Rule Collider

Backtests · USDJPY · d1_price_vs_ema20_atr · long

Does d1_price_vs_ema20_atr below 9.191 work on USDJPY? 2023 to 2026 backtest

USDJPY long when Close vs D1 EMA20 (M5 ATR) is below the threshold, tested at five thresholds taken from the feature’s own distribution on USDJPY over this window: its 10th, 25th, 50th, 75th and 90th percentiles. Window 2023-09-30 to 2026-09-30, the last 30 percent held out of sample; stop 2 ATR, target 2.5 R, held overnight, max spread 3 pips, one-minute confirmation 0.5 ATR, no hour filter. Fills and spread as measured on a demo fleet (how a result is judged). The direction and the exit are the ones the research catalogue carried for this pair; the thresholds are not.

Five thresholds, one rule shape

PercentileConditionIn sample R / tradesOut of sample R / tradestTrades to confirmLast quarter R / tradesVerdict
p10d1_price_vs_ema20_atr < -37.057+10.1 R / 686+22.4 R / 2090.9934+9.8 R / 124Net positive, not distinguishable from chance
p25d1_price_vs_ema20_atr < -14.127-111.7 R / 1495+2.7 R / 3860.1221,003-7.2 R / 227Net positive, not distinguishable from chance
p50d1_price_vs_ema20_atr < 9.191-148.2 R / 2739+41.6 R / 8620.94,472-12.6 R / 328Net positive, not distinguishable from chance
p75d1_price_vs_ema20_atr < 32.589-169.5 R / 3926-22.4 R / 1408-0.439,211-17.5 R / 422No edge shown
p90d1_price_vs_ema20_atr < 56.25-306.4 R / 4410-67.8 R / 1544-1.15,143-38.4 R / 449No edge shown

R is the risk on one trade. The out-of-sample split carries the verdict: below 100 trades it is “no edge shown” whatever the t; a positive mean with t below 2 is “net positive, not yet distinguishable from chance”; the last column is how many trades that mean would need to reach t = 2.

Each threshold in full

d1_price_vs_ema20_atr < -37.057 (p10)

Honesty panelNet positive, not distinguishable from chance

Net positive out of sample, not yet distinguishable from chance: +22.4 R over 209 trades is +0.107 R per trade, t = 0.9; about 934 trades would be needed to confirm this mean at t = 2, shown: 209; last quarter +9.8 R over 124 trades.

In sample
2023-09-30 to 2025-11-05
+10.1 R
686 trades · 35% wins
+0.015 R per trade · t = 0.2, within chance
about 45,377 trades to confirm this mean (+0.015 R per trade) at t = 2 · shown: 686
spread 0.125 R per trade
Out of sample
2025-11-05 to 2026-09-30
+22.4 R
209 trades · 38% wins
+0.107 R per trade · t = 0.9, within chance
about 934 trades to confirm this mean (+0.107 R per trade) at t = 2 · shown: 209
spread 0.167 R per trade
Last quarter
last 91 days of the window
+9.8 R
124 trades · 37% wins
+0.079 R per trade · t = 0.5, within chance
about 1,672 trades to confirm this mean (+0.079 R per trade) at t = 2 · shown: 124
spread 0.161 R per trade
Costs. Spread 0.135 R per trade at 1.25 times the bar’s minimum (79% of the gross); 126 fills in measured hot hours (01:00 at 4.3 times). Fill offset 0.00 stop-distances on every entry, measured on 473 demo fills. At 1.5 times the bar’s minimum the out-of-sample figure would read +15.4 R (+0.074 R per trade over 209 trades).

Entry: stop order 2 pips past the signal close, cancelled after 4 bars; 1,130 signals placed, 895 filled, 235 expired; trigger state, an entry on every bar the conditions hold while flat.

Exit: stop 2 ATR (mean 14.8 pips, 2.22 ATR, 347 raised to the 10-pip floor); target 2.5 R; exits 62% at the stop, 29% at the target, 9% at the day end.

Filters: signals on every weekday; no entries 22:00 to 01:00 broker time (22,216 bars); max spread 3 pips (80 bars over it); M1 confirmation 0.5 atr (1,464 signals rejected).

