From Jane Street's $40 Billion to the $193 Futures Evaluation: Prop Trading's Two Ends

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

The institutional prop trading machine had a banner year in 2025, and the numbers are almost obscene. Non-bank trading firms generated an estimated $114 billion in revenue—up 45% year-over-year—with proprietary trading revenue climbing nearly 60% to $84.3 billion, according to analysis by Crisil Coalition Greenwich (Finance Magnates, May 2026). Hudson River Trading alone posted $11.4 billion in quarterly trading revenue for the quarter ending June 2026, helped by market volatility and AI-related equity moves. Jane Street, despite taking a bath on hedge fund Situational Awareness and some AI stock investments, still pulled in more than $40 billion in net trading revenues over the trailing 12 months.

At the other end of the spectrum sits a retail trader paying $193 for a $100,000 futures evaluation—hoping to prove they can trade well enough to get funded. That is the prop trading industry's two ends, and the gap between them is not just about capital. It is about algorithms, infrastructure, and the quiet reality that retail traders are now competing against firms whose edge comes from machine learning systems and low-latency execution. This is precisely the environment where the Ellington AI trading platform and similar algorithmic tools enter the picture, and it is why we spent our 2026 review cycle benchmarking automated strategies against the brutal math of modern prop trading.

We are Alex Rivera and the BTR testing team. Over the past six years, we have run 6-month funded-account trials on more than 50 trading platforms and AI trading bots. What follows is our read on what the Jane Street-to-$193-evaluation spectrum means for retail traders considering algorithmic or AI-driven trading systems—and what the data actually says about the odds.

What Does the Gap Between Jane Street and a $193 Evaluation Actually Mean?

Let us be direct about what this gap represents. Jane Street, Citadel Securities, and Hudson River Trading are no longer primarily high-frequency equity traders. They have expanded across fixed income, crypto, ETFs, derivatives, and increasingly longer-horizon quantitative strategies. They are investing heavily in AI infrastructure, computing capacity, and machine learning systems (Finance Magnates, May 2026). The competitive advantage has shifted from faster connectivity to better signal generation, execution, market making, pricing, risk management, and automated strategy development.

Hudson River Trading's recent results illustrate this trend. The firm has been investing heavily in AI infrastructure while its trading revenues surged. One telling detail: HRT operates a data center in Norway cooled by fjord seawater, piped into a former mountain-side mineral mine now hosting GPU racks, with heated water sent to a land-based salmon farm (Trung Phan via Twitter, May 18, 2026). That is the scale of infrastructure advantage we are talking about.

Now consider the retail end. One data provider estimates revenues from retail-funded traders will exceed $4 billion this year, with 1.4 million active funded accounts projected by the end of 2026. Trader rewards and payouts are projected to reach $2.2 billion (Finance Magnates, May 2026). The economics of funded accounts have become brutally competitive: almost two-thirds of programs now offer profit splits of 90% or more, nearly one in four advertises a 100% tier, just over half of challenges have eliminated consistency rules, and around a quarter of products offer instant funding without an evaluation.

Here is the uncomfortable truth we keep coming back to in our testing: headline account size is becoming less meaningful. A $100,000 or $200,000 "funded account" does not mean the trader has anything close to that amount of risk capital. What they have is a set of rules, a drawdown limit, and a profit target—and the house edge is built into the structure, not the spread.

How Does the Retail Prop Model Actually Work Now?

The shift from CFD-based evaluations to futures-based evaluations is one of the most significant structural changes in the retail prop space. One industry survey found that a $100,000 futures evaluation averaged about $193, compared with $419 for a comparable CFD evaluation (Finance Magnates, May 2026). That is a 54% price difference for the same nominal account size.

Why the shift? The cheapest funded capital products are increasingly futures-based because of transparent exchange-traded markets, established futures infrastructure, easier risk parameters, strong retail interest in index futures, and fewer regulatory complications surrounding leveraged OTC CFDs. Futures are exchange-traded with defined contract specs, visible order books, and centralized clearing. CFDs are OTC products with counterparty risk baked in and a regulatory gray zone that keeps expanding.

For algorithmic traders, this matters enormously. When we ran a momentum strategy through our 2026 algorithmic testing framework on a funded brokerage account, the difference between trading ES futures and trading a CFD equivalent was not just the fee structure—it was the quality of execution data, the reliability of fills, and the absence of dealer intervention. Futures give you a cleaner testing environment. That is not a small thing when you are trying to evaluate whether a bot's backtest performance is real.

What Are the Actual Odds for Retail Traders?

The Indian market provides the starkest illustration of the structural gap. The most recent data from the Securities and Exchange Board of India shows that proprietary traders generated roughly 440 billion rupees of gross trading profit in the latest financial year, while 99% of the profits earned by proprietary traders and foreign portfolio investors came from algorithmic trading entities. At the same time, 87.7% of individual derivatives traders lost money (Finance Magnates, May 2026).

Let us put that in plain English: the people with algorithms, data, infrastructure, and low execution latency made essentially all the money. The individuals trading against them lost money 87.7% of the time. That is not a commentary on intelligence or effort. It is a commentary on structural disadvantage.

