Docs
Metrics Guide
What LeagueLab's metrics measure, how they are calculated, how to read them, and where they break down
This page explains what each metric on the dashboard measures, how it is calculated, and which pitfalls it carries. Every formula is kept in sync with the actual collection and aggregation code, and this document is updated whenever a definition changes. If a term is unfamiliar, the glossary is the faster place to start.
Lane Phase Stats (GD, CSD, XPD)
Gold difference (GD), CS difference (CSD) and experience difference (XPD) are measured against the opposing player in the same position. They are your value minus your lane opponent's, so a positive number means you were ahead in lane, and the two players' values always sum to zero. The cutoff is whichever comes first: the fall of the first tower, or the 15-minute mark. Once a tower falls the lane structure breaks down, and the gap that opens afterwards reflects team-wide macro play rather than lane skill. If no tower falls within 15 minutes, the 15-minute snapshot is used. Lane dominance is judged by looking at gold, experience and CS (including jungle camps) together to decide who came out ahead in each position. Solo kills only count kills taken alone in the lane area, without help from a teammate. Two things to keep in mind. First, games where the first tower falls early have a shorter measurement window, so the absolute numbers come out smaller — read averages over several games rather than comparing single matches. Second, a jungler's lane differential depends far more on pathing and gank outcomes than on individual laning.
Economy Stats (CSPM, GPM, gold share, DPG)
CS per minute (CSPM) counts minions and jungle camps together, divided by game length in minutes. Gold per minute (GPM) uses the value Riot provides directly. Both answer "how steadily did this player earn" rather than "how much did they end up with", which makes them far less sensitive to game length. Team gold share is a player's earned gold divided by the sum across all five teammates. It shows where resources were concentrated and which player the team was built around. Damage per gold (DPG) divides champion damage by gold earned, measuring how efficiently resources were converted into damage. Economy stats must always be compared within the same position. A support's CSPM being lower than an ADC's is a difference in role, not in skill. The same is true of gold share: doing your job on a low share is a good sign for a support or a tank, not a bad one.
Combat Stats (KDA, kill participation, DPM, damage share)
KDA is (kills + assists) divided by deaths. The denominator is adjusted so that games with zero deaths do not send the value to infinity. Kill participation (KP) is the share of the team's kills a player contributed to with a kill or an assist; it is computed per game first and then averaged, because dividing cumulative sums would let high-kill games dominate the result. Damage per minute (DPM) uses Riot's own value and counts only damage dealt to enemy champions, excluding minions and objectives. Damage share is a player's champion damage divided by the sum across all five teammates. Combat stats are tightly bound to champion and role. Tanks post low damage share but absorb a great deal, while supports show high kill participation with a KDA that is sensitive to deaths. Rather than ranking on a single stat, read damage share, damage taken and kill participation together to understand what a player's role actually was.
Vision Stats (vision score, wards, control wards)
Vision score is Riot's own value, combining the time your wards kept the map lit with your contribution to clearing enemy wards. LeagueLab also shows vision score per minute (VSPM), dividing by game length so that differences in match duration drop out. Wards placed, wards killed and control wards purchased are raw counts. Placement shows how actively vision was established, kills show how much enemy vision was stripped away, and control ward purchases show how much gold went into the vision battle. Vision is the axis with the widest gap between positions. A support recording two or three times what other roles do is normal, so a leaderboard that mixes positions simply fills up with supports and tells you nothing. Always compare within the same position.
Objectives and Early Game (first-objective rates)
First blood, first tower, first dragon and first herald rates are team-level stats: the share of games in which that team secured the objective first. A 60% first dragon rate means the team took the first dragon in six games out of ten. These four are useful for reading a team's early-game identity. High first blood and first herald rates suggest a team that initiates early skirmishes, while a high first tower rate suggests a team that converts lane advantage into structures. First tower is also the event that closes the lane-phase measurement window, so teams with a high rate here tend to have shorter lane-phase windows in their stats. Some tournaments do not record these four. When that happens the block is hidden entirely rather than shown as 0%, because "did not secure it" and "was never recorded" are completely different statements.
