LZLZL/Prediction markets/Kill file
FREEJUDGE IT E · MarketsKill file

Weather LP
about −$210, start to finish

2026-08-21 · the standard specimen of copying something that had already died

This line's cause of death is highly transferable: I copied a wallet that was performing well, and by the time I took it apart it had already stopped working. Tuition: about $210 — the cheapest lesson on this site, and the rule it produced was applied to every line afterwards.

1What it was betting on

Weather offers an appealing combination: the reward pool is real (paid on-chain at a fixed daily time, publicly verifiable) and the category's participants are relatively unsophisticated. So "post quotes and collect liquidity rewards" looks like a viable business.

I found a wallet performing well at the time, took its approach apart, and built a line copying it.

2Four stages

StageWhat happened
1hard stop triggered on day one
2redesigned, ran again
3stopped early
4replay returned FAIL → archived

Stage 4 is worth noting: a replay was run after the line stopped. "It lost money live" and "the method does not work" are different claims — live losses can come from execution, timing or luck. Only a failing replay licenses "this road is closed".

3★ The actual cause of death

The wallet I copied had already stopped working

When I took it apart I was reading its past results. Neither leaderboards nor historical records tell you whether an approach still works today. July's star was no longer valid in August.

This was not an execution problem or bad luck — it was a question I omitted at the verification stage: "is it still making money now?"

Expanded in leaderboards are lagging indicators.

4Two truths about the category

Conclusion
The poolreal. Paid daily, enumerable on-chain
Ladder arbitragenot real. See below

The common weather shape is a temperature ladder: a day's high cut into a dozen mutually exclusive, exhaustive bands — exactly one must be true, so the prices sum to $1.

Which makes "they add to 0.96, buy the set for four free cents" look obvious. I checked it, and it does not hold — falsified twice, independently.

5Why collecting the rewards is hard

The core tension is in the scoring formula: rewards are scored on distance from mid, with quadratic decay and zero beyond the board's maximum spread — the programme is built to buy tight quotes.

Quoting atRewardBad-fill $
join best×1×1
−2¢×0.26×0.29

So "quote deeper, stay safe, still collect" is structurally impossible: the safety is genuine, and you are paying three-quarters of the reward for it. Netting those two is the only way to know whether it pays — see LP rewards.

6Four rules this line produced

RuleFrom
Before copying, ask "is it still making money now?"the primary cause of death
Replay after stopping; do not conclude from live results alonestage 4
Size subsidy-dependent strategies as if the subsidy could be zeroit was 100% dependent on the pool
A real pool is not your share of itsplit by score; more participants, less each

7One line to keep

This line did not die of "wrong method". It died of "I verified the method and not whether it still works." What $210 bought was one sentence: before copying an approach, confirm it is still alive.

EvidenceCheck it yourself

Final about −$210, archived.
Sequence hard stop on day one → redesign → early stop → replay FAIL → archived.
Primary cause the wallet used as a reference had itself already ceased to work by the time it was analysed.
The pool is real batched transfers at a fixed daily time, enumerable on-chain. See the reward pool.
Scoring and step-back cost quadratic decay from mid, zero beyond maximum spread; stepping back 2¢ gives reward ×0.26 and bad fills ×0.29.
Ladder shape mutually exclusive and exhaustive multi-outcome boards whose prices must sum to $1.
Checked 2026-07 to 2026-08.

NextWhere to go

F · METHOD
Leaderboards are lagging indicators
B · REBATES
LP rewards: the scoring formula
C · ARBITRAGE
Ladder arbitrage: falsified twice
E · MARKETS
The most expensive line here
This is an educational and research record. It is not investment advice, promises no returns, and offers no personalised trading recommendations. Rules and API behaviour are per the official documentation; this page states when it was checked and both can change without notice. Prediction markets are restricted or unavailable in some jurisdictions — confirm your own before taking part.