Skip to content
TheChatArchiveThe Chat Archive
o3-mini· OpenAIHugging Face LMArena Human Preference 140K

there's a prediction market where each market has a probability and the final payout is…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Geminigemini-2.5-pro-preview-05-06’s answer.
  1. User

    there's a prediction market where each market has a probability and the final payout is proportional to the shares that are purchased. i want to arbitrage two markets that have the same resolution criteria but different percentages. the obvious thing to do would be to buy in each market in opposite directions so that the final probabilty converges, hwever different markets have different liquidity, such that buying the same amount in two different arkets might not yield the same amount of shares. how can I maximise profit ?

  2. Assistant

    Below is a high‐level roadmap for thinking about and setting up an optimal arbitrage strategy when you’re dealing with two markets on the same event that have different “prices” (i.e. implied probabilities) and – importantly – different liquidity characteristics. In short, you don’t just “flip a coin” by buying one share of A and selling one share of B; you have to size the positions carefully so that your final payoff is hedged (that is, independent of the event’s outcome) and so that you take full advantage of the mispricing given the amounts you can move at each market’s price.

    Let’s break this into steps.


    1. The Basic Idea of the Arbitrage

    Assume that both markets settle on the same binary outcome (e.g. “the event happens” gives a $1 payout per share, “it doesn’t” gives $0) but they offer different implied probabilities. For example, Market A might be pricing the event at 40% (so a share costs about $0.40) while Market B prices it at 50% (about $0.50 per share). In an ideal friction‐free world you could buy “undervalued” shares (those costing less than their “true” probability) and “sell” or hedge the “overvalued” side so that whatever happens, you have a net profit. With identical liquidity and no market impact this might mean “buy one share in A and sell one share in B” – but with liquidity differences the same dollar amount moved in A and in B does not translate into the same number of shares.


    2. How Liquidity Changes the Game

    Each market’s liquidity tells you how many shares you can buy (or sell) before your own orders start “eating into” the order book and moving the price. For example, if Market A is very liquid you might be able to buy a large number of shares near $0.40, but if B is not as liquid a large sell order might move the price from $0.50 upward (or “widen the spread” for the side you need).

    If you “buy one share” in Market A for $0.40, that might cost exactly $0.40; but “selling one share” in Market B might only get you $0.50 for a small volume, yet selling larger quantities may fetch less than $0.50 per share because you start filling higher‐priced (or “worse”) offers in the order book. So the profit you earn from the mispricing is not simply the difference in percentages; it depends on how many shares you can trade at those “good” prices.


    3. Setting Up a Hedged Position

    To earn a risk‐free arbitrage profit, you want your net payout to be independent of the resolution. Suppose you plan to:

    • Go long (buy) the “cheaper” side (say Market A), and
    • Short (hedge or sell) the “expensive” side (Market B).

    Let:

    • ( x ) = number of shares (or fraction thereof) you ultimately buy in Market A.
    • ( y ) = number of shares you sell (or short) in Market B.

    At resolution, if the event happens you’ll receive $1 per share in the market where you’re long and owe nothing on the short side (if structured as a “binary option” for which shorting makes sense), or vice versa if you set up complementary contracts. (Depending on market rules you might “sell” a contract and then later buy it back; the idea is to lock in a net cash flow that’s positive regardless of outcome.)

    To be fully hedged you want the final net “payoff” in each state to be equal. That is, if the event happens:

    • From Market A, you get roughly ( x ) dollars.
    • In Market B, your short position costs you ( y ) dollars since you will have to settle the difference.

    To “cancel out” the risk, you must choose ( x ) and ( y ) so that—after accounting for the costs of getting those shares—the outcomes are balanced. With no slippage a natural idea is to set ( x = y ). However, due to liquidity differences the cost to acquire ( x ) shares in one market and the price you get for selling ( x ) shares in the other may not be linear; you may need a different ratio.

    In practice:
    You can view the order book in each market as offering a price impact function (or a marginal cost function). Denote:

    • ( C_A(x) ) = the total cost to buy ( x ) shares in Market A.
    • ( R_B(y) ) = the total revenue (or cash inflow) from selling ( y ) shares in Market B.

    Your “arbitrage profit” if you end up with a settlement payoff of ( x ) dollars (if the outcome is “yes”) is something like: [ \text{Profit} = x + R_B(y) - C_A(x) ] if the hedge is arranged to cancel the exposure. (When the event does not happen, you get nothing from the long side and similarly settle the short side so that the net cash flow remains identical.)

