How to Quote Both Sides of a Polymarket Market with Split and Merge

Split USDC into a complete outcome set to make markets on both sides of a Polymarket question, manage inventory imbalance, and merge back to recover collateral.

Overview

To quote both sides of a market you need inventory in both outcomes. The obvious way to get it is to buy each on the book, which means paying the spread twice before you have quoted anything, and taking directional risk on both legs while you accumulate.

There is a better route, and it is specific to how prediction markets are built. Depositing collateral into a condition mints a complete outcome set: one YES and one NO for every dollar. Because exactly one outcome can be true, a complete set is always redeemable for the collateral that created it.

So you can acquire inventory on both sides at cost, with no spread paid and no market risk from the acquisition itself. This guide covers doing that, quoting from it, and managing the one real risk involved.

Read time: about 14 minutes. Assumes familiarity with order books and the difference between taking and providing liquidity.

TL;DR

• POST /v2/trade/positions/split turns collateral into one share of every outcome.
• Holding a complete set is market-neutral. It redeems for what you paid regardless of outcome.
• Quote by resting sells on each outcome. Prices summing above 1.00 is your spread.
• POST /v2/trade/positions/merge turns a complete set back into collateral, bounded by your smaller side.
• The risk is inventory imbalance: once one side fills and the other does not, you are directional.
• Wind down before resolution. An imbalanced book at settlement is a bet, not a spread.

What you will do

• Split collateral into a complete outcome set
• Quote both sides with static limits, or dynamically with pegged orders
• Understand exactly where the risk enters, and it is not the split
• Monitor inventory imbalance and act on it
• Merge unsold inventory back into collateral
• Wind the position down before the market resolves

What you will need

Knowledge

• Order book mechanics, and comfort with the idea of quoting a two-sided market
• How Polymarket works for outcome tokens and condition IDs

Tools and access

• A Bravado API key with trade.execute and trade.cancel
• USDC collateral
• A condition_id for the market you want to quote

export BRAVADO_API_KEY="your-bearer-token"
export BASE="https://bravado-api-k7kaq.ondigitalocean.app"

The mechanic

1 USDC  ──split──▶  1 YES + 1 NO
1 YES + 1 NO  ──merge──▶  1 USDC

That equivalence holds because the outcomes are exhaustive and mutually exclusive. Whichever resolves true pays $1; the other pays $0; together they always pay exactly $1.

Which means holding a complete set is not a position at all in any meaningful sense. It is your collateral, wearing a different shape.

Split

curl -X POST $BASE/v2/trade/positions/split \
  -H "Authorization: Bearer $BRAVADO_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "condition_id": "0xabc123...",
    "amount": "1000"
  }'

You now hold 1,000 YES and 1,000 NO, against 1,000 of collateral consumed.

condition_id identifies the market, not an outcome. Split and merge are among the few calls that take a condition rather than a symbol, because they act on the whole outcome set at once.

Quote both sides

Two approaches, and they suit different levels of attention.

Static, with limit orders

{ "type": "LIMIT", "symbol": "<YES token>", "side": "SELL", "price": "0.56", "size": "500" }

{ "type": "LIMIT", "symbol": "<NO token>",  "side": "SELL", "price": "0.48", "size": "500" }

Notice those prices sum to 1.04, not 1.00. That four-cent difference is your spread: if both sides fill you collect $1.04 for a set that cost you $1.00.

The wider you quote, the more you make per round trip and the less often you fill. That trade-off is the entire business.

Dynamic, with pegged orders

{
  "type": "PEGGED",
  "symbol": "<YES token>",
  "side": "SELL",
  "budget_usdc": "500",
  "execution": { "offset_ticks": 2, "price_floor": "0.50" }
}

offset_ticks is in CLOB ticks where one tick is 0.1 cents, so 2 quotes two ticks off the touch and follows it as the book moves. price_floor stops it chasing the market down past a level you are unwilling to sell at.

Pegged orders return a record_id and live on /v2/trade/strategies, not open orders. See Track order and strategy state.

Where the risk actually is

Splitting is riskless. Quoting is not. The moment one side fills and the other does not, you stop being flat.

State	You hold	Exposure
After split	1,000 YES + 1,000 NO	None. Complete set.
YES sells, NO does not	1,000 NO	Long NO
Both sell evenly	Collateral plus spread	None
Neither sells	Complete set	None. Merge back.

