# Stacking Indicators: Why Three Filters Beat One Signal One signal on its own fires too often and is wrong too often. Each filter you add removes a different kind of losing trade. This lesson walks through a real research case on Market Open, where a structure signal lost money on every market tested until two more filters were stacked on top of it. ## One signal, every kind of mistake Take a plain structure signal: every CHoCH the SMC chart draws, taken on the 1-minute chart, with the stop at the swing and a target at twice that distance. It sounds reasonable. In practice it makes three different mistakes at once: - **Noise.** On a fast chart, structure flips back and forth. Many trades are closed by the next opposite signal long before they reach either the stop or the target. - **Fighting the trend.** About as many signals point against the bigger move as with it, and those have the worst odds. - **Fees.** When the average win is around 1 % and every trade pays fees on the way in and out, a strategy that is roughly flat before fees turns clearly negative after them. No single setting fixes all three, because they have different causes. That is the whole argument for combining tools: each one should answer a different question. ## Filter 1: a stricter structure signal (IDM) The first filter asks *is this a real change of character, or just noise?* Price Action — SolAtlas can take only an **IDM CHoCH**, a change of character on a swing that was confirmed by an inducement sweep. That throws away most signals. On the 1-minute chart, the count drops from hundreds per market to roughly one a day, and the trades that remain have wider stops and are held for hours instead of minutes. ## Filter 2: trade with the trend (EMA 200) The second filter asks *which way is the bigger move?* A BUY is taken only when price closes above the EMA 200, a SELL only below it. A signal against the trend is skipped. In the research, most of the IDM trades that survived were already on the right side of the EMA 200; adding the filter formally removed the few that were not. ## Filter 3: make the winners pay for the losers (3 R) The third filter asks *is the reward big enough?* The target is set at 3 × the risk, measured from the entry close. With fewer, wider trades, the wins are large enough to cover the losses and the fees. In the research, a win rate under 40 % was enough, because the average win was much bigger than the average loss. Put together, this is exactly the rule Price Action — SolAtlas draws as BUY ▲ / SELL ▼: an IDM CHoCH, only with the EMA 200 trend, the stop at the structure level, the target at 3 × risk, and an opposite IDM CHoCH closes the trade. The AI Indicator bot in AI SMC IDM mode trades the same rule.  ## What the research showed The study tried 288 variants on 1-, 5- and 15-minute charts, picked candidates on 10 days of data, then re-tested only the finalists on 30 older days and on 4 coins it had never seen. Results are the sum of each trade's price move after fees, before leverage. | 1-minute chart, 30 days, 10 markets (out of sample) | Trades | Win rate | Net | Markets in profit | |---|---|---|---|---| | Plain signal (Classic CHoCH, target 2 R, flip on every signal) | 9,447 | 28.5 % | −901 % | 0 of 10 | | IDM only, target 2 R | 350 | 38.9 % | +15.7 % | 5 of 10 | | **IDM + EMA 200 + target 3 R** | 331 | 36.9 % | +58.9 % | 7 of 10 | Read the table as a lesson about filters, not as a promise of returns. Each layer removed a group of losing trades: IDM cut the noise and most of the fees, the trend filter and the 3 R target turned a small edge into a usable one. The same three filters did not work everywhere: on 5- and 15-minute charts nothing held up in both halves of the data, so the edge here is specific to the 1-minute chart. ## Applying it to altcoins, US stocks and gold Market Open lets you put the same rule on any market it charts: altcoin perps on Hyperliquid and Binance, US stock perps such as NVDA or TSLA, and gold. The research covered nine crypto markets and one stock perp. It did not test gold, and markets with different hours and volatility can behave differently. Before you rely on it anywhere new, watch the signals on that market in demo for a few weeks. ## Size by risk, not by margin IDM stops are wide and uneven, from under 1 % to more than 10 % away. With a fixed margin and fixed leverage, one wide stop can cost more than the margin itself. The research sized every trade so that a stop-out cost the same fixed share of the account, for example 1 %. The longest losing streak was 10 trades, so even at 1 % per trade the account fell about 15 % at its worst. Note that the bot today uses a fixed margin per trade, so keep its margin and leverage small enough that the widest stop is still a loss you accept. ## Honest limits - 288 variants were tried. When you test that many, some look good by luck. The out-of-sample check reduces that risk but does not remove it. - About 390 trades over roughly 40 days is a moderate sample, not long-term proof. - Slippage matters: with 0.10 % extra slippage each way, the edge shrank to roughly break-even. - Past results say nothing about the next trade. Combining filters improves the odds you start with; it does not make a strategy safe. The full research notes are public: [the strategy and its tests](https://claude.ai/artifact/AJePV9uFs24e7UPFcAcfv2) and [the bot backtest on ZEC and PUMP](https://claude.ai/artifact/FtQa6598SpGWAPz4gf8TpB). ## Try it on Market Open 1. Open a liquid market (BTC, ETH, SOL or ZEC) on the 1m chart and add Price Action — SolAtlas. 2. Open its **Signals** tab, turn the trend filter EMA to 0 for a moment, and count how many extra signals appear against the trend. Set it back to 200. 3. Follow the last five BUY / SELL to their result tags and note how many closed at TP, at SL, or on an opposite signal. 4. Try the same on a stock perp or gold in demo, and compare how often signals appear and how wide the stops are. _Educational content, not investment advice. Leveraged trading can lose more than your margin._