Quant Trading: What It Is, What Drives Results, and How to Verify a Track Record
Quant trading attracts professional investors for understandable reasons. The pitch is disciplined, data-driven, repeatable. That part is real.
The harder part is what the pitch leaves out. Most of what is sold as quant trading is not model-driven investing. It is judgment-based work with a model bolted on, or research that looks clean on paper and falls apart when real money moves through it.
The pitch is easy to make. The substance behind it is rare. The gap between the two is where most investment mistakes start in this field.
A professional investor cares about one question. Once costs, capacity, market stress, and weak controls start pushing back, what is left of the strategy. Strong managers separate themselves at that test, and most do not pass it.
- What quant trading is, and the real size of the category
- How quant trading behaves when markets break
- Where quant trading returns disappear, from research to scaled trading
- The strategy families inside quant trading
- Why much of ‘quant trading’ is not quant at all
- How accessible quant trading is
- Due diligence questions for a quant trading manager
- What a verified quant track record is, and what to require before you invest
- Quant performance verification: what each method costs, and what it proves
- How to check a quant track record you have been sent
- What not to accept: 5 claims that survive a quick look
- How to get your own quant track record verified
- What voids a track record, and what can still be saved
- What investable quant trading looks like, and why it stays rare
- Quant trading questions investors ask
What quant trading is, and the real size of the category
Quant trading uses data, statistical methods, and rules to turn investment ideas into signals, portfolios, and trades. The appeal is repeatability. The weakness is false precision.
A strategy can look rigorous in research and still prove fragile in live markets. The bigger question is what the category looks like in size and shape, and whether scale is a sign of depth or only of marketing reach.
These figures come from BarclayHedge’s industry AUM data and Graveyard Database. They say something specific about the field. Capital is real. Some of it sits in genuinely rules-based programs. And the historical record of funds that did not survive is many times larger than the population still trading.
Quant trading happens at the intersection of three things: research that produces a signal, portfolio construction that turns the signal into positions, and execution that gets those positions on at a sensible price. The harder work is the connection between them. Investors who only inspect the research piece miss the layers where most of the friction lives.
How quant trading behaves when markets break
The most informative time to look at a quant trading strategy is not when it is performing. It is when the system around it stops cooperating. Two events from the modern record do most of the teaching here: the August 2007 quant quake, and the March 2020 dash for cash.
- On the week of August 6, 2007, long/short equity quant strategies suffered sharp, fast losses, then partly rebounded by Friday.
- The cause was not a broken signal. It was that many funds were holding similar books, and one large player cut gross exposure quickly, dragging similar portfolios into losses together.
- What it reveals: in stress, the relevant question is not how good your model is. It is how many other people are running something close enough to yours that you all sell on the same day.
- A backtest cannot see crowding. The historical record contains the loss but not the cause, so a model rebuilt on the same data inherits the same hidden overlap with everyone else who built one.
The 2007 episode happened a year before the broader 2008 financial crisis, and is sometimes folded into the same story. They are different events. 2008 was a credit crisis with macro origins; 2007 was a quant-specific stress that did not depend on broader economic conditions.
The point of August 2007 is precisely that. A market does not need a recession to break a crowded set of similar quant trading books. Crowding, hidden overlap, and forced selling produce the same shape every time the market unwinds something: smooth, then violent, then a partial recovery in the names that survived.
March 2020 was the next test. The pandemic shock hurt more than risky assets. It revealed that even US Treasuries, the asset class quant models treat as safe, lost depth and width when investors needed to sell anything they could to raise cash.
- Sovereign bond markets including US Treasuries experienced severe selling pressure as investors sold even safe assets to raise cash. The pattern was not ‘flight to quality’. It was a flight from everything.
- Bid-ask spreads widened across high-quality fixed income, dealer balance sheets could not absorb the flow, and central banks had to step in with emergency facilities to restore basic market function.
- What it reveals: ‘liquid in normal times’ is not the same as liquid when it counts. Strategies that assumed exit costs in line with average days were running an assumption that did not survive the moment they needed to sell.
