Author: Jeff, IOSG
Abstract
Bots stay in their structural advantage zone; the human capital that runs bots will continue to migrate toward the next event; what ultimately wins is a platform that can host two types of traffic—bots and humans—in the right proportion across the greatest number of market types. The core data fueling the prediction market bot panic is quite straightforward: 5% of wallets on Polymarket—identified as likely bots—account for 75% of the platform's volume. Since January 2025, 823 wallets have each netted over $100,000, collectively withdrawing $131 million in profits from Polymarket. Of the top 20 profit-earning wallets, 14 are classified as bots (per Stacy Muur's leaderboard inspection). A University of Toronto study (covering 2.4 million users and $67 billion in volume since 2022) found that 68.8% of users are in the red, while the top 1% capture 76.5% of all profits.
The narrative that follows from this: prediction markets are a wealth transfer machine, and bots are its operators. The data is accurate, but the framework is half wrong.
Core Thesis
1. The central flaw in the bot narrative is conflating "volume concentration" with "capital extraction." The fact that 5% of wallets on Polymarket drive 75% of volume only speaks to the distribution of account activity; it doesn't directly imply that retail capital is being drained by bots.
2. Cohort-level data is more compelling. AI agent wallets have a win rate of about 37%, while human wallets hover at 7%-13%. This 3-4x gap at the cohort level is real evidence of a structural advantage. Meanwhile, the 14-out-of-20 robots in the profit leaderboard (Stacy Muur leaderboard inspection) is simply the right-tail projection of this distribution, not independent evidence.
3. Bots' advantage lies in structural dimensions, not judgment dimensions. The three market types bots dominate—price feed latency arbitrage, real-time sports state automation, and cross-platform portfolio arbitrage—share a common trait: no need to make judgments about real-world events themselves. Once market outcomes depend on synthesizing multi-source information, bots' advantage systematically weakens.
4. Polymarket's category mix has shifted over the past 12 months from "Politics 42%" to "Sports 50%," and the fastest-growing categories are precisely the long-cycle event markets where bots don's structurally dominate. The platform's retailification trend is clear.
5. Forward-looking judgment: The bot share will continue to rise as deployment costs fall, but the scale of capital extraction by bots from humans will peak before the bot share does—because bots cannibalize each other faster than they cannibalize human accounts.
6. Investment strategy: Equity opportunities at the platform layer (Kalshi + Polymarket, 97%+ combined share) are largely closed. Value is migrating upward to the L2 agent infrastructure layer (Olas / Valory model) and the venue-agnostic middleware layer. C-end bot products and the L3 data/pricing layer are not venture-fit.
1. Track Scale Outweighs Bot Panic
Three quantitative anchors define this report's scope of discussion.
First, on April 14, 2026, Bernstein revised its 2026E prediction market track size estimate to $240 billion, with the path to $1 trillion by 2030 becoming a consensus on the sell-side.
Second, the combined YTD volume of Kalshi and Polymarket surpassed $60 billion by mid-April 2026, already exceeding the entire track's total of $51 billion for all of 2025.
Third, Robinhood has listed 1,000+ Kalshi contracts, with 1 million+ clients trading 90 billion contracts cumulatively. Robinhood's prediction market business has an ARR of approximately $350 million, $150 million for all of 2025, and is projected at $586 million for 2026E—making it the company's fastest-growing product line.
These data points collectively point to one conclusion: prediction markets are no longer a purely crypto-native track; their character is closer to a TradFi distribution problem. The "retail being pillaged" group assumed in the bot narrative is primarily not crypto users, but retail investors entering through traditional brokerage channels.
From this, we can deduce a contextual bias in the bot panic: the track is not having value drained by automation; it's having traffic injected by mainstream finance at a pace far exceeding any automated extraction.
2. The Data That Really Matters: 37% vs. 10%
The most cited data point in the bot narrative suffers from sample selection bias.
The source data for "14 out of the top 20 profit wallets are bots" presupposes a small sample already sorted by profitability. This sample only reflects bots' presence in the distribution's right tail and cannot be used to infer cohort-level performance differences.
Cohort-level data (source: Polystrat / Valory disclosures, cross-validated with multiple Polymarket on-chain analytics):
A 3-4x win rate difference at the cohort level is the true reflection of bots' structural advantage. The 14/20 in the profit leaderboard should be understood as a downstream manifestation of that win rate distribution, not independent causal evidence.
3. Where Bots Win
Bot extraction is highly concentrated in three market types. Their commonality: they require no subjective judgment about real-world outcomes, relying instead on latency or pricing advantages relative to the platform's matching engine.
