Model

Quant researcher: 21 ways to use Ora

Signal ideas as exact rules, compared with forecasts, across a universe and over time. Each example is a complete, runnable request with the real answers it returned (42 requests, captured 2026-10-31), what they mean and, where it fits, the action it would send.

Run withAnswers were captured on ora-fin-0.1.1. Another model changes the request JSON and the Playground link; run it there to see its answers.

Rule functions and facts used by this player (83)

sma 7 offset 5 rsi 4 ref 3 roc 2 ema 2 zscore 2 highest 2 abs 2 log 2 sum 2 bblower 2 bbupper 2 crossup 2 linreg 1 rec 1 adx 1 barssince 1 diminus 1 diplus 1 avg 1 median 1 obv 1 percentrank 1 doji 1 engulfing_bear 1 engulfing_bull 1 evening_star 1 hammer 1 harami_bear 1 harami_bull 1 marubozu 1 morning_star 1 shooting_star 1 dark_cloud 1 dragonfly_doji 1 gravestone_doji 1 hanging_man 1 inverted_hammer 1 piercing_line 1 spinning_top 1 three_black_crows 1 three_white_soldiers 1 tweezer_bottom 1 greatest 1 hl2 1 hlc3 1 least 1 ohlc4 1 tweezer_top 1 between 1 iif 1 atr_pct 1 sqrt 1 std 1 beat_streak 1 eps_surprise_pct 1 last_earnings_gap 1 pre_earnings_close 1 revenue_surprise_pct 1 div_change_pct 1 div_count_ttm 1 div_yield_ttm 1 years_of_growth 1 bbmid 1 cum 1 sign 1 swinghigh 1 swinglow 1 cci 1 stochd 1 stochk 1 willr 1 rma 1 wma 1 highestbars 1 days_since_earnings 1 days_since_split 1 days_to_earnings 1 days_to_split 1 last_earnings_move 1 reverse_split_last 1 lowestbars 1

Examples 1–10 of 21 · page 1 of 3

Quant researcher · example 1

Momentum signal vs the forecast model on AAPL

Put a classic 12-1 style momentum rule next to the model's one-month forecast to see where the exact signal and the learned odds agree.

ora-fin-0.1.1AAPL5 questions

StateAAPL
Questions5 asked
Answersreal, captured
Actionsnone

State

{"symbol":"AAPL"} The server fills in the latest completed session.

Questions

`mom_12_1`	rule	`C[-21] > C[-252]` Price a month ago above price a year ago (12-1 momentum)
`mom_3m`	rule	`roc(C, 63) > 0` Positive three-month rate of change
`above_200d`	English → rule	Is the stock above its 200-day moving average?
`beat_spy`	forecast	Will it beat SPY over the next month? Beating SPY over a month
`up_week`	forecast	Will the stock be higher in a week? Higher in a week

Answers

Question	Answer	Reading
`mom_12_1` Price a month ago above price a year ago (12-1 momentum)	0.9463	true
`mom_3m` Positive three-month rate of change	0.9423	true
`above_200d` Is the stock above its 200-day moving average compiled to `C > sma(C, 200)`	0.9463	true
`beat_spy` Beating SPY over a month	0.531	53% · close to even
`up_week` Higher in a week	0.59	59% · close to even

What it means

For AAPL on the session of 2026-10-30: True now: Price a month ago above price a year ago (12-1 momentum); Positive three-month rate of change; Is the stock above its 200-day moving average. Forecasts: Beating SPY over a month 53%; Higher in a week 59%.

Rules carry a rule field and answer near certainty; forecasts carry none and read as odds. Agreement between the two is the interesting cell in a study.

Request JSON

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "AAPL"
  },
  "questions": {
    "mom_12_1": {
      "type": "yes_no",
      "ask": "C[-21] > C[-252]"
    },
    "mom_3m": {
      "type": "yes_no",
      "ask": "roc(C, 63) > 0"
    },
    "above_200d": {
      "type": "yes_no",
      "instructions": "Is the stock above its 200-day moving average?"
    },
    "beat_spy": {
      "type": "yes_no",
      "instructions": "Will it beat SPY over the next month?"
    },
    "up_week": {
      "type": "yes_no",
      "instructions": "Will the stock be higher in a week?"
    }
  }
}

Response JSON (AAPL, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "mom_12_1": {
      "yes_no": 0.9463,
      "rule": "C[-21] > C[-252]",
      "basis": "rule"
    },
    "mom_3m": {
      "yes_no": 0.9423,
      "rule": "roc(C, 63) > 0",
      "basis": "rule"
    },
    "above_200d": {
      "yes_no": 0.9463,
      "rule": "C > sma(C, 200)",
      "basis": "comparison"
    },
    "beat_spy": {
      "yes_no": 0.531
    },
    "up_week": {
      "yes_no": 0.59
    }
  },
  "context": {
    "symbol": "AAPL",
    "as_of": "2026-10-30"
  }
}

offset roc sma

Try it in the Playground

Quant researcher · example 2

A recursive filter signal with rec() on MSFT

Test an adaptive smoother built with rec(): an exponential filter with a 0.1 gain, compared with an EMA and a regression line.

ora-fin-0.1.1MSFT5 questions

StateMSFT
Questions5 asked
Answersreal, captured
Actionsnone

State

{"symbol":"MSFT"} The server fills in the latest completed session.

