Why Prediction-Market Arbitrage Disappears After Fees, Slippage, and Fill Risk

A three-cent prediction-market arbitrage gap can disappear before either order is placed. The headline calculation uses the best displayed prices; the real calculation uses equal executable size, every explicit cost, and the loss created if one independently submitted leg fails. This guide turns a gross price gap into a net-edge worksheet and explains where the apparent profit usually goes.

Start with gross edge

For complementary shares with a combined $1 settlement value:

gross edge per pair = $1 − YES cost − NO cost

Assume the best displayed asks are:

• YES: 45 cents;
• complementary NO: 50 cents.

The headline bundle costs 95 cents and appears to offer five cents of gross edge. That number is a screening signal. It does not yet include the size available at either price.

Replace best price with volume-weighted execution

Suppose you want 100 matched shares. The books contain:

The executable acquisition cost is $97.80, not $95. The gross edge falls from $5 to $2.20 before explicit fees.

average price = total cost across consumed levels ÷ shares filled

This is why available size matters more than the prettiest top-of-book quote.

Add venue fees

Fee models can vary by venue, market, side, price, and whether an order adds or removes liquidity.

For a worked model, suppose the two legs incur $1.20 in combined trading fees:

Do not substitute these hypothetical fees for a current venue schedule. Pull the market-specific fee or ticket estimate at the same time as the order-book snapshot.

Account for maker versus taker behavior

A trader can try to avoid or reduce taker cost with a resting limit order. That changes the problem:

• the order may not fill;
• only part of it may fill;
• the other venue can move while it rests;
• capital can remain locked;
• a later cancellation can arrive after a partial match.

Maker economics can improve explicit cost while worsening timing and completion risk. Compare strategies on expected completed bundles, not advertised fee alone.

Add funding and network costs

Cross-venue capital can require:

• a bridge;
• token conversion;
• network gas;
• deposit or withdrawal processing;
• pre-positioned collateral on two chains or venue wallets.

Assume another 40 cents of amortized funding and network cost. The modeled edge falls to 60 cents per $100 settled.

Pre-positioned capital reduces transfer delay but has an opportunity cost: funds remain fragmented across venues even when no trade exists.

Price capital lock-up

Headline edge is a dollar amount, not an annualized return. A 0.6% net edge that resolves tomorrow is different from the same edge that locks capital for a year.

A simple comparison is:

holding-period return = net edge ÷ capital committed

Annualizing a short holding period can produce absurdly large numbers and should not be used as a promise. Settlement can be delayed, disputed, or operationally difficult to redeem.

Model the second-leg move

Now assume the YES leg fills first. Before the NO order submits, its price moves from the modeled average of 51.2 cents to 52.2 cents.

That one-cent move adds $1 to the 100-share bundle. The expected 60-cent edge becomes a 40-cent loss before any emergency unwind cost.

Two-leg fill risk is therefore nonlinear: when the remaining gap is small, a modest move can flip the sign of the trade.

Partial fills create unmatched exposure

Suppose 100 YES shares fill but only 70 complementary NO shares execute. Seventy pairs are hedged; thirty YES shares remain directional.

Track the trade as two components:

Do not report the whole 100-share position using the bundle edge. The unmatched shares need their own current value and maximum-loss calculation.

Include unwind cost

If the second leg fails, the first leg may need to be sold. The unwind crosses the spread and can consume depth in the opposite direction.

Before submitting, estimate:

• current bid for the first leg;
• size available at that bid;
• loss after a one- or two-tick move;
• fee for selling;
• time required to cancel or unwind.

The maximum modeled unwind loss is more useful than the original gross edge when deciding position size.

Stale data can create a completely false edge

If one venue’s quote is several seconds or minutes older than the other, the apparent gap can already be gone. A robust screen should check:

• timestamp for each book;
• market accepting-orders state;
• quote refresh result;
• current available size;
• fee retrieval status;
• outcome-mapping verification.

Hunch uses verification and comparison states to communicate that distinction. A card with different venue prices should remain a comparison until the executable checks pass.

A complete net-edge worksheet

The original screen suggested $5. The amount-specific model loses 40 cents after one plausible move. This example is hypothetical, but the method should be applied to every live candidate.

Operational checklist

• Contract equivalence documented.
• Native outcomes mapped.
• Both books refreshed.
• Equal executable share size calculated.
• Market-specific fees retrieved.
• Funding and approvals ready before the first leg.
• Price bounds set on both legs.
• Maximum unmatched exposure defined.
• Cancellation and unwind route tested.
• Settlement horizon acceptable.

If the trade depends on ignoring one row, the edge is not fully modeled.