Check role coverage first — an uncovered lane matters more than any single matchup — then damage-type balance, and whether the team has frontline, engage, and disengage. Layer in behavioural signals like a duo forcing an unpracticed off-role pick, and re-check all of it after every new pick, since early reads mean less than late ones.
Role coverage first
A jungler locked into a pick that has never jungled, or two people fighting over mid with nobody willing to move, is a bigger problem than any individual champion matchup — and it's the cheapest read in champion select, if also the easiest to miss because it's boring.
Damage type split
Look at how your team's damage breaks down between physical and magic. A comp that's five champions dealing almost entirely one damage type is easy for the enemy to itemise against — one or two defensive items counter the whole team at once, instead of forcing a compromise. It doesn't have to be an even split, but a comp that's completely lopsided one way is giving something up for free. The same logic applies to the enemy team: a full-AD or full-AP enemy comp is a specific, actionable thing to build around, not just a stat to notice.
Frontline, engage, and disengage
Ask three questions about your team's teamfight shape: is there anyone who can walk in front and absorb damage, is there anyone who can start a fight on the team's terms, and is there anyone who can stop the enemy from doing the same to you. A team with five backline, immobile damage dealers has no way to start fights favourably and no way to escape when the enemy starts one on them. You don't need all three roles filled by dedicated picks, but if a comp has none of the three, that's worth knowing before the game starts, not after the third bad fight.
Obvious lane counters
Some matchups are lopsided enough that they're common knowledge — a lane bully into a scaling pick with a weak early game, a ranged laner into a melee pick with no way to close the distance. You don't need to memorise a matchup chart to catch the obvious ones. If a matchup looks bad on its face, it's worth a look at how the lane loser survives to the point where the rest of their kit matters — through jungle pathing, a different starting item, or just playing for scaling rather than lane dominance.
Behavioural signals
The picks themselves aren't the only information in champion select. A duo who insist on an off-role pick they clearly haven't practiced, or someone who locks in a pick right after being told directly it doesn't fit the comp, is telling you something about how the next 25 minutes are going to go. None of this is decisive on its own — people play unfamiliar roles fine all the time — but it's real information, and it's the kind no win-rate tool can see. Combine what you can observe here with a read on the comp's win condition before deciding whether the game is winnable.
Reading it as the picks come in
Champion select is sequential, not a finished list you read all at once. Early picks tell you less — a single pick two roles away from being assigned still leaves most of the comp's shape open. The read gets more reliable as more roles lock in, and it's worth deliberately re-checking role coverage, damage split, and engage/disengage after each new pick rather than forming an opinion in the first few seconds and holding onto it. A comp that looks fine after four picks can look very different after the fifth, especially if that fifth pick was supposed to bring engage or frontline and doesn't. Should I Dodge's own estimate gets more confident the same way — widening rather than narrowing while picks are still missing, not the other way around — see how many picks it takes before its number means much.
It's also worth noticing what's still open, not only what's already picked. If your team is missing frontline with two picks left, that's not yet a problem — it's a constraint on what those two picks should be. Reading champion select well is as much about tracking what the draft still needs as it is about judging what's already there.
What bans tell you
Bans happen before anyone locks a champion, but they're still information. A ban aimed at denying a specific pick from an opponent, rather than removing a generically strong champion, tells you something about what that opponent was expected to play — a scouting report, of a kind, before the game has properly started. It's weak evidence on its own, but combined with what actually gets picked afterward, it can round out the picture of what the enemy team was planning before it committed to it.
What this app adds
Should I Dodge automates one narrow piece of this: matching each pick against aggregate win rates and estimating a probability. Before it can do that, it has to read the screenshot correctly — see why a champ-select screenshot can get misread — and guess which role each unlabelled enemy pick is playing, since champion select never actually says. It can't see any of the behavioural signals above, and it doesn't try to. Use it as one input, not the whole read — see how it's built for exactly what it does and doesn't account for.