Whale Tracking and Copy Trading: Following Smart Money on Predite
Why follow smart money at all
In most markets, you can't see who is on the other side of your trade. Prediction markets are different. Because Polymarket settles on-chain and Kalshi publishes detailed trade data, the footprints of the largest, most consistent traders are visible if you know where to look. A wallet that quietly bought 80,000 shares of a "YES" contract at 31 cents three days before the price ran to 60 isn't a rumor — it's a record.
That visibility is the whole premise behind Predite's Whale Tracker and Copy Trading tools. The idea isn't to worship big accounts. Plenty of large wallets are reckless, lucky, or actively trying to shake out smaller traders. The idea is to *systematically separate signal from noise*: find the wallets whose edge survives scrutiny, understand what they're actually doing, and decide — deliberately — whether and how much to follow.
This guide walks through what the Whale Tracker detects, how Predite scores individual traders, how to evaluate a wallet before you risk a dollar behind it, how to configure copy trading, and the failure modes that separate people who profit from this from people who get dumped on.
What the Whale Tracker detects
The Whale Tracker is a real-time monitor sitting on top of Polymarket's CLOB and on-chain fills plus Kalshi's trade feed. It surfaces three distinct things: tiers, clusters, and moves.
Tiers
Not every large trade matters equally, so wallets are bucketed by size and behavior into tiers:
- •Minnow — sub-$1k positions. Mostly filtered out, but tracked for context.
- •Dolphin — roughly $1k–$10k per position. Often sharp retail or semi-pro.
- •Shark — $10k–$50k. Frequently the most *informative* tier, because these traders are big enough to have done real work but small enough to still be hunting for edge rather than just parking capital.
- •Whale — $50k–$250k. Capital that moves prices on entry.
- •Leviathan — $250k+ single positions. Rare, market-defining, and worth understanding even when too large to copy responsibly.
Tiers are about position size in a specific market, not total bankroll. A trader with a $2M bankroll taking a $5k position shows up as a Dolphin on that trade — which itself is information (low conviction or a small probe).
Clusters
A single big buy can be noise. Five independent wallets buying the same side of the same contract within a short window is a cluster — and clusters are where the real signal usually lives. The tracker flags when multiple high-scoring wallets converge on the same market and direction, especially when they don't normally trade together. Coordinated accumulation by uncorrelated smart-money wallets is one of the strongest pre-move tells available in these markets.
Predite annotates each cluster with how many distinct wallets are involved, their combined notional, the average trader score in the group, and how tight the time window was. A cluster of seven Sharks averaging an 80+ score, buying YES across 40 minutes, reads very differently from one Leviathan and a handful of minnows.
Real-time moves
The move feed is the live stream: large fills, sudden position flips (a wallet that was long flipping short), and aggressive sweeps that eat multiple levels of the order book. Each entry shows the wallet, tier, market, side, size, fill price, and the resulting order-book impact. Real-time move alerts (push/webhook) are a Pro feature; Starter sees the feed on a short delay. The Bot plan adds programmatic access to the same stream over the API so you can wire moves into your own logic.
How trader profiles are scored
Raw size tells you who is *loud*, not who is *good*. To find who's good, Predite builds a profile for every wallet with enough history and condenses it into a 0–100 Trader Score. Think of the score as a weighted blend of several independent components, each of which you can also inspect on its own.
- •Win rate — the share of *resolved* positions that closed profitable. Useful but easy to misread on its own (more on that below).
- •Profitability / realized P&L — actual money made, sized appropriately. A 55% win rate that compounds large EV-positive bets beats a 70% win rate that scalps pennies.
- •Consistency — how *stable* returns are over time. A wallet that grinds steady gains across 200 trades scores far higher here than one whose entire P&L came from a single lucky 2024 election bet. Consistency is essentially the antidote to survivorship bias.
- •Risk discipline — drawdown behavior, position sizing relative to bankroll, and whether the trader cuts losers or rides them to zero.
- •Recency — recent performance is weighted more heavily than results from a year ago. Edges decay; the score reflects that.
- •Sample size — a 90% win rate over 11 trades is statistically meaningless. The score is *shrunk toward the mean* until a wallet has enough resolved positions to trust, so small-sample flukes don't rocket to the top of the leaderboard.