That second row is the whole risk of the strategy, and it is not random. If one outcome keeps filling and the other does not, the market is telling you your prices are wrong, and you are accumulating precisely the side nobody wants.

Inventory imbalance is adverse selection, not bad luck. The fills you get are the ones informed traders wanted to give you. Track imbalance actively rather than assuming it averages out.

Monitor imbalance

from decimal import Decimal

MAX_IMBALANCE = Decimal("200")


def imbalance(yes_symbol, no_symbol):
    pos = read("/v2/trade/positions")["positions"]
    by_symbol = {p["symbol"]: Decimal(p["size"]) for p in pos}

    yes = by_symbol.get(yes_symbol, Decimal(0))
    no  = by_symbol.get(no_symbol,  Decimal(0))
    return yes, no, abs(yes - no)


def rebalance(condition_id, yes_symbol, no_symbol):
    yes, no, gap = imbalance(yes_symbol, no_symbol)

    if gap > MAX_IMBALANCE:
        # recover collateral from the complete portion, shrink the exposure
        write("POST", "/v2/trade/positions/merge",
              {"condition_id": condition_id, "amount": str(min(yes, no))})
        return "merged"

    return "ok"

Two responses when imbalance grows: re-price, because your quotes are wrong, or merge back, because you would rather hold collateral than a directional bet you did not intend.

Merge back

curl -X POST $BASE/v2/trade/positions/merge \
  -H "Authorization: Bearer $BRAVADO_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "condition_id": "0xabc123...",
    "amount": "300"
  }'

Merging requires a complete set, so you can only merge up to your smaller side. Holding 500 YES and 300 NO merges 300, leaving 200 YES outstanding and still directional.

That residual is the position you actually took, whether you meant to or not.

Resolution turns inventory into a bet

A complete set is safe through resolution: it redeems for collateral either way. An imbalanced book is not. The surplus side is worth $1 or $0, decided by an event you were not trying to forecast.

Wind down or rebalance before an event concludes. Resolution does not wait for your quotes to fill, and once UMA settles the outcome the losing side is worthless. See UMA resolution.

A sensible wind-down:

1. Stop quoting

   Cancel resting orders and any pegged strategies. Remember cancel-all does not stop strategies.

2. Merge the complete portion

   Recover collateral from whatever matched set you still hold.

3. Decide about the residual

   Sell it, or knowingly hold it as a directional position. Either is fine; drifting into it is not.

Venue constraints

• Resting orders need at least 5 shares, which floors your quote size.
• Prices are decimal probabilities from 0.001 to 0.999.
• Merges are bounded by your smaller side.
• Split and merge are mutating calls and take an Idempotency-Key.

Wrapping up

Split gives you two-sided inventory at cost, which is a genuinely better starting point than buying both legs on the book. The spread you quote is the business, and inventory imbalance is the cost of doing it.

The discipline that matters: know your imbalance at all times, and never carry an imbalanced book into resolution without deciding to. A market maker who accidentally becomes a directional trader usually finds out at settlement.

Frequently asked questions

Is splitting risky?

No. A complete outcome set redeems for the collateral that created it regardless of which outcome resolves true. The risk begins when one side sells and the other does not.

Why does split take condition_id instead of symbol?

It acts on the entire outcome set at once, so it needs to identify the market rather than a single outcome. Most other calls take the outcome token id as symbol.

Can I merge more than my smaller side?

No. Merging destroys a complete set, so it is bounded by whichever outcome you hold least of. The remainder stays as a directional position.

How wide should I quote?

Wider earns more per round trip and fills less often. Start wide enough that adverse selection does not eat the spread, then tighten while watching whether one side fills far more than the other.

My pegged order is not in open orders.

Pegged is a managed strategy and lives on /v2/trade/strategies. See Track order and strategy state.

What happens if I hold an imbalanced book through resolution?

The surplus side settles at $1 or $0. You took a directional bet on the outcome, which is a fine thing to do deliberately and an expensive thing to do by accident.

Resources

• Positions endpoints, split, merge, and redeem reference
• Place an order, limit and pegged order payloads
• Track order and strategy state, where pegged orders live
• Execute a large position, the taking-liquidity alternative
• UMA resolution, what settlement does to inventory
• Trade API reference, full field documentation