Every strategy on VoidOrigin shows a winrate, expectancy, equity curve, and drawdown. Underneath every one of those displays is a small disclaimer about hypothetical performance. It is there because the CFTC requires it, and also because it is the most important thing you can internalize as a subscriber.
A backtest is a story about the past
When we build a strategy, we run it over historical data with perfect hindsight about what the price did after each bar. We know every tick. We know where every fill would have happened if we had been there to place the order. We model commissions and slippage, but we model them based on assumptions. Those assumptions are not reality.
Backtests are the best tool we have for deciding whether a strategy is worth running. They are also the single biggest source of disappointment new traders experience. A strategy with a 1.7 profit factor in backtest might have 1.3 in live trading, and that difference is enough to make it not worth running.
Why live results differ
- Slippage: the difference between the price the backtest assumed and the price you actually got.
- Latency: your signal arrives milliseconds or seconds after the event. In a fast market that can be the difference between a win and a loss.
- Liquidity: backtests assume you can always fill at the displayed price. In a thin market, your order pushes the price.
- Partial fills: your order fills in pieces at different prices.
- Venue outages: the backtest does not know that CME stopped matching for 40 seconds the day your trade was supposed to enter.
- Commissions: your broker is not the broker the backtest assumed.
How we de-risk this
Every strategy we publish goes through out-of-sample testing on data the researcher never touched while tuning. Strategies that pass are promoted to paper-trade on live feeds for at least two weeks. Only then do subscribers see them in the catalog, and even then the stats are labeled based on which portion is hypothetical and which portion is live.
Paper trading catches some of the problems above, but not all of them. In particular, your commissions, your broker, your connection, and your execution speed are out of our control. The right mental model is to treat the published numbers as an upper bound of what you can expect.
What to do about it
Do not size your position based on a backtest expectancy. Size it based on a worst-case drawdown you can survive. If you cannot survive two back-to-back historical max drawdowns at your chosen position size, cut the size.
Track your own fills. After a month on a strategy, compare your live results to what the published stats would have predicted. If the gap is large, either the market regime has shifted, or the strategy has a fill problem on your setup that we did not model. Either way you want to know.