Reading a pick at a glance: Team Value

Here is the idea I kept coming back to. A draft is not five separate popularity contests. It is a checklist of jobs that a team needs done to win. Someone has to lock a target in place so the fight can start. Someone has to save a teammate who gets caught out. Someone has to catch a hero who wants to run. Someone has to break buildings once you win a fight, and someone has to clear waves so you are not drowning in creeps. If every job has an owner, you have a team. If two of your picks own the same job and nobody owns another, you have a hole, and the enemy will find it.

This reframes what "strong" means. A strong hero that duplicates a job you already have is worth less than an average hero that fills a hole you have left open. That is uncomfortable, because the flashy pick usually feels better. But games are lost on the missing job, not the doubled-up one. I wanted the tool to nudge me toward the hole instead of the highlight reel.

So every hero now shows its Team Value: the capabilities it brings, expressed as plain tags like lockdown, save, catch, siege, and waveclear. These show up on the hero draft guide and inside the draft itself, so the jobs a pick covers are legible right where you are choosing.

The part I care about most is where the tags come from. They are read from each hero's actual abilities on the live patch, not hand-typed by me from memory. A real stun reads as lockdown. An ally heal reads as a save. An innate gap closer reads as catch. That distinction matters because my gut is unreliable and it goes stale. Patches change what a hero does, and a list I maintain by hand drifts out of date the moment I look away. Deriving the tags from the abilities keeps the answer honest, and it means the tags describe the hero in front of you rather than the hero I remember from a season ago.

Tags tell you the jobs, but during a draft you also want the shape of a hero fast. So hovering a hero now opens a quick summary card, and on touch you tap to get the same thing. The card gives you star ratings for teamfight, scaling, pickoff, and push, the hero's early, mid, and late power curve, and the Team Value tags all in one place.

The point is speed. Drafting is a stream of small decisions under a clock, and the last thing you need is to leave the draft to go read a page about a hero. The power curve is the piece people underrate. A hero can be excellent and still be wrong for a draft that needs to end the game by minute twenty five, because a lineup that only comes online late is a lineup that has to survive to get there. Seeing early, mid, and late side by side makes that tradeoff something you can weigh in a glance instead of a thing you regret later.

There was an awkward inconsistency I had let sit for a while. The AI captain in the draft and the simulator were reasoning from a looser picture of heroes than the one I was now putting in front of players. So the AI would sometimes leave an obvious gap, like fielding a team with nobody who could reliably lock a target down, which is exactly the mistake the tags exist to prevent.

In 0.7.1 the simulator and the AI captain read the same verified hero capabilities that you see. That single source of truth means draft coherence and AI picks now account for what each hero truly brings, so the AI reliably fields lockdown and fills its missing pieces instead of leaving holes for you to walk through. It is a small thing to describe and it was overdue. An assistant that judges heroes by different data than it shows you is not really assisting, it is just guessing next to you, and I would rather it be wrong in the same honest way I am.

One smaller change came from the fact that I had crammed too much into too little time. The strategy phase timer now runs 120 seconds, up from 60, because I had added item builds to that phase and 60 seconds was not enough to set roles and builds both without feeling rushed. Continue still advances early, so if you are fast you are never held hostage by the clock.

This is the least glamorous line in the release and I am including it on purpose, because the honest version of a changelog admits when an earlier decision was simply too tight. I got the budget wrong, players felt it, and doubling the window was the fix.

I want to be clear about the edges. Reading capabilities from abilities is a real improvement over my gut, but it is still a model, and models flatten things. A tag tells you a hero can lock a target down. It does not tell you the cast range, the cooldown, the mana cost, or whether the hero can survive long enough to use it in the fight that matters. Two heroes can both wear the lockdown tag and be worlds apart in practice. The card is a starting read, not a verdict.

The match engine underneath all of this is an approximation too, and it improves in steps, and I get things wrong and fix them when reports come in. A lot of this release exists because players told me where the draft felt blind, which is the only reason I knew where to look. If you draft with these tags and the tool tells you a hero fills a hole it plainly does not, tell me. That feedback is what turns a rough model into a slightly less rough one, and it is genuinely how most of this got built.