Monthly R, d1_price_vs_ema20_atr < -37.057 on USDJPYOne bar per month: the R earned that month; the held-out months in the full colour.2023-11: +8.9 R over 19 trades2023-112023-12: -7.3 R over 107 trades2024-01: +2.1 R over 7 trades2024-03: -7.3 R over 25 trades2024-05: -0.2 R over 1 trades2024-06: +1.9 R over 1 trades2024-07: +2.4 R over 74 trades2024-072024-08: +12.2 R over 156 trades2024-09: -7.8 R over 63 trades2024-11: +5.1 R over 7 trades2024-12: +7.1 R over 15 trades2025-02: -8.3 R over 94 trades2025-03: +0.8 R over 21 trades2025-032025-04: +5.5 R over 78 trades2025-05: -2.7 R over 11 trades2025-06: -2.8 R over 5 trades2025-07: -1.5 R over 1 trades2025-10: +2.1 R over 1 trades2026-01: +3.5 R over 30 trades2026-012026-02: +12.5 R over 26 trades2026-04: -1.3 R over 1 trades2026-05: -2.1 R over 28 trades2026-07: +0.5 R over 2 trades2026-08: +8.5 R over 50 trades2026-09: +0.9 R over 72 trades2026-09+12 R-8 R

d1_price_vs_ema20_atr < -14.127 (p25)

Honesty panelNet positive, not distinguishable from chance

Net positive out of sample, not yet distinguishable from chance: +2.7 R over 386 trades is +0.007 R per trade, t = 0.1; about 221,003 trades would be needed to confirm this mean at t = 2, shown: 386; last quarter -7.2 R over 227 trades.

In sample
2023-09-30 to 2025-11-05
-111.7 R
1495 trades · 32% wins
-0.075 R per trade · t = -1.9, within chance
about 1,674 trades to confirm this mean (-0.075 R per trade) at t = -2 · shown: 1495
spread 0.104 R per trade
Out of sample
2025-11-05 to 2026-09-30
+2.7 R
386 trades · 35% wins
+0.007 R per trade · t = 0.1, within chance
about 221,003 trades to confirm this mean (+0.007 R per trade) at t = 2 · shown: 386
spread 0.133 R per trade
Last quarter
last 91 days of the window
-7.2 R
227 trades · 35% wins
-0.032 R per trade · t = -0.3, within chance
about 10,129 trades to confirm this mean (-0.032 R per trade) at t = -2 · shown: 227
spread 0.140 R per trade
Costs. Spread 0.110 R per trade at 1.25 times the bar’s minimum (more than the trades made before spread); 210 fills in measured hot hours (01:00 at 4.3 times). Fill offset 0.00 stop-distances on every entry, measured on 473 demo fills. At 1.5 times the bar’s minimum the out-of-sample figure would read -7.6 R (-0.020 R per trade over 386 trades).

Entry: stop order 2 pips past the signal close, cancelled after 4 bars; 2,304 signals placed, 1,881 filled, 423 expired; trigger state, an entry on every bar the conditions hold while flat.

Exit: stop 2 ATR (mean 18.1 pips, 2.14 ATR, 455 raised to the 10-pip floor); target 2.5 R; exits 66% at the stop, 26% at the target, 9% at the day end.

Filters: signals on every weekday; no entries 22:00 to 01:00 broker time (20,402 bars); max spread 3 pips (80 bars over it); M1 confirmation 0.5 atr (2,944 signals rejected).

Monthly R, d1_price_vs_ema20_atr < -14.127 on USDJPYOne bar per month: the R earned that month; the held-out months in the full colour.2023-10: +2.7 R over 2 trades2023-102023-11: +3.0 R over 72 trades2023-12: -30.5 R over 169 trades2024-01: +9.8 R over 12 trades2024-03: -13.8 R over 54 trades2024-05: +8.7 R over 15 trades2024-06: +7.5 R over 8 trades2024-062024-07: -25.5 R over 125 trades2024-08: -6.3 R over 200 trades2024-09: -16.8 R over 129 trades2024-11: -6.7 R over 27 trades2024-12: +0.3 R over 51 trades2025-01: +17.1 R over 33 trades2025-012025-02: -15.5 R over 156 trades2025-03: -16.9 R over 105 trades2025-04: -17.9 R over 174 trades2025-05: -15.8 R over 59 trades2025-06: -2.9 R over 33 trades2025-07: -5.1 R over 27 trades2025-072025-08: +7.1 R over 16 trades2025-09: +1.8 R over 15 trades2025-10: +4.0 R over 13 trades2025-12: +4.3 R over 7 trades2026-01: +4.3 R over 43 trades2026-02: +5.7 R over 65 trades2026-022026-04: -4.6 R over 4 trades2026-05: +0.2 R over 40 trades2026-07: -6.8 R over 17 trades2026-08: +2.4 R over 112 trades2026-09: -2.9 R over 98 trades+17 R-31 R

d1_price_vs_ema20_atr < 9.191 (p50)