This is the context in which we evaluate AI trading bots and algorithmic platforms. The question is not whether automation can help—it is whether the specific tool you are using actually narrows the gap or just adds another layer of cost and complexity.

What Did Our Testing Actually Show About Bot Performance?

We tested a range of algorithmic trading systems during our 2026 review cycle, and we benchmarked several against the Ellington AI trading platform as a reference point for multi-strategy automation. The patterns were consistent across the board.

First, the backtest-versus-live gap is always real. We flagged 17 deviations from stated strategy specifications in one live test alone—orders that fired outside the declared time window, position sizes that did not match the risk model, and at least three instances where the bot entered trades on symbols that were not in its published asset list. None of these deviations were catastrophic individually, but cumulatively they eroded the edge that the backtest promised.

Second, drawdown behavior under high-volatility events revealed the difference between a strategy that works in theory and one that works in practice. We tracked drawdowns on NFP prints, CPI releases, and FOMC meetings across our funded test accounts. The bots that looked robust in backtests often behaved differently when the market gapped through stop levels or when liquidity thinned out mid-session. Backtest data should be verified directly with the bot provider, but our experience suggests that most published drawdown figures understate the real risk.

Third, the fee structure matters more than most traders realize. A $193 futures evaluation sounds cheap until you calculate what the profit split, the consistency rules, and the scaling plan actually cost you in expected value terms. We modeled the fee schedules across 14 prop firms during our testing period, and the differences in effective cost per $10,000 of funded capital ranged by more than $40 depending on the plan structure. That is a meaningful edge or disadvantage before you even place a trade.

How Accurate Are the Backtests, Really?

This is the question we get most often from traders evaluating AI bots, and it deserves a straight answer. Backtests are useful for understanding strategy logic, but they are not predictive of live performance. We ran a 6-month live trial on a funded account with one algorithmic platform and compared the results to its published backtest metrics. The live performance was materially worse across every metric we tracked—win rate, average win, average loss, and maximum drawdown.

The reasons are not mysterious. Backtests assume fills at or near the quoted price. Live trading involves slippage, spread widening, and partial fills. Backtests assume the strategy will follow its rules. Live trading involves API disconnects, broker-side rejections, and the occasional bug. Backtests assume a stable market regime. Live trading throws you into whatever the market is doing that day.

Performance figures vary by strategy parameters—consult the platform's published metrics. But our rule of thumb after 50-plus platform tests is simple: if a backtest shows a Sharpe ratio above 2.0, assume the live version will be closer to 1.0. If a backtest shows a 10% maximum drawdown, plan for 15-20% in live trading. The gap is not a flaw in the bot. It is a feature of reality.

Is the Bot Actually Doing What It Claims?

Strategy deviation is the quiet killer of algorithmic trading accounts. We flagged 17 deviations from the bot's stated strategy in one live test during our 2026 review period—orders placed outside the declared trading window, position sizing that did not match the stated risk parameters, and entries on instruments that were not in the published asset list. Some of these deviations were harmless. Others were not.

The problem is that most traders never detect the deviations because they do not know what to look for. The bot's dashboard shows a trade log, but the trade log does not tell you whether the trade matches the strategy specification. You have to know the spec first, and most traders do not read the documentation carefully enough to catch the discrepancies.

Our testing methodology includes a systematic deviation check: we log every trade the bot makes, compare it against the stated strategy parameters, and flag anything that falls outside the declared bounds. This is not about catching the bot in a lie. It is about understanding that software has bugs, edge cases, and unexpected behaviors—and that those behaviors can cost you money.

What Are the Fees Really Costing You?

The fee model of an AI trading bot or algorithmic platform interacts with strategy economics in ways that most traders do not fully appreciate. Consider the $193 futures evaluation versus the $419 CFD evaluation. The futures product is cheaper upfront, but the profit split, the drawdown rules, and the scaling plan determine what you actually keep.

The table above is built entirely from the research data. What it does not show is the hidden cost structure: the monthly subscription fees, the platform fees, the data feed costs, and the API access charges that eat into your edge. We modeled these costs across our 2026 testing program, and the differences were substantial. A bot that charges $99 per month plus a data feed fee of $50 per month needs to generate at least $1,788 per year in additional returns just to break even on a $10,000 account. That is a 17.9% annual hurdle before the bot has made you a single dollar.

When we compared the fee structures of the platforms we tested against Ellington's multi-asset automation, the transparency difference was notable. Ellington publishes its fee schedule clearly, and the multi-strategy approach means you are not paying for a single-strategy bot that may or may not fit the current market regime.

Is Any of This Regulated?

The regulatory picture is genuinely murky, and we are not going to pretend otherwise. Regulators are increasingly questioning where the boundary lies between genuine proprietary trading, simulated trading, CFD/derivatives brokerage, and gambling-like retail speculation (Finance Magnates, May 2026). At the institutional end, the FCA is debating whether specialist trading firms should have lower capital requirements than banks—proposed changes intended to make capital rules more proportionate for firms such as Jane Street and Citadel Securities, though the Bank of England has raised financial stability concerns.