Questions

`rec_filter`	rule	`C > rec(C, self + 0.1 * (C - self))` Close above a recursive smoother (gain 0.1)
`ema_20`	rule	`C > ema(C, 20)` Close above the 20-day EMA
`linreg`	rule	`C > linreg(C, 20)` Close above the 20-day regression value
`rec_slope`	rule	`rec(C, self + 0.1 * (C - self)) > ref(rec(C, self + 0.1 * (C - self)), -5)` The recursive smoother is higher than 5 sessions ago
`up_week`	forecast	Will the stock be higher in a week? Higher in a week

Answers

Question	Answer	Reading
`rec_filter` Close above a recursive smoother (gain 0.1)	0.3041	false
`ema_20` Close above the 20-day EMA	0.3192	false
`linreg` Close above the 20-day regression value	0.2494	false
`rec_slope` The recursive smoother is higher than 5 sessions ago	0.608	true
`up_week` Higher in a week	0.574	57% · close to even

What it means

For MSFT on the session of 2026-10-30: True now: The recursive smoother is higher than 5 sessions ago. Not true: Close above a recursive smoother (gain 0.1); Close above the 20-day EMA; Close above the 20-day regression value. Forecasts: Higher in a week 57%.

Three smoothers usually agree in a trend and disagree at turns; disagreement is where a new filter earns or loses its keep.

Request JSON

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "MSFT"
  },
  "questions": {
    "rec_filter": {
      "type": "yes_no",
      "ask": "C > rec(C, self + 0.1 * (C - self))"
    },
    "ema_20": {
      "type": "yes_no",
      "ask": "C > ema(C, 20)"
    },
    "linreg": {
      "type": "yes_no",
      "ask": "C > linreg(C, 20)"
    },
    "rec_slope": {
      "type": "yes_no",
      "ask": "rec(C, self + 0.1 * (C - self)) > ref(rec(C, self + 0.1 * (C - self)), -5)"
    },
    "up_week": {
      "type": "yes_no",
      "instructions": "Will the stock be higher in a week?"
    }
  }
}

Response JSON (MSFT, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "rec_filter": {
      "yes_no": 0.3041,
      "rule": "C > rec(C, self + 0.1 * (C - self))",
      "basis": "rule"
    },
    "ema_20": {
      "yes_no": 0.3192,
      "rule": "C > ema(C, 20)",
      "basis": "rule"
    },
    "linreg": {
      "yes_no": 0.2494,
      "rule": "C > linreg(C, 20)",
      "basis": "rule"
    },
    "rec_slope": {
      "yes_no": 0.608,
      "rule": "rec(C, self + 0.1 * (C - self)) > ref(rec(C, self + 0.1 * (C - self)), -5)",
      "basis": "rule"
    },
    "up_week": {
      "yes_no": 0.574
    }
  },
  "context": {
    "symbol": "MSFT",
    "as_of": "2026-10-30"
  }
}

ema linreg rec ref

Try it in the Playground

Quant researcher · example 3

Short-term reversal across a mega-cap universe

Score a one-week reversal signal on six large caps in one batch and line it up with the next-week forecast.

ora-fin-0.1.16 symbols, one request each4 questions

State6 symbols
Questions4 asked
Answersreal, captured
Actionsnone

State

{"symbol":"AAPL"} then MSFT, NVDA, AMZN, TSLA, JPM, one request each. The server fills in the latest completed session.

Questions

`loser_5d`	rule	`C < C[-5]` Down over the last five sessions
`z5`	rule	`zscore(C / C[-5] - 1, 60) < -1` Five-day return more than 1 sd below its 60-day norm
`rsi2`	rule	`rsi(C, 2) < 10` Two-day RSI under 10
`up_week`	forecast	Will the stock be higher in a week? Higher in a week

Answers

Symbol	`loser_5d`rule	`z5`rule	`rsi2`rule	`up_week`forecast
`AAPL`	0.7392	0.3949	0.1464	0.59
`MSFT`	0.6253	0.0959	0.5812	0.574
`NVDA`	0.7931	0.3216	0.591	0.527
`AMZN`	0.8836	0.3761	0.5093	0.493
`TSLA`	0.0735	0.0537	0.1345	0.487
`JPM`	0.1512	0.0591	0.1602	0.558

true false not known for rules; forecasts are odds. Hover a column for its question.

What it means

Session of 2026-10-30, 6 symbols: Down over the last five sessions: true for AAPL, MSFT, NVDA, AMZN of 6. Five-day return more than 1 sd below its 60-day norm: true for none of 6. Two-day RSI under 10: true for none of 6. Highest odds for “Higher in a week”: AAPL at 59%.

In a reversal study the losers by rule should show higher next-week odds; if they do not, the effect is not present in this cross-section today.

