An intraday book with almost no exposure to the market's direction
Backtest results · hypothetical, not live
2020–today, blended intraday signals
Performance
Detail
Results from a single backtest on QuantConnect. Costs and slippage modeled per the strategy's deployment specification.
About
An ensemble of several intraday signals on a single liquid Nasdaq instrument, blended into one book that trades during the regular session and is flat by the close. The variants are combined so their individual quirks average out into a steadier whole.
Its standout property is that its returns barely move with the broad market; sensitivity to the index is close to zero. That makes it a genuinely independent return stream, valuable less for its standalone numbers than for what it does to a portfolio when combined with directional strategies.
Backtested from 2020 to present. The near-market-neutral profile pairs a Sharpe above 1.1 with a moderate drawdown. Included as a diversifying building block.
Execution costs
Intraday strategies live or die on execution. A model that holds for weeks pays its costs a handful of times a year. This one pays them on every entry and every exit, hundreds of times a year, so a cost assumption that is even slightly too kind can manufacture an edge that does not exist. Nothing shown above is a gross figure. Costs are charged inside the backtest, trade by trade, before any statistic is computed.
Each fill is charged an Interactive Brokers retail commission, per contract on futures and per share on ETFs, and charged per side, so a round turn pays twice. Retail pricing, not an institutional rate I do not actually get.
On top of commissions, every fill is moved against the strategy by a slippage adjustment before the trade is booked. Buys fill worse than the modeled price, and so do sells. The adjustment is sized to cover the bid-ask spread and queue position on the instruments actually traded.
The platform default for stop fills runs optimistic for breakout strategies, because stops trigger in exactly the fast conditions where fills are worst. I compare modeled fills against my own broker fills on the same signals and tighten the assumption until the backtest matches what I actually get.
The win rate, loss rate, average win, and average loss shown above are all post-cost measurements. Costs push each result toward the losing side: a trade that finished a tick or two ahead gross ends up recorded as a small loss, winners are booked slightly smaller than the price move, and losers slightly larger. So the win rate you see is lower than the raw signal produces, and the ratio between the average win and the average loss is narrower.
That matters most for a low win-rate breakout strategy, where the arithmetic depends on a small number of large wins covering many small losses. Shrink the wins and inflate the losses on every trade and the whole structure can invert. Reading these statistics as already-degraded numbers is the point: the margin between the average win and the average loss is what survives the cost model, not what the signal looks like before it.
This is also why some intraday strategies I have built are not on this site. They cleared every other test and then failed this one. The exact commission and slippage parameters, and the validation stages each strategy has to pass, are written up in full on the methodology page.