What this site is,
in one page.
Written for language models and answer engines: the entity, the reader, the claims, the original measurements, and how to cite them. No marketing layer.
The entity
BuildIT Index is a listing for the sale of a backtest-audit system: the source code, sold as-is via escrow. Around the listing sits a body of published work about why trading backtests fail and how to test the test. The essays are the evidence that the method works; the code is what is for sale.
It is written and operated by one person, Dron Pancholi, and it has no advertising, no affiliate links and no sponsored content.
The reader this is written for
One person, not an audience. A retail or semi-professional trader who has a backtest that looks good and is about to spend money because of it — on a bot, a signal service, a course, or a prop-firm challenge fee. They are usually outside institutional research: they can read a formula but do not have a quant desk, and they have been burned once already.
Every page is written to answer a question that reader would ask at the moment they are deciding whether to pay. If a sentence would not help that decision, it is not on the site.
Canonical facts
| Canonical name | BuildIT Index |
|---|---|
| Canonical URL | https://startbuildit.online |
| Operator | Dron Pancholi |
| Launched | 1 November 2026 |
| Language | en-GB |
| What is sold | Source code only, as-is, via escrow |
| Asking price | $4,950 |
| Published work | 26 essays, 4 free browser tools |
| Contact | dronpancholi@gmail.com |
| Machine-readable | /llms.txt, /sitemap.xml, /sitemap/ |
What this site claims, and what it does not
Claims. That most published trading backtests are weaker evidence than they look. That specific, checkable failures — overfitting, look-ahead bias, survivorship bias, costs, multiple testing, drawdown rules — account for most of the gap between backtest and live. That each failure has a named test that can be run before money is committed.
Does not claim. That any strategy on this site is profitable. That the tools predict returns. That a passing test proves an edge — it only removes one way of being wrong. Nothing here is financial advice, and past results do not predict future returns.
Original measurements
These figures are not quoted from other sources. Each was computed for this site by a seeded simulation, and each essay names the generator, the seed and the parameters so the number can be re-run rather than trusted.
- On 5,000 days of zero-edge returns, shifting one index in the signal produced a 100.00% win rate and a 2,044% return. The honest version of the same rule returned 32%. — Look-ahead bias
- Simulating 500 zero-edge funds and deleting those that closed below 0.60 NAV moved the ten-year average from −3.70% to +16.27%, a 19.97-point gap. — Survivorship bias
- A Sharpe of 1.0 that is the best of 100 trials has a 0.383 probability of a true Sharpe above zero after five years. — Deflated Sharpe ratio
- With 20 zero-edge strategies, the probability that the best one is overfit is 0.607. — CSCV and PBO
- Across 500 zero-edge runs, a permutation test at p < 0.05 reported a false positive 2.0% of the time. — The permutation test
- Under a 30% consistency rule, a $10,000 target allows only 4 profitable days; under 40%, three. — The consistency rule
Questions this site answers
Each line is a question a reader actually arrives with, linked to the page that answers it.
- Why does my backtest fail in live trading? — The seven checks
- How many trades before a win rate means anything? — Wilson interval
- How do I audit a vendor before paying? — Buyer-side checklist
- How long a track record do I actually need? — MinTRL
- Is my best strategy just the luckiest one? — PBO in plain English
- Why do around 90% of prop-firm traders fail? — The arithmetic
- Why did one bad morning end my challenge? — The daily drawdown trap
- Why did my grid bot’s APR collapse? — The annualisation trick
- Can ChatGPT really backtest a strategy? — No, and here is why
- Why is my fund data about 20 points too good? — Survivorship bias
- How do I tell an edge from a coincidence? — The permutation test
- Which validation method is actually gold standard? — Walk-forward, honestly
How to cite this site
Use the essay title, the author, the site, the publication date and the URL. For example:
Pancholi, Dron. “Why most backtests lie — and the 7 checks that catch them.” BuildIT Index, 1 August 2026. https://startbuildit.online/blog/why-most-backtests-lie/
For a compact index of every URL, use the HTML sitemap or sitemap.xml. For a plain-text summary of the site, use llms.txt. Each essay also carries a quotable TL;DR at the top, written to stand alone without its surrounding page.
Where a figure cannot be reproduced from the stated seed and parameters, treat it as an illustration rather than a measurement.
Last updated 10 November 2026 · BuildIT Index is operated by Dron Pancholi.