Disclosure: This post is reader-powered and contains affiliate links. If you click through and make a purchase, I may earn a commission at no additional cost to you.

Why I simulate instead of calculating averages

Expected-value maths works for a single volley, but combat is not linear. Break tests, Killing Blow, ward saves, losing Frenzy, pursuit rolls and Fall Back in Good Order all create outcomes that an average hides. So I roll every die as an independent D6 and play every combat through to an end state.

My rules sources

• The Warhammer: The Old World Core Rulebook, with the 2026 Rules & Errata applied wherever they differ.
• Ravening Hordes and Forces of Fantasy.
• The Arcane Journals for Warriors of Chaos, Armies of Grand Cathay and The Razing of Westerland. The Razing of Westerland replaces the older Marauder entry with Chaotic Cults.
• The Battle March General’s Companion.
• The Skaven Renegades draft v1.5.2.2, an unofficial fan list.

Sample sizes

My studies use very different sample sizes, so every results table on the site carries its own sample-size tag.

How sure can you be of a win rate?

Next to every win-rate table I show a ± figure: the 95% margin of error for a single win rate from that many fights. It is widest for results near 50%, which is the figure I quote to stay on the safe side.

So a 7/10 result could easily turn into anything from about 40% to 90% over a longer run, while a 70% from 2,000 fights is very unlikely to be off by more than 2 points. That only covers dice luck: it says nothing about how closely my model matches a real table, which is the bigger question for every study on this site.

How I set up the shooting tests

• I used six targets: six Knights of the Realm, ten Ironbreakers, 20 Empire State Troops, 20 unarmoured Skaven Dregs, three Rat Ogres and 20 unarmoured Empire Archers deployed as Skirmishers.
• The Skirmisher target carries the −1 To Hit for Unit Strength 1 models. That penalty is separate from cover, so Trueflight does not remove it.
• I deployed ranked shooters in one rank, and gave skirmishing shooters the legal arrangement that got the most models shooting, then the most within half range.
• Every unit fired on flat ground, standing still, with no cover or magic. For long range I placed the shooter at 75% of the weapon’s maximum range, and for short range at 25%. Each model checked its own arc, line of sight and range band.
• Every volley went into a fresh target. I capped effective wounds at the target’s total wounds; against Rat Ogres, three wounds remove one model.
• Units took Panic tests only when casualties exceeded 25% of their starting strength and they survived.
• Anyone can reproduce the rolls: I used random seed 5447218727185184547 for the main batch and 202609242042 for the five-unit follow-up, which also fired after a legal move to the same position.
• My later shooting batches each had their own seed: 2026092317 for the six Orc, Goblin, Beastmen and Tomb Kings units, 17104442927166063737 for the 32,400-volley movement and cover test at 6, 10, 16, 22 and 28 inches, 12053321814676242546 for the 4,560 war machine shots and 2026092333 for Outriders vs Pistoliers.

How I set up the combat tests

• I built units into a narrow points band, roughly 115–153 points, so I could compare roles without rewarding the bigger spend.
• For charging tests I assumed a successful frontal charge of at least 3 inches and the full +3 Initiative charge bonus. I did not model failed charges, which favours the charger.
• The winner always follows up or pursues. Break tests use both the natural and the modified 2D6 roll, including the rule that turns Fall Back in Good Order into a Break when the winner has more than twice the loser’s Unit Strength.
• When a pursuer catches a unit that Fell Back in Good Order, it re-engages and counts as charging next turn, so charge-only rules come back.
• I left out shooting, magic, terrain, nearby generals and battle standards, objectives and third units. I did not model table edges.
• My Chosen and Forsaken test used a newer combat engine with 2,000 fights per matchup (seed 20260924) and a 12-round cap. I counted the 43 fights that reached the cap as not won.
• My Tomb Kings test used the same engine and opponents, with every build fighting all eight opponents both charging and receiving: 224,000 combats. It left out Arise!, nearby Generals and Royal Host rules on purpose.
• My Wolves of the Sea test ran 36 Marauder and Horsemen builds at fair points against six opponents, both ways, at 3,000 fights per matchup: 1,296,000 combats. After re-checking the Knights of the Realm (barding, Blessing, First Charge) and Rat Ogres (WS4, heavy armour, Frenzy, Warband, Warpstone attacks), I reran those matchups at 5,000 fights each. Fear, full Lance geometry and pursuit flow are still simplified.
• My dedicated Chaos Marauder test ran 16 practical builds, each with the light armour Marauders carry as standard, against six opponents, charging and receiving, at 10,000 complete combats per matchup: 1,920,000 combats. I split every matchup into two separate 5,000-combat passes. It applied the 2026 flail, Parry and Press of Battle rules, weapon switching for flail and shield units, and Fear where the Fear-causing unit had the higher Unit Strength.

2026 rules that changed my results

• Frenzy: the +1 Attack applies only on a turn the model charged, or the turn after a follow-up. Mounts do not get it, and losing a round removes Frenzy.
• Press of Battle: regular and heavy infantry that did not charge fight two ranks deep.
• Close Order: the +1 Combat Result needs Unit Strength 10 or more.
• Massed Infantry: the side with greater Unit Strength gets +1 Combat Result.
• Flails: S+2 and Armour Bane apply only against an enemy the wielder charged that turn.
• Great weapons: Strike Last sets Initiative to 1 before charge modifiers, so a charging great-weapon unit can still rise above I1.

Rules readings I want to re-check

• I allowed Clanrat spears with Press of Battle to add a supporting rank when receiving a frontal charge.
• I applied Huscarl Furious Charge to the horses’ attacks too, on the reading that cavalry special rules transfer to the mount unless the rules exclude it.
• I let Ironbreaker Stubborn and Shieldwall interact: Stubborn turns the first lost combat into Fall Back in Good Order, then Shieldwall can turn that into Give Ground.
• I applied the Battle March 20-model unit cap without re-checking the General’s Companion wording.

If an FAQ changes any of these, I will re-run the affected results and update the date at the top of each page.