IMC Prosperity Trading Competition Guide

What is the IMC Prosperity Trading Competition?

Prosperity is IMC Trading's online market simulation, built around the decisions that quantitative traders make: valuing assets, quoting prices, managing inventory and finding profitable opportunities. Instead of trading real money, participants earn the game's virtual currency by competing against simulated markets and solving strategic games.

Each main round has two distinct parts. In algorithmic trading, your Python program places orders on a simulated exchange; in manual trading, your team analyses a separate numerical or strategic problem and submits a decision through the platform.

Prosperity also provides a practical introduction to market making, probability, financial modelling and programming. You do not need previous trading employment or advanced financial knowledge to learn from the simulation.

Who can enter IMC Prosperity?

The platform is designed for university students, particularly those studying science, technology, engineering and mathematics. Anyone can use the experience for learning, but competitive leaderboard recognition and prize eligibility have stricter rules.

Eligibility for rankings and prizes

For Prosperity 4, members of teams reaching the final top 25 had to be at least 18 at registration and able to prove that they were enrolled at a university when they registered. Students who graduated after registering could remain eligible.

• University status: all members of a qualifying team needed to meet the student requirement for competitive recognition.
• Previous winners: anyone who had been part of a top 10 team in an earlier Prosperity edition could still play, but their new team could not qualify for rankings or prizes.
• Employment conflicts: current IMC employees, people who had accepted IMC employment or internships, and employees of competing trading or investment organisations were excluded from rankings and prizes.
• Prize regions: cash prizes were restricted to eligible residents of EMEA, North America, South America, India, Australia and Hong Kong at registration.
• Verification: top five overall teams and the best manual trading team had to be available for a participation check.

Eligibility applied to the entire team. One ineligible member could cause every member to lose recognition or a prize, so agree on eligibility before combining accounts.

How to register and form a Prosperity team

When registration opens, create a participant account on the official Prosperity platform and either start a team or join one through an invitation. You can compete alone, but a coordinated team has more capacity to research data, code strategies, solve the manual problem and test submissions.

1. Register on the official platform

   Enter your accurate details and read the event terms. Keep your student enrolment information available if you expect to compete for ranked recognition or prizes.

2. Choose solo play or a team of up to five

   Recruit people with complementary skills in Python, statistics, optimisation and trading. Establish one shared place for code, assumptions and submission decisions.

3. Work through the tutorial and wiki

   Learn how the sample exchange represents order books, fills, positions and trading state. Run a minimal strategy successfully before adding a forecasting model.

4. Assign clear responsibilities each round

   Have someone own the algorithm submission, someone check risk and simulation results, and someone coordinate the manual entry. Everyone should understand the final decisions before the round closes.

5. Submit and verify

   Upload the Python program, inspect its simulation output and submit the manual answer separately. Confirm that the correct versions were successfully received before the deadline.

During Prosperity 4, participants could change teams through the first two rounds, after which membership was locked. That made early checks of team membership and prize eligibility especially important.

What makes an effective team?

A useful division of work is one researcher identifying pricing relationships, one or two programmers implementing and testing strategies, one person analysing manual decisions, and one reviewer monitoring risk and submissions. On a smaller team, combine roles without losing the independent review.

Agree on a final submission owner, a change cutoff and a rule for rejecting untested last minute revisions. A profitable backtest is less valuable than a valid program that actually runs in the competition environment.

How the Prosperity trading rounds work

Prosperity 4 had a tutorial followed by five competition rounds. Rounds 1 and 2 were qualifiers, and teams needed at least 200,000 XIRECs of net profit across that phase to advance; the leaderboard then reset for the final three rounds.

These tasks describe Prosperity 4 rather than a fixed syllabus for all future editions. The recurring skill is to understand a new market quickly, build a defensible model and use it in a constrained trading decision.

The schedule and qualification split

The most recent main competition ran from 14 to 30 April 2026. The first two rounds lasted three days each, followed by an intermission and three shorter final rounds from 24 to 30 April.

The 200,000 XIREC qualification target was specific to Prosperity 4. Profit from the qualifying phase did not carry into the final leaderboard, so teams had to optimise for reaching the threshold first and for final phase performance afterward.

How the algorithmic trading challenge works

The algorithmic challenge asks you to write a Python class called Trader with a run() method. The competition engine calls that method repeatedly, provides the latest market state and executes the orders your program returns.

What information does your algorithm receive?

• Order books: available bid and ask prices and their quoted volumes for each tradable product.
• Current positions: how many units of each product your strategy already owns or owes.
• Recent transactions: your own trades and other transactions exposed through the trading state.
• Product observations: additional data released for a round, such as underlying signals or conversion information.
• Persistent state: a traderData string for saving small pieces of information between calls.

