✕

My roster

Enter your player tag and the site hides everything you cannot play anyway: brawlers you are missing, and those below power 9 — below 11 from Mythic upwards. The tag goes to the official API once and then stays in your browser. It is neither stored nor logged here, and there is no account.

Load

You can find your tag in the game, under your name in your profile.

← Overview

How the draft helper calculates

For each brawler the helper shows a percentage and a few small increments like +11.3. This page says what each of them means — and which of the numbers you may take literally. Where the data comes from in the first place is on Where the numbers come from.

The row, taken apart

A real example, pulled from the database just now: Crystal Arcade (Gem Grab, 120,214 battles).

• 1The range: 58.6 to 60.6%. How Bolt does on Crystal Arcade in general — measured across 8,961 games, regardless of who else is playing. In the middle sits the measured win rate, 59.8% — the same number as on the map page. Left and right are the bounds between which the true value lies with 95% confidence: from 8,961 games the value cannot be pinned down more precisely, and the more games come in, the closer they move together. The calculation uses the minimum, 58.6 — a pick that is only good in the best case is no pick in a draft. This value is also the starting value of the points calculation: 58.6 points.
• 2An opponent already locked in. Amber on the enemy team is worth +6.8 points: against Amber things go better than is usual for Bolt.
• 3A team-mate of your own. The same calculation, just on your side — together with Belle it is +14.0 points.
• 4The big number: 85.6 P. Points, not percent. Simply everything above added up — 58.6 + 6.8 + 6.2 + 14.0 = 85.6. The higher the better the pick; it does not claim more than that. Why there is no percent sign on it is what the whole section below is about.

What an increment means

It is counted from real battles: how often does this brawler win when that one is on the enemy team? That is compared with what everyone else manages against the same opponent. What remains is the increment.

For Bolt against Amber it looks like this: 73.2% of those battles are measured as wins. His own average across all opponents is 66.5%, so that is +6.7 — but against Amber the whole field sits at -0.1. What remains is +6.8.

Both subtractions are the whole trick, and they answer two different questions. His own average has to go, because otherwise a strong brawler would show a big number everywhere and you could no longer tell whether it is down to the opposite number or simply to the brawler being good. And the field has to go, because some opponents are good against everyone: against those even the best counter still sits in the minus, and without the second subtraction that would read like a weakness. What remains is the question drafting is actually about: does this brawler cope with that one better than the rest of the field does?

So 0 does not mean “even”, it means “goes the same way it goes for everyone else”. And a minus does not mean you lose the duel — it means you come through it slightly worse than average. Against one of the hard opponents a row can carry a minus and still be the best answer on the board.

Watch for “with” and “vs”. The same two names give a completely different number depending on the side — 8-Bit with Nori on your own team is −2.6 in knockout, 8-Bit vs Nori on the other side is +4.3. That is why every tag says where the named brawler stands.

When the enemy is up, the list flips

Then it no longer sorts by your benefit but by his — it answers “what is he likely to take”, not “what would be good for me”. You can tell by the heading above it and by the rows carrying a border in the enemy colour. Your own Nori appears in those rows as “vs Nori”, because from the enemy's point of view that is exactly what he is.

The reason is the click: tapping a row always fills the slot whose turn it is — and on the enemy's turn that is his. A list that enters the enemy pick but sorts by your advantage would be the most dangerous display in the whole helper.

A pairing needs at least 60 shared battles, otherwise nothing is shown at all. Better no number than one from four games.

Why it shows points and not percent

Because the sum of base and increments is not a win rate. It only looks like one, and that is exactly the trap: read as a percentage it is wildly too optimistic. It can also go past 100 — at which point it is obvious that these cannot be percent.

The reason: three opponents and two team-mates partly explain the same thing. If your brawler is good against close-range brawlers and two of them are on the enemy team, you count the same advantage twice. Measured against 400,000 real battles, only 22% of an accumulated point actually comes through (the measurement).

So why does the list not simply calibrate the number? That was tried and dropped again. The calibration is measured on whole line-ups, on six brawlers — applied to a single row it squashes everything together. On an empty draft, first and sixth place ended up only 1.3 instead of 6.1 points apart, everything looked like 51%, and underneath it the base claimed a higher number than the recommendation above it. Arithmetically correct, useless for picking.

Points solve both: they do not look like a win chance, and they keep the full spread. Percent now appears only where it is true.

Rule of thumb if you do want to convert points into percent: take a fifth of them and add that to 50%. An edge of 45 points becomes roughly a 60% win chance — assuming the other side is average.

And the “answer” tag?

It says what the enemy would take in response to this pick and how that pairing goes for you. What is computed is not his hardest counter but his best overall pick — the hardest counter often costs him so much of his own strength that he never plays it. What is shown there is the pick that actually comes.

It does not go into the points. The points describe the board as it stands — locked picks, measured pairings. The answer, by contrast, is a guess about a move that does not exist yet. If both were in one number, a low score would no longer tell you whether the brawler is weak or merely easy to counter — and those are two very different reasons not to pick him. Whether the counter puts you off the pick is your decision.

If it shows (±0), the enemy has nothing special against you. On the first pick that is the most valuable piece of information in the whole row. A plus means his best pick actually suits you.

When it appears and when it does not

Drafting does not alternate but runs in double turns: if you start, the order is you · enemy · enemy · you · you · enemy; if the enemy starts, it is exactly the other way round. The tag only appears when the enemy moves immediately after you.

How does the helper know who starts? From your first click. Tap an enemy box at the top and the enemy has first pick; tap one of your own and you do — and the list immediately takes the answer into account, before the first brawler is even in. That is why there is no switch for it.

Where it says missing, either you are up again straight away — the enemy cannot answer before your next brawler is in — or it is the very last pick of the draft, after which nothing comes at all. In the latter case that is correct; in the former you are missing a warning that could exist, because the enemy can of course still respond later.

And where it only tells half the truth

Where two enemy picks follow — on your first pick, if you start. The tag then names the best single counter-pick, but the enemy places two brawlers before you are up again. So it covers only the first of them.

A second reason not to take the number there too seriously: the earlier in the draft, the more is still open. On the first pick five brawlers are unplaced, and any projection across five moves is a guess. The tag is most reliable on your last pick with exactly one enemy pick after it — there it computes exactly what is still to come.

What the numbers do not say

Why it is computed and not looked up. The value on the left of the range is the lower bound: the count is pulled towards the middle of the field and the safety discount for the sample is subtracted from it. Without that, an outlier from 40 battles leads the list — by raw win rate, three brawlers with around 50 games would occupy the top three places on Out in the Open in Mythic, ahead of Pearl and Brock. How far the minimum sits below the measurement depends purely on the number of games: Bo on Dueling Beetles in Mythic from 11,910 games comes to 55.8% measured and 54.8% computed; the same Bo in Gold from 108 games to 55.6% measured and 46.1% computed. Sorting uses the same bound as the tier list and the tables.

When the range is not shown. On a handful of rows the measured win rate lies outside its own range — the shrinking pulls thinly played outliers so far towards the middle of the field that the measurement no longer fits between the bounds. A range whose middle lies outside it would be visible nonsense; such rows therefore show the two named numbers base and map again. Of the 324 rows that appear in a recommendation list in Mythic at all, this affects three.