What PatternX can't do: limitations and honest caveats

Every trading tool has blind spots. This page lists PatternX's in plain terms, so you can judge how far to rely on it. Most of these limits follow directly from how the product works, as described in pattern detection and the other methodology pages.

The latest bar may still be forming

Every detector requires the breakout on the most recent bar, and that bar may not have finished. PatternX does not wait for it to complete. A breakout seen two minutes into a 15-minute bar can reverse before the bar ends, leaving a false breakout where the signal fired. The volume check runs on that same partial bar, so it measures volume so far, not the bar's final volume. In manual mode, look at the chart before approving; a signal expires after 15 minutes either way.

Fixed windows see only part of the chart

Each detector reads a fixed trailing window of the 200 bars fetched: 80 bars for a cup and handle or an ascending or descending triangle, 90 for head and shoulders, double tops and bottoms and rectangles, 110 for triple tops and bottoms, 75 for a symmetrical triangle, 70 for wedges, a 45-bar base for rounding patterns and up to about 41 bars for flags and pennants. A formation that took longer to build is not seen on that timeframe, though it may fit the window on a higher one. Swing points need two or three bars on each side, so peaks and troughs are confirmed late.

Because all 20 detectors run independently, one move can be reported as more than one pattern. A tight pause after a rally might match both a bull flag and a bullish pennant, and a flat-topped range both an ascending triangle and a bullish rectangle. Two signals on the same move are not two independent confirmations.

Geometric rules and fixed priors

The detectors apply geometric rules to price. They do not read news, earnings, fundamentals or macro events. Only the rectangle detectors check the trend before the pattern, the continuation and reversal labels are descriptive, and no detector has its own volume rule beyond the universal 1.2× check. In the PatternX Score, the per-pattern reliability priors (59 to 68) are fixed values, not measured statistics, the historical factor needs 10 closed trades on a market before it says anything, and the displayed "probability" is a heuristic.

Market filters are snapshots

The universal filters (volume, ATR, liquidity, spread, order-book imbalance and, for crypto, 24-hour volume) are measured at the moment of the scan. A spread can widen and an order book can thin out before your order arrives. The filters can also reject setups that would have worked. Passing them means the market looked tradable at that moment, not that the trade is safe. See market data for where these inputs come from.

Only your tracked symbols, at your plan's pace

PatternX scans your connected venues × tracked symbols × chosen timeframes, and nothing else. Tracked symbols are capped at 3 on Free, 20 on Starter and 100 on Pro, with no cap on Elite, and rescans run every 120 s on Free, 60 s on Starter, 30 s on Pro and 10 s on Elite. A pattern on a symbol you don't track is never seen. On short timeframes the interval matters: with 120-second scans on a 1m chart, some bars are never the latest bar when a scan runs, so a breakout on one of them can be missed.

Targets are capped by your settings

The take-profit is the nearer of your percentage target and the pattern's measured move, and the stop is the tighter of your percentage stop and the pattern stop, with a 0.2% buffer. The full textbook target is therefore often not used, and setups with reward:risk below 1 are dropped. There is no invalidation logic after entry: if the pattern fails, the stop is the exit.

Backtests are simpler than live trading

Backtests replay the same engine bar by bar, but with neutral market context: the liquidity, spread, order-book, funding and 24-hour-volume filters are effectively off, and there is no higher-timeframe or live win-rate input. They hold one position per series, use fixed sizing and percentage targets only, and do not model the risk engine, trailing stops, break-even, approval delay or partial fills. Costs are assumptions: 0.03% slippage, 0.05% fees per side and 0.01% funding per 8 hours, applied in every backtest. When one bar touches both stop and target, the stop is assumed to fill first. Fewer than 30 trades triggers a warning, because small samples say little.

Trailing stops and break-even run in PatternX's software

For live trades, the protective stop is placed at the venue right after the fill, using each venue's own mechanism. Break-even and trailing stop adjustments work differently: PatternX's position loop, which runs every 5 seconds, manages them, and they are not changed at the venue. They only work while PatternX is running and can reach the venue, and they exit with a reduce-only market close, which can slip. If a venue refuses the stop, PatternX closes the position at market and alerts you; if that close also fails, you get an UNPROTECTED alert. See risk controls.

Venues and brokers have their own limits

Seven brokers (eToro, Plus500, Webull, Robinhood, TradeStation, Charles Schwab and CMC Markets) are manual-only: PatternX sends signals and you place the trade. Eight crypto exchanges are spot-only, for data, paper and manual use. Each venue caps leverage, for example at 20 on Binance, 25 on Bybit and Hyperliquid, 4 on IBKR and 2 on Alpaca. Live trading needs a connection that passed its test and an API key not known to allow withdrawals. Paper trades fill at the live bid or ask with estimated fees, so live fills can be worse. On US margin brokers, the pattern-day-trader guard is on by default and blocks a fourth day trade within 5 business days while equity is under $25,000.

Data can be late, thin or missing

Signals are only as good as the candles, order books and funding data each venue or broker supplies, and quality and history vary between them. Funding exists only on perpetual futures. If a live position can't be priced, a watchdog alerts you after 3 minutes when no venue stop is resting, or after 10 minutes when one is. In Momentum Lab, data older than 10 seconds or from a delayed feed is rejected, IBKR supplies no VWAP, Alpaca's free IEX feed has premarket data only from 08:00 New York time, and relative volume needs IBKR data.

In Momentum Lab, a trail is not a maximum loss

Momentum Lab, an Elite feature, exits with a trailing stop and, by default, a 3% hard stop. Neither caps your loss. Gaps, wide spreads, slippage, partial fills and order rejections can all push an exit well past the trail. With default settings, positions use PatternX's Synthetic Trail, which the engine evaluates every cycle; in live mode an emergency STOP_LIMIT order also rests at the broker, but a stop-limit can fail to fill on a gap. Lab-wide breakers only block new entries; they do not close open positions.

Extended hours carry extra risk

Momentum Lab is built around the premarket, 04:00 to 09:30 New York time, and IBKR connections can also trade pre-market and after-hours. Outside the regular session, liquidity is usually thinner, spreads wider and gaps more likely. In extended hours the broker-side emergency stop rests only where the broker supports stops at that time: IBKR does, Alpaca does not. See extended-hours trading risks.

No guarantee of results, and no advice

Chart patterns describe tendencies, not certainties, and any setup PatternX shows can fail. Past results, live or backtested, do not guarantee future ones. PatternX is software, not a broker, an exchange or an investment adviser. It does not hold your funds, nothing it shows is financial advice, and its language-model features are instructed not to give buy or sell advice. You decide what to trade and how much to risk. Read the full risk disclaimer before trading with real money.