Request JSON (AAPL; the others differ only in the symbol)

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "AAPL"
  },
  "questions": {
    "loser_5d": {
      "type": "yes_no",
      "ask": "C < C[-5]"
    },
    "z5": {
      "type": "yes_no",
      "ask": "zscore(C / C[-5] - 1, 60) < -1"
    },
    "rsi2": {
      "type": "yes_no",
      "ask": "rsi(C, 2) < 10"
    },
    "up_week": {
      "type": "yes_no",
      "instructions": "Will the stock be higher in a week?"
    }
  }
}

Response JSON (AAPL, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "loser_5d": {
      "yes_no": 0.7392,
      "rule": "C < C[-5]",
      "basis": "rule"
    },
    "z5": {
      "yes_no": 0.3949,
      "rule": "zscore(C / C[-5] - 1, 60) < -1",
      "basis": "rule"
    },
    "rsi2": {
      "yes_no": 0.1464,
      "rule": "rsi(C, 2) < 10",
      "basis": "rule"
    },
    "up_week": {
      "yes_no": 0.59
    }
  },
  "context": {
    "symbol": "AAPL",
    "as_of": "2026-10-30"
  }
}

offset rsi zscore

Try it in the Playground

Quant researcher · example 4

ADX and DMI trend-strength features on XLE

Build a trend-strength feature set: ADX level, DI spread, and how long since the last 20-day breakout.

ora-fin-0.1.1XLE5 questions

StateXLE
Questions5 asked
Answersreal, captured
Actionsnone

State

{"symbol":"XLE"} The server fills in the latest completed session.

Questions

`adx25`	rule	`adx(14) > 25` ADX over 25
`di_spread`	rule	`diplus(14) - diminus(14) > 10` +DI more than 10 points above -DI
`di_minus`	rule	`diminus(14) > diplus(14)` -DI above +DI
`since_bo`	rule	`barssince(C > highest(C, 20)) < 10` A 20-day breakout in the last 10 sessions
`new_high`	forecast	Will it make a new 20-day high in the next week? A new 20-day high next week

Answers

Question	Answer	Reading
`adx25` ADX over 25	0.0708	false
`di_spread` +DI more than 10 points above -DI	0.3957	false
`di_minus` -DI above +DI	0.1748	false
`since_bo` A 20-day breakout in the last 10 sessions	0.9316	true
`new_high` A new 20-day high next week	0.865	87% · likely

What it means

For XLE on the session of 2026-10-30: True now: A 20-day breakout in the last 10 sessions. Not true: ADX over 25; +DI more than 10 points above -DI; -DI above +DI. Forecasts: A new 20-day high next week 87%.

Strong ADX with a wide positive DI spread and a recent breakout describes a trending regime; the new-high forecast tests persistence.

Request JSON

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "XLE"
  },
  "questions": {
    "adx25": {
      "type": "yes_no",
      "ask": "adx(14) > 25"
    },
    "di_spread": {
      "type": "yes_no",
      "ask": "diplus(14) - diminus(14) > 10"
    },
    "di_minus": {
      "type": "yes_no",
      "ask": "diminus(14) > diplus(14)"
    },
    "since_bo": {
      "type": "yes_no",
      "ask": "barssince(C > highest(C, 20)) < 10"
    },
    "new_high": {
      "type": "yes_no",
      "instructions": "Will it make a new 20-day high in the next week?"
    }
  }
}

Response JSON (XLE, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "adx25": {
      "yes_no": 0.0708,
      "rule": "adx(14) > 25",
      "basis": "rule"
    },
    "di_spread": {
      "yes_no": 0.3957,
      "rule": "diplus(14) - diminus(14) > 10",
      "basis": "rule"
    },
    "di_minus": {
      "yes_no": 0.1748,
      "rule": "diminus(14) > diplus(14)",
      "basis": "rule"
    },
    "since_bo": {
      "yes_no": 0.9316,
      "rule": "barssince(C > highest(C, 20)) < 10",
      "basis": "rule"
    },
    "new_high": {
      "yes_no": 0.865
    }
  },
  "context": {
    "symbol": "XLE",
    "as_of": "2026-10-30"
  }
}

adx barssince diminus diplus highest

Try it in the Playground

Quant researcher · example 5

Robust volume anomaly with median() and percentrank()

Compare a mean-based volume spike with median and percentile versions that resist outliers.

ora-fin-0.1.1AMD5 questions

StateAMD
Questions5 asked
Answersreal, captured
Actionsnone

State

{"symbol":"AMD"} The server fills in the latest completed session.

Questions

`mean_spike`	rule	`V > 2 * ref(avg(V, 20), -1)` Volume more than twice the previous 20-day average
`median_spike`	rule	`V > 2 * median(V, 20)` Volume more than twice its 20-day median
`rank90`	rule	`percentrank(V, 50) > 90` Volume in the top tenth of 50 sessions
`obv`	rule	`obv() > sma(obv(), 20)` On-balance volume above its 20-day average
`vol_spike_tmr`	forecast	Will volume spike tomorrow? A volume spike tomorrow

Answers

Question	Answer	Reading
`mean_spike` Volume more than twice the previous 20-day average	0.5394	on the line (true)
`median_spike` Volume more than twice its 20-day median	0.6441	true
`rank90` Volume in the top tenth of 50 sessions	0.6546	true
`obv` On-balance volume above its 20-day average	0.2203	false
`vol_spike_tmr` A volume spike tomorrow	0.101	10% · unlikely

What it means

For AMD on the session of 2026-10-30: True now: Volume more than twice its 20-day median; Volume in the top tenth of 50 sessions. Not true: On-balance volume above its 20-day average. On the line: Volume more than twice the previous 20-day average. Forecasts: A volume spike tomorrow 10%.

When mean and median versions disagree, one outlier day is driving the average; the median version is the cleaner feature.

