We built Potluck to be the most honest backtesting engine on the market.
Social media is full of trading setups that supposedly print money. Potluck tests them properly: on years of real prices, paying what trading really costs, raced against traders who pick at random. Then it tells you, in plain English, whether the setup works or whether it was luck.
Most backtests are built to make a strategy look good. Here is what happens to a typical one when you stop flattering it.
Entries picked with hindsight, no costs, the one lucky version out of many.
Every fill pays the spread, at the next minute's price, and every version tried is counted against it.
Two hundred traders entering at random, same number of trades, same costs.
An illustration of a typical result, not one specific strategy.
Pick a plain-English template, or write exact rules like rsi(2) < 10 and close > sma(200). Potluck knows 141 indicators: every price-and-volume indicator in TradingView's built-in list.
Minute by minute, over about ten years of real US stock prices. Each signal fills at the next price, never the one that triggered it, and pays the real cost of trading.
One straight answer, with the checks that show why: is there enough evidence, does it beat random entries, does it survive costs, does it hold up on years it was not tuned on.
These rules are on for every test. You can't switch them off to get a nicer number.
Each buy and sell pays that stock's actual bid–ask spread, measured stock by stock, plus a commission. Every report also shows the result before costs and the cost at which it breaks even.
A signal fills at the next price, never the one that triggered it. Every indicator must pass an automated test proving it can't see ahead.
Testing only today's winners flatters any strategy. Our stock list is rebuilt each year from the previous year's 500 most traded US stocks, including the ones that later went bankrupt or disappeared.
Every strategy races 200 random traders with the same trades and costs. We also shuffle the prices, delay the entries and double the costs. A real edge survives all four.
Try 70 versions of an idea and one will look great by luck. Every run is logged, and the more versions you try, the higher the bar gets.
The most recent prices are kept aside as a final exam that no ordinary test can touch, so a strategy can be checked on data it has truly never seen.
And we checked the engine itself: real spreads stock by stock, a limit so no order is bigger than what actually traded, short-selling rules, trading halts, known-answer tests, a second data source, and a second simulator written separately that has to agree with the first on hundreds of random strategies.
The Pot is where strategies go when they don't survive. These are some of the most-shared setups on YouTube, TikTok and in trading books, written down as exact rules from the way they are taught and run through the same test. None of them passed. A few made money, but not in a way that holds up: random entries held the same way did just as well, or there were too few trades to tell skill from luck.
Loading the results…
Each tested on 1-minute, 5-minute or daily prices from 2016 or 2017 to August 2022, after costs. "$1,000 became" is the whole period.
This is a real report from the engine, for one of the Pot's setups: the popular "liquidity sweep, structure shift, fair value gap" day trade on the two biggest index funds, 1,129 trades from 2016 to 2022.
| Trades | 1,129 |
| Win rate | 40.5% |
| Return before costs | +3.2% |
| Return after costs (1 bp a side) | −7.8% |
| Break-even cost | 0.29 bp a side |
| Chance it is luck, after 85 versions tried | can't rule it out |
Total return at each cost per trade, per side.
The full report adds the equity curve, drawdowns, results by year and market regime, every trade, and a snapshot that lets anyone rerun the exact test.
When Potluck opens, this is what you'll be able to use.
Start from a template like "buy the dip" in plain English, or write your own rules with any indicator, time of day, market context and exit.
Before and after costs, in-sample and out-of-sample, by year, time of day and weekday, without the best days, and against simply buying the market.
Each trade drawn on that day's candles, with the entry, exit, stop, target and your strategy's own indicator lines.
"Probably noise", "too thin after costs" or "worth a closer look", with the checks behind it. Most ideas land on the first.
Every test you run, saved, with the count of how many versions of each idea you've tried.
The famous setups and every failed idea, with the full report for each, so you don't have to lose money finding out.
We copy published academic strategies exactly and test them honestly. Five papers so far; most of the published results don't survive real costs.
Spreads, volume limits, short-selling rules, borrow fees, trading halts and leverage limits, all on by default.
An API and tool access so AI trading agents can test their ideas against the same honest engine.
Almost every day-trading setup we've tested, including the famous ones and the ones from published research, fails once real costs and chance are counted. Our own best ideas failed too, and we wrote that down instead of hiding it.
That's the point. A tool that tells you "no" before you risk real money is worth more than one that tells you what you want to hear. When something does pass every check, you'll know it earned it.
No. Potluck is a research tool that tests strategies on past prices. Nothing on it is financial advice, and past results don't predict future ones.
Getting a verdict on a strategy will be free. If we ever add paid extras, the honest answer to "does my strategy work?" stays free.
Real historical US stock and fund prices, minute by minute, from 2016 on, adjusted for splits and dividends.
We haven't set a date. Claim a username now and we'll email you the day it opens.
We use it to confirm your username and to tell you when Potluck opens. We don't sell it or share it. The details are in our privacy notice.
Use "Have a name?" at the top of the claim page and we'll email it to you.
Reserve your username before the site opens. It takes a minute.