arrow_back All field notes
// F.01 · Match analysis

The scrim timeline: finding the moment a Dota 2 game actually turned

A net worth graph tells you that you lost. The timeline tells you which ninety seconds you lost it in.

Published 2026-08-10Updated 2026-08-107 min read

Why the net worth graph alone is not enough

Every Dota analytics site shows you a net worth advantage curve. It is the single most recognisable chart in the game, and on its own it is close to useless for review. It tells you that at minute 28 you were nine thousand gold behind. It does not tell you that seven of those thousand arrived in one fight at minute 24, that the fight happened because a support died alone at 23:10, or that the actual mistake was a rotation at 22:30 that left the safe lane uncovered.

A graph is a summary of consequences. Review needs causes. The timeline exists to connect the two: it puts the curve and the events that produced it on the same axis, so a spike is not something you interpret — it is something you click.

Fights, not kills

Raw kill logs are the wrong granularity. A five-man wipe is one event that you should discuss for ten minutes; it is not five events. So kills that belong to the same engagement are clustered into a single fight, with the gold and experience outcome attached to the cluster rather than to the individual kills.

That changes what the game looks like when you scan it. Instead of forty scattered kill markers you get eight or nine fights, each with a result, and it becomes obvious which ones were actually expensive. A fight you won 3-for-2 but which cost you the Roshan window and a tier two shows up as a net loss, because the ledger includes what happened around it.

Gold and XP swing
What the engagement was worth, net, to each side — including the objectives that fell out of it.
Participation
Who was actually there. A five-man fight your carry missed is a different conversation from one they lost.
The lead-in
The fight is anchored to a tick, so you can jump into the mini replay thirty seconds earlier and watch the positioning that caused it.

The objective ledger

Towers, barracks, Roshan, tormentors, outposts and runes, stamped to the second and attributed to a side. This is the part of the game teams reconstruct badly from memory. Nobody remembers that the enemy took every single bounty rune between minutes 10 and 20; everybody remembers the fight they lost at 25.

Laid out on an axis, patterns that are invisible in the moment become obvious. Consistently losing the pre-fight rune. Never taking the second tier two after winning a fight. Taking Roshan at a time when you had no way to use the Aegis. These are habits, and habits only show up when you can see twenty of them next to each other.

1 TICK RESOLUTION · ~2,400 EVENTS / GAME · PER-SIDE ATTRIBUTION

The pivot

Most games have one. Not the moment the game became unwinnable — the moment the game stopped being even. The timeline flags the single largest sustained swing and marks it on the chart, because teams are systematically bad at identifying it themselves.

The reason is recency: the last twenty minutes of a loss feel like the loss. They usually are not. They are the consequence of something that happened while the game still felt fine, and which nobody flagged at the time precisely because it felt fine. A pivot marker at minute 19 on a game that ended at 44 reframes the entire review.

The most useful review question is not "why did we lose?" — it is "at what minute did we stop being able to win, and what were we doing thirty seconds before that?"

Reading a timeline in five minutes

A practical order of operations for a coach with a scrim block to get through and limited time:

1. Find the pivot
Start there, not at the beginning. Everything before it is a game you were in; everything after is a game you were losing, and those are different discussions.
2. Check the laning checkpoint
Look at where the curve sits at ten minutes. A game that was already behind at 10 has a laning problem, not a midgame problem, no matter what the pivot says.
3. Scan the objective ledger for one-sidedness
Any category the enemy took nearly all of is worth a sentence, even if it did not obviously cost you the game.
4. Open the two most expensive fights
Jump into the mini replay from each and watch the lead-in. Two fights properly understood beats nine fights skimmed.

Done in that order, a game takes five to eight minutes to review usefully, which means a five-game scrim block fits comfortably into an hour with time left for the drafting conversation.

Across a block, not just a game

A single timeline diagnoses a game. The value compounds when you have twenty of them from the same scrim block, because individual games are mostly noise. One game where you lost the 20-minute Roshan fight is a bad fight. Six games where the curve turns down between minutes 18 and 24 is a pattern about how your team plays the transition out of laning, and that is something you can actually train.

Because scrims live in a team-scoped vault rather than being scattered across five people’s drives, you keep the history to make that comparison. The analysis stays pinned to the patch it was played on, so a curve from three patches ago does not quietly get compared against a different game.

Questions

Does the timeline work on scrims played in private lobbies?

Yes. Upload the replay file through the parser and the timeline is built from that parse. Nothing requires the match to be public or to exist in Valve’s match history.

How are fights clustered?

Kills that fall inside the same engagement window are grouped into one fight card, with the gold and experience outcome attached to the cluster rather than to each individual kill.

Can I jump from the chart into the replay?

Yes — every event is anchored to a tick, so a spike on the chart links to the exact second in the 2D mini replay. That is the intended workflow: find it on the timeline, watch it on the map.

What is the pivot exactly?

The single largest sustained swing in the game — the point where the advantage curve stops oscillating and commits to a direction. It is flagged automatically so review starts at the moment that mattered rather than at minute zero.

Run this on your own scrims

Upload a replay and the whole workspace runs on your game. Create a team — visible to your roster only, never public.

Read next
© 2026 Scrimlog Labs · Built in ViennaField notesPrivacy