LZLZL/Prediction markets/Sports
FREEDO IT D · SportsEcosystem

The three winner archetypes

2026-08-21 · ordered by strength of evidence, not by how impressive they sound

Taken apart fill by fill, the lifetime winners fall cleanly into three shapes with completely different capital and capability requirements. The one with the strongest evidence is also the plainest and least infrastructure-dependent. The one that most resembles "professional market making" has the weakest.

1The three

B · Pregame value holdingA · In-play two-sidedC · Pregame semi-making
What they dopick a side before kickoff, hold to settlementhigh-frequency two-sided after kickoffquote both sides pregame
Two-sided rate0–28% (barely pairs)43–76%32–42%
Per-share margin+12.5 to +20.8¢+4 to +16¢−3.0 / −2.4¢
Earns frombeing more accuratespeed and volumerebates
Requiresodds data and judgementlive scores, low latency, sizescreen time
Evidencestrongeststrongweakest

2Archetype B: pregame value holding — strongest evidence

The signature is remarkably consistent: two-sided rate near zero (they do not pair, and do not want to), pregame dominant, buy and hold with sales near zero.

pregame realised margin: +12.5 to +20.8 cents per share spread capture is a fraction-of-a-cent business — this magnitude has only one source

That sentence is the entry's key inference: the magnitude of a margin is itself a fingerprint. Fractions of a cent means market making; ten-plus cents means judgement. Anyone claiming to "earn the spread" while showing double-digit-cent margins is describing something other than spread.

Two sub-shapesDetail
Concentratedone wallet made $1.75M across 24 tennis boards in 12 days (+15.6% margin)
Broadanother covered 986 boards in 19 days

Both work, which says the path does not depend on a particular betting cadence. It depends on the underlying ability: picking the underpriced side.

3Archetype A: in-play two-sided — highest ceiling

SignatureValue
In-play share of volume72–88%
Two-sided rate43–76%
Per-share margin+4 to +16¢ (thin)
Volumeenormous (one did $179k in 0.4 days)
★ One detail that says a lot

One wallet in this group has a lifetime of $1,685k and a pregame per-share margin of −18.9¢ — meaning even it loses money before kickoff, with every cent of profit made after.

Which shows A and B are not two versions of one capability but two different businesses. Being good at one implies nothing about the other.

The barrier is in the requirements row: live score and event feeds, low latency, and size. It is a latency race, and individuals do not have a seat in one.

4★ Archetype C: pregame semi-making — weakest, and the one most people want

This is the shape that most resembles "market making on a prediction venue": quote both sides before kickoff, wait to be filled, earn spread and rebates. It sounds the most professional and has the weakest evidence:

FindingValue
This group's pregame trading EV−3.0¢ / −2.4¢ (negative)
How they break evenon rebates
And notethese were selected from the lifetime winners
⚠ The losers running this style are heavily negative

A wallet eighth on the maker-rebate table with $254k collected is down $3,911,925 lifetime; another "model maker" with exact rebate reconciliation is down $254,149. Full account in profiling the winners.

This is the only archetype where the winners barely break even and the losers lose enormously.

5★ Why pregame market making does not work

I tested pregame two-sided quoting myself and found a pair rate near zero and almost no fills. My first assumption was queue position. Taking the winners apart revealed the real reason, and it is about the nature of the flow:

pregame flow is fundamentally one-directional value flow archetype B taking the side they believe is underpriced — not retail crossing back and forth

Market making needs two-way flow: someone buying, someone selling, you earning the middle. Pregame boards mainly have a group of people all pushing the same way into the side they think is cheap.

It was not that I could not get to the front of the queue. There simply was not enough two-way flow to pair against. That conclusion applies to anyone considering pregame market making — it is market structure, not an execution problem.

6Three things to take

ConclusionNote
Margin magnitude is a fingerprintfractions of a cent = making; ten-plus cents = judgement. A mismatch means someone has mislabelled what they do
Not one winner runs pregame passive two-sided as their enginezero of fifteen
Rebate farming is a losers' pathrebates are a consolation prize, not a business model

7One line to keep

The archetype with the strongest evidence is the plainest one: pick a side before kickoff, hold to settlement, never pair at all. It runs on judgement rather than infrastructure — while the one that most resembles "professional market making" has winners barely breaking even and losers on the leaderboard for the wrong reasons.

EvidenceCheck it yourself

Method screen on lifetime profit → match fills to kickoff time and settlement → attribute by period and side. Wallets anonymised; no identity attribution.
Archetype B two-sided rate 0–28%, pregame dominant, sales ≈0, pregame +12.5 to +20.8¢ per share; sub-shapes of 24 boards in 12 days for $1.75M (+15.6%) and 986 boards in 19 days.
Archetype A in-play 72–88% of volume, two-sided 43–76%, +4 to +16¢ per share on very large volume ($179k in 0.4 days); one wallet with a lifetime of $1,685k has a pregame margin of −18.9¢.
Archetype C two-sided 32–42%, pregame trading EV −3.0 / −2.4¢, breaking even on rebates; same-style losers at −$3,911,925 and −$254,149 lifetime.
No two-way flow pregame my own pregame two-sided test returned a pair rate near zero; the winner teardown shows pregame flow is one-directional value flow.
Checked 2026-07.
Not included any of my own sports selections, thresholds, leagues or timing.

NextWhere to go

D · SPORTS
Profiling the winners: the three reversals
D · SPORTS
The sportsbook baseline and vig
D · SPORTS
Devigging, four ways
B · REBATES
Why market making is this hard
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.