Signal Forge (LuxAlgo): review and backtest
What does Signal Forge claim?
It lets you switch on any of 11 signals (moving-average crosses, RSI, MACD, Supertrend and others) plus an ADX filter, shows each signal's standalone win rate, and backtests whatever combination you pick on the chart.
How we tested it
Every combination: 1,023 sets of signals with the ADX filter on and off, 2,046 in all, on gold, EURUSD, Nasdaq 100 and S&P 500 at 1, 5 and 15 minutes, each traded the same way (stop 1.5 times the average range, target twice the risk). We first confirmed our rebuild matches the indicator: 595 trades on gold 5-minute, the same count as its own backtest.
Its default setting (always in a trade, reversing at each cross) was flat before costs over 7,792 trades and clearly lost money after spreads and commissions. Its own stop and target defaults land just below break-even before costs (42.46% won, 42.86% needed).
Does Signal Forge work?
Not in our test: it did no better than chance. With 2,046 combinations to choose from, some will look good on any chart by luck alone. The fair question is whether the best combination beats what the same search finds on random trades. It did not: when a coin flip chose buy or sell at the same moments, the best of its combinations scored higher than the indicator's best in all 300 repeats. Adding more signals made results worse, not better.

Watch the 30-second version on YouTube
What are the best Signal Forge settings?
Not one of the 2,046 combinations we tested made money before costs, and none beat what coin-flip directions produce in the same search. Its default setting, a plain 10/20 moving-average cross, ranked 26th of 2,046 and was flat before costs over 7,792 trades.
Limits of this test
- Corrected 9 October 2026: the prices we had stored as Bitcoin were copies of the gold prices, so gold was counted two or three times and Bitcoin not at all. With the copies removed, 0 combinations made money before costs (we said 40), its default ranked 26th (we said 21st) over 7,792 trades (we said 9,576), and its own stop and target won 42.46% (we said 42.81%). The verdict is unchanged. Our first 30-second video of this test was made before the correction and said 40; we took it down, and the video on this page is the corrected one.
- All 2,046 combinations used one exit (stop 1.5x average range, target 2x risk) so they can be compared; the indicator's own exits were tested separately on its default setting (above).
- USDJPY is not in the figures: the prices stored for it were not USDJPY prices, and too short to use.
Measured results, not advice. We have no commercial link with LuxAlgo. If you made this tool and think a figure is wrong, write to corrections@trackedsignals.com; corrections are published within 7 days. How we test.
More strategy tests
- LuxAlgo YouTube strategiesNo edgeLuxAlgo · YouTube strategies127 videos, 22 new methods, 74 versions on 12 markets: about zero before costs, 0 passed.
- AI Adaptive Money Flow Index (Clustering)Edge too small for costsAlgoAlpha · TradingView indicatorIts trend labels lean the right way, by about half of one trade's cost; 0 of 75 settings held up after costs.
- HalfTrend Long/Short Signal EngineLost money before costsBigBeluga · TradingView indicatorIts dashboard shows 53% wins at 1:3; per trade, 22.7% reached the 1:3 target (25% needed).
- Multi-Timeframe Trend Signal Engine + Symmetry Reversal OscillatorNo edgeViprasol · TradingView indicatorIts two tuning inputs never changed a single trade; the trend signal lost after costs at every setting.