Request JSON

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "AMD"
  },
  "questions": {
    "mean_spike": {
      "type": "yes_no",
      "ask": "V > 2 * ref(avg(V, 20), -1)"
    },
    "median_spike": {
      "type": "yes_no",
      "ask": "V > 2 * median(V, 20)"
    },
    "rank90": {
      "type": "yes_no",
      "ask": "percentrank(V, 50) > 90"
    },
    "obv": {
      "type": "yes_no",
      "ask": "obv() > sma(obv(), 20)"
    },
    "vol_spike_tmr": {
      "type": "yes_no",
      "instructions": "Will volume spike tomorrow?"
    }
  }
}

Response JSON (AMD, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "mean_spike": {
      "yes_no": 0.5394,
      "rule": "V > 2 * ref(avg(V, 20), -1)",
      "basis": "rule"
    },
    "median_spike": {
      "yes_no": 0.6441,
      "rule": "V > 2 * median(V, 20)",
      "basis": "rule"
    },
    "rank90": {
      "yes_no": 0.6546,
      "rule": "percentrank(V, 50) > 90",
      "basis": "rule"
    },
    "obv": {
      "yes_no": 0.2203,
      "rule": "obv() > sma(obv(), 20)",
      "basis": "rule"
    },
    "vol_spike_tmr": {
      "yes_no": 0.101
    }
  },
  "context": {
    "symbol": "AMD",
    "as_of": "2026-10-30"
  }
}

avg median obv percentrank ref sma

Try it in the Playground

Quant researcher · example 6

A full candlestick feature panel on SPY

Record every candlestick pattern as a binary feature for one day, the raw input to a pattern study.

ora-fin-0.1.1SPY10 questions

StateSPY
Questions10 asked
Answersreal, captured
Actionsnone

State

{"symbol":"SPY"} The server fills in the latest completed session.

Questions

`doji`	rule	`doji()` doji candle: Body at most 10% of the day's range.
`hammer`	rule	`hammer()` hammer candle: Lower wick at least twice the body, upper wick at most 10% of the range, closing below the close five bars back.
`shooting`	rule	`shooting_star()` shooting star candle: Upper wick at least twice the body, lower wick at most 10% of the range, after a five-bar rise.
`engulf_bull`	rule	`engulfing_bull()` engulfing bull candle: A green body that engulfs yesterday's red body.
`engulf_bear`	rule	`engulfing_bear()` engulfing bear candle: A red body that engulfs yesterday's green body.
`harami_bull`	rule	`harami_bull()` harami bull candle: A small green body inside yesterday's red body.
`harami_bear`	rule	`harami_bear()` harami bear candle: A small red body inside yesterday's green body.
`morning`	rule	`morning_star()` morning star candle: A red bar, a small bar (half its body or less) that stays under its close, then a green bar closing above the middle of the first.
`evening`	rule	`evening_star()` evening star candle: A green bar, a small bar that stays above its close, then a red bar closing below the middle of the first.
`marubozu`	rule	`marubozu()` marubozu candle: Body at least 90% of the range: almost no wicks.

Answers

Question	Answer	Reading
`doji` doji candle: Body at most 10% of the day's range.	0.0579	false
`hammer` hammer candle: Lower wick at least twice the body, upper wick at most 10% of the range, closing below the close five bars back.	0.0941	false
`shooting` shooting star candle: Upper wick at least twice the body, lower wick at most 10% of the range, after a five-bar rise.	0.0537	false
`engulf_bull` engulfing bull candle: A green body that engulfs yesterday's red body.	0.0597	false
`engulf_bear` engulfing bear candle: A red body that engulfs yesterday's green body.	0.0604	false
`harami_bull` harami bull candle: A small green body inside yesterday's red body.	0.0604	false
`harami_bear` harami bear candle: A small red body inside yesterday's green body.	0.0597	false
`morning` morning star candle: A red bar, a small bar (half its body or less) that stays under its close, then a green bar closing above the middle of the first.	0.0579	false
`evening` evening star candle: A green bar, a small bar that stays above its close, then a red bar closing below the middle of the first.	0.1524	false
`marubozu` marubozu candle: Body at least 90% of the range: almost no wicks.	0.0537	false

What it means

For SPY on the session of 2026-10-30: Not true: doji candle: Body at most 10% of the day's range.; hammer candle: Lower wick at least twice the body, upper wick at most 10% of the range, closing below the close five bars back.; shooting star candle: Upper wick at least twice the body, lower wick at most 10% of the range, after a five-bar rise.; engulfing bull candle: A green body that engulfs yesterday's red body.; engulfing bear candle: A red body that engulfs yesterday's green body.; harami bull candle: A small green body inside yesterday's red body.; harami bear candle: A small red body inside yesterday's green body.; morning star candle: A red bar, a small bar (half its body or less) that stays under its close, then a green bar closing above the middle of the first.; evening star candle: A green bar, a small bar that stays above its close, then a red bar closing below the middle of the first.; marubozu candle: Body at least 90% of the range: almost no wicks..

Most features are zero on most days; that sparsity is why pattern studies need long histories and many symbols.

