The intuitive case for copy-trading rests on a chain of claims: bigger bets reflect more conviction, more conviction reflects more information, and more information predicts better resolved outcomes. The chain has a weak link. The first claim, that bigger bets reflect more conviction, is mostly true. The second claim, that conviction tracks information, is partly true. The third claim, that information predicts outcomes, is true in aggregate but not at the individual- trade level. This article works through what bet size actually predicts in the Rivo dataset, what it does not, and how to use the signal correctly in a copy-trading workflow.
Size as a filter on the noise floor
The most defensible use of bet size is as a noise filter. The population of small trades on prediction markets is dominated by three behaviors that do not carry directional signal: retail traders copying other retail traders, market makers rebalancing inventory across the bid-offer spread, and small-scale arbitrage bots running price discovery between related contracts or across platforms. None of these populations are useful for copy-trading purposes, and each can produce trade volume that looks superficially similar to whale activity.
Above the size threshold, the population thins to a smaller and more concentrated set of bettors. Retail traders rarely take single positions at that size because the position represents a meaningful share of a typical retail bankroll. Market makers occasionally place large single-sided positions, but the round-trip activity that follows is identifiable. Arbitrage bots do not operate at this size because the spreads on prediction markets are not wide enough to support the inventory cost. What remains is the population of directional bettors making deliberate, sized positions against specific binary outcomes. Size is the cleanest single filter that produces this population.
What size does not tell you
Size establishes that the trader cares about the position. It does not establish that the trader is correct.
The Rivo dataset shows substantial dispersion in resolved PnL across wallets that meet the whale-size threshold. The top decile of wallets, ranked by resolved PnL, produces substantial positive returns across categories. The bottom decile produces equally substantial negative returns. The middle of the distribution is roughly break-even before fees. Size identifies the population of whales; it does not identify which whales are skilled.
Size also does not tell the copy-trader the relative risk the whale is taking. A $200,000 position from a wallet operating a nine-figure book is a small percentage of working capital. A $200,000 position from a wallet operating a two-million-dollar book is a materially larger share. The copy-trader cannot read the wallet's total capital from the trade tape, which means the bet size, by itself, does not communicate the level of conviction the whale is expressing relative to their own bankroll. Sizing a copy-trade by ratio to the whale's ticket, without accounting for the whale's underlying capital base, is the most common sizing mistake in copy-trading. The corrective framework is set out in our common mistakes article.
What size does tell you
Three useful inferences are supported by the size of a trade, with appropriate qualifiers on each.
The first is that the trader made an active decision to deploy capital against the contract. The size threshold filters out passive, inventory, and arbitrage flow; the resulting population consists of traders who have actively committed capital to a specific directional view. The commitment alone is information, even before resolution.
The second is that the trade has likely affected the order book. Trades above the size threshold, particularly in less-liquid markets, push the price away from the entry level. The price movement is itself a signal to other market participants, and the follow-on flow that arrives in the next few minutes is often substantial. A copy-trader who acts quickly enters before the follow-on flow has compounded the initial move; a copy-trader who acts slowly enters after, at a worse price.
The third is that the trade is likely to attract further analytical attention. Other traders observe large positions and frequently allocate research time to the corresponding contracts, on the theory that the whale has done work worth examining. The downstream analytical effort can either confirm the whale's view or contradict it; in either case the additional work improves the market's pricing efficiency over the days following the trade.
How to weight wallets correctly
Once size has cleared the filter, the relevant axes for ranking wallets are not further refinements of size but other attributes of the wallet's history.
The first is the resolved-trade count. Wallets with more than twenty resolved positions allow for modestly confident skill inference; wallets below that threshold do not. The number is approximate but the principle is robust: a small sample of resolved positions, however positive, can reflect variance rather than skill.
The second is category-specific PnL. A wallet's aggregate PnL can mask a wallet that is genuinely sharp in one category and consistently wrong in another. Copy-traders should weight wallets by the category in which the copy-trade is being placed, not by their aggregate ranking across all categories. The category-fit framework is discussed in detail in our category-level breakdown.
The third is entry-price discipline. Wallets that consistently enter at contested prices, below 0.30 or above 0.70, and produce positive resolved PnL are operating with a more interesting and more replicable kind of edge than wallets that produce positive PnL primarily from positions near coinflip prices. The first kind of edge transfers to copy-traders more cleanly because the asymmetric payouts on contested- price contracts give the copy-trader more room to absorb the slippage cost of a delayed entry.
The fourth is holding pattern. Wallets that hold to resolution behave differently from wallets that scalp early moves in the price. Both can be profitable, but the strategies require different copy-trading behaviors. Copy-trading a holder by scalping the position is a contradiction; copy-trading a scalper by holding to resolution discards the whale's edge.
The wins that anchor this analysis
The list below shows the largest paying whale trades on record. Each trade started as a whale-sized entry in the Rivo feed, which is the necessary condition for the trade to have been surfaced at all. The trades did not pay because they were large; they paid because the underlying analytical view was correct. Size was the filter that surfaced them; skill was the condition that produced the payouts.
- 1Will United States win on 2026-06-12?Polymarket·culture·stake $920K·@ 46¢+$1.1M
- 2Will IR Iran win on 2026-06-15?Polymarket·politics·stake $790.8K·@ 49¢+$823.1K
- 3Will Spain win on 2026-06-15?Polymarket·culture·stake $78.6K·@ 9¢+$794.4K
- 4Will Spain win on 2026-06-15?Polymarket·culture·stake $47.4K·@ 9¢+$479.4K
- 5UFC 328: Sean Strickland vs. Khamzat Chimaev (Middleweight, Main Card)Polymarket·sports·stake $103.6K·@ 19¢+$441.8K
Reading the leaderboard correctly is the same exercise as reading the whale-size filter correctly: the figures reward size as a necessary condition without treating it as a sufficient one. Copy-traders who internalize the distinction produce materially different portfolios from copy-traders who do not.
Frequently asked questions
Is there an optimal whale-size threshold?
The defensible answer is no, with caveats. Thresholds in the low five-figure range produce broadly similar populations of trades; the choice within that range is more about volume than about quality. Rivo uses an internal default that produces a manageable number of trades per day across both platforms.
Do larger trades produce better outcomes within the whale population?
Modestly, but the effect is smaller than intuition suggests. Within the whale population, the largest ticket sizes are not meaningfully more likely to resolve correctly than the smallest tickets above the threshold. The variation across wallets, weighted by resolved history, is much larger than the variation across ticket sizes within a single wallet's activity.
What about cumulative position size on a single market?
Cumulative size, where a wallet accumulates a large position over multiple trades, is a stronger signal than a single large entry. The accumulation pattern indicates conviction sustained over time rather than a moment of decision. Rivo flags multi-trade accumulation patterns where the data supports the inference.
Can a small bet from a known sharp wallet be more informative than a large bet from an unknown wallet?
Yes, frequently. Wallet identity, weighted by resolved history, often carries more signal than ticket size alone. A small position from a wallet with two hundred resolved positions in the relevant category is typically more informative than a large position from a wallet making its first appearance in the feed.
How should I use the size filter in practice?
As an entry criterion, not as an output. Size determines which trades enter the universe of copy-trading candidates; the subsequent evaluation, based on wallet history, category fit, entry price, and book depth, determines which candidates are worth copying. Treating size as the primary signal for copy-trading decisions is the mistake; treating it as the filter that surfaces the candidates is the correct usage.