LZLZL/Prediction markets/Market making
FREEDO IT B · RebatesReal cost

Adverse selection
why you get the fills you did not want

2026-08-21 · the real cost of market making

Posting costs no fee and pays a rebate when filled. It sounds free. What the poster pays instead is adverse selection: the moment your quote is hit is usually the moment it should not have been. Here is that cost measured, and why it is an order of magnitude worse on short tenors.

1Who comes to hit you

You have a bid at 0.61. Under what circumstances does it fill?

CaseWhy they sold to youFor you
They are closing out / need cashnothing to do with valuethe fill you wanted
They know it is going downbecause 0.61 is already too expensivethe fill you did not want

The asymmetry: the second group is more urgent. Someone in no hurry can rest an order; someone with information must trade now — so informed traders are over-represented in the flow that reaches you relative to their share of the population.

quote never fills → price never moved → you earned nothing quote fills → the price is often moving against you this is adverse selection: the fill is itself bad news

2The arithmetic

NatureMagnitude
Maker rebatefixedabout +0.25¢/share
Adverse selectionvariable, and steep in tenor0.5¢ to 7¢+ /share

Income is a flat line; cost is not. So whether the business works depends almost entirely on which board you run it on. Measured on crypto up/down, tenor by tenor:

TenorNaked-leg shareAdverse selection /shareRebate /shareNet
1 hour3.5%−0.475¢+0.248¢+0.099¢, barely positive
5 minutes53%≈ −7.2¢+0.25¢an order of magnitude short

Same action, same rebate — change the tenor and it goes from barely positive to deeply negative. And +0.099¢ leaves essentially no tolerance: a slightly worse cost and it flips.

3Why short tenors are an order of magnitude worse

You are trying to earn the spread, typically a cent or two, while carrying whatever the price does while you hold. The ratio between those decides the business.

price movement ∝ volatility × √time the shorter the tenor, the faster the movement your fixed spread must outrun

Measured on five-minute boards — median mid-price movement over 30 seconds:

Period30-second median movevs the spread you want
across the window2–4¢about 1¢
final stretch27.5¢

That 27.5¢ is the point: the price can move twenty-seven times your target profit in half a minute. In that environment whether you make money has almost nothing to do with your quoting skill.

⚠ One transferable sentence

Market making a short tenor is roughly like permanently living in the last few minutes of a long one.

Everyone knows the close is dangerous and avoids it. A short-tenor board is in that state for its entire life — there is no safe stretch. I recorded several hours of book and checked every period: 2–4¢ throughout, no safe window.

4The other consequence: the leg that never pairs

The maker's ideal is both sides filling, merged into sets, spread locked. Adverse selection specifically destroys that: when the market trends, only one side fills — the side that loses you money.

The unmatched part is the naked leg, and I measured its cost: about −13.5¢ per unpaired share (−$111.51 spread over 825 shares). Against a +0.25¢ rebate — fifty-four times.

Which is why the business lives or dies on the pair rate, not on spread capture. See pair rate and naked legs.

5What the population actually earns

QuestionMeasured
Median return of the group−0.74¢/share (47 same-shape wallets)
What the best one earnsabout $95/day — the ceiling for the lane
Do rebates rescue it?Not necessarily. Wallets holding $100–380 in rebates stayed net negative

And one detail that says a great deal: the wallets at the top of the maker-rebate table are almost all in sports, with not one of their last thousand-odd fills in crypto up/down. The people best at this business have voted with their feet against this category.

6So what do you do

DoNote
Pick a longer tenorlonger holding, more chances to pair, less adverse selection
Quote further from midfewer sweeps — but LP rewards fall with you, measured at ×0.26 two cents out
Watch pair rate, not spreadpaired is locked; unpaired is the risk
Do not call the rebate profitit is the entry ticket. Compute net first
Read the population firstif the median is negative, why would you be the exception?

7One line to keep

A poster's income is fixed and their cost is variable — so "no fee plus a rebate" is where the analysis starts, not where it ends. Being filled is itself bad news: the person who took your quote usually knew slightly more than you did.

EvidenceCheck it yourself

Per-tenor arithmetic 1h: naked legs 3.5% (measured), adverse selection −0.475¢/share, rebate +0.248¢/share, net +0.099¢. 5m: naked legs 53% (from replay), cost ≈ −7.2¢/share, deeply negative.
30-second movement hours of tick-level book recording; five-minute boards median 2–4¢ throughout, rising to 27.5¢ in the final stretch, against a target spread of about 1¢.
Naked-leg cost −$111.51 ÷ 825 shares = −13.5¢/share, versus a +0.25¢/share rebate.
Population 47 same-shape wallets, median −0.74¢/share; best steady-state about $95/day; wallets collecting $100–380 in rebates still net negative.
Top makers avoid the category none of the leading rebate wallets' recent thousand-odd fills were in crypto up/down. No identity attribution is made here.
Checked 2026-07 to 2026-08. Fees and the ecosystem move; re-check against current data.

NextWhere to go

B · REBATES
Pair rate and naked legs: tenor decides
B · REBATES
Quoting wider: what it costs you
B · REBATES
Queue position and conservative assumptions
E · UP/DOWN
Up/down market making: the population is negative
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.