- Models trained on post-2010 data had not seen this regime. The historical record contained the GFC rather than a pandemic, so models that ‘worked through 2008’ were not necessarily ready for 2020.
Both events teach the same lesson in different vocabulary. These strategies do not protect against scenarios their data did not contain. The 1998 collapse of Long-Term Capital Management said it first, and February 2018 (when short-volatility products including the XIV ETN were accelerated and wiped out) said it again.
The strong managers we see take that argument to heart and stress-test against scenarios their data does not contain. The weak ones treat it as a marketing problem.
Where quant trading returns disappear, from research to scaled trading
The cleanest way to understand why most quant trading strategies disappoint is to follow what happens to the signal itself. A strong idea decays at every stage between the researcher’s screen and meaningful live capital. The shape of that decay is what most pitch decks skip over.
The pattern is consistent enough to draw, even if the exact numbers vary by strategy. A useful version looks like this.
The early steps look reassuring. An idea is formulated, the in-sample backtest confirms it (which is what backtests are designed to do), and out-of-sample testing knocks roughly a third off the apparent advantage. None of this is alarming yet, because most managers will report numbers from this part of the journey and stop there.
The real losses begin once capital becomes involved. Paper trading exposes timing assumptions and data delays that the simulation hid. Small live trading then exposes spread costs, fill quality, and the fact that the model was implicitly assuming someone else would always be on the other side of the trade.
The final step is the one that decides outcomes for professional investors. When the strategy is run at meaningful AUM, market impact arrives in earnest, the trades themselves move prices against the position, and the original advantage collapses to a fraction of what the research deck showed. Most of that advantage lives in this last step, in the sense that this is the step where most of it is lost.
This is why a backtest, however careful, cannot answer the investor’s actual question. The question is not ‘does this idea work in history.’ It is ‘how much of this advantage will still exist when my capital is in it, at the size I want, in the markets I will hold it through.’ Those are not the same question, and the gap between them is where the Two Sigma case in January 2025 (with $165 million repaid to clients and $90 million in penalties for failures to fix known model-oversight problems) ultimately landed.
“Most of what the industry calls ‘quantitative’ is not. It is judgment-based work with a model bolted on, and the bolt-on is usually the part being sold.”
The strategy families inside quant trading
Quant trading is not a single style. Different strategies fail in different ways, scale differently, and deserve different questions from investors. A description that fits ten strategies is not telling you what kind of bet you are making.
Statistical arbitrage looks for short-term pricing dislocations across related securities. It looks attractive on paper. In live trading, it depends heavily on turnover, hidden overlap with other firms, and execution quality at small margins.
Trend and momentum follow persistent price behaviour across securities or asset classes. The reasoning is easier to explain than most other model families, which is part of the appeal. The risk is sharp regime shifts where the strategies all turn at the same time.
Mean reversion bets that short-term extremes fade and prices move back toward a normal range. The advantage disappears quickly if trading costs are too high, or if the model’s idea of ‘normal’ is less stable than assumed.
Factor-based equity models rank and weight stocks using characteristics such as value, quality, momentum, or low volatility. They look simpler than they are. Portfolio construction discipline matters at least as much as the signal itself.
Microstructure and market-making strategies try to capture small margins over and over. They live or die on infrastructure quality, fee structures, and the ability to keep operating when markets become disorderly.
An investor does not need to master every model family. It does help to understand what kind of return is being claimed. If a manager cannot say whether returns come from behavioural persistence, liquidity provision, relative mispricing, or systematic ranking, the story is still too loose to judge.
- By 2016, the academic literature had already published at least 316 distinct factor signals for equity returns. Most fail to replicate out of sample, and a meaningful share are likely false positives once multiple-testing is accounted for.
- Cam Harvey, who co-authored that count, recommends a Sharpe ratio haircut of roughly 50% on backtested figures as a starting anti-overfitting rule, before any manager comparison even begins.
- Newer work on causality (Lopez de Prado, October 2025) goes further. It asks not whether a variable correlates with returns, but whether there is an economic reason it should. Most marketed factors do not survive that question.