Request JSON

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "SPY"
  },
  "questions": {
    "doji": {
      "type": "yes_no",
      "ask": "doji()"
    },
    "hammer": {
      "type": "yes_no",
      "ask": "hammer()"
    },
    "shooting": {
      "type": "yes_no",
      "ask": "shooting_star()"
    },
    "engulf_bull": {
      "type": "yes_no",
      "ask": "engulfing_bull()"
    },
    "engulf_bear": {
      "type": "yes_no",
      "ask": "engulfing_bear()"
    },
    "harami_bull": {
      "type": "yes_no",
      "ask": "harami_bull()"
    },
    "harami_bear": {
      "type": "yes_no",
      "ask": "harami_bear()"
    },
    "morning": {
      "type": "yes_no",
      "ask": "morning_star()"
    },
    "evening": {
      "type": "yes_no",
      "ask": "evening_star()"
    },
    "marubozu": {
      "type": "yes_no",
      "ask": "marubozu()"
    }
  }
}

Response JSON (SPY, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "doji": {
      "yes_no": 0.0579,
      "rule": "doji()",
      "basis": "rule"
    },
    "hammer": {
      "yes_no": 0.0941,
      "rule": "hammer()",
      "basis": "rule"
    },
    "shooting": {
      "yes_no": 0.0537,
      "rule": "shooting_star()",
      "basis": "rule"
    },
    "engulf_bull": {
      "yes_no": 0.0597,
      "rule": "engulfing_bull()",
      "basis": "rule"
    },
    "engulf_bear": {
      "yes_no": 0.0604,
      "rule": "engulfing_bear()",
      "basis": "rule"
    },
    "harami_bull": {
      "yes_no": 0.0604,
      "rule": "harami_bull()",
      "basis": "rule"
    },
    "harami_bear": {
      "yes_no": 0.0597,
      "rule": "harami_bear()",
      "basis": "rule"
    },
    "morning": {
      "yes_no": 0.0579,
      "rule": "morning_star()",
      "basis": "rule"
    },
    "evening": {
      "yes_no": 0.1524,
      "rule": "evening_star()",
      "basis": "rule"
    },
    "marubozu": {
      "yes_no": 0.0537,
      "rule": "marubozu()",
      "basis": "rule"
    }
  },
  "context": {
    "symbol": "SPY",
    "as_of": "2026-10-30"
  }
}

doji engulfing_bear engulfing_bull evening_star hammer harami_bear harami_bull marubozu morning_star shooting_star

Try it in the Playground

Quant researcher · example 7

Candlestick feature panel, part two, on QQQ

The remaining patterns as binary features, so a study can cover the full pattern set.

ora-fin-0.1.1QQQ10 questions

StateQQQ
Questions10 asked
Answersreal, captured
Actionsnone

State

{"symbol":"QQQ"} The server fills in the latest completed session.

Questions

`dragonfly`	rule	`dragonfly_doji()` dragonfly doji candle: Doji with a long lower wick (60%+ of the range) and almost no upper wick.
`gravestone`	rule	`gravestone_doji()` gravestone doji candle: Doji with a long upper wick and almost no lower wick.
`spinning`	rule	`spinning_top()` spinning top candle: Small body (10–30% of range) with wicks of at least a quarter of the range on both sides.
`hanging`	rule	`hanging_man()` hanging man candle: Hammer shape, closing above the close five bars back.
`inv_hammer`	rule	`inverted_hammer()` inverted hammer candle: Shooting-star shape after a five-bar decline.
`piercing`	rule	`piercing_line()` piercing line candle: After a red day, opens below its close and closes above the middle of its body (but under its open).
`dark`	rule	`dark_cloud()` dark cloud candle: After a green day, opens above its close and closes below the middle of its body (but above its open).
`soldiers`	rule	`three_white_soldiers()` three white soldiers candle: Three green bars, each closing higher and opening inside the previous body.
`crows`	rule	`three_black_crows()` three black crows candle: Three red bars, each closing lower and opening inside the previous body.
`tweezer_b`	rule	`tweezer_bottom()` tweezer bottom candle: Red then green, with lows within 0.1% of each other.

Answers

Question	Answer	Reading
`dragonfly` dragonfly doji candle: Doji with a long lower wick (60%+ of the range) and almost no upper wick.	0.0749	false
`gravestone` gravestone doji candle: Doji with a long upper wick and almost no lower wick.	0.0537	false
`spinning` spinning top candle: Small body (10–30% of range) with wicks of at least a quarter of the range on both sides.	0.0865	false
`hanging` hanging man candle: Hammer shape, closing above the close five bars back.	0.1466	false
`inv_hammer` inverted hammer candle: Shooting-star shape after a five-bar decline.	0.0537	false
`piercing` piercing line candle: After a red day, opens below its close and closes above the middle of its body (but under its open).	0.0537	false
`dark` dark cloud candle: After a green day, opens above its close and closes below the middle of its body (but above its open).	0.0537	false
`soldiers` three white soldiers candle: Three green bars, each closing higher and opening inside the previous body.	0.0537	false
`crows` three black crows candle: Three red bars, each closing lower and opening inside the previous body.	0.0537	false
`tweezer_b` tweezer bottom candle: Red then green, with lows within 0.1% of each other.	0.0537	false

What it means

For QQQ on the session of 2026-10-30: Not true: dragonfly doji candle: Doji with a long lower wick (60%+ of the range) and almost no upper wick.; gravestone doji candle: Doji with a long upper wick and almost no lower wick.; spinning top candle: Small body (10–30% of range) with wicks of at least a quarter of the range on both sides.; hanging man candle: Hammer shape, closing above the close five bars back.; inverted hammer candle: Shooting-star shape after a five-bar decline.; piercing line candle: After a red day, opens below its close and closes above the middle of its body (but under its open).; dark cloud candle: After a green day, opens above its close and closes below the middle of its body (but above its open).; three white soldiers candle: Three green bars, each closing higher and opening inside the previous body.; three black crows candle: Three red bars, each closing lower and opening inside the previous body.; tweezer bottom candle: Red then green, with lows within 0.1% of each other..