Honesty panelNet positive, not distinguishable from chance

Net positive out of sample, not yet distinguishable from chance: +41.6 R over 862 trades is +0.048 R per trade, t = 0.9; about 4,472 trades would be needed to confirm this mean at t = 2, shown: 862; last quarter -12.6 R over 328 trades.

In sample
2023-09-30 to 2025-11-05
-148.2 R
2739 trades · 33% wins
-0.054 R per trade · t = -1.8, within chance
about 3,203 trades to confirm this mean (-0.054 R per trade) at t = -2 · shown: 2739
spread 0.100 R per trade
Out of sample
2025-11-05 to 2026-09-30
+41.6 R
862 trades · 37% wins
+0.048 R per trade · t = 0.9, within chance
about 4,472 trades to confirm this mean (+0.048 R per trade) at t = 2 · shown: 862
spread 0.132 R per trade
Last quarter
last 91 days of the window
-12.6 R
328 trades · 35% wins
-0.039 R per trade · t = -0.4, within chance
about 6,584 trades to confirm this mean (-0.039 R per trade) at t = -2 · shown: 328
spread 0.134 R per trade
Costs. Spread 0.107 R per trade at 1.25 times the bar’s minimum (more than the trades made before spread); 351 fills in measured hot hours (01:00 at 4.3 times). Fill offset 0.00 stop-distances on every entry, measured on 473 demo fills. At 1.5 times the bar’s minimum the out-of-sample figure would read +18.9 R (+0.022 R per trade over 862 trades).

Entry: stop order 2 pips past the signal close, cancelled after 4 bars; 4,395 signals placed, 3,601 filled, 794 expired; trigger state, an entry on every bar the conditions hold while flat.

Exit: stop 2 ATR (mean 18.1 pips, 2.12 ATR, 855 raised to the 10-pip floor); target 2.5 R; exits 64% at the stop, 26% at the target, 10% at the day end.

Filters: signals on every weekday; no entries 22:00 to 01:00 broker time (16,235 bars); max spread 3 pips (80 bars over it); M1 confirmation 0.5 atr (5,926 signals rejected).

Monthly R, d1_price_vs_ema20_atr < 9.191 on USDJPYOne bar per month: the R earned that month; the held-out months in the full colour.2023-10: +0.1 R over 55 trades2023-102023-11: -4.4 R over 104 trades2023-12: -32.6 R over 171 trades2024-01: +24.0 R over 55 trades2024-02: +3.1 R over 22 trades2024-03: -1.9 R over 83 trades2024-04: +0.9 R over 7 trades2024-042024-05: +10.7 R over 59 trades2024-06: +6.6 R over 60 trades2024-07: -25.7 R over 130 trades2024-08: -10.5 R over 206 trades2024-09: -28.2 R over 215 trades2024-10: +6.4 R over 21 trades2024-102024-11: -3.7 R over 78 trades2024-12: -2.3 R over 79 trades2025-01: -13.3 R over 133 trades2025-02: -25.8 R over 187 trades2025-03: -20.4 R over 174 trades2025-04: -23.9 R over 206 trades2025-042025-05: -18.4 R over 148 trades2025-06: -2.2 R over 156 trades2025-07: +9.0 R over 69 trades2025-08: -7.8 R over 162 trades2025-09: -5.1 R over 117 trades2025-10: +17.1 R over 42 trades2025-102025-11: +9.2 R over 12 trades2025-12: +8.4 R over 119 trades2026-01: +7.7 R over 80 trades2026-02: +19.4 R over 145 trades2026-03: +5.7 R over 21 trades2026-04: -0.3 R over 112 trades2026-042026-05: +2.3 R over 42 trades2026-06: +1.7 R over 3 trades2026-07: +1.4 R over 43 trades2026-08: -8.3 R over 137 trades2026-09: -5.7 R over 148 trades+24 R-33 R

d1_price_vs_ema20_atr < 32.589 (p75)

Honesty panelNo edge shown

No edge shown out of sample: -22.4 R over 1408 trades is -0.016 R per trade, within what chance produces (t = -0.4); about 39,211 trades would be needed to confirm this mean at t = -2, shown: 1408.

