Tailing Polymarket's Smart Money — A 334-Trade Paper-Trade Backtest at +46.7% ROI

Quick glossary (read this if "copy trading" is new to you):

• Copy trading / tail strategy — you pick a trader with a good track record and mirror their positions automatically.
• Backtest — running a strategy against historical data to see how it would have performed. Not the same as live performance.
• Win rate (WR) — percentage of positions that end profitable.
• ROI — Return on Investment, your profit divided by what you staked. +46% ROI = you turn $100 into $146 over the measured period.
• Slippage — the gap between the price you see and the price you actually fill at. A real problem when you're 30 seconds behind the whale.
• Selection bias — when you pick the strategy based on recent winners, the future performance tends to regress to the mean. The 75% WR today may be 60% tomorrow.

The pitch is simple. Polymarket is a public market — every trade is on-chain, every wallet is visible. If a specific wallet has a 75%+ win rate on 10+ resolved positions at an average entry price of 0.60 or lower, that wallet has either domain expertise, inside information, or an extraordinary streak of luck. In any of the three cases, copying them ought to be profitable — you're buying into their edge for the price of execution latency.

We wanted to quantify it. Not pitch it, not speculate about it — actually measure, retrospectively, what would have happened if we'd copied these wallets' non-updown trades $100 at a time starting 60–90 days ago.

The result: 75.9% win rate, +$11,258 net P&L on $24,100 resolved stake, +46.7% ROI over 334 positions.

Here's exactly how it was computed and why the real number — if you tried to run it live today — would be smaller.

Building the watchlist

Every 15 minutes our cron recomputes a live watchlist with four filters:

1. ≥10 resolved positions on non-scalp, non-updown markets. This weeds out lucky one-shot streaks.
2. ≥$5,000 in total historical stake on those resolved positions. Ensures actual skin-in-the-game, not a tiny account flexing good aim.
3. ≥75% win rate on those resolved positions. The edge signal.
4. Average entry price ≤0.60. This is the critical filter. It excludes endgame scalpers — traders whose high win rate comes from buying at 0.94 once a market is nearly settled. We want wallets that predict outcomes before the market prices them in, not wallets that arbitrage the final minute.

At the snapshot date of this writing, that filter returns 7 wallets. They are not the same 7 every day — the list rotates as new positions resolve. Some weeks it's 4, some weeks it's 12.

The paper-trade rule

For each watchlisted wallet, every time they buy a position we haven't already mirrored:

• Insert a paper trade at the exact price they paid.
• Notional $100. Shares = $100 / entry_price.
• Log the trade with a unique key (their tx_hash) so duplicate cron runs can't double-enter.
• When the underlying Polymarket market resolves (MAX price on any outcome crosses 0.97), close the paper trade with the actual P&L: shares × 1.0 − $100 if their outcome won, −$100 otherwise.

Everything is zero-slippage, zero-latency, zero-gas. This is the upper bound of the strategy's performance, not a forward-looking expected value. Real execution will underperform.

The backtest, run retrospectively

We executed the rule against the full 90-day history in our database — a one-off backfill that mirrored every qualifying trade those 7 wallets made, then resolved the ones whose markets had already settled.

Headline numbers:

Metric	Value
Paper trades inserted	334
Resolved	241
Still open	93
Wins	183
Losses	58
Win rate (resolved only)	75.9%
Net P&L	+$11,258
Resolved stake	$24,100
ROI	+46.7%

If you had run this $100-per-trade for the 90-day window, you would have deployed $33,400 total (334 × $100), have $9,300 still open, and cleared $11.3k on the $24.1k that actually closed. ROI on resolved capital: +46.7%.

Scaled to $1,000 per trade, +$113k profit. $10,000 per trade, +$1.13M — on $2.4M resolved stake — is genuinely large-fund territory.

Why 75.9% ≠ the scalp trap

Earlier we published a piece showing that 92.1% win rate is a losing strategy when your average entry is 0.957. How can 75.9% win rate be wildly profitable here?

Payoff ratio.

• Scalp-at-0.94: win = +$6.38, loss = −$100. You need to win 15.7 times per loss to break even.
• Insider-tail at avg entry 0.50: win = +$100, loss = −$100. You need to win 1 time per loss to break even.

A 75.9% win rate means 3.15 wins per loss. At roughly equal payoff per trade, your edge is that extra 2.15 wins of clean profit. On $100 notional that's $215 for every 3 wins and 1 loss — or an average +$54/trade, which is ~+54% on $100 deployed (ignoring execution costs).