Paired with part one, every pattern function is recorded; label each row later with the realized next-week return.

Request JSON

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "QQQ"
  },
  "questions": {
    "dragonfly": {
      "type": "yes_no",
      "ask": "dragonfly_doji()"
    },
    "gravestone": {
      "type": "yes_no",
      "ask": "gravestone_doji()"
    },
    "spinning": {
      "type": "yes_no",
      "ask": "spinning_top()"
    },
    "hanging": {
      "type": "yes_no",
      "ask": "hanging_man()"
    },
    "inv_hammer": {
      "type": "yes_no",
      "ask": "inverted_hammer()"
    },
    "piercing": {
      "type": "yes_no",
      "ask": "piercing_line()"
    },
    "dark": {
      "type": "yes_no",
      "ask": "dark_cloud()"
    },
    "soldiers": {
      "type": "yes_no",
      "ask": "three_white_soldiers()"
    },
    "crows": {
      "type": "yes_no",
      "ask": "three_black_crows()"
    },
    "tweezer_b": {
      "type": "yes_no",
      "ask": "tweezer_bottom()"
    }
  }
}

Response JSON (QQQ, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "dragonfly": {
      "yes_no": 0.0749,
      "rule": "dragonfly_doji()",
      "basis": "rule"
    },
    "gravestone": {
      "yes_no": 0.0537,
      "rule": "gravestone_doji()",
      "basis": "rule"
    },
    "spinning": {
      "yes_no": 0.0865,
      "rule": "spinning_top()",
      "basis": "rule"
    },
    "hanging": {
      "yes_no": 0.1466,
      "rule": "hanging_man()",
      "basis": "rule"
    },
    "inv_hammer": {
      "yes_no": 0.0537,
      "rule": "inverted_hammer()",
      "basis": "rule"
    },
    "piercing": {
      "yes_no": 0.0537,
      "rule": "piercing_line()",
      "basis": "rule"
    },
    "dark": {
      "yes_no": 0.0537,
      "rule": "dark_cloud()",
      "basis": "rule"
    },
    "soldiers": {
      "yes_no": 0.0537,
      "rule": "three_white_soldiers()",
      "basis": "rule"
    },
    "crows": {
      "yes_no": 0.0537,
      "rule": "three_black_crows()",
      "basis": "rule"
    },
    "tweezer_b": {
      "yes_no": 0.0537,
      "rule": "tweezer_bottom()",
      "basis": "rule"
    }
  },
  "context": {
    "symbol": "QQQ",
    "as_of": "2026-10-30"
  }
}

dark_cloud dragonfly_doji gravestone_doji hanging_man inverted_hammer piercing_line spinning_top three_black_crows three_white_soldiers tweezer_bottom

Try it in the Playground

Quant researcher · example 8

Bar geometry features: wicks, bodies and price centers

Describe the latest bar with geometry features built from hl2, hlc3, ohlc4, least and greatest.

ora-fin-0.1.1NVDA7 questions

StateNVDA
Questions7 asked
Answersreal, captured
Actionsnone

State

{"symbol":"NVDA"} The server fills in the latest completed session.

Questions

`upper_half`	rule	`C > hl2()` Closed in the upper half of the bar
`typical_up`	rule	`hlc3() > sma(hlc3(), 20)` Typical price above its 20-day average
`ohlc_up`	rule	`ohlc4() > ref(ohlc4(), -1)` OHLC average above yesterday's
`long_lower`	rule	`least(O, C) - L > 2 * abs(C - O)` Lower wick more than twice the body
`long_upper`	rule	`H - greatest(O, C) > 2 * abs(C - O)` Upper wick more than twice the body
`tweezer_top`	rule	`tweezer_top()` tweezer top candle: Green then red, with highs within 0.1%.
`up_tomorrow`	forecast	Will the stock be up tomorrow? Up tomorrow

Answers

Question	Answer	Reading
`upper_half` Closed in the upper half of the bar	0.3742	false
`typical_up` Typical price above its 20-day average	0.0713	false
`ohlc_up` OHLC average above yesterday's	0.0685	false
`long_lower` Lower wick more than twice the body	0.1245	false
`long_upper` Upper wick more than twice the body	0.0714	false
`tweezer_top` tweezer top candle: Green then red, with highs within 0.1%.	0.0664	false
`up_tomorrow` Up tomorrow	0.485	49% · close to even

What it means

For NVDA on the session of 2026-10-30: Not true: Closed in the upper half of the bar; Typical price above its 20-day average; OHLC average above yesterday's; Lower wick more than twice the body; Upper wick more than twice the body; tweezer top candle: Green then red, with highs within 0.1%.. Forecasts: Up tomorrow 49%.

Geometry features are cheap, exact and plentiful; the up-tomorrow forecast is the benchmark any one-day feature must beat.