- A strong manager can answer two questions clearly: what does the signal economically capture, and how was it tested for the chance that the historical fit is a coincidence.
Why much of ‘quant trading’ is not quant at all
The marketing economics of the field are well understood inside it. Calling a product quant raises fees, attracts institutional interest, and lowers the burden on managers to explain themselves clearly. The description gets stretched as a result.
Three patterns are common, and worth naming. Each one is showing up in regulatory data.
- Model oversight failures inside elite firms. In January 2025, the SEC ordered Two Sigma to repay $165 million to clients and pay $90 million in civil penalties for failing to fix known access-control vulnerabilities in its investment models. In September 2025, the SEC and DOJ moved against a former Two Sigma researcher for allegedly manipulating models for personal compensation.
- AI-washing in retail-facing platforms. In March 2024, Delphia and Global Predictions paid combined penalties of $400,000 to settle SEC charges that they had made false and misleading claims about their AI capabilities. The ‘first regulated AI financial advisor’ was, on inspection, far from what it claimed.
- Idea crowding accelerating rather than differentiating. WorldQuant’s 2025 quant contest drew 80,000 university participants, roughly double the prior year. AI is lowering the barrier to generating ideas; that is not the same as creating new sources of advantage. More likely, it accelerates crowding.
- Across all three, the pattern is the same: the gap between marketing language and substance keeps widening, regulators are paying attention, and most of the corrective work is being done after capital has already moved into products that did not turn out to be what they were sold as.
How accessible quant trading is
Most retail-facing material implies that quant trading is becoming widely accessible. The reality is more layered. Idea generation has democratised; meaningful access to institutional-grade strategies has not.
Three patterns hold together. Capacity discipline at the strongest firms remains tight. The investable subset of large managers is much smaller than the total population of marketed programs. And the platforms offering retail ‘quant’ exposure are mostly running something different from what the elite franchises run.
| What you are looking at | Institutional access to quant trading | Retail-facing quant products |
|---|---|---|
| Manager population | The investable institutional subset is small. The SG Trend Index lists 10 large constituent managers; the broader Barclay CTA Index has 356 programs. | A subscription to an ‘automated quant strategy’ platform usually means a third-party model, often rebranded, with no direct manager relationship. |
| Capacity | Renaissance’s Medallion has been closed to outside capital since 1993; public reporting suggests assets stayed near $10 billion despite extraordinary returns through the 2010s. Capacity discipline is itself a quality signal. | Retail products take in flows continuously; the marketing model rewards growth over selectivity. |
| Minimums and structure | Allocations to elite quant funds typically require institutional minimums ($1 million to $25 million and up), accreditation, lock-up periods, and substantial due diligence work. | Brokerage-platform quant products are accessible at retail minimums, often with daily liquidity, often without disclosed model details. |
| Transparency | Sophisticated funds disclose return drivers, capacity, factor exposures, and stress-test results to qualified investors under NDA. | Retail products often disclose past performance and a high-level description, with proprietary methodology cited as the reason for opacity. |
| Oversight | Independent risk, model validation, and compliance functions are standard at firms managing institutional capital. | At smaller platforms, the same person often writes the strategy, runs the platform, and decides when to override the model. |
None of this argues that retail access is illegitimate. It argues that the words mean different things at different doors. A product called ‘quant’ inside a brokerage platform is not the same product as a quant fund running with billions in capital, an independent risk function, and a thirty-year live record.
In the manager conversations we have had with sophisticated firms, the most consistent feature is capacity discipline being treated as a planning constraint rather than a marketing line.
The harder filter is structural. Fewer than fifty managers globally run a recognisably institutional quant trading franchise with the team, infrastructure, and capacity discipline institutional capital requires. The marketed universe is many multiples of that.
Due diligence questions for a quant trading manager
Real diligence has a different feel from a polished pitch meeting. The goal is not to extract code. It is to find out whether the manager has real process integrity, where the strategy is vulnerable, and how the team behaves when results disagree with expectations.
