies twelve objective windows, records seven favorable setups and marks two deliberate trades. Another three windows appear unfavorable. A second pass discovers that one supposed unfavorable contest was actually abandoned before either team committed. The correction changes the attempted-contest sample, while the full list still contains twelve windows. Keeping both the original classification and the correction makes the resulting percentage understandable. Quietly deleting the awkward row would make the dataset easier to present but harder to trust.

The same discipline applies when a field cannot be observed. If the broadcast never shows a player's available gold before a recall, do not fill the gap with a likely purchase or a remembered item timing from another game. Mark the field unavailable, explain whether the missing value matters and continue with the observations you do have. A reader can evaluate a bounded claim about synchronized movement without knowing every inventory detail. This approach also makes collaboration easier: another reviewer can inspect the same timestamp and supply a missing observation without rebuilding the entire table. The result is a living record of evidence, rather than a polished ranking whose assumptions disappear behind the final score.

Turn the numbers into a cautious offseason verdict

A good final verdict names a strength, the conditions supporting it and the uncertainty that remains. Instead of declaring a hypothetical player the answer to every G2 problem, explain that the sample suggests effective weak-side management, while late objective coordination remains unresolved. Those are claims a future review can test. They are also more informative than attaching a transfer recommendation to a single KDA screenshot.

Keep team fit distinct from individual merit. Two excellent players can compete for the same resources, and a modest-looking statistic can reflect a role that enables another teammate. Riot's introductory guide frames League around a shared strategic objective; our interpretation should preserve that team context. The five measures in this guide are prompts for better questions, not replacement medals. Used carefully, they reveal why a number changed, what behavior it represents and whether that behavior addresses the actual problem an offseason move is supposed to solve.

Methodology and sources

This is an original analytical framework. All numerical scenarios are explicitly hypothetical, not measured G2 or player statistics. Definitions emphasize observable broadcast information and reproducible manual classification; official Riot sources establish game and data-system context.

Evidence limits

No proprietary scouting records, team voice communications, professional match API feed or verified current roster dataset was accessed. Manual categories can suffer observer bias; small samples and patch changes limit comparisons.

Primary references

Prepared with AI assistance under the G2 Fan Editorial Desk publication byline. Coverage date identifies the subject's editorial context; it is separate from publication. Report a correction.