Market Selection: Which Prediction Markets to Trade

The universe of prediction markets is enormous and most of it is not worth your capital. Combined monthly volume across Kalshi and Polymarket sat around $24 billion in April 2026, up from under $5 billion in September 2025, but that number hides a hard truth: a small set of markets carries almost all the liquidity, and the long tail is thin, wide-spread, or ambiguously worded. Before you automate anything, you need a repeatable screen that throws out the junk.

There are four filters that matter: liquidity, resolution clarity, time to resolution, and whether you actually have an edge. Run every candidate through all four. A market that fails any one of them is a market you skip, no matter how tempting the price looks.

1. Liquidity: can you get in and out at a fair price

Volume and liquidity are not the same thing, and confusing them is the most common mistake. A market trading at 72 cents on heavy volume is a completely different animal from one at 72 cents with almost no activity. Volume tells you how much trading has happened; depth on the book tells you whether you can fill your size right now without moving the price.

Concentration is the reality to plan around. Sports has accounted for roughly 80% of Kalshi's trading volume since July 2024 versus about 39% on Polymarket, and Kalshi's sports markets carry the deepest liquidity on either platform. Polymarket's political and geopolitical markets typically have more buyers and sellers than the same markets on Kalshi. So venue selection is partly category selection: trade where the book is deep for your topic, not where the brand is bigger.

Practical liquidity checks before you commit a strategy to a market:

• Look at the live spread, not just last price. Wide spreads quietly eat returns and are the signature of a thin market.
• Size your intended order against resting depth at the top few levels. If your position would sweep multiple ticks, the market is too thin for that size.
• Check whether liquidity is event-driven. Kalshi volume spikes around postseason schedules and marquee matchups, then drains away, so a market that looked deep last week may be a desert today.
• Prefer markets where multiple makers are present. Depth that depends on one market maker disappears the moment they pull quotes.

This is exactly where paper trading against the live order book earns its keep. Running a strategy in paper mode with Banger against real depth, rather than backtesting on mid-price fills, tells you whether your fills are realistic before a cent is at risk. A strategy that looks profitable at the midpoint often dies once you pay the real spread.

2. Resolution clarity: read the rules, not the title

The market title asks the question; the resolution rules decide how it pays. On Polymarket every market has pre-defined rules specifying the resolution source, the end date, and how edge cases are handled, and the platform's own documentation is blunt about it: always read the resolution rules before trading. Ambiguous wording is a risk factor you price in, not a detail you skip.

How a market settles is part of the trade. Polymarket's international book resolves through UMA's optimistic oracle, where anyone can propose an outcome and anyone can dispute it, and a twice-disputed question escalates to a token-holder vote whose ruling cannot be appealed. That path is rare relative to total resolutions, but it is real: Polymarket logged more than 1,150 disputed markets in the first five months of 2026, already past its full-year 2025 total. A Wall Street Journal investigation published in May 2026 found that in most disputed markets, more than half the votes came from the ten largest wallets.

Kalshi reaches settlement through opposite infrastructure: an exchange-cleared central-counterparty book, with disputes handled by exchange procedure under rules filed with a federal regulator rather than by anonymous token holders. Neither model is strictly better, but they carry different tail risks. For larger automated positions, prefer objectively verifiable questions with a single unambiguous data source, and treat vague or interpretive wording as a reason to shrink size or pass.

3. Time to resolution: match horizon to strategy

Time to resolution drives capital efficiency and the shape of your risk. A market resolving tonight ties up collateral for hours; one resolving after an election ties it up for months and exposes you to every headline in between. Neither is wrong, but they demand different strategies. Short-horizon markets suit high-turnover, mean-reversion, or news-reaction logic. Long-horizon markets suit conviction positions where you can tolerate mark-to-market swings.

Build the settlement timeline into the horizon math. On Polymarket, an undisputed resolution finalizes roughly two hours after an outcome is proposed, but a disputed one can run four to six days or longer once it reaches the token vote. If your strategy assumes capital frees up at the market's end date, a dispute breaks that assumption. Do not treat the first proposed outcome as final until the challenge window has expired.

4. Edge: why do you win this specific market

Liquidity, clean rules, and a workable horizon get you a market worth trading. They do not give you a reason to win it. Edge is the fourth filter and the one most traders skip. Before automating, state plainly where your advantage comes from: faster reaction to a data feed, a calibration edge on a category you know well, a structural spread you can capture as a maker, or an inefficiency in how the crowd prices a specific event type. If you cannot name the edge, you do not have one.

Deep, liquid markets are efficient precisely because everyone is watching them. Your edge is often easier to find one level down, in markets liquid enough to trade but not so crowded that the price is already perfect. That is a narrow band, and finding it is the actual work of market selection.

Putting the screen into code

The four filters translate directly into a pre-trade gate. A strategy should refuse to touch a market that fails any of them, before position sizing ever runs:

def tradeable(mkt):
    # 1. liquidity: real depth for our size, tight spread
    if mkt.top_of_book_depth < MIN_DEPTH:
        return False
    if mkt.spread > MAX_SPREAD:
        return False
    # 2. resolution clarity: single objective source, no open dispute
    if not mkt.has_objective_source or mkt.in_dispute:
        return False
    # 3. time to resolution: fits our horizon
    if not (MIN_HORIZON <= mkt.time_to_resolution <= MAX_HORIZON):
        return False
    # 4. edge: our model must actually diverge from the book
    return abs(model_prob(mkt) - mkt.mid) > EDGE_THRESHOLD

Wire that gate into a strategy, paper-trade it against the live book to confirm the fills are real, then run it under a declarative risk envelope: a per-trade cap, a daily loss stop, a max number of open positions, and a kill switch. Banger runs exactly this loop, and it never custodies funds, so you bring your own venue keys and keep control of your capital. The install is one line.

pip install bangertrades
banger run strategy.py --paper

A note on venue status

Venue access changes fast, so re-verify before you route real orders. As of mid-2026, Kalshi operates as a CFTC-registered exchange for US users. Polymarket's international book remains geoblocked for US IP addresses under its 2022 CFTC settlement, while a separate CFTC-regulated venue, Polymarket US, launched in late 2025 as a Designated Contract Market. Polymarket filed with the CFTC on April 28, 2026 seeking to reopen its main offshore exchange to US traders, and a CFTC investigation into Polymarket was reported in June 2026. The picture is still moving, so check your state's current status and the last 30 days of federal news before depositing anywhere.