Continue reading: Quantitative Investment Managers: What Investors Should Check
The strongest managers we have spoken to answer these questions clearly without giving up protection of their actual IP. Weaker managers tend to dress up the lack of an answer.
Start with the signal itself. The questions worth pressing on are simple. What does the signal economically capture, in language a sensible non-specialist understands?
How many candidate models were rejected before this one was selected? Has the variable selection been tested for causal logic, or is it pattern fit? How does the signal behave across regimes, regions, and liquidity conditions?
Trading reality comes next, and weaker managers often slow down here. What cost model was used in research, and how does live slippage compare?
How is order scheduling, venue selection, and market impact handled at the strategy’s actual size? What happens operationally when data arrives late, incomplete, or wrong?
Portfolio risk and oversight close the review. They are where most of the post-mortem material on bad outcomes lives.
What are the main hidden factor exposures the strategy is running? How does capacity analysis change at two or five times current assets? Who approves model changes, and how are changes reviewed and documented?
If answers stay vague exactly where the strategy is most vulnerable, that vagueness is the answer. The 2025 Two Sigma case showed what happens when the answers are not pressure-tested in advance: model changes that should have been controlled were not, and the resulting losses became investor losses rather than firm losses.
- Where do the data come from? Industrial DDQs ask explicitly about input-data sourcing. Strong managers can document each major source. Weak managers describe data as ‘proprietary’ without saying what.
- How is the model validated? Investment-model validation is now a formal expectation rather than a research-team courtesy. The CFA Institute’s 2024 framework treats unvalidated models as a category of fiduciary risk that reaches beyond research hygiene.
- What is the reporting lag, and what does an investor see in real time? Time between trade and reporting is a quiet operational signal. Strong managers minimise it. Weak managers describe lag as a feature.
- How are models retired? A strategy that has only ever been added to has not been controlled, only optimised. Strong managers can name signals they have decommissioned, and explain why.
What a verified quant track record is, and what to require before you invest
A verified quant track record is one where an administrator or prime broker struck the returns, meaning priced the positions and calculated the result, and an external firm audited them. Correcting 2 database biases cuts the average reported hedge fund return from 12.6% to 6.3%.
The survey figure comes from the AIMA Emerging Manager Survey 2026, which polled 180 managers and 50 investors. Require these 6 things before you spend a day on diligence. Each is a document or a number the manager either has or does not.
- Who struck the returns. Administrator or prime-broker statements. A manager spreadsheet is not evidence of anything beyond arithmetic.
- Who signed them. An audit firm you can find on a public register, auditing this entity, in these years.
- 1, 5 and 10 years, shown with equal prominence. The SEC Marketing Rule has required this of SEC-registered advisers since 4 November 2022. A single since-inception number does not meet the rule.
- Net of fees, beside gross. Same rule: equal prominence, same periods, same method.
- Every account run the same way. Related performance must include all similar portfolios, which is the rule that stops a manager showing you the good one.
- What the record lived through. Ask which months cover a liquidity shock, a rate repricing or a crowded unwind. 18 months spanning all 3 tells you more than 3 calm years.
Those 6 are a filter that runs before diligence. Anything clearing them is worth a diligence budget. We hold the managers and strategies in our reviews to that same list, starting with who struck the numbers.
Quant performance verification: what each method costs, and what it proves
Quant performance verification runs on 7 methods, costing from $0 to more than $100,000 a year. Each proves something different. A read-only broker feed shows the trades happened. GIPS verification, from about $7,900 a year, checks the firm’s process and certifies no single track record.