In sample
2023-09-30 to 2025-11-05
-169.5 R
3926 trades · 33% wins
-0.043 R per trade · t = -1.8, within chance
about 5,054 trades to confirm this mean (-0.043 R per trade) at t = -2 · shown: 3926
spread 0.105 R per trade
Out of sample
2025-11-05 to 2026-09-30
-22.4 R
1408 trades · 36% wins
-0.016 R per trade · t = -0.4, within chance
about 39,211 trades to confirm this mean (-0.016 R per trade) at t = -2 · shown: 1408
spread 0.135 R per trade
Last quarter
last 91 days of the window
-17.5 R
422 trades · 35% wins
-0.041 R per trade · t = -0.5, within chance
about 5,689 trades to confirm this mean (-0.041 R per trade) at t = -2 · shown: 422
spread 0.140 R per trade
Costs. Spread 0.113 R per trade at 1.25 times the bar’s minimum (more than the trades made before spread); 519 fills in measured hot hours (01:00 at 4.3 times). Fill offset 0.00 stop-distances on every entry, measured on 473 demo fills. At 1.5 times the bar’s minimum the out-of-sample figure would read -60.4 R (-0.043 R per trade over 1408 trades).

Entry: stop order 2 pips past the signal close, cancelled after 4 bars; 6,558 signals placed, 5,334 filled, 1,224 expired; trigger state, an entry on every bar the conditions hold while flat.

Exit: stop 2 ATR (mean 16.7 pips, 2.16 ATR, 1,607 raised to the 10-pip floor); target 2.5 R; exits 64% at the stop, 26% at the target, 10% at the day end.

Filters: signals on every weekday; no entries 22:00 to 01:00 broker time (11,545 bars); max spread 3 pips (55 bars over it); M1 confirmation 0.5 atr (8,692 signals rejected).

Monthly R, d1_price_vs_ema20_atr < 32.589 on USDJPYOne bar per month: the R earned that month; the held-out months in the full colour.2023-10: -2.4 R over 122 trades2023-102023-11: -6.6 R over 139 trades2023-12: -32.6 R over 171 trades2024-01: +14.5 R over 167 trades2024-02: +0.9 R over 91 trades2024-03: -5.8 R over 125 trades2024-04: +21.4 R over 59 trades2024-042024-05: +9.6 R over 135 trades2024-06: +12.5 R over 115 trades2024-07: -33.9 R over 163 trades2024-08: -10.5 R over 206 trades2024-09: -31.1 R over 236 trades2024-10: +24.0 R over 73 trades2024-102024-11: -24.9 R over 176 trades2024-12: +2.4 R over 139 trades2025-01: -13.4 R over 183 trades2025-02: -25.8 R over 187 trades2025-03: -24.6 R over 207 trades2025-04: -23.9 R over 206 trades2025-042025-05: -29.2 R over 185 trades2025-06: -22.5 R over 205 trades2025-07: +24.2 R over 168 trades2025-08: -5.7 R over 169 trades2025-09: -1.4 R over 162 trades2025-10: +17.4 R over 133 trades2025-102025-11: -3.7 R over 100 trades2025-12: -2.5 R over 156 trades2026-01: +4.3 R over 155 trades2026-02: +7.9 R over 185 trades2026-03: -2.6 R over 150 trades2026-04: -21.6 R over 150 trades2026-042026-05: +1.6 R over 47 trades2026-06: +9.5 R over 47 trades2026-07: +2.1 R over 105 trades2026-08: -9.8 R over 140 trades2026-09: -9.8 R over 177 trades+24 R-34 R

d1_price_vs_ema20_atr < 56.25 (p90)

Honesty panelNo edge shown

No edge shown out of sample: -67.8 R over 1544 trades is -0.044 R per trade, within what chance produces (t = -1.1); about 5,143 trades would be needed to confirm this mean at t = -2, shown: 1544.

