I'm Ayan Jain, publishing as Finance Broski. I build trading strategies and data tools, run them through hostile tests, publish every failure with its cause, and sell only what survives. Everything below is built to be checked: free builds to poke at, protocols complete enough to reimplement, and a named human liable for every verdict.
See everything I do ↓ Walk the graveyard →One person runs every test. One person signs every verdict. Named, and liable for it.
Two paid desks. Both exist because most quant work fails the same few checks, and it is cheaper to hear that from me than from the market.
You send a backtest. I spend 48 hours trying to break it. You get a written verdict, check by check.
Six forensic checks cover the failure modes that turn a paper edge into a live loss: survivorship in the data, look-ahead in the plumbing, costs that flatter, overfitting and selection, statistical validity, and whether the fills could physically have happened. You do not need to share code or reveal your alpha; a trade list and an equity curve are enough for all six.
The verdict is per-check, in writing, signed by me. If nothing is wrong, you get an independent audit you can show a funder, a prop firm, or yourself before going live.
Recent findings: a reported t-statistic of 404 that was roughly 15 once the test was computed correctly. A liquidity filter that quietly knew 2024's liquid names back in 2015, worth +1.6 points a year of fake alpha.
Before you publish quantitative research, I referee it the way a hostile reviewer would, so your readers do not do it for you.
Methods, claims and numbers checked line by line: does the data support the sentence, does the test support the claim, does the number survive its own error bar. Written notes you can act on, delivered before your deadline, under an engagement letter when the work is confidential.
None of these people are clients. They are researchers and founders who engaged with the method in the open, corrected it where it was wrong, and adopted it where it held. That is a more useful thing than a testimonial, and a harder one to manufacture.
"Effectively audits PhD-level quantitative work from very little information. Given a chart and a few hundred words, without the underlying code or data, he independently derives the figures and identifies what does not reconcile. His review has materially improved the published version of several analyses in my series."
Dr. Heather Dempsey, quantitative researcher, amended a published piece after an exchange about a discrepancy, and noted the change in the body of the post rather than editing quietly. Correcting a published piece in public is rarer than the catch that prompts it.
A simulation showing that ranking on the same window that selected a candidate can manufacture apparent mean reversion was published as a follow-up piece with named credit, and the disjoint-window recompute became stated practice. The useful test is whether a method travels beyond the person who proposed it.
On multiple testing under correlated search, we each corrected the other in the same thread: an equicorrelation figure that did not survive checking, and a threshold approximation of mine that ran high at practical sample sizes. Both corrections sit in the public exchange.
Two trading-platform founders have taken methodology suggestions from public threads into their own products, and one invited me in as an independent voice on data honesty for their users. The checks are not specific to my own book.
"He has reproduced my statistics from the posted material alone, caught a mechanism error I had already published, and retracted one of his own corrections unprompted when his model of my simulation turned out wrong. ... He's fast and very good."
US equities, 1.7 million pair combinations examined every week. It is the screener that publishes the test it failed.
A pairs screener finds two stocks that historically move together, so that when their gap stretches, a trader can bet on it closing. Mine screens roughly 1,900 liquid US names every week and shows its work.
Before selling anything I pointed the screener at itself: recomputed every accept/reject decision inside each test year so nothing could know the year it was judged on. The apparent edge died, 13.9 points a year of it was pure hindsight, so this desk refuses to sell hindsight. No column ranks the year ahead; the register ranks on what a pair is and where it stands today, and the full autopsy is public.
A complete free build lives on the page, aged on purpose, with a measured decay curve that says exactly what its age costs. The weekly refreshed desk and covering brief are paid, with tiers priced on that same curve.
"The graveyard is the most informative database I own, and almost nobody keeps one."
A public ledger of the strategies I built and then killed: what each idea was, which test ended it, and the exact cause of death.
It exists because refinement is not amnesty. A strategy that gets quietly reworked after failing is a strategy whose track record is fiction, so every kill is written down and git-timestamped, which means the record cannot improve after the fact. The kills are being published batch by batch; the count on this page is the running total.
Point-in-time data is the difference between a backtest and a story. These are the pieces I needed and could not find, so I built and published them.
150 sourced corporate exits behind the S&P names that free price data cannot see: who disappeared, when, and why, with a citation for each.
Roughly one in four members of the 2010 S&P 500 is invisible in standard free data today. This registry documents where they went, acquisition, delisting, bankruptcy, so a backtest can stop pretending they never existed.
The index as it actually stood in January 2010 and January 2015, rebuilt from the public changelog, so you can test on the stocks that were really there.
Ships with the visibility probe: for every historical member, whether today's free sources can still see it. Use it to measure the survivorship hole in your own pipeline before it flatters your results.
A live record of auction dislocations, updated nightly, with standing bars drawn as they happen and CSVs anyone can download.
Ayan Jain. Independent quantitative researcher, based in Jaipur, India, publishing online as Finance Broski.
I run my own capital through the same discipline I sell: preregistered gates written before the results exist, kill criteria written while calm, and every claim carrying its error bars. The self-test that killed my own screener's ranking is published in full, because the standard has to bind me first or it binds nobody.
The methods writeups are on SSRN (survivorship, selection look-ahead), the code that can be public is on GitHub, and the survivorship tooling is on PyPI with an archived release on Zenodo.