Pair Plays
When two indicators move unusually in the same direction within a few days of each other, what does the market do next? We graded 1,422 pairs across 6 years of out-of-sample data.
Two indicators "fire" together when both make unusually large moves (above their σ thresholds) in the same direction within a short window of each other. We then measure what the S&P 500 does over the next horizon_days. A pair is Reliable only if both halves of history (2010–2019 test, 2020+ held-out) agree on the same side of 50/50. Below 8 fires per period: Too new to judge.
Sample-size floor: we publish nothing graded on fewer than 8 fires. We publish the misses too. Nobody asked. We ledger anyway.
Top pairs — top 50 by absolute deviation from 50/50
When btc_volume_z and stl_precip both have unusually large moves in the same direction within 3 days of each other, the market has followed in that same direction 100% of the time over the next 21d in our 2010-2019 records (14 examples), and 91% of the time in held-out 2020+ data (11 examples). Held up both periods.
When slc_precip and triple_witching both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 126d in our 2010-2019 records (15 examples), and 90% of the time in held-out 2020+ data (10 examples). Held up both periods.
When mastercard_stock and slc_precip both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 252d in our 2010-2019 records (16 examples), and 100% of the time in held-out 2020+ data (37 examples). Held up both periods.
When house_trades_buy_count and slc_precip both have unusually large moves in the same direction within 7 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (11 examples), and 88% of the time in held-out 2020+ data (17 examples). Held up both periods.
When btc and slc_snowfall both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (13 examples), and 100% of the time in held-out 2020+ data (28 examples). Held up both periods.
When btc and slc_precip both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (28 examples), and 95% of the time in held-out 2020+ data (43 examples). Held up both periods.
When btc and stl_precip both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 252d in our 2010-2019 records (50 examples), and 100% of the time in held-out 2020+ data (31 examples). Held up both periods.
When btc and triple_witching both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 126d in our 2010-2019 records (14 examples), and 100% of the time in held-out 2020+ data (10 examples). Held up both periods.
When btc and den_precip both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (14 examples), and 100% of the time in held-out 2020+ data (8 examples). Held up both periods.
When den_precip and den_temp both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 90d in our 2010-2019 records (12 examples), and 100% of the time in held-out 2020+ data (13 examples). Held up both periods.
When den_precip and mastercard_stock both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (16 examples), and 87% of the time in held-out 2020+ data (23 examples). Held up both periods.
When btc and new_moon_day both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 126d in our 2010-2019 records (9 examples), and 100% of the time in held-out 2020+ data (8 examples). Held up both periods.
When mastercard_stock and sd_precip both have unusually large moves in the same direction within 3 days of each other, the market has followed in that same direction 100% of the time over the next 90d in our 2010-2019 records (8 examples), and 92% of the time in held-out 2020+ data (12 examples). Held up both periods.
When den_precip and us_national_temp both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 5d in our 2010-2019 records (9 examples), and 89% of the time in held-out 2020+ data (9 examples). Held up both periods.
When btc and fomc_meeting_eve both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (13 examples), and 100% of the time in held-out 2020+ data (10 examples). Held up both periods.
When dal_precip and us_national_temp both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 45d in our 2010-2019 records (8 examples), and 100% of the time in held-out 2020+ data (9 examples). Held up both periods.
When btc and dal_precip both have unusually large moves in the same direction within 3 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (28 examples), and 100% of the time in held-out 2020+ data (15 examples). Held up both periods.
When dal_precip and den_temp both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 45d in our 2010-2019 records (8 examples), and 100% of the time in held-out 2020+ data (11 examples). Held up both periods.
When btc and hou_precip both have unusually large moves in the same direction within 3 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (20 examples), and 100% of the time in held-out 2020+ data (19 examples). Held up both periods.
When den_temp and hou_precip both have unusually large moves in the same direction within 7 days of each other, the market has followed in that same direction 100% of the time over the next 45d in our 2010-2019 records (9 examples), and 100% of the time in held-out 2020+ data (18 examples). Held up both periods.
When hou_precip and mastercard_stock both have unusually large moves in the same direction within 7 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (21 examples), and 100% of the time in held-out 2020+ data (8 examples). Held up both periods.
When btc and zodiac_season_change_day both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 21d in our 2010-2019 records (10 examples), and 91% of the time in held-out 2020+ data (11 examples). Held up both periods.
When btc and dc_precip both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 252d in our 2010-2019 records (10 examples), and 100% of the time in held-out 2020+ data (11 examples). Held up both periods.
When dc_precip and den_temp both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 126d in our 2010-2019 records (9 examples), and 100% of the time in held-out 2020+ data (14 examples). Held up both periods.