| Method | What it proves | What it does not prove | Cost and time |
|---|---|---|---|
| Self-reported | The arithmetic on the manager’s own spreadsheet or platform export. | That the account exists, that the trades happened, or that this was the only account being run. | Free. Immediate. |
| Broker feed | The trades happened in that account, on those dates. Access is read-only, so the platform can see the account but cannot trade it. | That the account was the only one running. No platform checks this. | FundSeeder free; Kinfo $0 to $44.90 a month; Darwinex Zero €45 to €50 a month. FundSeeder ranks you only after 120 trading days with $5,000 in the account. |
| Signal platform | A signal was recorded before the outcome was known. | That anyone traded it. Collective2’s published return is a simulated model account. | Collective2 $19 to $99 a month. Immediate. |
| Timestamping | A signal, position or return existed at a stated minute and was never edited afterwards. | That the money was real or the order was filled. vBase records portfolio outputs rather than executions. | vBase $0 to $200 a month. AuditZK $0 to $299 a month, and says its own metrics mean little before 2 to 3 months. |
| Fund administrator | A third party priced the positions and calculated the return, so the manager did not author the numbers. | That the statements follow GAAP, or that the pitch deck matches what the administrator produced. | No administrator publishes a price. Outside estimates run $25,000 to $120,000 a year. Apex quotes 5 weeks to launch. |
| External audit | A registered accountant signed the fund’s annual financial statements. | Any monthly return, any composite, or any number in the deck. The audit is annual and covers the entity. | $10,000 to $50,000 a year for a small fund, with estimates reaching $100,000. No source publishes how long one takes. |
| GIPS verification | The firm has a written process for calculating and presenting performance, applied across the whole firm. | Any single track record. In verifier Longs Peak’s own words, verification “does not guarantee the accuracy of any specific performance presentation”. | From $7,900 a year at EVIV; under $30,000 for emerging managers per Absolute Verification. ACA says a few months or less. |
Read each row across before comparing rows. The cheap paid methods show trades happened in one account. The expensive ones check the firm’s process. Neither does the other job, and no single method closes both gaps.
- Collective2’s headline number is simulated. Its own support page says the results “do not represent actual trading” and that “the results you see posted do not match any specific real-life account”.
- Myfxbook runs two verifications and they are not the same thing. “Verified Track Record” means the data matches what the broker reports. “Verified Trading Privileges” means the person holds the master password. A record can carry the first and lack the second.
- Kinfo removes the Verified mark from demo accounts. A prop account running on a simulator does not get verified status there.
- A firm can claim GIPS compliance without ever being verified. Verification is a CFA Institute recommendation, not a condition of claiming compliance.
- None of them addresses the second account. Myfxbook, Kinfo and MyVeridex make no claim about whether the verified account is the only one the manager runs.
Buyers do not publish what they accept. We found no prop desk, family office or institutional consultant stating a minimum in public. The nearest published bar is an index construction rule from 2015: the Dow Jones Credit Suisse Hedge Fund Index admitted only funds with at least $50 million under management, a 12-month record and audited financial statements.
What buyers do publish is exclusions. DV Trading’s experienced-trader posting rules out Forex traders and TopStep funded accounts by name, so at least one desk discounts a platform-verified record entirely. Ask your specific counterparty which row of this table they want before paying for any of it.
- The evidence this section tells you to demand from a quant trading track record is exactly what a profile should lay out: when the record went live, on which venue, at what size, and how live results tracked the out-of-sample expectation.
- Our profile of a systematic gold (XAU/USD) strategy in The Review organises a developer’s live-capital quant trading record against the five-dimension framework: live since September 2024, run on a real venue, with the deepest drawdown in the whole live record documented at 2.87%.
- The figures are developer-reported and organised against the framework; they are not independently audited, which is the exact limit this section tells you to check for. A profile shows you what to ask; where a developer’s word still needs a third party, it says so.
How to check a quant track record you have been sent
7 checks, all free, none needing the manager’s cooperation. Form ADV gives the assets they filed with the regulator, Schedule D names their auditor, the PCAOB register says whether that auditor is real, and Form D dates the first outside dollar.
- The assets they filed. Search the adviser on the SEC’s public adviser database, IAPD, then open Form ADV Part 1, Item 5.F.(2)(c). Regulatory assets are defined differently from marketing assets, so the two never match exactly. A gap of 10 times is a signal. A gap of 10% is normal.