Your output maps each product to orders with a price and signed quantity. Positive quantity means buy, negative quantity means sell; the exchange matches orders against available opposing quotes and applies position limits.

Crossing the spread versus placing a quote

If the best ask is 99 and you submit a buy order at 99, you can trade immediately against the seller at that price. If the best ask is 99 and you instead bid 97, your order rests for the iteration and will fill only if another participant trades against it.

Crossing the spread gives more certain execution but pays for immediacy. Resting quotes offer the chance to earn the spread, at the cost of fewer fills and the risk that your quote is taken when your valuation is wrong.

Position limits and hidden simulation risk

Before submitting, calculate the largest possible final position if every order fills. The engine can reject orders that collectively breach the permitted long or short limit, even when individual orders look safe.

In Prosperity 4, the public test environment simulated 1,000 iterations while final evaluation ran 10,000 iterations. That makes a robust model more valuable than one tuned to a single favourable public run.

During Round 2 of that edition, an additional bid() method set a Market Access Fee. The highest half of bids earned access to 25% more order book quotes, and accepted bidders paid their bid from their round profit.

How the manual trading challenges work

The manual component is a separate economic decision made through the Prosperity interface rather than a Python trading script. Typical problems combine probability, auctions, game theory, pricing or resource allocation, with profits determined after the round ends.

Prosperity 4 included a clearing price auction, a resource allocation challenge, options decisions and a market information game. The right approach changes with the payoff rule, so start by expressing the scoring formula before deciding what to submit.

A practical method for solving manual rounds

1. Translate the rules into a payoff function

   Identify what you choose, what the other teams influence, and exactly when a reward or fee is applied.

2. Separate guaranteed returns from uncertain returns

   List the outcomes you control directly, then assign probabilities or plausible scenarios to the remaining factors.

3. Search the decision space

   For a few discrete choices, calculate each one by hand. For many possible allocations, use a short Python enumeration or simulation outside the manual submission interface.

4. Stress test your assumptions

   Repeat the calculation with conservative and aggressive estimates of competing bids, prices or future events.

5. Submit a single justified decision

   Document the input values, the payoff model and the reason for the chosen action so the team can review it quickly.

The manual result does not alter the simulated order book used by the algorithmic challenge. Treat it as its own opportunity to earn profit rather than an afterthought once the code is complete.

Scoring, leaderboards and competition rules

Teams compete on simulated trading profit, shown in XIRECs. Algorithmic and manual performance contribute to the overall score, while the final ranking for Prosperity 4 followed the simulated account balance after the final phase.

What is PnL?

PnL is profit and loss. In a market making strategy, it depends on money earned from executed trades together with the value of your remaining inventory; a large cash balance alone does not mean you made a profit if you are carrying a costly position.

Watch your inventory, transaction prices and mark value during tests. A strategy that wins many small spreads but suffers one severe inventory loss can finish behind a simpler, more disciplined algorithm.

Why rankings can change after the round closes

Final evaluation takes place on the trading engine after teams submit their strategies and manual answers. A local backtest or public test is feedback on a strategy, not a guaranteed final leaderboard result.

IMC Prosperity prizes and career opportunities

The total cash prize pool for Prosperity 4 was USD 50,000, including USD 45,000 for the top five overall teams and USD 5,000 for the top manual trading score. The figures below describe that completed edition.

Team prizes were divided equally among eligible members. Winners also had to complete IMC's verification and payment process, including the response deadline specified in the prize notification.

Can Prosperity help you get a job at IMC?

Prosperity is an opportunity to demonstrate Python, mathematical judgement, teamwork and an interest in quantitative trading. IMC has featured former contestants who subsequently joined the firm, but placing well is not an internship or job offer.

Keep your submission code, research notes and a short summary of the strategies you tested. They can become a concrete project to discuss in a future trading or engineering application, including what failed and how you improved it.

Worked examples: a trading algorithm and a manual decision

These two EverythingQuant practice exercises use fictional prices, limits and payoffs. They demonstrate how to turn an estimated edge into a valid order and how to choose a manual bid using expected profit.

Example 1: buy an undervalued offer without exceeding your limit

Assume a fictional asset has a best ask of 98, an estimated fair value of 101, and six units available at the ask. You already own seven units and your position limit is ten, so you can buy only three additional units.

The 9 XIREC edge is an estimate, not guaranteed cash profit. If the asset's value falls to 95, those three newly purchased units instead lose 9 XIRECs when marked at that lower value.

