Drafting is the one phase you can prepare completely
Almost everything else in Dota is executed live. The draft is not. It is a sequence of decisions made under time pressure from a state you can enumerate in advance — and it is therefore the single phase where preparation converts most directly into advantage.
The practical problem is that draft preparation usually lives in someone’s head or in a spreadsheet that is one patch out of date. The draft lab is that preparation made concrete: a board you can drag heroes onto, a score for both sides, a ranking of what to take next, and a history of what your opponents actually do.
Scoring a draft on two different scales
Drop heroes into two line-ups and the board scores both sides continuously — a partial draft is valid, which matters, because the interesting questions happen at two picks each, not at five.
Each side gets two separate tallies, and keeping them separate is the whole design:
- Traits
- Fifteen properties a draft can have — the things you say out loud in a draft conversation, like sustained damage, initiation, or the ability to break a defensive stance. Scored per hero and summed.
- Roles
- Eight role coverages on Valve’s own levels. This is the "do we actually have a draft" check: it catches the line-up that looks strong and has nobody who can hold a lane.
The two run on different scales, so adding them together produces a number that means very little — a fact the interface takes seriously. The combined total is shown as a muted footnote and nothing is decided on it. Comparison happens within a scale, where it is meaningful.
The verdict counts metrics won, not points won. One blown-out trait should not outweigh five metrics lost by a nose — that is how you talk yourself into a draft that wins one number and loses the game.
Edges: who wins what
The default readout is not a single score but a split: what your side wins, what theirs wins, and what is tied. Each column is grouped into traits and roles, biggest gap first, showing both the gap and the raw pair behind it.
Bars are scaled within their own kind, because a three-point lead means completely different things on the two scales — on the role tally it is a rout, on the trait tally it is close to noise. Scaling them together would make the smaller scale unreadable and the larger one look decisive.
This is deliberately closer to how a captain thinks than a single "draft rating" would be. You are not looking for a verdict; you are looking for the two or three axes on which the game will be decided, so you know what the plan has to be. A draft that loses the teamfight metric and wins the pickoff metric is not a bad draft — it is a draft with a specific way to win, and the point of scoring it is to name that before minute zero.
Ranking what is left on the board
Scoring a finished draft is retrospective. The harder question is the live one: it is your pick, here is your side, here is theirs, here is what is banned — what should you take?
The suggestion ranking answers exactly that. It scores every hero still available by what it would add to *your* side against *this* opponent, and each suggestion carries its reasons, biggest first, in one of three kinds:
- Trait
- A strength your draft is missing. Independent of what the enemy has taken — this is about your own line-up being complete.
- Counter
- An answer to something the enemy draft has exposed. These only exist once they have picked; against an empty enemy side there are none, which is itself informative about how much of a first pick is a real read.
- Role
- An uncovered position. The unglamorous one, and the one that loses games when ignored.
The score behind each suggestion is marginal — it is what that hero adds to this specific board, not a rating of the hero. It is comparable within one ranking and meaningless across two, which is why it lives in a tooltip and the visible bar is relative to the best option in the same list.
Bans are part of the question. The ban shelf holds a full Captains Mode ban count, and a banned hero leaves the pool the same way a picked one does — because nobody can draft it, regardless of who removed it.
What your opponents keep doing
The other half of draft prep is not about heroes in the abstract — it is about the specific five people you are about to play. Pick and ban trends across your scrims and officials answer the questions that actually come up the night before a series:
- What do they take first phase, and how often does it repeat?
- What gets banned against us, every single time, in every single game?
- Which heroes does their carry reach for when the draft has gone badly for them?
- Which of our comfort picks have they stopped respecting?
The board is a dense list — one line per match, with each side’s picks shown as a hero strip under its team colour — and clicking a line unfolds the full pick and ban sequence. Filters are patch-aware, because a trend from two patches ago is a fact about a game that no longer exists.
When you filter on a hero, that hero keeps its slot in the strip even when the row matched on a ban rather than a pick, so a row that appears in the results always shows you why.
Using it as a drill, not an oracle
The failure mode of any draft tool is treating its output as an answer. A scoring sheet does not know your players’ hero pools, does not know that your position three has been practising one specific hero all week, and does not know that the enemy captain panics in the last ban. It knows what the line-ups contain.
Used as a drill, though, it is genuinely valuable. Replay last week’s draft on the board and try the alternative you argued about afterwards. Score both. Usually you find that the disagreement was about a metric neither of you had named, and naming it is the entire benefit — the number is just what forced the conversation to be specific.