- Their auditor’s name, from the filing rather than the deck. Form ADV Schedule D, Section 7.B.(1), Question 23(b). The same question set asks whether the auditor is registered with the PCAOB, at 23(e), and whether it is inspected by the PCAOB, at 23(f). Those are the manager’s own sworn answers.
- Whether that auditor exists. Search the name on the PCAOB Registered Firms list. The SEC custody rule requires the auditor to be both registered and regularly inspected, which is why Form ADV asks those as 2 separate questions. A name missing from the register means one of 3 things: the fund did not meet the audit requirement, the filing is wrong, or you have the wrong spelling. Ask which.
- Whether the auditor has been criticised. PCAOB inspection reports are public. Part II, the quality-control section, stays private unless the firm fails to fix the problems within 12 months, at which point the PCAOB publishes it. A published Part II says the audit was signed by a firm the regulator had to chase.
- When the first outside dollar arrived. Search the fund on EDGAR, the SEC’s filing archive, and open its Form D. The filing is due within 15 days of the first sale and states that date. A record claiming to run since 2014 whose earliest Form D shows a first sale in 2019 has 5 years belonging to something else.
- Whether the people are registered. NFA BASIC covers CTAs, CPOs and named individuals, with registration dates and disciplinary history from NFA, the CFTC and every US futures exchange. FINRA BrokerCheck covers principals with a broker-dealer past. A 15-year program at a firm registered 3 years ago was built somewhere else.
- Offshore, whether the fund and its auditor are both real. Use the Cayman regulator’s entity search for the fund, then its public approved-auditor list for the auditor. A Cayman fund must be audited annually by an approved auditor and file within 6 months of its year end.
If the manager is a registered commodity trading advisor, ask for one more document. CFTC Regulation 4.35 prescribes its format: date trading began, number of accounts, assets, largest monthly drawdown and worst peak-to-valley drawdown across the last 5 calendar years and year to date, all current within 3 months. Rule 4.35(b) matters most, because it requires the performance of every account the firm and each of its trading principals direct, not the best one.
The catch is Regulation 4.7. An offering sold only to qualified eligible persons is exempt from 4.35, and most institutional quant managers take that exemption. Ask for the document in the CFTC’s format anyway, and treat a refusal as an answer.
What not to accept: 5 claims that survive a quick look
5 claims pass a first read and fail a second. Selling a backtest as live results cost F-Squared $35 million. Advertising one investor’s returns as the whole fund’s cost Twenty Acre $100,000. Each of the 5 has a single question that breaks it.
| What you are shown | The question that breaks it | What it cost someone |
|---|---|---|
| A long smooth record with no bad year | Which months are live and which are simulated? In writing. | F-Squared sold a 2001 to 2008 backtest as “actual performance of real investments for real clients”. SEC penalty, December 2014: $35 million. |
| A record that starts before the strategy did | When was this strategy first run with money? | Wellesley built an index back to January 2000, then told clients on webinars it was composite returns. SEC penalty, September 2023: $1 million. |
| One flawless account | How many accounts run this strategy, and can I see them all? | Twenty Acre advertised one investor’s returns as the fund’s. That investor was exempt from restrictions the others faced. SEC penalty, June 2024: $100,000. |
| A record carried from a previous firm | Is every position inside it the kind of investment the strategy describes? | Old Ironsides showed a legacy holding as an early-stage drilling investment returning 10.9 times capital. It was a stake in another firm’s fund. SEC penalty, April 2020: $1 million. |
| A very large annualised return | How many months does that annualise from? | Titan advertised up to 2,700% annualised, projected from the strategy’s first 3 weeks. SEC penalty, August 2023: $1,042,454. |
- Gregory Zandlo traded through a single omnibus account and decided afterwards which client got which fill.
- 91% of the dollars traded into his own, his family’s and the firm’s accounts ended the day in profit. Of the dollars traded into the other 78 client accounts, 31% did.
- The favoured accounts gained about $105,820. The 78 client accounts lost about $112,667.
- None of this shows up in a return series. It shows up in the order timestamps and the account list, which is why asking for every account is the check that keeps recurring above.