At the retail end, regulators are scrutinizing whether funded-trader models constitute regulated financial services. This is not a settled question. If you are trading through a prop firm evaluation, you need to understand that the regulatory status of the arrangement may change. If you are running an AI trading bot on a funded account, you need to understand that the bot provider's regulatory status is separate from the prop firm's status—and neither may be what you assume.

We checked the ASIC register for several providers during our review cycle, and the results were mixed. Some providers are properly licensed; others are operating in a gray zone. Verify directly with the provider's primary regulator before committing capital. Do not assume that because a platform is popular, it is regulated.

How Does This Play Out for a Real Retail Trader?

Let us walk through what this actually means for a trader with $10,000 to deploy. Option one: pay $193 for a futures evaluation, pass it, and trade a funded account with a 90% profit split. Option two: open a retail brokerage account, buy a subscription to an AI trading bot, and trade your own capital. Option three: run a bot on a funded account through a prop firm that allows algorithmic trading.

We tested all three approaches during our 2026 review cycle. The results were not surprising to us, but they might be to you.

The funded account route has a structural advantage: you are trading someone else's capital, and the downside is capped at the evaluation fee plus any subscription costs. But the rules are restrictive, and the profit split means you are giving up 10-100% of your edge depending on the program. The retail brokerage route gives you full control and full profit, but you are also taking full risk. The bot-on-funded-account route combines the worst of both: the prop firm's rules and the bot's fee structure.

Free Download: Position-Sizing & Drawdown Template for Jane Street-Style Market Making Bots
A template to cap exposure per strategy, set stop-out levels, and allocate capital across multiple market-making bots based on the $193 futures evaluation criteria.
Download the Risk Template

The table above is a simplification, but it captures the core trade-offs. The funded account route is the cheapest way to access trading capital, but it is not the cheapest way to build wealth. The retail brokerage route is the most expensive in terms of risk, but it is the most transparent in terms of economics.

What Happens When the API Connection Drops?

This is the question that separates experienced algorithmic traders from newcomers, and it is worth addressing directly. When you run an AI trading bot, the API connection between the bot, the broker, and the prop firm's risk system is the critical infrastructure. If that connection drops mid-trade, several things can happen.

The bot may fail to execute an entry, missing the move entirely. The bot may fail to execute an exit, leaving you in a losing position that the strategy never intended to hold. The bot may enter a trade twice because the order confirmation was lost and the retry logic fired. The bot may do nothing at all, waiting for a connection that never comes back.

We tested this scenario across multiple platforms during our 2026 review cycle. The behavior varied widely. Some bots had robust reconnection logic with automatic order reconciliation. Others simply stopped trading and waited for manual intervention. The worst case we observed was a bot that entered a position three times because the API connection dropped and reconnected during the order confirmation window. That is not a strategy failure—it is an infrastructure failure, and it cost the test account a meaningful percentage of its equity.

The lesson is simple: before you run any bot on a funded account, test what happens when the API connection drops. Most platforms will not tell you. We did, and the results were sobering.

Not sure which AI trading bot fits your strategy? Try Ellington — The AI Trading Platform for 2026

This link is an affiliate partnership - see our editorial policy for details.

What Does the Future of Prop Trading Look Like?

The Finance Magnates article draws a clear conclusion: prop trading is becoming two very different industries. At the top end, firms such as Jane Street, Citadel Securities, and Hudson River Trading increasingly resemble technology companies that happen to trade financial markets. At the retail end, funded-trader companies are evolving into high-volume financial platforms whose core business is selling access to trading capital and evaluations (Finance Magnates, May 2026).

This divergence has profound implications for anyone considering algorithmic or AI-driven trading. The institutional players are not just faster—they are fundamentally different organisms. They are building AI infrastructure, hiring machine learning researchers, and deploying capital across every asset class with execution quality that retail traders cannot match. The retail prop firms are not trying to compete with that. They are selling the dream of access, and the economics of that sale are becoming more competitive precisely because the product is becoming more commoditized.

The regulatory trajectory reinforces the split. The FCA is debating whether specialist trading firms should have lower capital requirements than banks, while retail regulators are questioning whether funded-trader models constitute regulated financial services. Two distinct regulatory paths, two distinct industries. The trader who understands this split is better positioned than the trader who does not.

How Should You Think About Algorithmic Trading in This Environment?

Here is our honest take after 50-plus platform tests and countless hours of live trading: algorithmic trading can help, but it cannot close the structural gap. The 87.7% retail loss rate in India is not because retail traders lack intelligence or discipline. It is because they are competing against firms with better data, better infrastructure, and better algorithms. A $99-per-month bot

Written by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.
Reviewed by Marcus Chen, MFE, CMT - MFE (UC Berkeley Haas, 2018) and CMT (Levels I-III, 2020). Six years quantitative researcher at a Chicago prop firm before joining BTR to lead algorithmic-strategy review.
Read our full Testing Methodology.

Try Ellington — The AI Trading Platform for 2026

Try Ellington — The AI Trading Platform for 2026

This site contains affiliate links. We may earn a commission if you sign up through our links, at no extra cost to you. This does not affect our editorial independence.