A concrete example. Wallet A shows a 78% win rate but only 14 resolved trades, all in one category, with a single position responsible for most of the gains. Wallet B shows a 58% win rate across 240 resolved trades, positive P&L in 9 of the last 12 months, modest drawdowns, and clear specialization. Despite the lower headline win rate, Wallet B will score meaningfully higher — and Wallet B is the one actually worth following.
The score is a starting filter, not a verdict. Use it to narrow a leaderboard of thousands down to a shortlist of a dozen, then do the real work by hand.
How to evaluate a wallet before copying it
This is the part most people skip, and it's the part that matters most. Before you attach real money to a wallet, open its full profile and work through five questions.
1. Is the win rate real, or is it an artifact?
Look at win rate *together with* average odds. A trader who only buys heavy favorites at 90 cents will have a gorgeous win rate and almost no edge — winning 90% of bets that pay 11 cents on the dollar is roughly break-even before fees and slippage. Conversely, a 45% win rate on contracts bought at 30 cents can be enormously profitable. Win rate without average entry price is close to useless. Check that the wallet's profitability is driven by EV, not by cherry-picked near-certainties.
2. What's the average position size — and can you match its shape?
A wallet's average size tells you how it sizes conviction and whether you can realistically follow it. If a Whale routinely takes $120k positions and you're copying at $300, you'll be a rounding error — fine for you, but understand that your fills will often be worse than theirs because they get the good part of the book and you get the leftovers. Also watch size *variance*: a trader who bets $5k on probes and $60k on high-conviction plays is telling you which signals to weight, and a good copy setup should preserve that proportionality rather than flattening every trade to the same dollar amount.
3. How bad do the drawdowns get?
Pull up the equity curve and the maximum drawdown. A wallet up 300% on the year that endured a 60% peak-to-trough drawdown is a very different psychological and financial commitment than one up 80% with a 15% max drawdown. When you copy a wallet, you inherit its drawdowns — including the next one, which hasn't happened yet. Ask honestly: if this wallet immediately entered its worst historical drawdown the day you started copying, would you stick with it or panic out at the bottom? If the answer is panic, the wallet is too volatile for you regardless of its score.
4. Is there genuine category specialization?
The strongest edges are narrow. A wallet that is 65% on resolved political markets across 180 trades but barely break-even on sports and crypto has a *real, identifiable* edge in politics — and you should consider copying only its political trades. Predite breaks P&L and win rate down by category (politics, sports, crypto, economics, pop culture, and so on) precisely so you can see whether someone is a specialist or a generalist getting carried by one good vertical. Copy the edge, not the whole wallet. Use category filters to follow a trader only where they've demonstrably earned it.
5. How is it making money — and does that style fit you?
Skim the recent trade history. Is this a patient accumulator who builds positions over days and holds to resolution? A fast scalper flipping intraday on news? A contrarian who fades crowded markets? An arber exploiting price gaps between Polymarket and Kalshi? Each style implies different demands on *you*. A scalper's edge can evaporate in the seconds between their fill and yours; a multi-day accumulator is far more forgiving of copy delay. Match the wallet's clock speed to what your setup — and your attention span — can actually keep up with.
Setting up copy trading
Once you've shortlisted wallets that survive that scrutiny, copy trading automates following them. Live copy trading is a Pro and Bot feature; the Bot plan adds API-driven execution and higher position limits. Paper trading is available on every plan, and you should use it first.
To configure a copy stream:
- Open the wallet's profile and select "Copy this trader." You can copy more than one wallet at once; each becomes its own independent stream with its own rules.
- Set your copy percentage. This is the core control. A copy % of 1% means that for every $10,000 the source wallet commits to a position, you commit $100. The percentage preserves the trader's *relative* sizing — their big-conviction bets stay proportionally bigger than their probes — while scaling the absolute dollars to your bankroll.
- Set a max position size. A hard ceiling per trade, independent of copy %. This caps the damage from any single move and protects you specifically from the moment a whale takes an outsized, out-of-character swing. For example, copy % might call for $400 but a $250 cap holds you there.