The math works because the insiders enter before the market agrees with them. They buy at 0.50 what settles at 1.00. That 2x payoff on correct calls is what generates the edge.

The caveats — please read these

We are showing a backtest, not a forward-looking live run. Four reasons the real number will be lower:

1. Selection bias. The watchlist is filtered on current data — meaning the wallets we're following are the ones that have already demonstrated a 75% WR in the sample. Some of them will regress. Some are lucky streaks that the 10-position threshold didn't filter out. Forward WR on the same cohort might be 65–70% instead of 75.9%.

2. Zero-slippage assumption. The paper-trade enters at the whale's exact entry price. In live execution, by the time you see the signal (our cron tick + browser poll, ~10–90s), the market has moved. If the whale's entry caused the movement (likely in thin markets), you'll pay 0.50 + 0.05 = 0.55 instead of 0.50. That's a 10% haircut on every trade, which drags a 46.7% ROI down to roughly 30% ROI.

3. Capacity limit. A wallet buying $25,000 at 0.50 leaves little book depth behind them. If 50 people are all tailing the same wallet, they're competing for remaining liquidity. The first 5 copycats get close to the whale's price; the 50th gets skin the second outcome of a 5-outcome market. The strategy has finite capacity.

4. 93 open positions. Those 93 will resolve over weeks. Some will win, some will lose. Naïvely assuming the same 75.9% WR applies projects ~$4k additional profit on them, but the same selection-bias caveat applies.

The honest expected live ROI — try it yourself

Instead of trusting our caveat math, play with the calculator below. Start with the defaults (zero slippage, zero WR regression — that's our backtest). Then:

• Drag Slippage to +3¢. That's a realistic haircut for copying after the whale's entry. Watch ROI collapse from +46.7%.
• Now add WR Regression of −10 pp. That's selection bias — you picked these wallets because they already printed 75% WR, but forward performance tends to regress. Watch ROI collapse further.
• Now set bet size to $500 and N to 300. That's a live-scale deployment.

The takeaway: with 3¢ slippage and 10 pp WR regression (a conservative, realistic haircut), the +46.7% backtest ROI lands around +15 to +25% — still genuinely positive alpha, but a fifth to half of the paper number.

With 5¢ slippage and 20 pp regression, you're near breakeven. That's what a crowded tail-trade eventually looks like: everyone copies the same wallets, the market front-runs the signal, and edge evaporates.

Most prediction-market strategies are −5% to +3% after costs. +15–25% is genuine alpha. +46.7% is the upper bound. Know the difference before you deploy capital.

What Polyloly actually does with this

The live watchlist + paper-trade infrastructure run continuously in production. Every minute, a Vercel cron recomputes the watchlist, scans for fresh insider entries, and records paper trades. When markets settle, they're resolved and P&L is booked.

The public endpoint polyloly.com/insider-picks surfaces the recent 24h entries from currently-watchlisted wallets — filtered to open positions you can still tail. Free, no signup.

If you want to go deeper: - /wr-leaderboard ranks all traders within a category by P&L - /speed-traders ranks wallets by 15-min price movement after their entries (a different edge signal — latency/news speed) - /alpha shows the cohort ROI heatmap so you understand which categories are worth tailing in the first place

The thesis of the whole stack: Polymarket edge is not a secret. It's public data if you run the analysis. The edge sits in being willing to build the analytic infrastructure and then disciplined enough to act on it.

The trade we're still refining

One open question: how does forward WR compare to backfilled WR for the same watchlist? We need 2–4 weeks of clean live-cron data to answer that. If the live number lands around the backfilled 75.9%, the strategy is real and the caveats we listed are mostly academic. If live lands at 50–60%, it's a nothing-burger and the backtest was selection bias.

We'll publish the live-vs-backtested comparison in a follow-up post in 3–4 weeks. Until then, treat the +46.7% number as an upper bound and size accordingly.

---

About the author

Poly Loly — Prediction Markets Expert

Lead analyst behind Polyloly, a real-time analytics platform tracking whale positions across $1B+ in monthly Polymarket volume. Focus areas: on-chain data aggregation, insider-detection heuristics (80%+ win-rate flags on resolved markets), and market microstructure across political, sports, crypto, and esports prediction markets. Published daily trading-terminal intel, trader leaderboards, and automated alerts via @PolylolyHi.

🌐 polyloly.com · 𝕏 @PolylolyHi · ✉ hi@polyloly.com

---

This article is for informational purposes only and does not constitute investment advice. Prediction markets carry a risk of capital loss.

This article contains affiliate links. If you sign up through our links, Polyloly.com may earn a commission, which helps us produce free analytics. It does not influence our analysis.