LZLZL/Prediction markets/Mechanics
FREEDO IT A · MechanicsResolution

Resolution sources
differ by category

2026-08-21 · this one cannot be inferred, only read

On one venue, sports, crypto, politics and weather use four entirely different resolution mechanisms — and markets inside one category can differ too. "My last market resolved this way so this one probably will" is one of the more expensive assumptions available here. I have paid for it.

1Four mechanisms

CategoryDecided byMain risk
Crypto pricea price oracle reading by rulewhich price, which timestamp — often unintuitive
Sportsofficial results plus human submissionpostponement, abandonment, extra time, later reversals
Politics / eventshuman submission plus disputewording: what counts as "officially announced"
Weathera named station's official readingwhich station, which period, do revisions count

Their failure modes are unrelated, so experience in one transfers almost not at all. That is the entire reason "read it per category" is a rule rather than advice.

2Crypto: everything is in the reading rule

The most objective-looking category is the easiest to misjudge, because "did BTC rise over this period" contains three words with multiple legitimate readings:

WordCould mean
"price"one exchange's trades? a median of several? an oracle aggregate?
"over this period"which second? the close? or an average of the final stretch?
"rise"against the open? against the previous close? strictly greater, or greater-or-equal?
⚠ Measured: two reasonable readings differed by 10.1%

Mine used exchange spot close vs open; the market used the oracle's final-60-second time-weighted average vs open. Both are defensible.

Judging all 2,051 windows under both: 207 disagreed — a whole-sample unconditional mislabel rate of 10.1% — and the disagreement had a direction (125:82 my way).

The consequence is not "slightly less profit", it is the research conclusion being void. Full account in the broken ruler.

A rule of thumbthe shorter the tenor, the worse this gets. The final 60 seconds is 20% of a five-minute window and 0.4% of a four-hour one — audited across four tenors, the mislabel rate falls from 10.1% to essentially zero, tracking that ratio exactly.

3Sports and events: the edge cases are the point

A normal match resolves without argument. What is worth reading is the abnormal:

CaseConfirm
Postponedcancelled and refunded? or held open until it is replayed?
Abandoned mid-matchsettled on the score at abandonment, or voided?
Extra time / penaltiesincluded in the final result or not
Reversed after the factwhich moment's official result governs
An event "happening"announced? in force? completed?
⚠ I ate the postponement case

A position of mine hit a postponed fixture. The market was not cancelled — it stayed open awaiting the replay. Meanwhile the bid side collapsed to 0.01, so a hundred-odd shares could be neither sold nor settled. The capital was simply locked until the fixture was played.

The lesson is not "avoid postponable fixtures". It is: know how long your money can be trapped in the worst case before you bet. "Postponed means refunded" is a natural assumption that frequently does not hold.

4Why you cannot infer

Looks inferableActually
"one venue, so one rulebook"no — it varies by category and sometimes by market
"one category, so the same rule"not necessarily; the source or timestamp may differ
"the displayed price is the settling price"no. The display is a market price; settlement comes from elsewhere
"it worked this way last time"terms get revised, and you are not notified

5A checklist worth copying

#Question
1Which specific source decides? (not "official" — which official)
2Which timestamp? An instant, or an average over a stretch?
3How do edge cases resolve? (exact ties, postponement, cancellation, revisions)
4Can I query that source myself? If not, you cannot verify your own outcome
5Worst case, how long is my capital locked?

Number 4 gets skipped most often and decides something important: whether you can independently tell that a settlement was correct. If you cannot reach the source, you can only accept the result — you would not even have grounds to dispute it.

6One line to keep

Only one thing is uniform on this venue: the winner is $1 and the loser is $0. Who rules, and on what, varies by category and sometimes by market. Two minutes on the terms is the highest-return two minutes available here.

EvidenceCheck it yourself

They differ open markets across categories and read their resolution terms side by side.
The 10.1% all 2,051 windows judged under both rulers; 207 disagreed, direction 125:82 in my favour. A whole-sample unconditional rate, not the knife-edge subset's.
Shorter tenor, worse audited across four tenors at equal granularity; the rate declines monotonically toward zero as tenor lengthens, tracking the final-segment share of the window.
Locked by postponement a position of mine: fixture postponed, market not cancelled, best bid 0.01, unable to exit, held until the replay settled.
The chain is final disputed windows reviewed on-chain 207:0, and 31:0 on a second batch.
Checked 2026-07 to 2026-08. Terms are revised without notice — re-confirm every time.

NextWhere to go

A · MECHANICS
Who decides: the optimistic oracle
F · METHOD
The broken ruler: 10.1% in full
G · ENGINEERING
The redeemable trap
A · MECHANICS
Glossary · EN ⇄ 中文
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.