Your best setup is the one with the highest expectancy over a meaningful sample - and you find it by segmenting your trades, not by memory. Aggregate profit and loss blends everything into one average, hiding the fact that one great setup is carrying two mediocre ones. Break the data apart and the truth appears.
Step 1: tag every trade by setup
You cannot segment what you did not label. Each trade needs a setup tag - a short, consistent name like "breakout retest" or "London reversal". Consistency matters more than cleverness: the same setup must always get the same label, or the grouping falls apart. This tag is the key that unlocks everything below.
Step 2: group and compare expectancy
Sort your trades by setup tag and compute expectancy in R for each group. Expectancy, not win rate and not total profit, is the fair comparison, because it accounts for both how often a setup wins and how much it wins relative to risk. Line the setups up side by side and a ranking emerges.
Overall the account looks mildly profitable. Segmented, the news scalp is a leak and the London reversal is the real edge - do more of one, cut the other.
Step 3: add a second dimension
Setup alone is a good start, but the sharpest insights come from crossing it with another variable - usually session or market condition. A breakout retest might be flat overall yet strongly positive only during the London session. That is not a setup to cut; it is a setup to filter. Segmenting by two dimensions at once turns "this does not work" into "this works here, not there".
Important: respect sample size. A setup with +0.9R expectancy over 6 trades is not your best setup - it is unproven. Require at least 30 trades before trusting a ranking and closer to 100 before making big commitments. Small samples produce confident nonsense.
Step 4: act on the ranking
Once the data is clear, the decisions are simple: do more of your highest-expectancy setup, apply filters to rescue the borderline ones, and cut the negative ones you executed correctly. This single exercise - concentrating on what works and removing what does not - often improves results more than learning any new strategy, because it makes your existing edge purer.
Why backtesting makes this faster
Live trading produces data slowly, so it can take months to gather 100 trades per setup. Backtesting compresses that. When you replay charts in a simulator and tag each trade, you can generate hundreds of trades across several setups in days, then use the report's tag analysis to see which setup carries your edge - long before you commit real money to any of them. It is the fastest honest way to answer "which of my setups actually works?"
Finding your best setup FAQ
How do I know which of my trading setups is best?
Tag every trade by setup, group your history, and compare expectancy in R over a decent sample. The highest-expectancy setup with enough trades is your best - total profit hides this.
How many trades do I need before trusting a setup?
At least 30 for a rough read, closer to 100 for big decisions. Below 30, a couple of lucky trades can reverse the ranking.
Should I drop a setup that loses money?
Usually yes if it shows negative expectancy over a reasonable sample and you executed it correctly - but first check whether a session or condition filter rescues it.