The backtest numbers on the home page didn't come from one lucky run. Here's the short version of how Level Reactor was built, what broke along the way, and how the final rules beat the alternatives.
I already traded US30 off daily levels: the previous close, yesterday's range, the open. Drawing them was easy. The hard part, every single day, was knowing in the moment whether a level was holding or giving way.
So the goal was never "more lines on a chart." It was a tool that answers that one question, clearly, as it happens.
An earlier version of my levels indicator printed a sell arrow. I took the trade, it lost, and when I looked back the arrow was gone. The indicator had been repainting: quietly rewriting its own history so the chart looked better than it really traded.
Level Reactor tracks every level through a set of states: away, testing, respected, breaking, broken, retesting, retest held, failed break. Each state is judged on closed candles only.
Before it was finished, I checked it against real chart days from September. That caught two flaws worth fixing:
I traded with it on a Topstep futures account and logged every day: screenshots of the session, every trade, and whether each one actually matched a Level Reactor signal.
The honest finding: most of my bad days came from me, not the indicator. Taking trades with no signal, re-entering a setup that had already stopped out, flipping direction straight after a win. That's what pushed me to stop trading on feel and test a fixed set of rules properly.
I pulled the signal history from TradingView and replayed it, changing one decision at a time: chart timeframe, trading session, start time, target, trades per day, and breakeven.
| Choice tested | Sep – Oct | May – Aug | Verdict |
|---|---|---|---|
| 15-minute chart | +$17.6k | +$19.2k | Best |
| Hourly chart | +$14.2k | +$3.9k | Inconsistent |
| London + New York | +$17.6k | +$19.2k | Best |
| London only | +$2.5k | +$10.5k | Weakest |
| Asia session | +$1.1k | −$13.9k | Avoid |
| Target 1:1 (150 pts) | +$11.0k | +$20.2k | OK |
| Target 1:2 (300 pts) | +$9.7k | +$14.8k | Weaker |
| First level past 150 pts | +$17.6k | +$19.2k | Smallest drawdown |
September–October was an easy market: roughly 9 out of 10 rule combinations made money. Testing that many combinations means some will look great by chance.
Exits were also checked against 1-minute data for the 5–16 October window, the only 1-minute history available: all 349 trades in that window came out the same. When a single candle touched both the stop and the target, it was counted as a loss.
What should you do if an opposite signal fires while you're still in a trade? Four answers, same rules, same dates:
| When an opposite signal fires | Trades | Net | Worst drawdown | $3k drops |
|---|---|---|---|---|
| Ignore it | 209 | +$40,127 | −$4,411 | 4 |
| Close the trade, stay flat | 222 | +$41,356 | −$2,528 | 0 |
| Close and reverse into it | 222 | +$34,946 | −$2,600 | 0 |
| Close only if losing | 217 | +$36,070 | −$4,976 | 1 |
Closing the trade made about the same money as ignoring the signal, with roughly half the drawdown, and the smaller drawdown showed up in both the May–August and September–October periods. That became the final rule.
Where it landed
15-minute chart. First two trades of the day from the London open. Target the first level past 150 points (300 if there isn't one). Stop to breakeven at +100. Close if an opposite signal fires.
About 5½ months of 15-minute data, which is all TradingView provides. It can't be tested further back.
The test assumes exact fills at entry and breakeven, with no slippage. Live trading will be somewhat worse.
The rules were picked from the same period they're tested on. The two-period filter reduces that risk but doesn't remove it.
The rules are now being traded forward. Daily recaps, losing days included, go up on Instagram so you can watch it live.