are the army lists for the event around somewhere (as I guess this isn’t in the Companion)?
In the pre-tournament excel report, last tab, there should be a data dump of the lists. There were companion issues so I had to convert from txt files rather than the standard pdf. Several people did not appear to submit lists so it’s a partial set.
1 Like
In the meantime, I’m trying to build out the 4e dataset. If anyone sees tournaments that have (1) lists and (2) results (ideally either Mantic Companion or tabletop.to for Australia, but I can work with most datasets), please let me know.
2 Likes
Too late, tomorrow I will play a tournament with elves and the lists are final already. 
1 Like
I’ve been busy parsing and adding in tournament data to beef up the 4e dataset. I’m up to about 1100 games:
Kings of War 4th Edition Meta Analysis – Insights from 1,100 Real Games – Data & Dice
Still too early to draw real conclusions. At a high level, balance looks pretty good between factions, and it seems to me that archetype matters more than faction. Notwithstanding blogs, podcast, and fanatics obsessing over lists, I still don’t see many connections between list-building and on-table performance.
You can see the full data project here at kow-dataset.web.app.
7 Likes
I mean, Rats and Nature at the top entirely checks out with my experience of the meta 
I’m surprised by Abyss being so high (I’ve only seen Bag of Hammers style armies do well) and Ogres and Nightstalkers so low, tho I wonder if the latter case is a space marine situation, where they have good rules but so many people play them that results get dragged down due to varying player skill (they do have two of the highest N values). Undead dragging the bottom makes me sad, because I think the faction has a lot going for it, but you’re right, their generic archetype isn’t great (I think Bag of Hammers is a solid option but it has poor core unlocks for that approach). I also think Orcs are going to become more of a problem soon, as people paint new Orc armies to take advantage of Thonaar’s Rage being borderline broken …
2 Likes
no no Undead love having Commands for the Shambling Core only, that’s fine for another year, its fine.
I think Revenants are pretty good, but yeah its just going to end up getting Alpha Strike’d off the table most times.
1 Like
I’ve played against Abyss and they certainly feel strong.
Combat is slower, so regeneration comes into play more and they have units that hit pretty hard. Molochs and horsemen (TC3) are not nice.
My wights didn’t hit as hard as they should have and my opponent managed to avoid multi-charges well, tough.
Nightstalkers have the same command, but butchers are core while zombie trolls are not. 
5 Likes
Nightstalkers get the Lifeleech on Butchers and Fiends, but Undead can’t Command their Soul Reavers … oh and of course the Pathfinder Command works on everything. That’s nice.
Who signed off on this, and can I get what they’re smoking?
Is there a definition of the “archetypes” somewhere?
[Updated]: Nevermind, found it!
1 Like
Also, if it helps some data from this weekend’s Loaded Dice GT in the US.
Lists:
1 Like
Thanks for sharing, I was hoping to see the full results! As nice as all the top finishers are as dudes, I’m going to give my golf clap to Ed’s Ogres in third place. Rats and Goblins in one and two is less interesting to me 
Looks like Ed took highest attrition too 
EDIT: Lists are here if y’all need them >> 2026-Loaded Dice-Lists – Google Drive
I was just typing out a response and you beat me to it. If you have other ways to define lists, please let me know. I’m not thrilled with the current approach, but I haven’t seen any better suggestions out there. It’s based on percentiles and not on absolutes because the dataset is still relatively small but I’d love to dial it in more.
Perfect. Thanks. Let me see how I can go about parsing that and incorporating the data.
If you’d like another recent data set, March of Death happened Feb 28 + Mar 1 up in Brantford, ON. The lists and results are here: 2026-March of Death-Lists – Google Drive
Worth noting that third place Ben T’s 2500 point list was generated by rolling on a series of charts 
2 Likes
2 Likes
For various reasons, Google docs spreadsheets shared publicly aren’t something that can be readily scraped (I can screenshot and then feed to AI and hope it gets all the numbers right, but that’s a bit dicey). Are you able to email it to me? Would love to incorporate the data but can’t in this format.
For those that didn’t dig in, Ratkin won this tournament too
At the start of 4th edition, I didn’t have enough list data to define archetypes from the real tournament environment, so I used relative baselines. In practice that meant broad, percentile-style rules inside the current pool of lists: top-end shooting became Gun Line, very fast high-pressure lists became Alpha Strike, very durable lists became Grind, high unit-count lists became Swarm/Spam, and everything else fell into more general buckets. That worked well enough as a first pass, but it was always meant to be provisional because the labels were based on “compared to this sample” rather than “what this list actually is.”
Now that I have a meaningful historical list database, I updated the classifier to use absolute values normalized to a 2300-point standard and classify lists from observed build traits instead of relative percentiles. The system looks at a few independent style axes like ranged pressure, tempo, durability, board control, and concentration, using fixed bands from the actual dataset. So instead of saying “top 20% shooting is a gunline,” I can now say “40+ ranged attacks at 2300 is a gunline profile,” “7.5+ average speed with enough damage is alpha strike,” “400+ shots-to-six-nerve is grind,” and so on. From those axes I assign a primary archetype, but I also keep style tags, because a lot of real lists are hybrids. That gives me something much better for ML too, since the model can learn from the actual structural traits of winning lists rather than only from one broad label.
2 Likes