Original Python practice strategy

This short example follows the Prosperity Trader.run() interface and makes at most one side of a trade per call. Replace the fictional product, fixed valuation and limit with the values and model for a real practice round.

from datamodel import Order

class Trader:
    def run(self, state):
        symbol = "EXAMPLE_ASSET"  # Fictional practice symbol
        fair_value = 100
        position_limit = 10
        position = state.position.get(symbol, 0)
        book = state.order_depths.get(symbol)
        orders = []

        if book and book.sell_orders:
            best_ask = min(book.sell_orders)
            available = -book.sell_orders[best_ask]
            capacity = max(0, position_limit - position)
            quantity = min(available, capacity)

            if best_ask <= fair_value - 2 and quantity > 0:
                orders.append(Order(symbol, best_ask, quantity))

        if not orders and book and book.buy_orders:
            best_bid = max(book.buy_orders)
            available = book.buy_orders[best_bid]
            capacity = max(0, position_limit + position)
            quantity = min(available, capacity)

            if best_bid >= fair_value + 2 and quantity > 0:
                orders.append(Order(symbol, best_bid, -quantity))

        return {symbol: orders}, 0, state.traderData or ""

This is a basic execution example, not a winning trading model. A stronger submission estimates fair value from market observations, tests the quality of fills, handles every permitted product and considers how risk changes as inventory grows.

Example 2: choose a manual auction bid using expected profit

Suppose you value an item at 130 XIRECs and must submit one bid in a fictional first price auction. A competing bid is equally likely to be 60, 80 or 100, and you pay your own bid only when you win.

Of these three bids, 90 has the highest expected profit because it balances a higher win probability against a smaller gain when successful. In a real manual round, derive the calculation from that round's actual payoff rule rather than assuming every auction uses first price bidding.

How to prepare for IMC Prosperity

Prepare the algorithmic and manual tracks together. A strong trading strategy needs reliable code, a reason to believe its estimated prices, and controls that keep it within the rules; strong manual decisions need the same discipline with probability and payoffs.

A practical preparation sequence

1. Build the Python foundation

   Review dictionaries, classes, loops, basic statistics and reading structured market data. Write and test a function that identifies the best bid, best ask, available size and current inventory.

2. Learn market making and trading risk

   Practise the bid and ask spread, fair value, passive versus aggressive orders, inventory skew and position limits. Work through examples where a seemingly profitable quote becomes a loss when the estimated value changes.

3. Backtest one simple strategy

   Use sample prices to compare a fixed valuation strategy with a moving average. Track total PnL, inventory, number of fills, worst loss and sensitivity to the valuation threshold.

4. Practise auctions, expected value and options

   Solve small bidding games, resource allocation problems and binary payoff examples. For options, learn intrinsic value, payoff at expiry and how changing volatility affects a price estimate.

5. Run a complete simulated round

   Give your team a limited research window, submit one algorithm and one manual answer, and then review the result. Record one fix to test and one assumption that remains uncertain.

EverythingQuant practice benchmarks

• Market reading: interpret 10 fictional order books, identify the best bid and ask, and calculate permissible order size without a position error.
• Programming: implement a strategy that handles missing products, empty books and both long and short limits before moving to statistical models.
• Backtesting: compare at least three parameter choices on separate datasets rather than selecting a winner solely from training data.
• Manual decisions: solve five expected value or allocation problems, showing the objective and the constraints before choosing the answer.
• Submission review: rehearse the final upload and manual entry with another teammate checking input values, output format and file version.

These are preparation targets created by EverythingQuant, not thresholds imposed by IMC. The aim is to make basic mechanics automatic so you can spend the competition analysing the new products and incentives.

Common mistakes that cost teams points

• Overfitting the sample market: choosing parameters that win on one price path but fail when the final simulation behaves differently.
• Ignoring inventory: accumulating a large position for small apparent spread profits and taking an outsized loss when the price moves.
• Skipping manual optimisation: spending the entire round on the Python model while leaving a separate profitable decision unexplored.
• Misreading a fee: treating additional market access as free liquidity rather than comparing its estimated benefit with the price of obtaining it.
• Submitting too late: replacing a working version with an untested algorithm just before the final processing deadline.

IMC Prosperity frequently asked questions

Official resources and further reading

Use the competition's own materials for eligibility, technical interfaces and the rules of a particular edition. These sources also provide the best foundation for recreating historical challenges as practice exercises.

• IMC Prosperity official website: latest event status, overview, participation information and prize details.
• IMC Prosperity terms and conditions: eligibility, ranking, disqualification and cash prize rules.
• Prosperity 4 wiki: tutorial material, round descriptions, game concepts and community resources.
• Writing an Algorithm in Python: trading state structure, order submission and simulation mechanics.
• Prosperity 4 Round 2 specifications: qualifying threshold, position limits, market access fee and resource allocation mechanics.
• Independent participant strategy writeup: a completed team's implementation and reflections across algorithmic and manual rounds.

For a useful next project, reproduce one historical market in a local simulator and measure how your results change when the price path, spread, fill rate and position limit change. Keep the focus on a strategy you can explain and test rather than copying a leaderboard submission unchanged.