    Your task is to choose ( x ) and ( y ) so that:

    • The net exposure (the difference between the number of “yes” shares you have long and those you’re short) is zero.
    • The arbitrage profit (the difference between what you earn and what you spend) is maximized.

    Because the liquidity functions ( C_A(\cdot) ) and ( R_B(\cdot) ) are not necessarily linear, you end up solving an optimization problem.


    4. Maximizing Profit: The Optimization Step

    Step A. Model the Order Books

    1. Quantify the liquidity:
      For each market, determine how the effective price changes with the order size. That is, if you want to move ( dx ) shares, what is the marginal price? For a simplified model, you might assume:

      • Market A: Effective cost per share is ( p_A(x) ) (which might start at the quoted price and then slowly increase as you buy more).
      • Market B: Effective revenue per share is ( p_B(y) ) (starting at its quoted price and then decreasing as you sell more).
    2. Express the cumulative cost/revenue functions:
      [ C_A(x) = \int_0^x p_A(u),du, \quad R_B(y) = \int_0^y p_B(v),dv. ]

    Step B. Set Up the Hedging Equation

    You want the net “exposure” (in terms of final payout) to be zero. One way to do this is to choose ( x ) and ( y ) so that the amount of “winning shares” purchased in A exactly offsets the “selling exposure” taken in B. (Depending on whether you have to buy back the short or if it’s a contract that settles at a fixed amount, the details may vary.)

    Step C. Maximize the Net Cash Profit

    Your net arbitrage profit might look something like: [ \text{Net Profit} = R_B(y) - C_A(x) ] subject to the constraint that your final exposures match (e.g. ( x = y ) or more generally, a linear relationship ( x = \alpha, y ) based on payout kinetics).

    Because the functions ( C_A ) and ( R_B ) come from the order book data, you can increase ( x ) (and ( y )) gradually and check how the incremental cost of buying in Market A compares to the incremental revenue of selling in Market B. In practice, you’ll find that as you push larger orders, the marginal cost and revenue “convergence” might eventually erode your arbitrage margin.

    Step D. Find the Optimal Trade Sizes Given Liquidity

    The idea is to “stack” your orders incrementally:

    • Start with a small trade size and compute the risk-free payoff.
    • Increase the size while monitoring how the order books adjust (i.e. monitor the slippage).
    • Continue until the incremental trade would no longer add to the overall risk-free profit.

    Mathematically, if you could differentiate the cumulative functions, the optimal ( x^* ) and ( y^* ) often satisfy a condition like: [ \frac{dR_B}{dy}\bigg|{y=y^*} \approx \frac{dC_A}{dx}\bigg|{x=x^*} ] subject to the exposure-matching condition. In words, you push orders up to the point where the marginal revenue in the “expensive” market equals the marginal cost in the “cheap” market.


    5. Taking Market Impact and Execution Strategy Into Account

    In the real world:

    • Market Impact: Your own orders will change the prices as you trade. You need to use sufficiently “small” orders or a smart algorithm that slices orders into smaller pieces.
    • Liquidity Constraints: You might ultimately be limited by the less liquid market. In other words, even if Market A can absorb huge orders at a good price, you can only hedge as many shares as Market B will allow at or near the quoted price (and vice versa).
    • Transaction Costs/Fees: Always include fees or other costs in your calculations.

    A good execution strategy is to use order book information to simulate your total execution cost on both sides before committing to a full arbitrage order. Often, you’ll use a computer algorithm that continuously monitors the available liquidity and dynamically adjusts the order sizes in each market to maximize the net profit while keeping your overall position hedged.


    Summary

    To maximize your arbitrage profit when the two markets have different liquidity:

    1. Model the order books: Determine how many shares you can trade at each incremental price level.
    2. Set up your hedge: Choose trade sizes (say ( x ) in Market A and ( y ) in Market B) so the final net exposure is zero.
    3. Optimize the trade sizes: Incrementally increase your orders and compute the net profit:
      • The profit is given by the difference between the cumulative revenue from the “overpriced” side and the cumulative cost on the “underpriced” side.
      • Stop when the marginal benefit from an additional share on one side equals the marginal cost on the other.
    4. Take execution issues into account: Use smart order-slicing strategies to minimize market impact and account for transaction costs.

    By solving this “mini‐optimization” problem with the actual liquidity data in hand, you can identify the optimal amounts to trade in each market to extract the maximum risk-free profit.

    This framework gives you the idea of how to approach the problem. The exact numbers and ratios will depend on your high‐frequency access to the order book and the dynamic liquidity available in each market.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.