When traders first get access to Trade Ideas' scan builder, they usually skip straight to configuring filters. They want to test an idea immediately. That's backwards. Before you build a single filter, you should answer fundamental questions about what you're actually trying to find. Most traders skip this work. They build scans without clarity about the edge they're trying to capture. Then they wonder why the results disappoint.
The first question nobody asks is: Why would this pattern work? Not "what does historical data show"—that comes later. Before backtesting, you need to understand the mechanism. If you're building a scan for stocks that bounce off their 200-day moving average, the mechanism would be that the 200-day average represents an important long-term trendline, and institutions that missed buying lower are willing to buy when the stock returns to that level. That's a credible mechanism. A pattern based on bouncing off the 50-day average has less mechanism because it's shorter-term and less likely to represent institutional support.
If your pattern mechanism is weak or nonexistent, the backtest won't save you. You might find that the pattern worked 55% of the time historically. But 55% isn't an edge. That's barely above random. You probably overfit finding those results. A pattern with a clear mechanism—multiple ways it could work logically—is more likely to have genuine edge than a pattern that just happens to have worked statistically.
The Unasked Gatekeeping Questions
The second question is: How often should this pattern occur? If you're building a scan for stocks that break above the 20-day high on volume above 1.5x average and RSI between 50-70, that's a fairly specific setup. It might occur 10-20 times weekly in a market of 3,000+ stocks. That's good—you have multiple opportunities. But if your scan fires 60+ times daily, you've made it too broad. If it fires once weekly, you might be too specific and you're not getting enough data to evaluate whether the pattern actually works.
Most traders don't think about this. They build a scan and see how often it fires. If it fires too frequently, they add more filters. If it fires too infrequently, they remove filters. They're using alert frequency as the signal that their scan is configured right. But alert frequency and setup quality aren't necessarily correlated. The right alert frequency depends on your trading style and your ability to execute. A day trader should expect 5-15 alerts daily. A swing trader should expect 2-5 alerts daily. A position trader might expect 1-3 alerts weekly. Build with those expectations in mind.
The third question is: Can I actually execute this scan's alerts? Some patterns look great in theory but create execution problems in practice. Maybe your scan identifies consolidations that are about to break but only fires alerts on stocks that are very illiquid. You can't get filled at the right price. Maybe it identifies momentum setups but only in penny stocks where spreads are huge and fill quality is poor. Or maybe it identifies setups that require entry exactly at a specific price level—but it takes you 30 seconds to get your order in, and in 30 seconds the price has moved too far. Before backtesting, think through whether you can actually execute what the scan tells you to do.
This filters out a lot of ideas before you waste time backtesting. A scan for stocks breaking above resistance on very high volume sounds good until you realize most of those stocks are illiquid and you can't get quality entry fills. A scan for earnings surprises sounds good until you realize earnings happen after hours and you can't trade pre-market. A scan for sudden volume spikes on specific indicators sounds good until you test it and realize that by the time you get your order in, the spike is already fading. Execution filtering saves you time wasted on backtesting patterns you couldn't trade anyway.
The Backtesting Readiness Check
Once you've passed the mechanism test, the frequency test, and the execution test, then you can backtest. But before you run the backtest, ask yourself: How much historical data should I test? If you test just the past year, you get recent results but you're not accounting for different market regimes. If you test 10 years, you get diverse market conditions but the oldest data might not reflect modern trading dynamics. Most traders should test at least 3-5 years. That captures different bull markets, bear markets, and sideways markets. Test longer if possible.
What win rate are you looking for? This matters because it drives expectations. A pattern that works 60% of the time is excellent. A pattern that works 55% of the time might still have edge if the winners are larger than the losers. A pattern that works 65% of the time is very strong. Most traders think they need 70-80% win rates. Those are unrealistic for any non-trivial pattern. Set realistic expectations. 55-65% is actually quite good for a mechanical pattern.
Another unasked question: Will this pattern work during the types of market conditions I'm actually trading? If you're backtesting a trend-following pattern during 2010-2015, when markets were generally bullish and trending, the pattern will look excellent. But if you trade it in 2022-2023 when markets are choppy and choppy, it'll underperform because the market regime is different. The solution is to test your pattern separately on bull markets, bear markets, and sideways markets within your backtest period. If it works reasonably well in all three, it has a chance. If it only works in one regime, you need to understand that limitation before trading.
Finally, the question traders should always ask but rarely do: How much capital do I need to actually trade this pattern? If your scan identifies 50 setups daily but you only have enough capital to trade 3-5 daily, you're leaving opportunities on the table. Or you're being too selective and degrading the mechanical execution that gives the pattern its edge. Better to build a scan that fires less frequently so you can trade all of them with proper position sizing. A scan that fires 5 times daily that you execute 5 times is much better than a scan that fires 50 times daily that you execute 5 times after careful selection.
These questions don't get asked because they're not glamorous. Everyone wants to jump to backtesting and parameter optimization. But the traders who build the best scans are the ones who answer these foundational questions first. The scan that emerges from that thinking is usually simpler than the scan built through pure optimization. But simpler scans tend to work better because they're harder to overfit. They're based on mechanism rather than lucky parameter combinations. They're designed to be executed consistently rather than optimized for maximum backtest returns. That's how you build scans that Trade Ideas Review actually work.