Price Feed Latency Arbitrage
Representative case: wallet 0x8dxd turned $313 into $437,600 in January 2026, trading only 15-minute BTC up/down contracts with a 98% win rate.
Strategy: monitor spot prices on Binance and Coinbase, and open positions when Polymarket's quotes lag behind CEXs. On January 7, 2026, Polymarket introduced a taker fee for 15-minute crypto contracts (peaking at ~3% near 50% probability), specifically to neutralize this strategy. The wallet's cumulative win rate has since dropped to 54.7%.
Conclusion: bots' advantage in feed-based markets is real but confined to a very narrow time window and is significantly compressed once the platform introduces friction costs.
Real-Time Sports State Automation
Data source: cancun2026 team's Polymarket wallet classification (Dune query 6648075, https://dune.com/queries/6648075, past 7 days, as of May 11, 2026).
Source of advantage: bots react to in-game events significantly faster than retail users relying on live streams (30-second delay). Additionally, trading terminals like Kreo and PolyCop open this advantage to non-programmers through copy-trading and auto-follow features, meaning the measured bot share includes human capital routed through bots.
Cross-Platform Portfolio Arbitrage
Data source: IMDEA Networks paper "Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets" (AFT 2025, dspace.networks.imdea.org/handle/20.500.12761/1941).
The study covers approximately $40 million in arbitrage extraction on Polymarket between April 2024 and April 2025, primarily via two modes: (1) rebalancing YES/NO shares within the same market, and (2) cross-platform portfolio trades (buying YES on Polymarket, buying NO on Kalshi, entering when the sum of implied probabilities is less than $1). This mode has rigid requirements for multi-platform infrastructure and compresses as the matching engines of various platforms converge.
4. Where Human Accounts Win and Their Constraints
The categories with the lowest bot share are not due to "retail picking better picks" but because "profitability in these markets depends on the ability to synthesize multi-source real-world information"—a domain where automation is structurally disadvantaged.
Two independent studies confirm this.
Joshua Della Vedova (University of San Diego) on-chain behavior study (jdellavedova.com) notes: retail users pick winning outcomes more often than bots; bots' advantage lies in execution—when retail buys YES at $0.72, bots have already entered at $0.55, securing an unrealized gain of $0.17 per share.
A working paper from University of Toronto / HEC Montréal / ESSEC (Akey et al., SSRN 6443103, March 18, 2026) finds that 56% of losing users' order prices fall in extreme ranges (<10¢ or >90¢), while only 28% of the top 0.1% of profit earners place orders in extreme ranges. Losing users typically "chase 5-cent bets for 20x" or "chase 95-cent near-certainty," while profitable users build positions in the middle of the probability curve.
Both studies point to the same conclusion: retail judgment is generally underestimated, but their execution timing and order structure are systematically weak.
5. Forward Path: Four Forces Shaping the Bot/Human Landscape
The key variable over the next 12-24 months is not the current bot/human ratio, but its evolutionary direction. This report identifies four forces whose directions are not aligned.
Bot Deployment Costs Collapse Further
Coding agents like Claude Code and Codex, open-source frameworks like Hermes, and Polymarket's own MIT-licensed Polymarket Agents framework have collectively lowered the engineering barrier for strategies like 0x8dxd from "serious project" to "weekend prototype." Copy-trade services further channel human capital into bot infrastructure, mechanically inflating the measured bot share.
Individual Bot Yields Are Cannibalized by Peers
The 823 profitable bot wallets are the right tail of a much larger population of losing bots. As the number of wallets employing similar strategies rises, each bot's profitable window narrows. 0x8dxd's 98% win rate is structurally unrepeatable—not because inefficiencies disappear, but because of competition from peers plus platform fee adjustments. The scale of bot capital extraction from humans is likely to peak before the bot share does.
Platform Category Mix Shifts Toward Retail
Polymarket's category mix in April 2026: Sports 50%, Crypto 24%, Politics 16%, Other 10%. In April 2025: Sports 29%, Crypto 12%, Politics 42%.
Sports volume in absolute terms grew 11x YoY. New volume primarily falls into long-cycle event markets where retail dominates. Bernstein expects sports' share of track volume to decline from the current 62% to 31% by 2030, filled by economic, political, and corporate event contracts—a structural migration that further expands the category exposure where bots don't dominate.
Different Platforms Naturally Segregate by Category
Hyperliquid's HIP-4 went live on May 2, 2026, offering daily BTC binary contracts with zero opening fees, USDH collateral, unified perpetual/spot markets, and a validator-slashable market deployment mechanism (1 million HYPE per slot, approximately $42.76 million at current prices).