- Apply filters. Restrict the stream by category (e.g., politics only), by minimum trade size on the source side (ignore their sub-$2k probes), by platform (Polymarket only, or include Kalshi), by side (some traders are sharp on the buy but sloppy unwinding), and by minimum source-wallet score if you're copying a basket. Filters are where copy trading goes from blunt to surgical.
- Choose paper or live. Paper trading mirrors every action with simulated fills, modeled slippage, and modeled delay so you see realistic — not idealized — results. Run any new copy configuration on paper for at least a few weeks spanning real volatility before committing capital. The point isn't to confirm the wallet is good; it's to confirm *your specific settings* behave the way you expect.
- Set notifications and a kill switch. Decide how you're alerted on each copied entry and exit, and predefine the condition — usually a drawdown threshold — that automatically pauses the stream. Set the kill switch when you're calm, not mid-drawdown.
A worked example. You shortlist a Shark scoring 84, specialized in economic-data markets, $18k average size, 22% max drawdown. You set copy % to 0.5%, max position $300, filter to economics-only and source trades above $5k, and run it on paper for a month. The simulation shows you'd have caught 9 of their 11 qualifying trades, missed 2 to delay, and netted a modest gain after modeled slippage. Now you have an evidence-based reason to go live — and realistic expectations about the fills you'll miss.
Slippage and delay: the gap between their fill and yours
Copy trading is never a perfect mirror, and pretending otherwise is how people lose money. Two frictions always sit between the source trade and yours.
Delay. There is unavoidable latency between a wallet's fill, Predite detecting it, and your order reaching the book. In slow-moving markets this is negligible. In a market reacting to breaking news, the price can move several cents in that window — and you may be buying *after* the very move the whale's entry caused. This is exactly why fast-scalper wallets are dangerous to copy and patient accumulators are forgiving.
Slippage. You rarely get the source wallet's exact price. They may have swept the best levels of the order book on entry; you're filling against what's left, at a worse average price. The thinner the market and the larger your (or their) order, the worse this gets. Predite's paper mode models both effects, and live mode lets you set a max slippage tolerance — if the price has run past your limit by the time your order arrives, the copy is skipped rather than chasing a trade whose edge is already gone. Skipping a trade is frequently the correct outcome. Consider using TWAP-style splitting on larger copied entries in thin markets to reduce your own footprint.
The risks nobody should gloss over
Copy trading packages real risk in a convenient wrapper. Be clear-eyed about it.
- •Past performance is not predictive. Every score and curve is backward-looking. A wallet's edge can be structural and durable, or it can be a regime that's already ending. Recency weighting and consistency scoring reduce the odds you're chasing a fluke — they don't eliminate them.
- •They may dump on you. This is the sharpest risk in any market where positions are visible. A whale that knows it's being followed can build a position, let copiers pile in and push the price, then sell *into that demand* — leaving followers holding a worse entry as the price reverts. Diversifying across several uncorrelated wallets, capping per-trade size, filtering out tiny probe trades, and being skeptical of wallets whose entries are suspiciously easy to front-run all reduce this exposure.
- •Crowding kills edge. The more capital that copies a wallet, the worse everyone's fills get and the faster the edge decays. An edge that worked when one person followed it can vanish once a thousand do.
- •Correlated baskets aren't diversified. Copying ten wallets that all trade the same political markets the same way is one bet wearing ten costumes. Diversify across *styles and categories*, not just wallet count.
- •Automation compounds mistakes. A bad filter or an oversized copy % executes flawlessly and repeatedly until you notice. Start small, watch the first live trades closely, and keep the kill switch armed.
Treat copied trades as part of a portfolio you own, not a hands-off subscription. The wallets do the analysis; the risk, and the responsibility for sizing it, stay with you.
Start on paper, then scale in
The disciplined path through all of this is short: shortlist a few wallets that survive real scrutiny, copy them at a small percentage with a hard max-position cap and tight category filters, run the whole thing on paper through a stretch of genuine volatility, and only then route real capital — scaling in slowly as the live results confirm what the simulation promised. Open the Whale Tracker, sort the leaderboard by Trader Score, pick one specialist whose drawdowns you could actually stomach, and spin up a paper copy stream today. A month of simulated following will teach you more about smart money — and about your own risk tolerance — than any amount of reading, and it costs you nothing but patience.