The category stopped being a crypto curiosity and became something people check like a scoreboard. That shift was about access, not cleverness. Here is what changed, where the volume actually sits, and the specific slice Predict Bay is built for — including a plain statement of what is not live yet.
Predict Bay ·
For most of their history, prediction markets were an idea economists liked more than the public did. The mechanism was elegant — put money behind a belief and the price becomes a forecast — but the products were thin, slow and awkward to reach. That is no longer the shape of the category. Prediction markets now sit inside mainstream finance surfaces, get quoted in news coverage, and clear volumes that put them alongside serious venues rather than experiments.
ROUND CLOSES IN
--:--
LAST PRICE
—
STRIKE
—
NO POSITIONS YET THIS ROUND — BE THE FIRST
Practice mode · no wallet, no deposit, free to start
This post is about that shift, and about the specific slice of it Predict Bay is built for. It is worth saying up front what this post is not: it is not a claim that Predict Bay has captured any part of that growth. Predict Bay runs practice mode alongside a Sui mode that trades real Circle USDC on DeepBook Predict's Sui mainnet deployment, and real-money EVM trading on Polygon is paused. Category growth is context for a design decision, not evidence about a product, and treating it as the latter is the most common dishonest move in writing like this.
The numbers moved by roughly an order of magnitude in about a year. TRM Labs' analysis of on-chain prediction market activity puts monthly trading volume at around USD 1.2 billion in early 2025, and over USD 20 billion by January 2026, with a single-day record of USD 425 million on 28 February 2026. The same analysis reports unique participating wallets more than tripling to about 840,000 in the six months to February 2026 — which matters more than the volume figure, because it says the growth came from new people rather than from existing traders sizing up.
The causes were mostly boring, which is the tell that the change is structural. TRM attributes the acceleration to regulatory clarity, mainstream platform integration and media visibility — Robinhood distribution, ICE/NYSE investment, CFTC relief, and prediction-market odds appearing inside Google Finance. None of that is a better mechanism. All of it is better access. The instrument did not get smarter; it got reachable.
Here is the part that argues against the easy version of this post. The dominant categories are not crypto. TRM's breakdown finds geopolitics, macroeconomics and politics driving the majority of trading activity, with sports close behind, and crypto price questions a consistently smaller share across every user tier — despite the whole thing running on blockchain infrastructure. Predict Bay lists crypto price markets and commodity price markets. By volume, that is not where the category's centre of gravity is.
We would rather write that down than route around it. A venue that tells you it is riding the biggest wave in the category, while listing the markets with the smallest share of it, has told you something false in the first paragraph. The honest framing is narrower and more defensible: the category proved that ordinary people will trade a well-posed question if you make it easy to reach, and there is a specific class of well-posed question that the large venues serve least well.
An event market's horizon is set by the world, not by the venue. An election resolves in November because the election is in November. That is fine, and it is the whole point of an event market, but it means your collateral is committed for weeks and the answer arrives long after the decision to trade felt interesting.
A price question has no such constraint. "Where is BTC in three minutes" is a complete, well-posed question with a canonical answer, and it can be asked continuously. Predict Bay's practice venue runs three-minute and ten-minute rounds: a strike is set at the open, you pick above or below, and the round resolves against an on-chain price feed at expiry. There is no proposer, no challenge window and no vote, because there is nothing to adjudicate — the number is published. That is what makes a three-minute market mechanically possible at all, and it is covered in more depth in oracle resolution and on-chain settlement.
The trade-off is real and runs both ways. Predict Bay cannot list "will the central bank cut rates in September", because that question has no canonical price to read. A general event venue cannot settle a three-minute round in seconds without a price oracle doing the work anyway. These are different instruments, not competing versions of one — Predict Bay vs Polymarket works through the design differences properly.
If access is what moved the category, then the interesting engineering is at the entrance, not the matching engine. Three things in this product exist for that reason and no other.
None of these is a novel idea. All of them remove a step that used to happen before a first-time user reached the actual product. The category's own growth suggests that removing those steps is worth more than any improvement to the thing behind them.
Predict Bay lists BTC, ETH, SUI, gold and oil. The two commodities are there because a price question does not care what the underlying is — it needs a continuously published, canonical price and nothing else, and gold and oil have had one for longer than crypto has existed. They also carry a property the crypto assets do not: they close. Gold and oil follow the CME clock, which means a market that is genuinely shut for a third of the week, and a venue that ignores that will happily open rounds against a frozen price. Gold and oil prediction markets covers what listing them actually required.
The bet Predict Bay is making is narrow and easy to state: the category's growth came from access, the shortest-horizon price questions are the ones the large venues serve least well, and those two facts point at the same product. Whether that is the right bet is not something a blog post settles. What we can do is be exact about which parts of it exist today, and this page is the current answer.