One more thing about F-Squared. Virtus Investment Advisers advertised the same record, in the SEC’s words “accepted it at face value and ignored red flags”, and kept selling it until the strategy held $11.5 billion. SEC penalty, November 2015: $16.5 million. A record that already cleared one firm’s diligence has not cleared yours.
How to get your own quant track record verified
5 situations, 5 routes. Running a fund, an administrator and an audit together cost roughly $35,000 to $220,000 a year on published estimates. Without a fund, a read-only feed plus timestamping costs $0 to about $350 a month and proves less. Choose by who you are asking for money.
| Your situation | The route | Cost and time | What it will not do |
|---|---|---|---|
| Running a fund | Administrator NAV plus an annual audit. Add GIPS if you sell to institutions. | Administrator $25,000 to $120,000 a year; audit $10,000 to $50,000, with some estimates reaching $100,000; GIPS from $7,900. Apex quotes 5 weeks to launch. | Make sense below roughly $5m to $10m in the fund, where running costs swallow the management fee. |
| Separate accounts | Read-only broker feed, plus an administrator where the mandate allows one. | FundSeeder free, ranked only after 120 trading days; Kinfo $0 to $44.90 a month. | Say anything about the other accounts you run. |
| Proprietary capital only | Cryptographic timestamping plus a read-only feed. | vBase $0 to $200 a month; AuditZK $0 to $299, with 2 to 3 months before the metrics carry weight. | Show that the capital was ever at risk, or that an order was filled. |
| Pre-launch | Incubator fund, converted to a fund structure when outside money arrives. | No published fee. Capital Fund Law Group quotes on consultation. | Produce a record anyone else has seen while it is running. |
| Crypto | Exchange read-only API, aggregated across venues. | AuditZK covers Binance, Bybit, Coinbase, Kraken, OKX, Bitget and Gate from $0 to $299 a month. | Close the account gap: one verified account out of an unknown number. |
Every route on that table shares one ceiling. A feed proves the trades in the account it watches. Only the full account list, from an administrator or a broker, shows there was no second account, which is why it belongs at the top of a request list rather than the bottom.
What voids a track record, and what can still be saved
4 defects kill a record. A backfilled composite, a restart shown as the inception date, the best of several accounts, and client money mixed with your own are all fatal. 3 things people assume are fatal are not, including a gap when you changed firms.
Dead, in every case.
- A backfilled composite, where accounts join the history after their results were known.
- A restart after a bad year presented as the inception date.
- The best of several accounts shown as the record.
- Client capital and personal capital reported as one series, which cannot be unpicked once the trades sat in the same account.
Survivable, with the right paperwork.
- A broker change mid-track, if both sets of statements still exist.
- A gap between firms. GIPS provision 1.A.33 lets you use the earlier performance, shown as separate periods and never linked to the later record.
- A move to a new firm. GIPS provision 1.A.32 sets 4 tests: substantially all the investment decision makers moved, the process stayed substantially intact and independent, the new firm holds the supporting records, and no break interrupted the track. The SEC Marketing Rule sets its own version, requiring appropriate similarity in personnel and accounts, disclosed clearly and prominently.
- In managed-futures databases, the years a manager backfilled on joining average 11.3% a year. The years reported live as they happened average 4.9%.
- Across hedge funds, correcting for both backfill and survivorship cuts the average annual return from 12.6% to 6.3%.
- Deleting a fixed number of early months, the usual correction, still leaves 70% to 75% of backfilled returns in the data. Only dropping everything before the listing date removes them.
- Across 582 managed-futures programs from 1994 to 2007, returns above cash averaged 5.4% a year before fees and 0.85% after them.
One question does most of this section’s work: what is the earliest date on which this record existed in the form you are showing me? How a record is verified feeds the research-discipline and operations dimensions our review framework scores.
What investable quant trading looks like, and why it stays rare
The strongest quant trading franchises share a few features. None of them are about being clever. All of them are about being honest about what is hard.
