How our backtests and data work
This page explains where Tactfolio’s data comes from and how our backtests work. A backtest replays a set of investing rules on past prices, as if you had followed them. Every backtest makes assumptions. We list ours here so you can judge how much to trust a result.
- Prices are daily prices of US-listed stocks and funds, with dividends reinvested.
- Crypto perp strategies use Hyperliquid’s own prices.
- Each trade pays a cost of 0.05% of the amount traded.
- Rules only see prices up to the day they run.
- Young funds can be extended back with simulated history, which is always labeled.
- Every backtest is hypothetical. It is not a record of real trading.
Where does the price data come from?#
Tactfolio uses daily prices for stocks and funds listed in the United States. Prices are adjusted for splits and dividends. That means a backtest’s return includes dividends, as if they were reinvested.
We update prices after each US market close. Big one-day jumps are checked against a second price source before we store them. If the data for a day is missing, the backtest stops with an error. We never fill a gap with a made-up return.
The list of tickers holds stocks and funds that trade today. Companies that failed or were delisted are not in it. This is called survivorship bias, and it flatters any rule that picks single stocks. Funds that track broad indexes are much less affected. Why past results mislead explains this and other traps.
Where does the economic data come from?#
Rules can read 22 US economic series from the Federal Reserve Bank of St. Louis (FRED). Examples are unemployment, inflation and interest rates. Net liquidity, the Fed’s balance sheet less the Treasury’s account and reverse repos, is worked out from three of them and counts only once all three are out.
Economic numbers often get revised after they first come out. For monthly series, Tactfolio keeps each version as it was first published, using FRED’s archive (ALFRED). A rule reads each number the way it looked on that day. It never sees a later revision early.
When does a backtest trade?#
A rule reads the day’s closing prices and trades at that same close. No rule can see a price from a later day.
Real trading cannot quite do this, because you only know the exact close once the market has closed. So backtest charts also show a Next-day open line. It runs the same rules but trades at the next morning’s opening price. If the two lines drift far apart, the result depends on trading right at the close.
When a strategy runs in Trade, it makes its decision at 3:45 pm New York time, 15 minutes before the close. It retries at 3:50 and 3:55 if something fails.
What costs does a backtest include?#
Every trade pays 0.05% of the amount traded. This is called slippage. It stands for the small gap between the price you see and the price you get. The cost is the same for every strategy, so results compare fairly.
A Reality checks table repeats the same trades at 0.10%, 0.20% and 0.40%. You can see how much a result depends on cheap trading.
Some costs are not included:
- commissions,
- taxes,
- fund fees beyond those already in the fund’s price.
Cash that is not invested earns nothing. To earn a Treasury bill return, a strategy can hold a cash-like fund such as BIL.
What is simulated history?#
Simulated history is a modelled price series that extends a young fund back before its first real trading day. For example, GLD’s real prices start in 2004. Its simulated history reaches back to 1968.
Simulated values are estimates. No one could have traded at those prices. When simulated history is used, the backtest names every ticker that used it and switches to real prices from each fund’s first trading day. Simulated history lists every fund it covers and how each one is built.
How do crypto perp backtests work?#
A crypto perp strategy trades perpetual futures on Hyperliquid, a crypto exchange. Its backtest uses Hyperliquid’s own prices. It never swaps in a US fund with the same name.
Perps trade all day, every day. So these backtests differ in a few ways:
- Bars. A strategy can run on hourly or daily bars. Every day counts, weekends included. Daily bars close at midnight UTC.
- Funding. Holders of a perp pay or receive a small fee called funding. Each bar’s price includes it, so returns show what holding the perp earned.
- Yearly figures. Volatility and Sharpe use 365 days or 8,760 hours a year.
- No score yet. Crypto perp strategies have no Tactfolio Score and are not ranked in Discover.
Crypto perp strategies can run with pretend money on Hyperliquid’s test network. Real-money trading on Hyperliquid is off.
How are returns measured?#
| Measure | What it means |
|---|---|
| Total return | Growth over the whole backtest |
| CAGR | Average yearly growth, compounded |
| Volatility | How much daily returns swing, scaled to a year |
| Max drawdown | The biggest drop from a high point to a low point |
| Sharpe ratio | Yearly return divided by volatility |
| Sharpe over T-bills | The same, after subtracting the 13-week Treasury bill return |
| Calmar ratio | CAGR divided by the size of the max drawdown |
| Turnover | How much of the portfolio is traded each year |
You can also switch a backtest to “Real” returns. Real returns take out inflation, using the US consumer price index from the Bureau of Labor Statistics. Before 1913, they use Robert Shiller’s monthly price index.
How to read a backtest explains how to use these numbers together. What a backtest simulates has the full list of settings.
How is each benchmark chosen?#
Every backtest is compared with a benchmark over exactly the same dates. The default benchmark is SPY, a fund that tracks the S&P 500. You can pick another ticker or another saved strategy.
Safer strategies usually earn less and fall less than the benchmark. So the Reality checks also show the benchmark scaled to the same risk as the strategy. If a strategy cannot beat that, holding the benchmark at that size would have done as well.
How does the Tactfolio Score rank strategies?#
The Tactfolio Score ranks public strategies in Discover from 1 to 100. It tries to remove what a backtest may owe to hindsight. Then it compares strategies on three expected measures at once: Sharpe ratio, Calmar ratio and growth.
The score takes points away for complexity and for results that break when the numbers are nudged. It gives weight to live results once a strategy has them. A high score is not a forecast and not a recommendation. How the Tactfolio Score works explains every step.
How are the classic strategy pages built?#
Each of our classic strategies is built from its original paper or book. The article names that source and says how our version follows it. It also lists the funds our version holds.
The results on each classic page are recomputed after every market close. The page shows the date the figures are as of.
What can a backtest not tell you?#
A backtest cannot tell you what will happen next. Rules that look great on the past often do worse later. This happens because some of the rule fits noise that will not repeat.
Some things a backtest also leaves out:
- your taxes and your own trading costs,
- how you would feel and act during a big drop,
- delays or failed orders in real trading,
- events that never happened in the test period.
Read why past results mislead before you rely on any result.
When does this page change?#
We update this page when the way Tactfolio computes results changes. The date at the top shows the latest version. Changes to the Tactfolio Score are also noted on its own page.
Read the full risk disclaimer before you trade.
Sources
- Federal Reserve Economic Data (FRED), Federal Reserve Bank of St. Louis
- Archival Federal Reserve Economic Data (ALFRED), Federal Reserve Bank of St. Louis
- Consumer Price Index for All Urban Consumers (CPI-U), not seasonally adjusted, U.S. Bureau of Labor Statistics, via FRED
- Online data: stock market and CPI since 1871, Robert J. Shiller, Yale University, accessed October 16, 2026