Position Sizing for Polymarket Copy Trading Bots: The Math

Why Sizing Beats Selection

Ask ten copy traders what they spend most of their time on and nine will say "finding good wallets to follow." That is understandable — the narrative of discovering the perfect trader is compelling. But the mathematics of compounding tells a different story.

Consider two traders both copying the same wallet with a 58% win rate and a 1.6× average payout on wins. Trader A bets 15% of capital on every trade. Trader B bets 3%. After 100 trades, Trader A is in serious drawdown territory — a 10-trade losing streak, which a 42% loss rate makes nearly inevitable over 100 trials, wipes out 80% of capital at 15% per bet. Trader B ends that same losing streak down 26% and recovers quickly once the wins resume.

Same signal. Same underlying edge. Completely different outcomes — driven entirely by how much was bet, not what was bet on.

This is the central insight of position sizing theory, and it applies directly to Polymarket copy trading. The four models below give you a framework for translating that insight into actual bot configuration.

Model 1: Fixed-Size Copying

How It Works

The simplest possible approach: every copied trade deploys the same absolute dollar amount, regardless of what the tracked trader spent. You set a single number — say, $50 per trade — and the bot uses that number every time.

Pros

• Zero configuration complexity — one number controls everything.
• Predictable capital deployment: you always know your maximum exposure per trade.
• Easy to reason about drawdown: losing 10 consecutive trades costs exactly 10× your fixed size.

Cons

• Ignores conviction. When a tracked trader puts 15% of their capital into a market versus 1%, they are signaling something. Fixed copying treats both identically.
• Scale mismatch. A $50 fixed size on a $500 account means 10% per trade — aggressive. On a $10,000 account it means 0.5% — negligible.
• No self-adjustment. As your account grows through profits or shrinks through losses, the fixed amount doesn't adapt.

When It Works

Fixed sizing works best as a starting point for brand-new accounts where you have no data on the tracked trader's capital base, or when you want strict, predictable budget control above all else. It also works well as a cross-check against more complex models — if proportional sizing returns a number wildly different from your fixed baseline, that discrepancy is worth investigating.

Model 2: Proportional Copying

The Core Formula

Proportional copying solves the conviction problem by treating each trade as a percentage of the tracked trader's estimated capital, then applying that same percentage to your own allocation. The formula is straightforward:

The challenge is estimating the tracked trader's capital. The bot cannot read their bank account. What it can read is their on-chain history: total USDC inflows to their Polymarket wallet, observable position sizes, and current open positions. From these signals, the platform builds an estimate — typically a rolling 90-day average of peak observable capital.

Notice what happened: when the tracked trader expressed higher conviction (5% vs. 2%), your bot automatically scaled up proportionally. The conviction signal is preserved.

Capital Estimation Accuracy

The reliability of proportional copying depends heavily on how accurately the platform estimates the tracked trader's capital. Underestimation makes your trades too large (you end up more aggressive than intended). Overestimation makes them too small (you leave money on the table). Good platforms surface this estimate transparently so you can validate it against public on-chain data and adjust the multiplier accordingly.

Pros and Cons

Proportional copying is the right default model for most users. It scales correctly with account size, preserves the tracked trader's conviction signal, and adapts automatically as both your account and the tracker's positions evolve. The main risk is error propagation: if the capital estimate is wrong, every trade inherits that error. A hard max-cap (discussed later) is essential as a backstop.

Model 3: Kelly Criterion

The Theory

The Kelly Criterion, developed by Bell Labs mathematician John Kelly Jr. in 1956, answers a specific question: given a known edge, what fraction of your capital should you bet on each wager to maximize the long-run geometric growth rate of your wealth?

In prediction market terms, the Kelly formula is:

A Concrete Example

Suppose the tracked trader has a 62% win rate over 300 markets, and their average winning trade returns 1.55× their stake (they buy at around $0.40 on average and the position resolves at $1.00, netting $0.60 profit on a $0.40 bet — odds of 1.5×).

Full Kelly says to bet 36.7% of your capital on each trade. That is almost certainly too aggressive for copy trading, where your edge estimate is inherently imprecise.

Why Full Kelly Is Too Aggressive

The Kelly formula assumes exact knowledge of p and b. In practice, win rates are estimated from a finite sample, and the average payout varies across markets. Any overestimate of your edge causes Kelly to recommend an oversized bet. The result is a mathematically optimal strategy in theory that produces catastrophic drawdowns in practice when the inputs are slightly wrong.

Fractional Kelly: The Practical Version

The solution is to use a fixed fraction of the Kelly recommendation — typically 25% (quarter-Kelly) or 50% (half-Kelly). This dramatically reduces variance while retaining most of the geometric growth benefit.

