Polymarket Liquidity Risk

Which Polymarket reward markets have historically been most toxic for liquidity providers — measured, not guessed.

Data snapshot: 2026-08-25T19:03:38Z · 75 reward markets analysed · free · research snapshot, not a continuously updated feed

Read this before you use the numbers. A high toxicity score does not mean a market will lose money. It means that historically, liquidity providers in that market were filled shortly before the price moved against them. This is a measurement of past adverse-selection risk, not a prediction of your results, and it says nothing about whether a market's reward pool compensates you for that risk. This is analytics and risk information — not financial advice, not a recommendation, and not a signal to trade.

Most toxic reward markets right now

#Market Toxicity (7d)
$ per $1k		 Prior 7d Markout
¢/share		 Persistence Reward pool
$/day		 Risk
1 Will Oura's market cap be less than $7.5B at market close on IPO day?
Over the last 7 days makers lost 383.54 per $1,000 of volume they filled, across 44 trades. In the previous 7-day window makers lost 331.87 per $1,000. Elevated in both 7-day windows — this is the pattern that persisted out-of-sample in our testing.
 -383.54 -331.87 -3.13 Persistent $30.00 Extreme
2 Will Oura's market cap be between $7.5B and $10B at market close on IPO day?
Over the last 7 days makers lost 308.66 per $1,000 of volume they filled, across 74 trades. In the previous 7-day window makers lost 265.13 per $1,000. Elevated in both 7-day windows — this is the pattern that persisted out-of-sample in our testing.
 -308.66 -265.13 -2.871 Persistent $30.00 Extreme
3 Will Sal Stewart lead the MLB in RBIs for the 2026 regular season?
Over the last 7 days makers lost 259.3 per $1,000 of volume they filled, across 39 trades. In the previous 7-day window makers lost 56.91 per $1,000. Elevated in both 7-day windows — this is the pattern that persisted out-of-sample in our testing.
 -259.3 -56.91 -4.59 Persistent $34.00 Extreme
4 Clarity Act (H.R.3633) signed into law in 2026?
Over the last 7 days makers lost 116.06 per $1,000 of volume they filled, across 1980 trades. In the previous 7-day window makers lost 24.56 per $1,000. Elevated only in the most recent window — treat as a weaker, less-confirmed signal.
 -116.06 -24.56 -1.966 Newly elevated $500.00 Extreme
5 Will XRP reach $2.80 by December 31, 2026?
Over the last 7 days makers lost 94.97 per $1,000 of volume they filled, across 31 trades. No comparable prior-window sample. Not enough history in the prior window to judge persistence.
 -94.97 — -1.293 Insufficient data $50.00 Extreme
6 Will BNB dip to $500 by December 31, 2026?
Over the last 7 days makers lost 92.52 per $1,000 of volume they filled, across 41 trades. In the previous 7-day window makers gained 36 per $1,000. Elevated only in the most recent window — treat as a weaker, less-confirmed signal.
 -92.52 36 -1.681 Newly elevated $50.00 Extreme
7 Will the UK’s 2026 inflation be between 2.5% and 2.9%?
Over the last 7 days makers lost 89.61 per $1,000 of volume they filled, across 54 trades. No comparable prior-window sample. Not enough history in the prior window to judge persistence.
 -89.61 — -3.663 Insufficient data $42.00 Extreme
8 Will OpenAI’s market cap be $1.5T or greater at market close on IPO day by December 31, 2027?
Over the last 7 days makers lost 87.53 per $1,000 of volume they filled, across 21 trades. In the previous 7-day window makers gained 6.4 per $1,000. Elevated only in the most recent window — treat as a weaker, less-confirmed signal.
 -87.53 6.4 -1.6 Newly elevated $41.00 Extreme
9 Will the next Prime Minister of Romania be an Independent or a Technocrat?
Over the last 7 days makers lost 84.3 per $1,000 of volume they filled, across 48 trades. In the previous 7-day window makers lost 4.02 per $1,000. Elevated only in the most recent window — treat as a weaker, less-confirmed signal.
 -84.3 -4.02 -2.305 Newly elevated $221.00 High
10 Will the number of Democratic House members who retire in 2026 be between 20 and 23 inclusive?
Over the last 7 days makers lost 71.08 per $1,000 of volume they filled, across 55 trades. In the previous 7-day window makers lost 113.19 per $1,000. Elevated in both 7-day windows — this is the pattern that persisted out-of-sample in our testing.
 -71.08 -113.19 -3.55 Persistent $35.00 High
11 Will Zcash reach $900 by December 31, 2026?
Over the last 7 days makers lost 56.46 per $1,000 of volume they filled, across 26 trades. No comparable prior-window sample. Not enough history in the prior window to judge persistence.
 -56.46 — -4.924 Insufficient data $50.00 High
12 Will the UK’s 2026 inflation be between 3.5% and 3.9%?
Over the last 7 days makers lost 47.9 per $1,000 of volume they filled, across 38 trades. In the previous 7-day window makers lost 201.51 per $1,000. Elevated in both 7-day windows — this is the pattern that persisted out-of-sample in our testing.
 -47.9 -201.51 -5.149 Persistent $40.00 High
13 Will Alexandria Ocasio-Cortez win the 2028 Democratic presidential nomination?
Over the last 7 days makers lost 44.15 per $1,000 of volume they filled, across 720 trades. In the previous 7-day window makers lost 14.31 per $1,000. Elevated only in the most recent window — treat as a weaker, less-confirmed signal.
 -44.15 -14.31 -0.408 Newly elevated $43.00 High
14 Jack Lowden announced as next James Bond?
Over the last 7 days makers lost 41.45 per $1,000 of volume they filled, across 215 trades. In the previous 7-day window makers lost 71.23 per $1,000. Elevated in both 7-day windows — this is the pattern that persisted out-of-sample in our testing.
 -41.45 -71.23 -1.455 Persistent $217.00 High
15 Will The Odyssey get the most Oscar nominations at the 99th Academy Awards?
Over the last 7 days makers lost 38.95 per $1,000 of volume they filled, across 34 trades. In the previous 7-day window makers gained 6.28 per $1,000. Elevated only in the most recent window — treat as a weaker, less-confirmed signal.
 -38.95 6.28 -1.118 Newly elevated $44.00 High
16 Will the Iranian regime fall before 2027?
Over the last 7 days makers lost 34.82 per $1,000 of volume they filled, across 384 trades. In the previous 7-day window makers gained 74.36 per $1,000. Elevated only in the most recent window — treat as a weaker, less-confirmed signal.
 -34.82 74.36 0.195 Newly elevated $70.00 High
17 No one announced as next James Bond?
Over the last 7 days makers lost 32.65 per $1,000 of volume they filled, across 268 trades. In the previous 7-day window makers lost 66.39 per $1,000. Elevated in both 7-day windows — this is the pattern that persisted out-of-sample in our testing.
 -32.65 -66.39 -0.952 Persistent $160.00 High
18 Will Morgan Wallen be the Billboard #1 top artist in 2026?
Over the last 7 days makers lost 28.42 per $1,000 of volume they filled, across 43 trades. In the previous 7-day window makers lost 250.41 per $1,000. Elevated in both 7-day windows — this is the pattern that persisted out-of-sample in our testing.
 -28.42 -250.41 -10.031 Persistent $50.00 High
19 Will the Republicans win the Wisconsin governor race in 2026?
Over the last 7 days makers lost 25.89 per $1,000 of volume they filled, across 346 trades. In the previous 7-day window makers gained 2.07 per $1,000. Elevated only in the most recent window — treat as a weaker, less-confirmed signal.
 -25.89 2.07 -0.142 Newly elevated $40.00 High
20 Will the next Google Gemini Pro model added to the Arena Leaderboard debut at a score of at least 1480?
Over the last 7 days makers lost 21.31 per $1,000 of volume they filled, across 23 trades. In the previous 7-day window makers lost 55.71 per $1,000. Was elevated previously but is not in the recent window.
 -21.31 -55.71 -1.822 Improving $50.00 Medium
How to read the toxicity column. It is maker profit or loss in dollars per $1,000 of volume you filled. A value of -15.00 means that for every $1,000 of orders that got hit, makers were on average $15 worse off one hour later. Compare that against the reward pool you would be sharing.

