The battle report now reads like the game's own Combat Log — same cards, same colours, same effect names — and still shows the math behind every hit
Version 2.2 redesigns the detailed battle report after the Combat Log the game itself shows, rule for rule, and makes every damage formula start from where the number comes from. It also fixes several rules that we only learned by sending more attacks to the game and comparing every action of the reports. Everything from 2.1 and 2.0 is still here below.
The round-by-round report is now laid out like the Combat Log in the game: one card per action, the attacker and its target side by side, then DMG, Units and Health, then the effects that shaped the hit. The rules were measured on real in-game reports, action by action, not copied from a screenshot: gold marks your units and purple the enemy's, and an effect's dot shows whose numbers it changed — which is why an enemy Revenant's aura shows up gold, on the unit it weakens. Effects use the game's own names (Weakness, Quick Strike, Killer Instinct, Paralyzing Fear, Wailing Rend, Stink Bomb, Death Cheater) and appear in the order the damage is worked out. A Rotmaw overkill gets a second line for the target it spills onto, exactly like the game, and a unit slowed by your Tier-6 keeps its speed pill on every later action it takes part in. There is also a direct See a full report button in the 2.2 welcome note, so you can open a real battle and look for yourself. Checked against 91 real battle reports taken from the game: in 66 of them every single action — effects, colours, order and numbers — matches, and so do 91% of all 1,618 actions; what is left is mostly which of two equally fast units acts first.
The red ⓘ guide next to Find Optimal now has a See a real example button under every single passive — yours and the enemy's — and under the two speed-tie exceptions (Bonewing/Wraith, Bonewing/Spiderling). Click one and it loads that exact army, closes the guide, runs the report and scrolls straight to it, so you watch the rule happen instead of just reading about it. The two order-tie examples are the literal armies from the real battle reports that proved those rules, flip and all.
The math under each hit no longer begins with a finished number. It starts from the unit's attack × how many of them are fighting, then every modifier on top — so 9 * 7 = 63 DMG instead of a bare “63 = 63”. The explanation the game does not give you is still there, just quieter, under the game-style card.
We sent a batch of new attacks to the game and compared each report with our engine action by action. What that showed, now fixed: the Blood Witch's even-round splash is weakened by an enemy Revenant (8 Blood Witches hit for 24, not 28) and adjusted by type for every rear unit it hits; the Necromancer's Soul Harvest grows with every unit group still alive on both sides, not with the dead ones as its description says (up to +120% in the reports we checked); the Vampire Thrall's bonus against slow targets looks at the target's own speed, not the one your Gargoyle slowed; a Rotmaw overkill that runs out of targets in one row carries on into the other, and a display glitch that could misdraw a revived Zombie's numbers when that overkill triggers a revival is gone; the Lich's splash is exactly half of its kill; the Undead Cultist hands out its +10% on every hit it survives, not once per round; and when two units tie on speed the higher tier acts first — except for two pairs the game breaks its own way: Bonewing and Wraith, and Bonewing and a Spiderling, each pinned down only after a dedicated round of attacks built specifically to flip the outcome, including one that disproved our first guess before the real rule came out. The report draws the Blood Witch turn and the Lich splash the way the game does: one action, a line for every unit hit. Measured on our 4,550 confirmed real battles: the model now reproduces 4,534 of them, up from 4,495 — and action by action, it now matches 336 of the 337 real battle reports we've checked it against.
Give the Test Bench your enemy scenarios for a layer and it finds one reusable ally formation that wins every combination — because the ruins re-roll their enemies every few hours and you want a single army that copes with whatever you get. Rank the results three ways: Most free (most 0-loss sets), Lowest average, or Lowest worst single battle. Each result shows its per-scenario cost, opens a full battle report, and can be sent straight to the Dynamic Lab.
Choose how the planner treats the Cultist's random buff: One shot, one kill keeps only formations that wipe the whole Cultist stack in round one (no surprises); Cultist Proof keeps formations whose losses are the same no matter who gets buffed; Allow buffs also considers leaner armies that win while letting the Cultist buff, costed at their worst case. When One-shot hides a formation you expected, the results now explain why and point you to the looser modes.
The layer box used to stop at 101, because that is where the ruins currently end. It now accepts far deeper layers, and the search tuning finishes with an open-ended 91+ band instead of a hard 91–101 one. If Gameforge ever extends the ruins, the planner keeps working from day one — there is nothing here you would have to wait for.
A player reported Zombies returning from their revival at full speed, even though a Tier-6 Gargoyle had already slowed them to a standstill. The revival was resetting the group's speed, which quietly cancelled the Gargoyle's permanent slow. The same bug explained the second oddity in that report — the Tier-4 Banshee striking Skeletons instead of Zombies, because with the Zombies back at full speed the slowed Skeletons had become the slowest target on the field. The slow now carries through the revival, in the battle report and the optimizer alike. Across our 589 confirmed real battles the engine went from 564 to 573 exact matches, and not one battle it already got right changed.
