Position Sizing for Copy Trading

Copy the direction, not the size. A leader's percentage is calibrated to the leader's bankroll, concentration and tolerance for loss. Yours is not. In a binary market, size every position by the dollars you lose if the outcome is zero, then cap how much those losses can add up to across related markets.

The most common sizing error is treating a copied percentage as a recommendation. It is not. It is an artifact of someone else's account.

Fixed versus proportional sizing

Proportional sizing gives each copied trade the same share of your bankroll that it had in the leader's. If the leader puts 5 percent into a market, you put 5 percent in. This preserves the relative shape of the leader's portfolio, which is useful when both accounts are similar in size, holdings and risk tolerance.

It breaks when they are not. Suppose a leader with a $500,000 account puts $100,000, or 20 percent, into one market. You have $10,000 and mirror the 20 percent, putting $2,000 into a single binary outcome. If that outcome loses, you lose 20 percent of your bankroll in one resolution. Two such losses leave you down 36 percent. The leader may be able to absorb that; you may not.

Fixed sizing puts the same dollar amount, or the same maximum loss, into every copied trade regardless of the leader's size. It caps the damage from any one position and makes the risk budget easy to see. Its weakness is that it ignores conviction and price. A fixed $100 at $0.05 buys 2,000 shares with a large potential payout; the same $100 at $0.80 buys 125 shares with a small one. The risk is the same in dollars, but the payoff and probability are not.

A workable compromise is a fixed fraction of your bankroll per position, capped by a maximum loss per market and per event. The leader decides the direction and timing. You decide the dollars.

The binary-payoff math

A Yes share bought at price p costs p and pays $1.00 if the event happens and $0.00 if it does not. The CFTC describes event contracts as typically carrying a fixed payout, usually $1.00. The maximum loss per share is p. The maximum gain is 1 minus p. If you risk L dollars, you buy L divided by p shares. A win produces a profit of L times (1 minus p) divided by p.

The same dollar risk produces very different returns:

• At $0.25, a $100 risk buys 400 shares. A win pays $400, a profit of $300, or 3 times the risk.
• At $0.50, a $100 risk buys 200 shares. A win pays $200, a profit of $100, or 1 times the risk.
• At $0.80, a $100 risk buys 125 shares. A win pays $125, a profit of $25, or 0.25 times the risk.

The price is the market's probability estimate, and the same share count carries very different risk at different prices. Buying 1,000 shares at $0.20 costs $200. Buying 1,000 shares at $0.80 costs $800. If you size by share count, you take four times as much risk on the expensive position without deciding to.

Bankroll math and correlation

Bankroll math

Start with a risk budget. A workable framework for a $10,000 bankroll is 1 to 2 percent of maximum loss per position, 5 to 10 percent per event, and 25 to 50 percent total at risk across all open positions. The rest stays in cash for drawdowns and new opportunities.

At a 2 percent cap, the maximum loss per position is $200. At a price of $0.35, that buys 571 shares, costing $200. If the position wins, it pays $571, a profit of $371. If it loses, you lose $200. At a price of $0.70, the same $200 buys 285 shares. A win pays $285, a profit of $85, and a loss still costs $200.

The dollar risk is identical. The reward is not, because the market thinks the second outcome is more likely.

Correlation is the hidden multiplier

Five positions of 2 percent each are not five independent bets if they all depend on the same event. They are one 10 percent bet with extra steps. A candidate winning the nomination, the same party winning the presidency, and the same candidate winning a state are correlated markets. A team winning a match and the same team winning a tournament are correlated. A weather market for one city and a market for a neighboring city can be correlated.

Before copying, write down the scenario in which every position loses. Sum the maximum loss across that scenario. If the sum exceeds your event cap, reduce the individual sizes or skip the cluster. This is more useful than counting positions, because the number of positions is not the number of risks.

Opposite positions need the same treatment: know your net exposure, not just your gross activity.

A worked example with two traders

You have $10,000. You want to follow two leaders with different concentration habits.

Leader A runs four concurrent positions, each about 20 percent of their account. If you mirrored the percentages, each position would be $2,000 and the four would put $8,000, or 80 percent of your bankroll, at risk. Instead, you cap each position at 2 percent maximum loss, or $200. Four positions put $800 at risk, or 8 percent. If all four win at an average entry price of $0.40, each $200 position returns a $300 profit, for $1,200 total, or 12 percent of the bankroll. If all four lose, you lose $800, or 8 percent.

Leader B runs ten concurrent positions, each about 8 percent of their account. Mirroring would put $800 into each and $8,000 at risk. At a 2 percent cap, ten positions would put $2,000, or 20 percent, at risk, which is above the 12 percent total you set for copied exposure. So you cut each position to $120, or 1.2 percent, for $1,200 total. If all ten win at $0.40, each returns $180, for $1,800, or 18 percent. If all ten lose, you lose $1,200, or 12 percent.

Same bankroll, same leaders, different concentration. The second portfolio has more positions, but its total risk is capped at 12 percent instead of the 80 percent a naive mirror would have produced.

Rules that survive contact with a real order book

• Size by maximum loss, not notional. The notional of a binary position is the stake; the maximum loss is also the stake. Do not double-count it as if it were a leveraged position.
• Cap per market, per event and in total. A per-trade cap alone is not enough when trades are correlated.
• Reduce size when the book is thin. If your order is large relative to visible depth, your exit will be worse than your entry. Polymarket's order book has no fixed size limit, but large orders can move the price significantly.
• Round down. The difference between a $200 cap and a $210 order is not worth the breach.
• Keep cash. A fully invested copy portfolio has no room to survive a drawdown or take a better price later.
• Rebalance with bands, not reflexes. If a position grows to 1.5 times its target, trim it. If it falls to half, do not automatically add; that is averaging down into a loser unless you have a separate reason.
• Treat conviction as your overlay. If you want to overweight one of the leader's trades, that is your decision and your risk, not an instruction from the leader.

What sizing cannot fix

Sizing controls the magnitude of a loss, not its direction. It cannot turn a negative-edge strategy positive, remove resolution risk, make an illiquid market liquid, or protect you from a leader who changes strategy without notice. IOSCO's review of copy trading flags excessive risk taking, unsuitable products and leverage as sources of investor harm; position sizing is the part of those risks you can control directly.

The test is simple. Before you copy, state the maximum dollars you can lose on the position, the maximum you can lose on the event, and the maximum you can lose if several things go wrong at once. If any of those numbers makes you uncomfortable, the position is too large, regardless of what percentage the leader used.