For most Polymarket copy traders, quarter-Kelly applied to a well-validated tracked wallet is a reasonable starting point — aggressive enough to compound meaningfully, conservative enough to survive estimation error. As your data set on the tracked trader grows, you can migrate toward half-Kelly with higher confidence.

Model 4: Volatility-Adjusted Sizing

The three models above treat every market as equally certain — they size based only on the tracked trader's behavior and your capital, not on the inherent uncertainty of the specific market being copied. Volatility-adjusted sizing fixes this.

The idea: scale your position size inversely to the market's implied uncertainty. A market trading at 0.50 (pure coin flip) carries maximum uncertainty. A market at 0.92 or 0.08 carries much lower uncertainty — the outcome is nearly decided. Betting the same dollar amount on both treats them as equivalent risks, which they are not.

In this formulation, the volatility factor adds a size bonus on markets where the outcome is more decided (lower uncertainty), and applies no bonus on pure coin-flip markets. You can also implement this as a reduction rather than a bonus — shrinking bet size on high-uncertainty markets rather than growing it on low-uncertainty ones. The reduction approach is more conservative and generally preferred.

The Compounding Effect Across 100 Trades

Abstract comparisons between models are hard to evaluate. Numbers are not. The table below shows the simulated terminal capital for each model applied to a $3,000 starting allocation, copying a tracked trader with a 58% win rate and a 1.5× average payout, over 100 trades. The sequence uses the same fixed win/loss pattern for all models.

The numbers are illustrative — real returns depend heavily on the specific win/loss sequence and market payouts — but the pattern is consistent across simulations: proportional and Kelly models compound faster but draw down harder. The right choice depends on your risk tolerance and the statistical reliability of the tracked trader's edge estimate.

Edge Cases: When the Tracked Trader Bets Big

Every sizing model breaks down at the extremes. The specific edge case that trips up most copy traders: the tracked wallet makes an unusually large bet relative to their historical behavior.

Imagine a trader who typically bets 3–6% of capital suddenly deploys 25% in a single market. Under proportional copying, your bot would automatically attempt to deploy 25% of your allocation on that single trade. That might be exactly the right call — the trader saw a rare high-conviction opportunity. Or the wallet was compromised. Or they made an error. Or they have private information you don't.

The bot cannot distinguish between these scenarios in real time. This is precisely why a hard max-cap override is non-negotiable.

The Max-Cap Override

Regardless of which sizing model you use, every copy trading configuration needs a hard ceiling: the absolute maximum USDC the bot will deploy on a single trade, no exceptions.

Setting the right max-cap is a function of your total allocation and risk tolerance. A reasonable starting heuristic: cap each single trade at no more than 7–10% of your total copy trading allocation. On a $3,000 account, that is a $210–$300 ceiling. This ensures that even a perfect storm — an oversize bet from the tracker, a faulty capital estimate, and a string of losses — cannot wipe out your account.

Model Comparison: Six Dimensions

The table below summarizes how each sizing model performs across the dimensions that matter most for Polymarket copy traders.

Implementing in PolyCopyTrade

The sizing models described above map directly to settings in the set up automated Polymarket trading configuration panel. Here is how each option translates:

• Fixed Amount — Set a single dollar value in the "Trade Size" field. The bot ignores the tracked trader's size entirely and uses your fixed number on every copy.
• Proportional (%) — Select "Mirror Percentage" mode. The bot reads the tracked trader's estimated capital, calculates the trade ratio, and applies it to your allocation. You can view and override the capital estimate per tracked wallet.
• Fractional Kelly — Available under "Advanced Sizing." Input the tracked trader's historical win rate and average payout (or let the platform calculate these from their on-chain history), then set your Kelly fraction (0.25 for quarter-Kelly, 0.5 for half-Kelly).
• Max-Cap Override — The "Maximum Per Trade" field applies universally and overrides every sizing model. Set this before going live with any configuration.

Start Proportional, Graduate to Kelly

Position sizing is not exciting. There are no stories about the trader who found the perfect Kelly fraction. But the mathematics are unambiguous: given two traders with identical access to good signals, the one with better sizing will outperform over 100 trades. Every time. The edge compounds.

The practical path forward is clear. Start with proportional copying — it is robust, intuitive, and adapts automatically to your account size. Apply a hard max-cap from day one and do not remove it. Track the tracked trader's actual win rate and average payout over your first 50 trades. When those numbers stabilize, you have the inputs you need to calculate a meaningful fractional Kelly and migrate to it.

Volatility-adjusted sizing can be layered on top of either model as a refinement once the basics are solid. The hierarchy is: get the model right first, then tune the parameters, then add modifiers.

The goal is a sizing engine that you can leave running for months without manual intervention — one that grows with your account when the tracked trader's edge is real, and protects capital when it isn't. All of these controls are available when you configure sizing in PolyCopyTrade. The math is built in.