What you can test

• The market: one of 46 U.S. futures markets, E-mini Nasdaq 100 (NQ), E-mini S&P 500 (ES), E-mini Russell 2000 (RTY), E-mini Dow (YM), Crude Oil (CL), Natural Gas (NG), Gold (GC), Silver (SI), Micro E-mini S&P 500 (MES), Micro E-mini Nasdaq 100 (MNQ), 10-Year T-Note (ZN), 30-Year T-Bond (ZB), Euro FX (6E), Japanese Yen (6J), Canadian Dollar (6C), Bitcoin (BTC), Micro Bitcoin (MBT), Copper (HG), Platinum (PL), Heating Oil (HO), Corn (ZC), Soybeans (ZS), Wheat (ZW), Soybean Meal (ZM), E-mini MidCap 400 (EMD), Micro Dow (MYM), Micro Russell 2000 (M2K), 5-Year T-Note (ZF), 2-Year T-Note (ZT), Ultra 10-Year Note (TN), Ultra T-Bond (UB), British Pound (6B), Australian Dollar (6A), Swiss Franc (6S), New Zealand Dollar (6N), Mexican Peso (6M), Micro Euro FX (M6E), RBOB Gasoline (RB), Brent Crude (BZ), Micro Gold (MGC), Micro Silver (SIL), Soybean Oil (ZL), KC HRW Wheat (KE), Live Cattle (LE), Lean Hogs (HE) and Feeder Cattle (GF).
• The bars: 1, 2, 3, 5, 10, 15, 20, 30, 60, 120 or 240 minutes, a day or a week, each anchored at the session’s open.
• Bars that are not time bars: point-and-figure, kagi and line break, drawn from 1-minute closes on every market’s sessions at any size; and range, Renko, volume and tick bars, built from the tick data, only at the markets, sessions and sizes the data hold: NQ’s 16-, 40-, 260- and 340-tick range, 16-, 40- and 160-tick Renko, 31,500-contract volume and 233-, 500-, 1,000- and 1,500-trade tick bars and ES’s 4- and 8-tick range, 4- and 8-tick Renko, 10,000-contract volume and 233-, 500-, 1,000- and 1,500-trade tick bars, in the regular session. Their signals come from the bars and their fills from the underlying 1-minute prices.
• The session: the regular session or, on NQ, ES, CL, BTC and MBT, where the 24-hour data exist, the whole Globex session.

The gates measure each market on its sessions from January 2018, the sample their pass lines were set on, or from where its data begin: RTY from May 2023, YM from September 2021, CL’s Globex session from November 2021, NG from February 2023, GC from March 2022, SI from August 2023, MES from May 2019, MNQ from December 2022, BTC from January 2022, BTC’s Globex session from January 2022, MBT from May 2021, MBT’s Globex session from May 2021, MYM from May 2019, M2K from May 2019 and SIL from February 2020. A rule that reads another market or a Globex level is measured from where those data begin, if they begin after January 2018. The most recent sessions are held back from every research path and never scored.

Any series can be read on a higher timeframe than the rule’s own bars, longer minute bars, a day or a week: the hourly trend under a 5-minute entry, or last week’s high. A rule sees only the last higher-timeframe bar that has closed, never one still forming.

Any series can also be read on another of the 46 markets, such as ES’s close as a filter on an NQ entry: any of them by a rule on the regular session, and NQ, ES, CL, BTC or MBT by a rule on the whole Globex session, where the 24-hour data exist. The other market’s bar is read where it ends when the rule’s bar ends; where it has none then, outside its hours or on its holidays, the value is missing and a condition that reads it is not met.

The NYSE’s market internals, $TICK, $ADD, $VOLD and $TRIN, can be read the same way on the regular session, as data beside the market a rule trades.

The indicators, 62 of them, each read with the settings you give it, on the rule’s own bars or a higher timeframe:

• Moving averages: SMA, EMA, WMA, Hull, VWMA, DEMA, TEMA, Wilder’s smoothed average, zero-lag EMA, least-squares and Kaufman’s adaptive.
• Oscillators: RSI, Stochastic, Stochastic RSI, Williams %R, CCI, rate of change, MACD, PPO, TRIX, Chande momentum, the Ultimate Oscillator and the Awesome Oscillator.
• Volatility and bands: true range, ATR, ATR as a percent, standard deviation, z-score, realised volatility, Bollinger Bands, Keltner Channels, Donchian Channels and the squeeze.
• Volume and VWAP: VWAP by session, week or month, with bands, anchored VWAP, OBV, MFI, Chaikin Money Flow, accumulation/distribution and volume against the same time of day.
• Trend studies: ADX, Aroon, Supertrend, Parabolic SAR, Ichimoku, Vortex, Choppiness, linear regression and Heikin Ashi.
• Community lines, in their non-repainting versions: UT Bot, the Chandelier line, HalfTrend, Range Filter, SSL, the Hi-Lo Activator, WaveTrend, QQE, Connors RSI, IBS, the VIX Fix, the Alligator and Nadaraya-Watson.

• Session levels (today’s open, high and low, the prior day’s, week’s or month’s, an opening range of any whole number of bars, a clock window’s range, the opening gap, and on NQ, ES, CL, BTC and MBT the overnight range).
• The bar itself: its open, high, low and close, its range, body and wicks, its volume and number of trades, its delta and its bid and ask volume; the highest or lowest value of the last N bars; the change over N bars, sums and running totals; and arithmetic on any of them.

Compare any series with a number, with price or with another series: above, below or between; crossing or touching; rising or falling; how many of the last bars met a condition; how many bars since one last did; the first time today. Combine them with all, any and not.

