Automated momentum trading: what software can and can't do

What software does well

Premarket momentum trading is an attention problem before it is anything else: moves start at 04:00, play out within minutes and happen in many stocks at once. Software is good at that kind of problem.

Monitoring many symbols at once. A person can watch a handful of charts; a program can re-evaluate every tracked candidate on each cycle. PatternX's Momentum Lab runs a cycle about once a second, handling open positions first so exits take priority, then its circuit breakers, then the strategies. It refreshes the top 50 gainers and 50 losers from your data source every 15 seconds. That is broad coverage, not the entire market: a stock outside those lists isn't seen.

Applying the same rules every time. The trigger, filters and sizing at 09:25 are the ones you set before 04:00. There is no hesitation after a loss and no "just this once" exception. The flip side is that a bad rule is applied just as consistently, which is why testing matters.

Deterministic state. Each symbol a strategy tracks is in exactly one state at a time: WATCHING, ARMED, TRIGGERED, ORDER_PENDING, PARTIALLY_FILLED, OPEN, EXIT_PENDING, CLOSED, COOLDOWN, or SKIPPED. That matters because some of the costliest automation mistakes are duplicates: two orders for one signal, or a re-entry nobody intended. Explicit states make it clear at every moment whether a stock is being watched, has an order working or is already held. See deterministic trading.

No attention limits. Software doesn't tire at 05:30, miss a breakout while looking at another screen, or loosen its rules out of boredom on a slow morning.

Persistence across restarts. Momentum Lab stores candidate states, the high- and low-water marks its trailing exits use, and its orders. After a restart, orders are re-polled from the broker, never resent, so a restart doesn't duplicate them.

A kill switch. One control blocks new entries and cancels working entry orders; closing open positions is a separate, optional choice. Stopping and manual exits work on every PatternX plan, not only Elite.

Circuit breakers. Lab-wide breakers stop new entries when something goes wrong. Defaults: a daily loss of 3% of equity, 4 consecutive losses, 20 trades in a day, 3 execution errors, 30 seconds of data failure or 60 seconds of broker failure; a dollar loss limit and an aggregate exposure cap exist but are off by default. Breakers only block entries. They don't close positions. See risk controls.

What software can't do

Fix its own data

Every decision rests on the feed. Feeds differ in coverage and start time (Alpaca's free IEX feed starts its premarket at 08:00 ET), in what they report (Alpaca data has no halt flag; IBKR data has no VWAP) and in reliability. Momentum Lab refuses delayed, untimestamped or crossed data and anything older than 10 seconds, but refusing bad data is not the same as having good data. During an outage, entries stop, and the software's trailing exits wait too, because price-based exits act only on fresh data; time-based exits still fire on stale data, and a stop resting at the broker, where there is one, is unaffected. The scanner guide covers data quality in more detail.

Control fills

Signals are decisions; fills are outcomes. Momentum Lab enters with a marketable limit priced 0.5% through the ask or bid, and an entry that hasn't filled within 10 seconds times out. Exits use marketable limits re-priced on every attempt: the first allows up to 1% through the quote, each retry widens that by 0.5 points, up to 6 attempts, after which it keeps retrying at the widest buffer and raises a critical alert. For example, against a $15.00 bid the first exit attempt's limit is $14.85 and the sixth's is about $14.48, if the bid hasn't moved. A market order is used only in the regular session, when fresh data isn't available. Partial fills happen, which is why PARTIALLY_FILLED is a state of its own, and brokers reject orders, for instance a short sale with no shares to borrow.

Change extended-hours order rules

Brokers restrict orders outside the regular session, and software has to work within those rules. At the time of writing, Interactive Brokers accepts only limit and stop-limit orders outside regular hours, and its trailing orders work only in regular hours; Alpaca accepts extended-hours orders only as limit orders with day time-in-force. So a premarket trailing exit can't rest at the broker. Momentum Lab handles it with the PatternX Synthetic Trail: the engine evaluates the high- or low-water mark every cycle, stores it, and sends an exit when the trail is hit. A broker-native trailing stop is used only when a strict set of conditions all hold, among them live mode, an entry filled in the regular session and no hard stop; with the default 3% hard stop, positions always use the synthetic trail. In live mode the hard stop also rests at the broker as a good-till-cancelled stop-limit order, with its limit 1% beyond the stop, and in extended hours only where the broker supports stops then: IBKR does, Alpaca doesn't. The guide's protection section has the full rules.

Turn a trail into a maximum loss

For example, suppose a long position fills at $10.00 in the premarket with the default 0.5% trail ($9.95) and 3% hard stop ($9.70); trading live through IBKR, a stop-limit also rests at the broker with its limit near $9.60. Unexpected news drops the next fresh price to $9.40. The trail and the stop have both been passed, the stop-limit's limit is above the market so it doesn't fill, and the software's exit goes into a falling book. If it fills at $9.38, the loss is 6.2%, not 0.5% and not 3%. Gaps, spreads, slippage, partial fills and rejections can all do this. Trailing stops for momentum explores the problem further.

Make paper fills realistic

Paper trading in Momentum Lab runs the same engine as live, from a $25,000 virtual balance. Paper orders fill at the ask or bid within your limit, partial fills are capped at the displayed size, fees are zero and queue position isn't modelled. Real orders queue behind others at the same price, pay fees and can move a thin book themselves. Expect live results to be worse than paper; see paper trading.

Make backtests predictive

Momentum Lab's backtests replay your data source's 1-minute extended-hours bars for up to 25 symbols and 30 trading days, with a grid of up to 150 parameter combinations. They assume a 0.2% spread, 0.05% slippage, $1,000 per trade and no commission; they don't replay quote size, halts, the AI filter or RVOL; and they apply today's market cap to past days. Where history is missing, the backtest fails rather than inventing bars. Testing many combinations on a short history also makes it easy to pick settings that merely fit the past. See backtesting.

Exempt you from day-trading rules

Momentum Lab never holds overnight, so every round trip is a day trade. In a US margin account, the pattern day trader rule can limit how many you make; FINRA's day-trading page explains the requirements.

How to roll out safely

1. Start in paper. The engine is the same as live; only the fills differ.
2. Begin with the defaults. Change one parameter at a time and compare results by parameter version, which Momentum Lab's analytics report separately.
3. Check your data. Confirm the feed is real-time and covers the hours you plan to trade.
4. Read the skip reasons. When a stock you expected is skipped, find out why before loosening a filter.
5. Go live deliberately. Live trading needs an enabled IBKR or Alpaca live connection and the typed phrase "ENABLE LIVE TRADING".
6. Start small and know the off switch. Keep the default size until live fills look the way you expected, and know where the kill switch is before the session starts.

Momentum Lab is part of the Elite plan, and the limitations page collects what PatternX can't do. It is software, not a broker or adviser; read the risk disclaimer before trading with real money.