When dc_precip and gold_macd_hist both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 90d in our 2010-2019 records (8 examples), and 100% of the time in held-out 2020+ data (10 examples). Held up both periods.
When btc and pit_snowfall both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (36 examples), and 100% of the time in held-out 2020+ data (87 examples). Held up both periods.
When btc and fomc_meeting_today both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (13 examples), and 100% of the time in held-out 2020+ data (13 examples). Held up both periods.
When btc and full_moon_day both have unusually large moves in the same direction within 3 days of each other, the market has followed in that same direction 100% of the time over the next 252d in our 2010-2019 records (11 examples), and 89% of the time in held-out 2020+ data (9 examples). Held up both periods.
When btc and den_snowfall both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (10 examples), and 100% of the time in held-out 2020+ data (16 examples). Held up both periods.
When breakeven_inflation_10y and nyc_snowfall both have unusually large moves in the same direction within 7 days of each other, the market has followed in that same direction 100% of the time over the next 10d in our 2010-2019 records (15 examples), and 100% of the time in held-out 2020+ data (8 examples). Held up both periods.
When btc and sac_precip both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (17 examples), and 91% of the time in held-out 2020+ data (22 examples). Held up both periods.
When house_trades_buy_count and sac_precip both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (14 examples), and 90% of the time in held-out 2020+ data (10 examples). Held up both periods.
When btc_volume_z and sac_precip both have unusually large moves in the same direction within 3 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (13 examples), and 89% of the time in held-out 2020+ data (9 examples). Held up both periods.
When slc_snowfall and visa_stock both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (12 examples), and 100% of the time in held-out 2020+ data (17 examples). Held up both periods.
When btc_volume_z and visa_stock both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (9 examples), and 100% of the time in held-out 2020+ data (10 examples). Held up both periods.
When sac_precip and spy_sma_20 both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 10d in our 2010-2019 records (14 examples), and 100% of the time in held-out 2020+ data (9 examples). Held up both periods.
When nyc_snowfall and spy_sma_20 both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (16 examples), and 100% of the time in held-out 2020+ data (8 examples). Held up both periods.
When pit_snowfall and spy_sma_20 both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 252d in our 2010-2019 records (15 examples), and 100% of the time in held-out 2020+ data (9 examples). Held up both periods.
When phl_precip and spy_sma_20 both have unusually large moves in the same direction within 7 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (16 examples), and 100% of the time in held-out 2020+ data (10 examples). Held up both periods.
When btc and phl_precip both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (13 examples), and 100% of the time in held-out 2020+ data (13 examples). Held up both periods.
When btc_volume_z and phl_precip both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 252d in our 2010-2019 records (17 examples), and 90% of the time in held-out 2020+ data (10 examples). Held up both periods.
When bos_snowfall and spy_sma_20 both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (21 examples), and 100% of the time in held-out 2020+ data (9 examples). Held up both periods.
When bos_snowfall and btc both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (23 examples), and 100% of the time in held-out 2020+ data (46 examples). Held up both periods.
When bos_snowfall and stl_precip both have unusually large moves in the same direction within 7 days of each other, the market has followed in that same direction 100% of the time over the next 252d in our 2010-2019 records (28 examples), and 100% of the time in held-out 2020+ data (9 examples). Held up both periods.
When lv_precip and visa_stock both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (8 examples), and 100% of the time in held-out 2020+ data (10 examples). Held up both periods.
When btc and lv_precip both have unusually large moves in the same direction within 10 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (12 examples), and 93% of the time in held-out 2020+ data (44 examples). Held up both periods.
When lv_precip and mastercard_stock both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 180d in our 2010-2019 records (10 examples), and 100% of the time in held-out 2020+ data (11 examples). Held up both periods.
When phx_precip and visa_stock both have unusually large moves in the same direction within 1 days of each other, the market has followed in that same direction 100% of the time over the next 252d in our 2010-2019 records (9 examples), and 100% of the time in held-out 2020+ data (8 examples). Held up both periods.
When den_precip and phx_precip both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 252d in our 2010-2019 records (17 examples), and 90% of the time in held-out 2020+ data (10 examples). Held up both periods.
When btc and phx_precip both have unusually large moves in the same direction within 5 days of each other, the market has followed in that same direction 100% of the time over the next 63d in our 2010-2019 records (11 examples), and 100% of the time in held-out 2020+ data (15 examples). Held up both periods.
Data: 1422 pairs · regenerated 5/6/2026 · 2010–2019 in-sample · 2020+ out-of-sample