Request JSON

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "NVDA"
  },
  "questions": {
    "upper_half": {
      "type": "yes_no",
      "ask": "C > hl2()"
    },
    "typical_up": {
      "type": "yes_no",
      "ask": "hlc3() > sma(hlc3(), 20)"
    },
    "ohlc_up": {
      "type": "yes_no",
      "ask": "ohlc4() > ref(ohlc4(), -1)"
    },
    "long_lower": {
      "type": "yes_no",
      "ask": "least(O, C) - L > 2 * abs(C - O)"
    },
    "long_upper": {
      "type": "yes_no",
      "ask": "H - greatest(O, C) > 2 * abs(C - O)"
    },
    "tweezer_top": {
      "type": "yes_no",
      "ask": "tweezer_top()"
    },
    "up_tomorrow": {
      "type": "yes_no",
      "instructions": "Will the stock be up tomorrow?"
    }
  }
}

Response JSON (NVDA, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "upper_half": {
      "yes_no": 0.3742,
      "rule": "C > hl2()",
      "basis": "rule"
    },
    "typical_up": {
      "yes_no": 0.0713,
      "rule": "hlc3() > sma(hlc3(), 20)",
      "basis": "rule"
    },
    "ohlc_up": {
      "yes_no": 0.0685,
      "rule": "ohlc4() > ref(ohlc4(), -1)",
      "basis": "rule"
    },
    "long_lower": {
      "yes_no": 0.1245,
      "rule": "least(O, C) - L > 2 * abs(C - O)",
      "basis": "rule"
    },
    "long_upper": {
      "yes_no": 0.0714,
      "rule": "H - greatest(O, C) > 2 * abs(C - O)",
      "basis": "rule"
    },
    "tweezer_top": {
      "yes_no": 0.0664,
      "rule": "tweezer_top()",
      "basis": "rule"
    },
    "up_tomorrow": {
      "yes_no": 0.485
    }
  },
  "context": {
    "symbol": "NVDA",
    "as_of": "2026-10-30"
  }
}

abs greatest hl2 hlc3 least ohlc4 ref sma tweezer_top

Try it in the Playground

Quant researcher · example 9

Regime-dependent thresholds with iif() on IWM

Test an oversold rule whose threshold changes with the trend regime, and a band rule written with between().

ora-fin-0.1.1IWM5 questions

StateIWM
Questions5 asked
Answersreal, captured
Actionsnone

State

{"symbol":"IWM"} The server fills in the latest completed session.

Questions

`regime_os`	rule	`rsi(C, 14) < iif(C > sma(C, 200), 40, 30)` RSI under 40 in an uptrend, under 30 in a downtrend
`neutral_band`	rule	`between(rsi(C, 14), 40, 60)` RSI between 40 and 60
`regime`	rule	`(C > sma(C, 200) and rsi(C, 14) < 40) or (C < sma(C, 200) and rsi(C, 14) < 25)` Oversold, with the line set by the trend
`trend`	English → rule	Is the stock above its 200-day moving average?
`rally5`	forecast	Will it rally at least 5% at some point in the next month? A 5% rally within a month

Answers

Question	Answer	Reading
`regime_os` RSI under 40 in an uptrend, under 30 in a downtrend	0.1346	false
`neutral_band` RSI between 40 and 60	0.95	true
`regime` Oversold, with the line set by the trend	0.1364	false
`trend` Is the stock above its 200-day moving average compiled to `C > sma(C, 200)`	0.5564	on the line (true)
`rally5` A 5% rally within a month	0.285	28% · unlikely

What it means

For IWM on the session of 2026-10-30: True now: RSI between 40 and 60. Not true: RSI under 40 in an uptrend, under 30 in a downtrend; Oversold, with the line set by the trend. On the line: Is the stock above its 200-day moving average. Forecasts: A 5% rally within a month 28%.

Regime-aware thresholds trade less in downtrends; compare the hit rate of each version against the rally forecast over many days.

Request JSON

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "IWM"
  },
  "questions": {
    "regime_os": {
      "type": "yes_no",
      "ask": "rsi(C, 14) < iif(C > sma(C, 200), 40, 30)"
    },
    "neutral_band": {
      "type": "yes_no",
      "ask": "between(rsi(C, 14), 40, 60)"
    },
    "regime": {
      "type": "yes_no",
      "ask": "(C > sma(C, 200) and rsi(C, 14) < 40) or (C < sma(C, 200) and rsi(C, 14) < 25)"
    },
    "trend": {
      "type": "yes_no",
      "instructions": "Is the stock above its 200-day moving average?"
    },
    "rally5": {
      "type": "yes_no",
      "instructions": "Will it rally at least 5% at some point in the next month?"
    }
  }
}

Response JSON (IWM, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "regime_os": {
      "yes_no": 0.1346,
      "rule": "rsi(C, 14) < iif(C > sma(C, 200), 40, 30)",
      "basis": "rule"
    },
    "neutral_band": {
      "yes_no": 0.95,
      "rule": "between(rsi(C, 14), 40, 60)",
      "basis": "rule"
    },
    "regime": {
      "yes_no": 0.1364,
      "rule": "(C > sma(C, 200) and rsi(C, 14) < 40) or (C < sma(C, 200) and rsi(C, 14) < 25)",
      "basis": "rule"
    },
    "trend": {
      "yes_no": 0.5564,
      "rule": "C > sma(C, 200)",
      "basis": "comparison"
    },
    "rally5": {
      "yes_no": 0.285
    }
  },
  "context": {
    "symbol": "IWM",
    "as_of": "2026-10-30"
  }
}

between iif rsi sma

Try it in the Playground

Quant researcher · example 10

Comparing volatility estimators on TSLA

Close-to-close, ATR-based and range-based volatility measures side by side, with the model's volatility forecasts.

ora-fin-0.1.1TSLA6 questions

StateTSLA
Questions6 asked
Answersreal, captured
Actionsnone

State

{"symbol":"TSLA"} The server fills in the latest completed session.

