Nine areas. One of them is your problem this month.
Hindsight doesn't grade you on everything at once. Every report lands in one of these areas, sets one target inside it, and measures that target on the games you play afterwards.
The curriculum
These aren't categories we invented for a marketing page — they're the units the product actually classifies against. A report's findings are tagged with one, the assignment binds to a metric inside it, and your Coaching page tracks you area by area.
Good economy is not a big GPM. It's a GPM that's big for the hero you're on, at the point in the game where that hero is supposed to be converting gold into pressure. A position-one Anti-Mage and a position-four Rubick are playing different games, and a single number can't grade both.
Hoarding through a fight
Dying with enough gold for the item that would have won the fight is one of the most expensive habits in the game, and one of the easiest to fix. If you can afford a component, buy it — a courier delivery is faster than a respawn timer.
There's no single right number, and anyone quoting one is quoting it for a core in a lane that went their way. What matters is the number for your position, your hero, and your bracket — a support with 14 last hits at ten may have had an excellent lane, and a mid with 40 may have had a poor one.
Winning the lane and losing the map
Staying to farm a lane you have already won, while the rest of the map is being decided somewhere you aren't, is a way to have great numbers at ten minutes and a losing game at twenty.
Most deaths that feel unlucky aren't. Reconstructed from the replay, they tend to fall into a small number of repeating shapes: alone in ground nobody has vision of, first into a fight that hasn't opened yet, or too far forward for a hero with no way out.
Stepping in on beat one instead of beat three
The fight has an order. Someone commits, the enemy answers, and then the follow-up lands. Arriving during the first beat means eating the answer that was meant for your initiator.
The count is the least interesting thing about your vision. A support who places eleven observers that see nothing has worse vision than one who places six in places the enemy actually walks. Wards are a bet on where the game is about to happen.
Warding where the game was, not where it's going
The most common wasted observer is placed on the objective that just finished. Ward the ground the next fight will be fought over, not the one you're standing on.
Participation is a floor, not a ceiling. Being present for most of your team's fights is the minimum; the question is what you contributed while you were there, and whether you were able to contribute at all.
Spending the ultimate on the fight before the fight
Using a long-cooldown ability to win a skirmish that didn't matter, minutes before the fight that decides the game, is a participation number that flatters you and a contribution that doesn't.
Most itemisation mistakes aren't picking a bad item — they're picking a fine item late. A timing is a promise about when your hero becomes dangerous, and a build that arrives two minutes after the window has closed is a build that did nothing.
Committing to the build you planned in the draft
You picked your items before you knew how the game would go. Fifteen minutes in, you know things you didn't — and carrying on with the original plan ignores all of them.
Mechanics are the most over-blamed part of most players' games and occasionally the real one. Missed abilities, spells cast into magic immunity, an item never used in the fight it was bought for — these are visible in the replay and countable.
Buying the answer and never pressing it
The item that counters their carry only counters them if it comes out. An unused active in a lost fight is a common and very fixable finding.
Good movement is mostly about the cost of being somewhere. Every rotation gives something up, and the players who move well are the ones whose rotations reliably buy more than they cost — not the ones who rotate most.
Not answering the call
Your team commits somewhere, you have a teleport available, and you keep farming. The fight was five-on-four before you decided anything.
Kills are a means. The scoreboard that decides the game is towers, Roshan, and the map control they buy — and a team that converts every won fight into an objective beats a team with a better kill count almost every time.
Contesting an objective a player short
The most expensive objective mistake isn't skipping one. It's starting one while somebody is dead or out of position, so the enemy arrives into a fight you're already losing.
Why one at a time
Being told five things to fix produces a list you read once. Being told one thing, and then shown whether it moved across your next five games, produces a change. That's the entire argument for the assignment loop, and it's why these areas exist as separate units rather than one undifferentiated score.