They run capacity discipline. Renaissance’s Medallion has been closed to outside capital since 1993; public reporting suggests assets stayed near $10 billion through the 2010s despite extraordinary returns. AQR, in its public material, describes systematic investing as grounded in economic theory rather than data-mined pattern finding, and treats that distinction as foundational.
They integrate human judgment and machines, rather than pretending one or the other has disappeared. Bridgewater describes its process in those terms publicly.
D. E. Shaw says it launched as a pioneer in systematic investing and later built fundamental and judgment-based work alongside the machines. Two Sigma frames AI in investing as dependent on channelling capability wisely rather than on AI itself producing alpha.
They hold oversight to a high bar. Model validation is a separate function with the authority to block deployment rather than a quarterly afterthought. The cost of getting that wrong has been written into recent enforcement orders, including its own.
The result is a category that looks the way the word should imply: rule-based, evidence-driven, and small, distributed so unevenly across the marketed universe that any honest count weighting live infrastructure, capacity discipline, and oversight alongside returns leaves fewer than fifty firms globally meeting the standard at scale.
Quant trading questions investors ask
No. Algorithmic trading is the broader category, and often refers to the automated execution of trades or orders. Quant trading is a narrower idea that includes model-driven signal generation, portfolio construction, and systematic decision-making.
A strategy can be quantitative without being fully automated, and automated without much quantitative depth. Most judgment-based funds use automation in their order routing and reporting, but that does not make them quant funds. Most quant funds rely on automation, but the meaningful work is the modelling and oversight behind it.
It depends on whether you mean an institutional fund or a retail-facing product. Elite quant trading funds typically require institutional minimums starting at $1 million and often well above, plus accreditation, lock-up periods, and meaningful due diligence work before access is granted. Some of the strongest, including Renaissance’s Medallion, are closed to outside capital entirely.
Retail-facing ‘quant’ platforms have much lower minimums and offer near-daily liquidity. The trade-off is that the products are usually different in nature: third-party models, less transparency, smaller research teams, and less oversight. Both can be legitimate. They are not the same product wearing different price tags.
No. A backtest is hypothetical performance built from historical data, useful for research and easy to make flattering. A track record is realised profit and loss on real capital that traded.
Under the SEC Marketing Rule, in force since November 2022, presenting one as the other is close to the exact behaviour the regulation exists to stop. When a manager shows a curve, the first question is whether any real money was behind it.
They are not the same assurance. An audit is an accountant checking a firm’s financial statements. Verification, in the GIPS sense, is an independent check that the firm’s process for calculating and reporting performance follows the Global Investment Performance Standards.
A firm can be GIPS verified and still have a weak strategy, because verification covers the reporting process rather than the quality of the results. Neither, on its own, proves live skill. Ask for both, plus the administrator statements underneath them.
For proving a record is real, yes. A broker-connected feed or a tamper-proof timestamp does cheaply what a reference call never can. For proving a quant trading strategy is good, no tool delivers that. A verification service answers authenticity; the capacity, regime behaviour and durability of a quant trading record stay the reviewer’s work.
Yes, and it is the most common trap in the category. A verified quant trading record can still be survivorship-selected from many accounts, run at a size the strategy cannot hold, lucky across a single market regime, or too short to carry weight. Verification confirms the record is real. Durability is a separate question it never answers.
The proof lives in the record while the quant trading model stays private. Third-party administrator statements, a broker or exchange API feed, a tamper-proof timestamp, or GIPS verification each confirm the returns are real without exposing a line of code. A manager can show results and still protect the intellectual property underneath.
Most quant trading managers do not pass the standard. The few that do are worth knowing.
The marketed universe of quant trading is large and growing. The investable subset, weighted by capacity discipline, oversight quality, and a real live record across stress periods, is much smaller. Fewer than fifty firms globally meet a demanding institutional standard at meaningful scale.
The Review is where this site applies that standard to specific quant trading strategies, managers, and platforms, scored across five risk dimensions so you can see which records hold up under stress and which fall away. It is the first independent directory that puts capital protection before returns, and where the few quant trading names that pass are added through 2026.