The built-in search schedule was re-measured layer band by layer band and unit count by unit count, against roughly 600 real in-game scenarios. Quick now comes back in under five seconds everywhere it used to stall — layers that took more than fifteen seconds are down to a couple — and Deep stays inside about a minute even at the deepest layers. Wherever a search finished early the spare time was spent scanning more finely rather than pocketed. Measured against the previous schedule, the formation it found was cheaper on 24 layers, identical on 25, and worse on none.
A new Search Settings page lets you steer how "Find Optimal Formation" explores — set the capacity floor, the per-tier count limit and the per-tier scan step, separately for every 10-layer band, every unit-count and for Quick vs Deep. Leave a box empty to use the built-in value. The built-in schedule is now tuned out of the box for faster, leaner results, and you can export the whole table as CSV.
The results list has a Filter & sort by unit panel: keep only formations that must include or must exclude a tier, and sort by any tier's count (most or fewest first). Your choices now stick across searches and page reloads. Nothing is picked for you: the panel starts empty, and "Clear all" puts it straight back. Left alone, equal-cost formations are ranked to prefer higher-tier units and then more unit types. When a layer has thousands of equally cheap formations, what you see is no longer a block of near-twins: the shown set favours formations that use more different unit types and spreads its picks as far apart as it can — and the same result cap now applies in every Cultist mode, so switching modes never silently changes how many you get.
The filter panel carries a badge showing how many filters are active, and dropdowns you have not set are greyed out, so a single glance tells you what is actually on. The results panel folds away when you want the screen back. A banner now warns you when a non-default engine setting is quietly affecting every report and every search, with one button to put it back. Modals, the results header and the "X found" line were reworked to sit properly on a phone.
Running a large batch of tests no longer crawls — the bench paces its screen updates instead of repainting after every single row. Switching between Verify, Mass testing and Find formation is now near-instant with a clear loading indicator, and scrolling or hovering the big results table stays smooth even with hundreds of rows. The table is now virtualised — only the rows you can actually see exist on the page — so a bench holding four thousand tests opens and changes mode as briskly as one holding fifty. You can import enemy formations in bulk or pull them straight from Verify, export predictions as CSV, and every mode carries a short in-app explainer.
The Undead Cultist hands a +10% damage buff to a random enemy the instant it survives an attack. Its buff stacks for every round it stays alive, and it can even buff itself.
Targeting, type advantages, the T2 Ghoul shield, rear-bypass attackers, zombie revival, splash abilities and explosion reflects have all been reviewed and tightened against real battle reports, so predictions line up far better with what actually happens.
On its every-second-round special, your Tier-7 unit no longer makes a normal attack — instead it splashes 25% of its raw attack onto every enemy rearguard unit. Testing against real fights showed this matches the game far better, so it is now the default for both the report and the optimizer.
Predictions are far closer to the real game than before, but the model still isn't perfect — a few rare edge cases and deep-layer interactions can differ from a live fight. We keep tightening it against real reports, so always treat a result as a very good estimate rather than a guarantee. Spotted a mismatch? It helps us improve.
When the Undead Cultist is in the fight, "Find Optimal Formation" shows each suggested army's losses as a best-to-worst range — because the Cultist buffs a random enemy, the real loss sits somewhere in that band. Prefer absolute certainty? Tick 🎯 one-shot the Cultist and the search keeps only armies that kill it with the very first hit it takes — it never gets to buff, so the losses become exact.
The "Find Optimal Formation" search now spreads the work across all of your CPU cores with a persistent worker pool, so it explores far more compositions in less time — even on deep layers.
"Find Optimal Formation" now shows a live count of how many army combinations it has tested, and keeps the final total on screen when it finishes — so you can see exactly how much deeper a Deep search digs than a Quick one.
For players who like to dig into the mechanics, the Dynamic Lab opens a full override editor: change any unit's speed, attack, HP, type, position (vanguard/rearguard), cost or army strength, switch its passive off or make it ignore typing / your shield, edit the revived-Zombie form, retarget or retime the Cultist's buff, and tweak the Zombie's revive & attack timing, damage rounding and attack/target sort order. Every change applies live to both the report and the optimizer.
Build a suite of battles — layer, enemy line-up and your army — and run them all at once. In Verify vs reality mode you add the real in-game losses and the bench tells you whether the engine reproduces them; switch to Mass testing and it simply shows each battle's predicted losses. Inspect any row's full battle log in a click, and import/export whole suites as JSON or CSV.
A red ⓘ button next to "Find Optimal Formation" opens a full, plain-English guide to how a fight is decided — who attacks first, how targets are chosen, the type triangle, and every unit's passive with how the algorithm uses it. No more black box.
Hover any unit image in a battle report (or the override editor) and a rich tooltip appears with the unit's larger portrait, name, full stat chips — speed, attack, HP, position, type, army strength — and its passive.
Because the Cultist buffs a random enemy, a single number can't tell the whole story. Whenever an Undead Cultist is in the fight, the report now automatically checks every enemy it could possibly buff and shows each of your units' losses as a [min–max] range instead of one figure — no toggle to remember. You can even open the dropdown in the Cultist Buff Range card to replay any single buff path and read its exact losses. (No Cultist in the line-up? You just get the single exact result.)