Backtest vs Forward Test: Why You Need Both

A backtest runs a strategy's rules over past prices to see how it would have behaved, while a forward test runs the same fixed rules on new prices as they arrive, usually with pretend money. The backtest gives you years of history quickly but is built by someone who already knows how that history ended. The forward test is slow and short but cannot be fitted to data that did not exist yet. You need both because each covers the other's main weakness.

What a backtest can and cannot tell you

A backtest is a computer test on past prices. You write the rules, feed in years of data, and the software records every trade the rules would have made. If you want the basics, start with what backtesting is.

What it does well:

• Shows how a strategy behaved across many market conditions, including crashes you would not want to wait for in real time.
• Gives enough trades to see patterns, such as how often the strategy is wrong and how large its losses get.
• Lets you compare against a benchmark over exactly the same dates.

What it cannot do:

• Prove the rules were not tuned to the past. The person building the test has seen the data, and every tweak that improved the result was a choice made with hindsight. This is the core of overfitting.
• Capture real execution. Backtests assume fills at certain prices with certain costs. Real fills can be worse.
• Tell you how you will feel. Reading "fell 30%" in a table is not the same as watching it happen.

What a forward test can and cannot tell you

A forward test, sometimes called paper trading or out-of-sample testing, fixes the rules first and then lets time pass. Every trade is decided on prices that did not exist when the rules were written.

What it does well:

• Removes hindsight completely. Nobody can fit rules to tomorrow's prices.
• Tests the plumbing. Does the strategy actually place the orders it should, on the days it should? Do the fills look like what the backtest assumed?
• Shows you the strategy at human speed, one day at a time, so you learn whether you can follow it.

What it cannot do:

• Cover enough history. A few months of forward testing might include no crash at all, or only one unusual period.
• Prove a strategy works. A short good run can be luck; a short bad run can be noise.
• Give a statistically meaningful sample for slow strategies. A strategy that trades once a month makes only three decisions in three months.

Side by side

A middle ground: out-of-sample history

Between a pure backtest and a pure forward test sits a useful step: testing on historical periods the rules were not built on. If a strategy was designed using 2011-2026 data, running it on 2000-2010 is a kind of forward test into the past. The designer may still have known roughly what happened in those years, so it is not perfect, but it is a strong check.

Our own strategies show why this matters. All figures here come from historical backtests, computer tests on past prices; nobody earned them and future results will differ.

Leveraged Growth rotates between leveraged growth funds, volatility funds and bonds. In its main backtest from 2011-01-03 to 2026-11-03 it returned +129.1% a year with a worst fall of -53%, compared with +13.9% a year and -34% for the S&P 500, and +18.2% and -35% for the Nasdaq 100. Its deepest fall, in the 2020 Covid crash, recovered in about three months. Its caveat: its rules may be fitted to the years after 2010, and leveraged funds can lose value quickly.

When a simulated three-fund version was run back through the dot-com crash of 2000-2002, a period outside the main test, it fell 91% against 47% for the S&P 500, and took 115 months to get back to its pre-crash value against 78 months for the index. In the 2007-2009 financial crisis, the simulated version did better than the index, falling 48% against 55% and recovering in 19 months against 53.

Market Switch holds one fund at a time and switches when the market looks stretched. Its main backtest, 2011-11-01 to 2026-11-04, shows +71.9% a year with a -42% worst fall, against +15.0% and -34% for the S&P 500 over the same period. Its caveat: nearly half of its tested profit came from UVXY, and it failed outside the US. In the simulated three-fund version run through the dot-com crash, it fell 100%, against 47% for the S&P 500, and took 259 months to get back to its pre-crash value against 78 months for the index. Its main-period deepest fall, 42% in March 2020, recovered about three months after the bottom.

Those older-period results are exactly the kind of information a single backtest window hides. They do not mean the strategies will fail in the next crash, and they do not mean they will succeed. They show that the headline numbers come from one particular period, and that behaviour outside it can be very different.

What to look for when you forward test

A forward test is most useful when you know what you are checking. Before you start, write down what the backtest says you should expect. Then compare.

1. Trade timing. Does the strategy trade on the days and at the frequency the rules describe? A monthly strategy should not be trading every week.
2. Holdings. Are the positions the kind the backtest held? If Steady Growth, for example, should be holding some combination of its seven fixed names, anything else is a red flag.
3. Fill quality. Are the prices realistic? A forward test that fills at idealised prices is just a slower backtest. Good paper systems fill at a real market price and take off a cost. For more on this, see why paper trades should include costs.
4. Size of moves. If the backtest shows daily swings of a few percent and the forward test shows swings of 15%, something is different.
5. Your own behaviour. Did you want to override it? Did you check it constantly? That tells you something about whether you could run it with real money.

What you should not do is judge the strategy mainly on whether the forward test made money. Three months is not enough for that.

How long should a forward test run

There is no magic number. A rough way to think about it: the forward test should be long enough to include a reasonable number of the strategy's decisions. A strategy that checks gold every hour makes many decisions in three months. A strategy that checks once a month makes three. For slow strategies, the forward test is mostly a check on mechanics and behaviour rather than on results. Our guide on how long to paper trade goes deeper.

How AlgolaneBots combines the two

AlgolaneBots publishes each strategy's historical backtest with its benchmark, its worst falls in past crashes, and the specific caveat that could make the numbers look better than reality. That is the backtest half.

The forward-test half is the free 90-day paper trial. You get a $10,000 paper account. The trial starts when you press start, not at sign-up, and no card is needed. Paper fills happen at the next real market open, with a 0.05% cost taken off every fill and splits adjusted, so the record is built from real prices that arrived after the rules were fixed. You can run all five strategies, including Gold Breakout, on paper.

After the trial, nothing charges you automatically. If you decide to continue, it is $29 a month; see pricing. Live trading stays off until you subscribe, consent and connect a live broker account.

Key takeaways

• A backtest replays the past; a forward test runs fixed rules on new prices as they arrive.
• Backtests cover many years but carry hindsight; forward tests have no hindsight but cover little time.
• Testing on historical periods outside the design window is a useful middle step, and can show very different results.
• Use a forward test to check behaviour, timing, holdings and fills against what the backtest predicts, not to prove profitability.
• Neither test guarantees future results; together they give a more honest picture than either alone.

This is general education, not personal financial advice. Strategy figures are from simulated historical backtests.

Run your own forward test

You have read the backtests; the next step is watching the rules work on prices nobody has seen yet. Start a free 90-day paper trial with all five AlgolaneBots strategies and a $10,000 paper account, and check the safety features that apply when you later go live.