Historical test result
What doubling costs did to a historical Brent test
Is it true that costs matter little on intraday oil trades? If you double modeled costs, the same Brent hour rule drops from -10% a year to -16% a year. I kept the symbol, feed, test period, signal times and size fixed on a product that tracks the Brent oil price without buying oil, from 2017-2021, so you see what costs alone did.
What many believe
Many traders start from this thought: Costs matter little on intraday oil trades.
I get why it sticks. One intraday move, one move within the day, looks bigger than the fee to open and close it. A backtest is a historical simulation, so it lets you replay the same rule with one change and see what that change did.
What I kept fixed
I kept five things fixed for a clean read. The symbol, the market name, stayed a product that tracks the Brent oil price without buying oil. The price feed, where the prices come from, stayed the same. The test period, the years replayed, stayed 2017-2021. The signal times, when the rule checks, stayed the same. The size, how much per trade, stayed the same.
Baseline is the first run with normal modeled costs. Modeled costs are the cost number I use in the simulation to stand in for what trading would charge. The second run doubled only that number. Nothing else moved.
What doubling changed
Doubling modeled costs turned -10% into -16% a year on the Brent hour rule.
Expressed as a constant rate per year, the first run lost 10.05% and the run with double modeled costs lost 15.91%. That is a 5.86 percentage point deterioration, and it does not explain away an already negative baseline. Both legs, the two parts of the rule, were negative even before costs.
On a hypothetical $100,000 starting balance, -10.05% is -$10,050 for one year and -15.91% is -$15,910 for one year. This is one year of arithmetic at the equivalent yearly rates. To check a challenge loss limit, you need the equity path and the limit in force at each point.
Why I look at the loss path too
The yearly net, what is left after costs for the year, tells you where you ended. The loss path tells you how deep the dip ran before the end. A rule can end at the same yearly net with a shallow dip or a deep one, and the deep one hurts more to hold.
Here both runs ended negative, and the doubled-cost run ended more negative. I drop a rule that already loses before costs because doubling costs only made this loss worse.
How to run the same check on your report
Take your own report from a historical simulation you trust. Copy it, change only the cost number, and run it again. Double only costs in your own report and compare the yearly net and the loss path with the baseline.
Keep the symbol, feed, test period, signal times and size fixed like I did. If the yearly net turns clearly worse and the loss dip runs deeper, costs are not a side note for that rule. If the rule loses before costs, better fees alone cannot make that run profitable.
Common questions
Do you share the entry times or the rule?
No. I keep signal times fixed in the test but I do not publish them, and this note does not give you a rule to trade.
Does this say what live trading would pay?
No. This is a historical simulation on a product that tracks the Brent oil price, from 2017-2021 with modeled costs. It shows what doubling that cost number did in the earlier run.
Your next step
Run the one-change check on a rule you already use. If you want a second pair of eyes, start with the free Prop Rule Guard (https://stratcorealpha.com/rule-guard?ref=website).
Sources and limits
Original question with the historical run (https://www.instagram.com/reel/DfUo7DOmDOW/)
This historical run covers 2017-2021 and says nothing about your feed or future results.