A massive data-fetching feedback Analysis of each character's skill-tree update, based form various platform.

[NNOUS]

After I have read seveal internet debate i was kinda curious. therefore I scored every skill tree by what ACTUALLY changes your gameplay.

I will note here i have proof readed everybits of those declears so IT IS NOT SLOP!!!!
 

The setup We all throw around "this tree is filler" / "this one's amazing," so I tried to actually measure it across all 12 reworked survivors. I ended up doing it twice, and the gap between the two passes is the whole point of this post. here is the methods of how it turned to 💩

 

Spoiler

---

Attempt #1 — just count the filler First I went node-by-node and tagged each skill as either:

• A = a new ability / item / mechanic
• B = pure stat-stacking (+5% / +10% / +15% type nodes)
• C = conditional / minor passive

Then I ranked trees by their % of "A" nodes — i.e. how much of the tree is "real new stuff" vs padding. That gave me this:
…and it immediately felt wrong. Wendy and Wilson at the top? Winona dead last? Anyone who's actually played them knows that's backwards.

---

Why it was wrong The mistake: counting "nominally new" over-credits gimmicks. A node the wiki calls a "new ability" but that nobody ever uses — Torch Toss, the Abigail "commands," beard storage — is still filler wearing a costume. Meanwhile a tree like Winona's looks like a wall of boring +% nodes, but those numbers are what make her catapults actually function. So node-counting measures "how many new labels are stuck on," not "how many of them cash out in real play."

---

Attempt #2 — score by actual impact + reception So I re-classified every node into four buckets instead of three:

• Real keystone — genuinely changes how you play, and people use it
• Gimmick — nominally new, but the experience doesn't change / nobody uses it
• Scaffolding filler — a stat node, but one that makes a keystone usable
• Dead filler — pure stacking that props up nothing

Then I scored each tree /10 on real gameplay change + community reception, not label-count

Winona jumps from last to #3; Wendy falls from #1 to #10; Wilson from #4 to rock bottom. Why:

• Winona — lowest share of "new" nodes, but the few that matter (Handy Remote, Energy-Saver, remote Strikes) turn her from a broken half-engineer into a real turret commander. The +% triples everyone calls filler are actually scaffolding that makes the turrets work.
• Wendy — highest share of "new" nodes, but the headline additions (the three Abigail commands) are widely considered redundant gimmicks, and the tree never touches her real combat weakness. Max novelty, minimum payoff.
• Wilson — transmutation is a convenience engine, torch toss + beard storage are flavor. Tons of "new," almost zero change to how you play.

---

Takeaway What separates a great tree from a padded one isn't node count. It's:

1. how many "new" nodes actually cash out into real play (call it A-grade purity), and
2. whether the filler at least props up a keystone instead of just inflating numbers.

By that lens Winona / Wortox / Wormwood / WX-78 / Walter are the strong ones; Wendy / Wilson / Wolfgang are where the padding shows.

---

Where the data came from / how I scored

• Node lists & mechanics: the wiki (dontstarve.wiki.gg) cross-checked against the dstcraft skill-tree simulator.
• Reception / "do people actually use it": these forums (skill-tree feedback threads + the community polls, e.g. the Walter V2 poll and the Woodie mastery poll), Steam discussions & guides, tier-list write-ups, and Chinese guides (GamerSky, huiji wiki, Bilibili) for a second community's read.
• Honesty bit: the /10 numbers are my synthesis, not hard telemetry, and the real-vs-gimmick line has subjective edges (I leaned on "do experienced players actually slot this," not on theory). Reddit was hard to pull from, so the English reception is mostly Klei-forum + Steam.

So I re-classified every node into FOUR buckets and scored by real gameplay change + community reception instead of label-count:

• Real-A (keystone) — a new mechanic that genuinely changes how you play, and people use it
• Fake-A (gimmick) — nominally new, but the experience doesn't change / nobody uses it
• Support-B — a stat node, but one that makes a keystone actually usable
• Dead-B — pure stacking that props up nothing

Two metrics fall out: A-purity = realA / (realA + fakeA), and Bloat = (fakeA + deadB) / total nodes.
Data sources: node lists from the wiki (dontstarve.wiki.gg) cross-checked against the dstcraft skill-tree simulator; reception pulled from these forums (feedback threads + the community polls, e.g. the Walter V2 poll and the Woodie mastery poll), Steam discussions & guides, tier-list write-ups, and Chinese guides (GamerSky, huiji wiki, Bilibili) for a second community's read. Honesty bit: the /10 scores and the bucket counts are my own judgment (±1–2 per tree), not telemetry, and the real-vs-gimmick line has subjective edges. Reddit was hard to pull from.

---

Conclusion from Table 1: Bloat% is the cleanest predictor of where a tree lands — the top four sit at 21–35% bloat, the bottom four at 48–65%. A-purity catches the "fake-A trap" (Wilson 21%, Wendy 40% — lots of "new" nodes that don't cash out). But purity alone lies: Wolfgang's 67%* looks fine until you see it's 67% of only 3 A-nodes, drowned by 15 dead-filler. So a good tree needs enough real keystones AND low bloat — neither metric works on its own.

---

 

arguebly I cant be sure if wilson SHOULD even get a skill tree cuz he is the basic character. For I do like torch tossing but 6 point for that only isnt gonna cut it. To boost that this tablet isnt just pure data, there is the summery of keynote sheet for each character summed form the InternetTM

 

Spoiler

 

those are comments / posts summed form klei, steam, several chinese platform (wegame) and forums. 

It very clear that players are more favorable to new contents than buff/nerf to their stats. (might as well just stop making suggesting posts saying nerf who buff who okay?) new core mechanical is likely to be well received.

I am curious to see how people respond to this finding. I am also aware people would call this AI slop anyway XD
however. the analitical model i have injected is my own. 

 

 

 

 

 

Edited 7 hours ago by NNOUS
remove imagines

[AliceShiki]

5 hours ago, NNOUS said:

I am curious to see how people respond to this finding. I am also aware people would call this AI slop anyway XD

I think it's kinda pointless to call it "massive data fetching" without putting any numbers on how many player opinions you got during the research.

Because right now, it seems like you were only looking at highly enfranchised player opinions based on what platforms you said you looked at.

Also... Trying to use old polls that talk about things more broadly, as opposed to using polls made for the specific purpose of gathering data looking at each individual node would naturally lead to a pretty strange result IMO.

So, overall... I feel like there are no findings there to speak about? It feels far too subjective and the method to obtain the data looks quite questionable.