The clock and the calendar: a time of day or a window of hours, the weekday, the month, the trading day of the month, the days of FOMC, CPI, payrolls, EIA reports and expirations, and the minutes around a scheduled release.

A condition that reads a value that has not formed, such as an opening range before it completes or an indicator before its lookback fills, is not met: the rule waits.

• Long, short or both: each side is its own rule, with its own condition, order and exits, and the short side is never taken to be the long side’s mirror unless you say so.
• Every rule is decided at a bar’s close. A market order fills at the next bar’s open; a stop order when price trades at or through it; a limit order only when price trades through it, at its price; each with an expiry and a condition that cancels it; and on day and week bars an order can fill at the close.
• Stop and limit orders of several rules can wait together, the first to fill cancelling the others: a stop above and a stop below make a bracket.
• One position at a time, besides adding to one.
• Limits: a cap on trades a session, a pause after an exit, stop and reverse, and a flat time; no new entries for the rest of a session once its realised profit or loss, net of costs, reaches a daily loss or profit limit, or after a run of losses in a row; and the prior trade in an entry’s rule (whether the last trade closed was a win or a loss).
• Positions held overnight and over weekends, or not, as the rule says.

A position held overnight or over a weekend can open well beyond its stop. The screen fills that stop at the opening price, and a real order can fill worse, so a loss can be larger than the stop you set.

• A position can be ended by a stop or a target (in points, ticks, percent, dollars, ATR or R, or at a price level); a trailing stop or a breakeven; a time limit in bars, minutes or sessions; or a condition.
• Part of it can be taken off at nearer targets, and a unit added each time price moves a set step in its favour.
• Fills are pessimistic: when one bar reaches both the stop and the target, its 1-minute bars decide which came first, and a minute that reaches both counts as the stop.
• Costs: every contract entered pays the market’s round trip, commission and spread together, from the screening engine’s own cost table, the same one every screen pays. A screen trades one contract and reports per contract, in R and in dollars.

Every screen, from the menu or in the Strategy Language, faces the same seven gates, written down before it runs:

• Causality: Could the signal have read a bar it could not have seen live?
• Edge vs. noise: Is the average trade clearly above zero: over enough trades, not carried by a handful of sessions, and after counting every idea screened here?
• Drift null: Does it beat random entries with the same trade count, hold and long/short mix?
• Walk-forward: Does it hold on sessions the selection never saw?
• Mirror test: Does the opposite trade, taken on the same bars, do clearly worse?
• Cost: Does the edge clear real per-fill commission and slippage, with room to spare?
• Era split: Did it also pay in the years before the sample it was judged on? Reported on survivors; it never decides a verdict.

The first check that fails decides the verdict: candidate, inconclusive, or not this one. An idea with too few trades to judge is inconclusive, not a verdict either way. A rule written in the Strategy Language is also replayed on its data cut short at several points, and every decision, fill and value must come out the same, so a rule that reads ahead fails the causality gate.

The screen tests the rules as written and does not search: the desk reads them back, clause by clause, before anything runs. How the gates work.

A rule that can hold a position overnight is a swing screen. It is graded candidate, promising, doubtful or dead rather than judged gate by gate, and it is never inconclusive. The grade says how likely its edge over random entries on matched days, held the same way, is to be real and to clear a set hurdle. What each grade means.

Its own trades, the same rule on the other markets it runs on, and what the record of swing screens knew before it ran all count as evidence, and the report says how much of the answer each carried.

How likely the edge must be for each grade, and the hurdle it must clear, were set by a calibration study on rules with no edge and rules with a planted one.

On day or week bars, RTH or the whole Globex session, it reads the market’s daily tape, one bar a Globex session closing at the settlement: ES’s runs from September 1997 and NQ’s from June 1999, and each market’s page gives its own.

A swing screen tests one futures market at a time: not stocks, and not a book of markets ranked against each other.

A position held overnight or over a weekend can open well beyond its stop. The screen fills that stop at the opening price, and a real order can fill worse, so a loss can be larger than the stop you set.

The Strategy Language keeps these out, whatever is built:

• Data the site does not hold: options (gamma exposure, put/call ratios, max pain), COT positioning, news and sentiment, fundamentals, fund flows, auction and closing imbalances, and open interest.
• The order book, the tape and footprints: a rule reads each bar’s delta and its bid and ask volume, not the ladder.
• Machine-learning models, and searches over settings: the screen tests the rules you give it and does not search.
• Discretionary methods, such as Elliott counts, Gann angles, harmonic patterns fitted by eye, Wyckoff phases or reading the tape: none is a rule a bar’s close can decide. Such an idea can be screened once it is written as rules a bar’s close decides.
• Portfolios, spreads, rankings across markets, position sizing, grids and martingales: a screen trades one contract of one market and reports per contract.

And, in the rules themselves:

• Notes or free text inside the rules.
• Code, or a formula typed as text: arithmetic is written as parts of a rule.
• Anything that reads the future, such as news that was not scheduled.
• A date range of your choosing. A recurring window, such as a month or the days around a release, is a rule and can be tested.
• Costs below the market’s own.
• Bars shorter than a minute.
• A decision inside a bar: an “at the open” rule on intraday bars acts at the first bar’s close.
• Sessions other than the regular and the Globex session. A window of hours inside them is a condition.
• More than one position at a time, besides adding to one.
• Markets the site holds no data for, such as the Nikkei, SOFR, fed funds, palladium, micro crude and ether futures, cash bonds, stock baskets and foreign indexes, and trading the VIX futures.