Fight Initiation

The Fight Initiation tab — sometimes called "Going First" — answers a simple but important question: who started each teamfight, and did going first actually pay off? Initiating a fight is a deliberate choice with real risk and reward, and this tab shows you, fight by fight, whether your team's openers were winning you fights or feeding the enemy. It sits on the Initiation tab of a scrim's map page.

What It Answers

Going first is not always good. A clean, coordinated commit that catches the enemy out of position wins fights; a hasty one that gets answered instantly loses them. Fight Initiation grades each teamfight by who threw the first punch and whether they came out ahead, so you can tell the difference between proactive, profitable aggression and just dying first.

Crucially, initiation is treated as a team decision, not a single player's moment. The detector looks for breadth of commitment — several players engaging the enemy together — rather than one big damage spike. A lone Widowmaker headshot is a damage spike, but it is not your team going first; a two-player dive (say, Ball and Tracer diving in together) is. This is a deliberate choice: it keeps the stat honest, so the "who initiated" call reflects an actual team commitment you can hold the team accountable to, not a flashy pick that happened in isolation. A hard commit — an offensive ultimate, or multiple abilities fired off together — also counts as going first, because that is the team spending resources to open.

The Per-Map Inspector

The tab opens with a pair of summary cells, one per team, showing each team's win rate going first and how many fights they initiated, along with a note on how many of the map's fights could be labeled. The win rate counts only fights with a clear initiator; contested starts are still listed in the timeline but left out of the win-rate math, since neither side truly went first.

The inspector is read-only — there is no way to override a label by hand. That is by design: it exists so you can audit every call and see the evidence the detector used, fight by fight.

Below that is a chronological timeline of every fight on the map. Each fight is a row with:

• The fight number and the time it happened.
• A headline describing the start: one team going first, or a contested start when both teams committed at almost the same moment. When no clear pre-fight commit is found (for example, an instant pick out of neutral), the call falls back to whichever team scored first and is marked low confidence so you know it is a best-effort read.
• Whether the initiator won or lost the fight, when both are known. If a team went first and still lost, the row flags it as losing the opener — exactly the situation worth reviewing. Parsertime keeps these "went first, lost the opener" fights on purpose, because they are the ones coaches most want to find.

Confidence and Evidence

Detecting who went first is a judgement call, so every labeled fight carries a confidence level — high, medium, or low — shown with a small indicator. Confidence is built from three things: how cleanly the initiator committed before the enemy responded (a bigger head start reads cleaner), how many independent signals agree (damage breadth plus an ult plus an ability burst all pointing the same way), and a reactive-healing cross-check (the team being engaged on should show a healing spike just after — agreement raises confidence, contradiction caps it low rather than guessing).

Treat the confidence chip as how much weight to put on a single fight. High confidence reads can be taken at face value in review. Low confidence reads — a near-simultaneous start, an instant pick with no clear opener, or signals that disagree — are worth opening in the Replay Viewer to confirm with your own eyes before you draw a conclusion. Confidence does not change the aggregate numbers; it just tells you how trustworthy each individual call is.

Each fight also shows the supporting evidence behind the call, as small chips:

• Response gap — how long the initiating team committed before the other team responded (for example, "1.2s before response"). A bigger gap is a cleaner initiation.
• Ability and ult commit signals — whether the opener was backed by an ultimate or a burst of offensive abilities fired together.
• Players committed — how many of the initiating team pushed in together.

Behind the scenes the grader also cross-checks reactive healing — if the healing pattern contradicts the engagement read, the fight is graded with less confidence — so the labels stay honest about what the log can actually support.

Round Markers

A Rounds toggle in the section header drops round-boundary markers into the timeline, showing where each round started, changed, and ended. This makes it easy to read initiation in the context of the map's structure — for example, whether your team consistently goes first on attack but plays reactively on defense.

Team-Level Views

The per-map inspector is the detailed view, but initiation is also rolled up so you can see the trend across a whole block or season:

• The Teamfights tab on the team stats page aggregates your win rate going first, how often you initiate, and your win rate going second across every map with detailed logs.
• The scrim overview has a Fight Initiations section that summarizes initiation across all the maps in a single scrim.

How to Use It

Read the per-map timeline to spot openers that lost — fights where your team went first and still lost the fight. Check the confidence and evidence to make sure the read is solid, then use the Rounds toggle to see whether the problem is tied to attack, defense, or a specific point. When you find a costly opener, jump into the Replay Viewer to see the commit itself, then bring the Teamfights trend to your review to check whether it is a one-off or a habit.