Which Obvious Trading Ideas Fail When You Actually Test Them?

Most trading ideas that sound obviously right fail when you test them on real tick data. Behind the three gold systems I have shipped sit 158 pre-declared experiments, and 140 of them never became a product. Taking profit earlier, never giving the day back, cutting losses at a daily line: each one made intuitive sense, and each one lost to the plain version of the system.

MasterVP · Systematic Trading Journey

Written by the founder of KenKem, a software engineer of twenty years who traded gold by hand before building the research stack behind MasterVP. This is for anyone who has a clever improvement in mind for their own strategy and wants to know how to check it before it costs money. It is knowledge sharing, not a recommendation. Every figure below is a backtest or research measurement on real tick data from my release 1.26 write up of 15 September 2026, and none of it is a live result.

How many of my ideas actually survive testing?

Very few, and the count is the point.

The research record for release 1.26 lists 158 pre-declared experiments. Eighteen were promoted past the first stage and three shipped. Of the 140 that did not become a product, 52 were rejected outright, 5 were stopped mid-campaign, and 83 stayed research-only. Along the way the engine searched 7,990 distinct configurations, and 169 runs were refereed on MetaTrader 5 with real ticks.

I keep that number in view because a survivor only means something if the selection was harsh. A tool that passed a soft test is barely different from one that was never tested. Master Volume Profile is one of only two edges that came through the whole process in my engine. That is a literal count, not modesty dressed up as a number. The default answer in my research is no, and almost everything gets it.

Does taking profit earlier reduce risk?

It is the most intuitive idea on the list, and in my testing it made things worse.

The reasoning is easy to follow: bank a gain sooner, expose less, sleep better. I tested nine variants of partial profit-taking on the shipped system. They cut the winners that pay for the losers. On this system a few big winners carry the result, and one trade that reaches its full target pays for about five full losses. Clip those winners early and the losses stay the same size while the thing paying for them shrinks. Drawdown got worse, not better.

The full measurement is in should you take profit early or let winners run. The short version is that a risk reduction idea has to be tested as a risk reduction idea. "It feels safer" is not a result.

Should you never give the day's profit back?

I wrote it down as a rule and tested it. It cut profit and worsened drawdown.

A giveback stop sounds like discipline: once you are up for the day, protect it. My read of why it failed: the trades that come after a good morning look like the ones before it. A stop that ends the day early does not remove losing trades. It removes the next winners along with them, including the recoveries that tend to follow a bad stretch.

Is a daily loss cutoff a good way to improve results?

Used as an optimisation, no. Used as a safety rail, yes, and those are different jobs.

I tested per-session loss caps and win-streak caps as ways to improve the system: 35 variants, and none of them beat the control on both of the ten-minute legs. The cutoffs clipped the bad days and also blocked the recoveries that followed them. I still run a daily loss limit, 5 percent, because account rules and common sense call for one. I just do not pretend it makes the strategy better. Its job is to stop a bad day from becoming the last day.

What about a hard drawdown halt, or trading crypto through the weekend?

Both failed clearly, one of them badly.

A hard drawdown halt, where the system simply stops after a set loss and waits, was outright catastrophic in testing. My read is that stopping at the bottom locks in the loss and skips the climb back. That is different from the account guardian I do ship, which stops at a prop-style loss line to protect the account, not to improve results. Trading crypto through the weekend made profit clearly worse and drawdown clearly deeper, which tells me the rules were not built for that market. The shipped system holds nothing over a weekend.

The pattern in both is the same. The idea was reasonable in a sentence and expensive in a measurement.

Can a filter fail even when its numbers look positive?

Yes, and this is the failure I find most instructive.

One lever I tested refused entries during three opening hours. On the 1,448 setups it refused, the average result looked fine in isolation. But when I replayed the whole sequence with the filter on, the benefit disappeared. A refused trade does not leave an empty slot. It frees the slot for a different trade, so the filter has to be judged by what replaces the trade it removed. In isolation the setups looked positive; in sequence they were near zero or negative. The idea was closed.

Other levers died more quietly. An anti-chase ceiling, which declines entries far from the start of a move, was tried in four formulations on three feeds, with no passes. The best arm failed the deflated Sharpe test and rested on only 20 trades. A volatility floor improved one broker's results, but the result fell off a cliff one step from the chosen value, which I treat as the signature of a lucky setting, so it was not adopted. A VWAP-drift strategy family was built and refereed in four days, and shelved when none of 840 cells was positive on 2021 to 2023.

Why do I show the setups the system turned down?

Because a tool that only shows its hits is telling half the story.

The MasterVP indicator can mark near-misses: places where a setup began to form but did not clear the filters, tagged with a short plain reason such as the trend disagreeing or the move looking like a chase. It is optional and off by default, because it makes the chart busier. I find those rejected setups more instructive than the winners. The discipline of a system lives in what it refuses, and I wanted that half to be visible as well.

What does a bad idea cost if you never test it?

Real money, usually in the form of a rule that feels like protection.

The ideas above are all ones a sales page would happily sell. Each sounds like risk management. Each was written down as a rule before it was tested, and each failed on my own data. Had I trusted the sentence instead of the measurement, I would have shipped a system with a worse drawdown and called it careful.

I cannot tell you these ideas fail on your system. I can only say they failed on mine, on gold, on the clocks I trade. What transfers is the method: write the rule down first, set the bar before you see the result, compare against the plain version re-sized to the same drawdown, and keep the graveyard visible. How the pass mark is set is covered in how do you turn a trading pattern into a rule you can test, and the wider question of trusting a result in what should you ask before you believe a backtest.

Frequently asked questions

How many trading ideas should I expect to reject? More than feels comfortable. For my gold systems, 140 of 158 pre-declared experiments never became a product, and only 3 shipped. A method that never rejects anything is a story with numbers attached.

Does taking partial profits reduce drawdown? Not on my system. Across nine variants it cut the winners that pay for the losses, and drawdown got worse. It may behave differently on a system with a different profit distribution, so test it on yours.

Is a daily loss limit worth having? As a safety rail, yes. As a way to improve results, my testing says no: 35 variants of session and streak caps did not beat the control on both legs. I keep a daily limit to stop one bad day from becoming the last one, not to make the strategy better.

Why test a filter in sequence and not in isolation? Because refusing a trade changes which trade comes next. A filter that looks positive on the setups it removes can still make the whole system worse once the freed slots are filled by other trades.

What is a sign that a parameter was lucky? A cliff next to it. If the result collapses one small step away from the chosen value, the setting probably fit noise rather than structure.

Can I trust a result from only a few trades? Not much. One arm I tested looked best but rested on 20 trades and failed the deflated Sharpe check. Sample size is the first thing I check, before the headline number.

Does MasterVP tell you what to buy or sell? No. It is software and education, not advice. The indicator shows the read and leaves every decision with you.

If you want to see where this ended up, MasterVP is my own intraday method built on public volume profile theory. It ships as the Master Volume Profiler indicator on TradingView and MetaTrader 5, which shows the read and leaves the order with you, and as Master Volume Sniper on MQL5 market, an Expert Advisor that executes the rules and risk settings you configure. Which of your own ideas have you tested against the plain version?

---

This article was composed from the rejected-ideas posts of the MasterVP build log series. Educational purpose only. Not financial advice. Figures cited are backtest and research results on real tick data, and past performance does not guarantee future results. Every trader remains responsible for their own risk settings and decisions.