Would this information change your liquidity decision?

Anonymous, one click, no email and no account. If you actually provide liquidity, this is the single most useful thing you can tell us — including if the answer is no.

Methodology

What is measured

Every public trade has a maker on the other side. We take the maker's side of each trade and measure where the market's midpoint sits one hour later.

Toxicity

Total maker profit/loss divided by total notional filled. Normalised so big and small markets are comparable. Negative = adverse for LPs.

Data

Polymarket's public CLOB and data APIs. 10-minute midpoint series, trade tape paginated per market. No private or paid data.

Risk buckets

Percentiles of this universe, not absolute thresholds. Extreme = bottom 10%, High = bottom 25%, Medium = bottom 50%, Low = above median.

What we deliberately do not do. There is no composite "opportunity score" here. We tested combined reward/volume/toxicity formulas out-of-sample and they overfit — one scored t=8.9 in-sample and t=0.16 out-of-sample. Past toxicity on its own was the signal that survived. So that is all we publish.

What the testing showed. Across 1,207 market-days over 32 days, toxicity persisted from one day to the next (rank correlation 0.39, positive in 30 of 31 day-pairs) and did not meaningfully decay over 14 days. Importantly, markets that were bad tended to stay bad; markets that were good did not reliably stay good. That is why this page is a list of what to be careful with, not a list of what to buy.

Get the daily report

One email a day: what changed, which markets newly turned toxic, and which stopped being toxic. Free.

No spam. One click to unsubscribe.

Pro — planned, not yet available

PROTOTYPE · NOT CHARGING YET

$19/month — if enough people want it, this is what it would include:

We are not taking payment. Join the waitlist and you will be asked once, when it exists, whether you actually want it.

Why trust the numbers

Every figure on this page is reproducible from public APIs with no key. The collector and the exact methodology are documented, and the raw per-market-day data is kept separate from the calculated scores so you can check the arithmetic yourself. If you find an error, tell us and we will fix it and say so.