About Pred Scout

A build-and-counter companion for Predecessor. It turns each hero's kit math and real match data into builds, lane counters, and coaching — and labels every number honestly so you know what's measured and what's still a model.

How builds are made

A simulator scores every viable item set against your kit's ability ratios, cooldowns, and the current patch's gold economy — then ranks them for burst, sustained fights, survivability, and the other things that actually win lanes. That's the Sim Build: a kit-math second opinion that doesn't care what's popular.

Alongside it, Meta Builds show what the field actually wins with — the item sets real players are taking this patch and their win rates. Two opinions: what the math likes, and what the data shows.

What the simulator models

The sim plays out a fight on paper from each kit's real numbers, then scores the build for the engagements its kit wants to win.

What it reads, the engagement types, and the gaps it can't yet model

For a given build it reads:

The kit — every ability's base damage (or heal/shield), its scaling per rank, which stat it scales off (physical or magical power), cooldowns, and any AoE or execute (missing-health) behaviour; plus basic-attack scaling and the hero's per-level base stats across all 18 levels.
The build's stats — physical & magical power, attack speed, crit chance, flat and percent penetration, armor shred, ability haste, health, physical & magical armor, mana, and lifesteal / omnivamp.
Defense & mitigation — damage taken = damage × K / (K + armor), where armor is reduced in order by shred, then percent pen, then flat pen.
Crit, attack speed & haste — average crit multiplier from crit chance, attacks-per-second from base × (1 + attack-speed%), and cooldowns shortened by ability haste.
The patch economy — median gold curves per role, so each item gets a spike minute (when the median player affords it) and the build is scored at the level it's online by.

It then scores that build for each engagement type: one-combo burst vs a squishy, teamfight AoE, skirmish uptime (a 10-second rotation), extended fights (20 seconds vs a bruiser), sustained auto DPS, effective HP vs physical and vs magical damage, and for supports heal/shield output, enchant/peel utility, and drain sustain. Each hero's build is the one that best fits the engagements its kit wants to win.

Gaps we still need to fill

It's a stat-and-damage model, not a full combat AI — so it's honest about its edges:

Unverified constants (the THEORY label): the exact armor/mitigation constant, the ability-haste and attack-speed formulas, the attack-speed cap, the crit multiplier, and whether ability scaling counts only bonus power. These await a practice-mode measurement.
Not yet modeled: crests and consumables; augments without tractable math (the ones we can model are simulated, the rest are listed as field evidence); and a few Eternal majors.
No head-to-head field data — matchups and counters are 1v1 kit-math simulations, not real head-to-head win rates (that data isn't published).
Positioning, ability order, terrain and team composition aren't simulated; the sim assumes standard builds and primary-role income, so a flexed/off-role pick's gold may run leaner than modeled.

That's why every combat number is THEORY until calibration, and why we always show the field's real win rates and builds next to the sim's opinion.

What THEORY means

Some combat constants — the armor/mitigation formula, ability haste, the crit multiplier, the attack-speed cap — haven't been measured directly in-game yet. Anything that rests on them is labeled THEORY until it's confirmed in practice mode.

We never fill an unverified number with a guess. If we're not sure, we say so and lean on the field data instead. Honesty about uncertainty is the whole point.

Where the numbers come from

Hero kits, item stats and play rates come from the current patch's game data and thousands of current-patch matches via pred.gg.

How small samples and evidence deltas are handled

Win rates are pulled toward the average for small samples, so a hero with a handful of games can't fluke its way to the top of a list.

Evidence deltas (an item's win-rate swing) are comparative, not absolute — they carry finished-inventory bias, so read them as "better than the alternative here," not as a guarantee.

How the writing is checked

The explanations, item reasoning, and coaching are written and then machine-checked: every number a sentence cites has to exist in its source data, and an independent reviewer passes back over the copy for anything misleading. Lines that can't be backed up get cut.