LZLZL/Prediction markets/Sports
FREEBUILD IT D · SportsMethod

Devigging, four ways
and the odds supply

2026-08-21 · strip the margin before you can compare anything

Odds contain vig, so they cannot be used as probabilities directly. Stripping it is called devigging. The awkward part: there are four mainstream ways to redistribute that margin, and they diverge at long odds — which is precisely where any spread would appear.

1Why it is necessary

A fixture's implied probabilities sum to 104.7%; the extra 4.7 points are the bookmaker's margin. To compare against a prediction market price you must first remove those 4.7 points and recover a set summing to exactly 100%.

devigging = compressing "sums to 104.7%" back to "sums to 100%" the question is: which side gives up more of the 4.7?

2The four methods

MethodHow it allocatesCharacter
Multiplicativescale every outcome proportionallysimplest, most common
Additivesubtract the same amount from eachhits longshots harder
Powerraise to a power, then normalisebetween the two, tunable
Shinassume the book is defending against informed traders and solve that outstrongest theoretical basis
⚠ They diverge most at long odds

On the favourite the four barely differ. On the longshot they differ substantially.

Which is the dangerous part, because spreads between a prediction market and bookmaker odds tend to appear on the longshot side. Meaning your choice of devig method decides whether you "find" the opportunity at all.

Put differently: an edge discovered on the longshot side may be nothing more than an artefact of the method you picked — structurally identical to the instrument failure documented in the broken ruler.

3So which one

There is no universally correct answer, but there is an operable discipline:

DoWhy
Pick one, write it down, do not change itotherwise you will unconsciously select the one that shows an opportunity
Use the other three as a sensitivity checkif the edge disappears under another method, it is not an edge
Report the divergence"all four agree" and "only one works" are completely different strengths of evidence

The second is the cheapest self-check available: run your finding through all four. Anything that only holds under one is almost certainly a method artefact rather than a market bias.

Basisthis site does not publish which method I use or how I combine sources — that is strategy. The methods themselves are public knowledge, in any textbook.

4The odds supply

Before you can devig anything you need odds. The practical situation:

IssueReality
Coveragegood for major leagues, poor for niche sports
Esportsmainstream odds APIs have zero coverage
Update frequencytiered by subscription; faster costs more
Historical datausually charged separately, and frequently has gaps
Free sourcesexist, but laggy, narrow, and liable to break
⚠ The data supply is the real barrier on this path

Devig algorithms are public knowledge and a few lines of code. Obtaining timely, clean, sufficiently broad odds is the hard part — and it costs money.

So when evaluating this path, compute the data cost first: if a month of odds subscription exceeds your expected monthly return, there is no point continuing. That judgement can be made before writing a line of code.

5What the devigged probability is for

fair probability (post-devig) vs the prediction market price the gap is a candidate opportunity — a candidate only

Turning a candidate into an opportunity requires three gates:

GateQuestion
Costdoes the gap cover the taker fee plus the spread?
Executableis there size, or only a quote with no depth?
Robustdoes it survive another devig method? another odds source?

The third is this entry's point: anything that disappears under a reasonable change of basis was never an edge.

6One line to keep

Devigging is mandatory, but "how" has four defensible answers that diverge most at long odds — exactly where spreads live. So run any longshot-side finding through all four; one that holds under only one is a method artefact, not a market bias.

EvidenceCheck it yourself

The four methods multiplicative, additive, power and Shin are all public, documented in textbooks and the literature. They differ noticeably at long odds.
Sensitivity check compute the same finding under all four; holding under only one marks it a method artefact.
Esports gap measured: mainstream odds APIs have zero esports coverage.
Compute data cost first subscription against expected return can be compared before any development.
Checked 2026-08. The data-source landscape moves quickly; verify current coverage and pricing.
Not included which devig method I use, how I combine sources, and any selection thresholds.

NextWhere to go

D · SPORTS
What vig is
E · UP/DOWN
Esports: the odds feed is the gap
F · METHOD
Findings that vanish under a change of basis
G · ENGINEERING
Whether a candidate is executable
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.