Questions

`hv_cc`	rule	`std(log(C / C[-1]), 20) * sqrt(252) > 0.5` 20-day close-to-close volatility above 50%
`atr_pct`	rule	`atr_pct(14) > 3` ATR over 3% of price
`range_vol`	rule	`sum(log(H / L), 20) / 20 > 0.03` Average log high/low range above 3%
`vol_ratio`	rule	`std(log(C / C[-1]), 5) > std(log(C / C[-1]), 60)` 5-day volatility above 60-day volatility
`vol_up`	forecast	Will volatility be higher over the next month? Higher volatility next month
`hv30`	forecast	Will annualized volatility over the next 21 sessions exceed 30%? Next-month volatility above 30%

Answers

Question	Answer	Reading
`hv_cc` 20-day close-to-close volatility above 50%	0.1042	false
`atr_pct` ATR over 3% of price	0.848	true
`range_vol` Average log high/low range above 3%	0.3554	false
`vol_ratio` 5-day volatility above 60-day volatility	0.5391	on the line (true)
`vol_up` Higher volatility next month	0.446	45% · close to even
`hv30` Next-month volatility above 30%	0.441	44% · close to even

What it means

For TSLA on the session of 2026-10-30: True now: ATR over 3% of price. Not true: 20-day close-to-close volatility above 50%; Average log high/low range above 3%. On the line: 5-day volatility above 60-day volatility. Forecasts: Higher volatility next month 45%; Next-month volatility above 30% 44%.

Estimators that agree describe the regime robustly; the forecasts test whether today's regime predicts next month's.

Request JSON

{
  "model": "ora-fin-0.1.1",
  "state": {
    "symbol": "TSLA"
  },
  "questions": {
    "hv_cc": {
      "type": "yes_no",
      "ask": "std(log(C / C[-1]), 20) * sqrt(252) > 0.5"
    },
    "atr_pct": {
      "type": "yes_no",
      "ask": "atr_pct(14) > 3"
    },
    "range_vol": {
      "type": "yes_no",
      "ask": "sum(log(H / L), 20) / 20 > 0.03"
    },
    "vol_ratio": {
      "type": "yes_no",
      "ask": "std(log(C / C[-1]), 5) > std(log(C / C[-1]), 60)"
    },
    "vol_up": {
      "type": "yes_no",
      "instructions": "Will volatility be higher over the next month?"
    },
    "hv30": {
      "type": "yes_no",
      "instructions": "Will annualized volatility over the next 21 sessions exceed 30%?"
    }
  }
}

Response JSON (TSLA, real)

{
  "model": "ora-fin-0.1.1",
  "answers": {
    "hv_cc": {
      "yes_no": 0.1042,
      "rule": "std(log(C / C[-1]), 20) * sqrt(252) > 0.5",
      "basis": "rule"
    },
    "atr_pct": {
      "yes_no": 0.848,
      "rule": "atr_pct(14) > 3",
      "basis": "rule"
    },
    "range_vol": {
      "yes_no": 0.3554,
      "rule": "sum(log(H / L), 20) / 20 > 0.03",
      "basis": "rule"
    },
    "vol_ratio": {
      "yes_no": 0.5391,
      "rule": "std(log(C / C[-1]), 5) > std(log(C / C[-1]), 60)",
      "basis": "rule"
    },
    "vol_up": {
      "yes_no": 0.446
    },
    "hv30": {
      "yes_no": 0.441
    }
  },
  "context": {
    "symbol": "TSLA",
    "as_of": "2026-10-30"
  }
}

atr_pct log offset sqrt std sum

Try it in the Playground

Information, not investment advice. Rules answer from the latest completed daily session; forecasts are probabilities. Actions ran as dry runs: nothing was sent. More in Buy side: Financial advisor / RIA, Wealth manager, Family office, Hedge fund PM, Portfolio manager (long-only), Risk manager, Private equity, VC / growth investor, Fund-of-funds allocator.

Create an account Open the Playground Everything here also lives in the console, next to your keys and usage.

Individuals

Buy side

Sell side

Corporates

Builders

Institutions

Momentum signal vs the forecast model on AAPL

State

Questions

Answers

What it means

A recursive filter signal with rec() on MSFT

State

Questions

Answers

What it means

Short-term reversal across a mega-cap universe

State

Questions

Answers

What it means

ADX and DMI trend-strength features on XLE

State

Questions

Answers

What it means

Robust volume anomaly with median() and percentrank()

State

Questions

Answers

What it means

A full candlestick feature panel on SPY

State

Questions

Answers

What it means

Candlestick feature panel, part two, on QQQ

State

Questions

Answers

What it means

Bar geometry features: wicks, bodies and price centers

State

Questions

Answers

What it means

Regime-dependent thresholds with iif() on IWM

State

Questions

Answers

What it means

Comparing volatility estimators on TSLA

State

Questions

Answers

What it means