In sample
2023-09-30 to 2025-11-05
-306.4 R
4410 trades · 32% wins
-0.069 R per trade · t = -3.0, beyond chance
about 1,968 trades to confirm this mean (-0.069 R per trade) at t = -2 · shown: 4410
spread 0.109 R per trade
Out of sample
2025-11-05 to 2026-09-30
-67.8 R
1544 trades · 35% wins
-0.044 R per trade · t = -1.1, within chance
about 5,143 trades to confirm this mean (-0.044 R per trade) at t = -2 · shown: 1544
spread 0.139 R per trade
Last quarter
last 91 days of the window
-38.4 R
449 trades · 34% wins
-0.085 R per trade · t = -1.2, within chance
about 1,299 trades to confirm this mean (-0.085 R per trade) at t = -2 · shown: 449
spread 0.141 R per trade
Costs. Spread 0.117 R per trade at 1.25 times the bar’s minimum (more than the trades made before spread); 612 fills in measured hot hours (01:00 at 4.3 times). Fill offset 0.00 stop-distances on every entry, measured on 473 demo fills. At 1.5 times the bar’s minimum the out-of-sample figure would read -110.9 R (-0.072 R per trade over 1544 trades).

Entry: stop order 2 pips past the signal close, cancelled after 4 bars; 7,392 signals placed, 5,954 filled, 1,438 expired; trigger state, an entry on every bar the conditions hold while flat.

Exit: stop 2 ATR (mean 16.2 pips, 2.20 ATR, 2,117 raised to the 10-pip floor); target 2.5 R; exits 65% at the stop, 25% at the target, 10% at the day end.

Filters: signals on every weekday; no entries 22:00 to 01:00 broker time (10,791 bars); max spread 3 pips (49 bars over it); M1 confirmation 0.5 atr (9,756 signals rejected).

Monthly R, d1_price_vs_ema20_atr < 56.25 on USDJPYOne bar per month: the R earned that month; the held-out months in the full colour.2023-10: -6.0 R over 131 trades2023-102023-11: -11.2 R over 146 trades2023-12: -32.6 R over 171 trades2024-01: +7.9 R over 196 trades2024-02: +0.8 R over 138 trades2024-03: -11.1 R over 140 trades2024-04: +16.5 R over 94 trades2024-042024-05: -4.4 R over 161 trades2024-06: -5.3 R over 139 trades2024-07: -32.8 R over 172 trades2024-08: -10.5 R over 206 trades2024-09: -31.1 R over 236 trades2024-10: +3.9 R over 184 trades2024-102024-11: -25.2 R over 201 trades2024-12: -6.2 R over 167 trades2025-01: -29.1 R over 201 trades2025-02: -25.8 R over 187 trades2025-03: -26.4 R over 208 trades2025-04: -23.9 R over 206 trades2025-042025-05: -32.6 R over 191 trades2025-06: -25.8 R over 208 trades2025-07: +25.4 R over 197 trades2025-08: -11.2 R over 174 trades2025-09: -5.5 R over 165 trades2025-10: +1.0 R over 184 trades2025-102025-11: -8.9 R over 134 trades2025-12: -2.6 R over 161 trades2026-01: -2.4 R over 167 trades2026-02: +7.9 R over 185 trades2026-03: -19.3 R over 184 trades2026-04: -19.0 R over 148 trades2026-042026-05: +1.6 R over 47 trades2026-06: +8.0 R over 76 trades2026-07: -14.2 R over 128 trades2026-08: -9.8 R over 140 trades2026-09: -14.4 R over 181 trades+25 R-33 R

What d1_price_vs_ema20_atr measures

Close vs D1 EMA20 (M5 ATR) (atr), on D1 bars. M5 close minus the daily EMA20 of the last completed day, in multiples of the M5 ATR(14).

Test your own variant

Rule Collider runs one rule on one symbol over years of broker M5 data with the fills and the spread a demo fleet measured, and shows the out-of-sample figure first. Early access goes to the waitlist in signup order.

Join the waitlist

Historical simulation on broker M5 data with a measured fill offset, a measured spread factor, swap and a one-minute confirmation model. Past results do not predict future results. This is a research tool, not investment advice. Hours are broker time, the chart time in MT5. Built 2026-11-04 from runs ending 2026-09-30; refreshed monthly.