Champion Tier Score
The tier score combines two axes: strength and presence. Strength does not use raw win rate; it uses an empirical Bayes posterior mean — (wins + κ × expected performance) ÷ (picks + κ), with κ=8. Pro play often gives a champion only a handful of games per patch, and taking two wins in three games at face value would let noise flip the tiers. Expected performance is a prior derived from the champion's actual lane-phase results, centred on 0.5 within its position and clamped to the 0.35–0.65 range. Presence is (picks + 0.5 × bans) ÷ total games. Bans count for half a pick because a ban signals that teams respect the champion, but it is not evidence that the champion performed in a game. The final score standardises strength within the position and adds standardised presence multiplied by a coefficient λ. That λ is asymmetric: 0.30 for strong champions (strength of 0.5 or above) and 0.12 otherwise. This stops a champion with a poor win rate from climbing the tiers on pick-ban priority alone, and the influence of presence is capped at both ends as well. Tiers 1 through 5 are cut by percentile within the position, with a few common-sense guardrails on top. Champions with fewer than 3 picks are left as "—" rather than tiered, fewer than 8 picks cannot reach tier 1, and a champion picked 20 or more times is never pushed down to tier 5. Conversely, 10+ picks with a win rate of 55% or better plus high pick-ban priority or first-pick rate locks in tier 1. Tiers are computed per patch or per tournament stage, and the tier shown always matches the scope you are currently viewing.
Draft Stats (first pick, second pick, ban rate, pick rate)
Draft tables are split by first-pick team and second-pick team rather than by side (blue and red). Depending on tournament rules, side and pick order do not always coincide, so grouping by side would mix different draft situations into the same cell. Bans are counted in two phases: phase 1 (bans 1–3) and phase 2 (bans 4–5). Phase 1 bans remove top-meta champions or an opponent's signature picks, while phase 2 bans target the last gap in the opposing composition — the two serve different purposes. Picks are split by slot (pick 1, picks 2–3, picks 4–5, pick 6, pick 7, picks 8–9, last pick) to show which champions appear where in the draft. The win rate attached to a ban card is based on games where the champion was actually picked, not on ban count. A champion that was only ever banned has no win rate sample and shows "—". For tournaments that do not record draft order at all, the slot tables are replaced with pick distributions by side and by position.
Pro Solo Queue Stats
Solo queue stats are calculated from pro players' public solo queue accounts, collected through Riot's official API. These are not tournament games, so there is no team strategy or draft constraint, and what a player is currently practising tends to show through fairly directly. Solo queue stats are retained for the most recent 14 days only. Older records cannot be queried, and the date range selector on screen does not extend past that window. The value of pro solo queue lies in "what are they working on right now"; account activity from six months ago says little about current form. Champion builds (runes, starting items, boots, core items, skill order) are aggregated from pro accounts together with verified one-trick accounts. On the champion analysis screen you can separate pro match builds from solo queue builds by source. The account list is maintained through manual verification, so transfers and new accounts may take some time to appear.
Sample Size and Confidence
When the sample is small, the number reflects luck rather than skill. Between a player at 100% across two games and one at 60% across forty, the second is the more trustworthy figure. LeagueLab handles this in two ways. First, player rankings have a minimum sample threshold. Rather than a fixed game count, the threshold is 20% of the average number of games teams played in that tournament stage, with a floor of 3 games — split lengths differ, so a fixed number would not travel. Players below the threshold still have their values shown, but are excluded from ranks and percentiles and are flagged on screen as low sample. Second, champions with fewer than 3 picks are not given a tier and show "—". They are also excluded from the denominator of the ranking, so a one-off champion cannot skew the percentile distribution and inflate everyone else's tier. Percentiles and ranks are always computed within a single position. For display, percentage metrics are rounded to one decimal place and all other figures to two.
Data Source Limits
Even for official matches, how much detail survives varies by tournament. LeagueLab draws on three kinds of source, and which metrics can be offered depends on which one filled the record. The first is Riot's official match data. Both participant records and the timeline are available, so lane differentials, time-sliced snapshots, item purchase order and skill order can all be derived. The second is tournament scoreboards. End-of-game figures such as kills, deaths, assists, gold and CS are there, but without a timeline there is no way to build lane-phase stats or build order. The third is end-of-game snapshots from the broadcast system. Final items and runes survive, but game duration is sometimes missing, which makes per-minute metrics impossible. Metrics that cannot be filled are hidden rather than padded with zero. A zero has to mean "this really was zero"; once it blurs into "never recorded", the whole table reads wrong. Comparing across leagues also calls for care. Leagues differ in game counts, meta and patch timing, and international events run short enough that samples stay small. Before placing numbers from different leagues side by side, check that they came from comparable conditions.
If the metric names or broadcast terms are unfamiliar, the glossary is a good place to start. Go to Glossary →
Definitions are maintained against the calculation code. If you spot an error, please report it via the contact address on the about page. Go to About →