Upward-Move Scorecard

Credit-scorecard methodology applied to price direction: daily trend features are coarse-classed into bins, converted to Weight of Evidence, ranked by Information Value, fitted with logistic regression and scaled into points — the same pipeline used to build a credit scorecard.

Read the numbers with credit-scoring instincts inverted. A credit scorecard separates a rare event (~3% default) from a common one, so IVs of 0.1+ and AUCs of 0.75+ are normal. Predicting whether a stock rises is a near coin flip (~50% base rate), so IVs here are one to two orders of magnitude smaller and an out-of-time AUC of 0.53 is a genuine result, not a broken model. The split is strictly by date, with a gap so no training row's outcome overlaps the test window — a random split would leak the future and produce a flattering, false AUC.

Score a stock today

Model performance — fitted 4 Aug 2026 11:29

Out-of-time AUC

0.5647

in-sample 0.5949

Gini

0.1294

0.1151

Overfit gap

0.0302

Base rate

49.91%

rise in 126d

Panel

165,532

148 stocks

Verdict: Genuinely useful separation for equity direction — check the lift table holds across deciles before believing it.
Train 2021-08-04 → 2024-09-26 (115,054 rows) · Test 2025-04-04 → 2026-02-04 (31,090 rows, strictly later).

Lift by score decile (out of time)

Decile	Rows	Rose	Lift
1 (best)	3,109	57.16%	1.143
2	3,109	57.48%	1.15
3	3,109	56.61%	1.132
4	3,109	57.54%	1.151
5	3,109	46.25%	0.925
6	3,109	47.76%	0.955
7	3,109	46.0%	0.92
8	3,109	44.97%	0.899
9	3,109	41.78%	0.836
10 (worst)	3,109	44.45%	0.889

The practical test: does the top decile actually rise more often than the bottom? A monotonic gradient down this table is worth more than the headline AUC. Lift is the decile's up-rate divided by the 49.91% base rate.

Feature information value

Feature	Group	IV	Bins	Coefficient
vol_63d	Volatility	0.07115	7	-0.6663
beta_63d	Relative strength	0.04547	6	-0.6523
vol_21d	Volatility	0.04473	8	-0.0222
ret_252d	Momentum	0.04308	5	-0.6319
atr_pct	Volatility	0.03565	9	0.2478
ma50_vs_ma200	Trend	0.03256	5	-0.7054
px_vs_ma200	Trend	0.02224	4	0.0393
ret_126d	Momentum	0.01543	5	0.3412
drawdown_from_high	Mean reversion	0.0135	4	-0.6776
dollar_volume_log	Volume	0.01309	4	-0.9942
px_vs_ma50	Trend	0.00827	3	-0.2063
ret_21d	Momentum	0.00719	4	0.0789
trend_slope_21d	Trend	0.00719	4	0.0789
px_vs_ma20	Trend	0.00659	4	-0.0712
ret_63d	Momentum	0.00633	3	-0.3153
ret_5d	Momentum	0.00596	4	-0.2753
rel_ret_21d	Relative strength	0.00584	4	-0.3592
rel_ret_63d	Relative strength	0.00387	3	0.2854
range_position	Mean reversion	0.00347	4	-0.782
ret_1d	Momentum	0.00329	4	-0.0085
volume_trend	Volume	0.00293	4	-0.2926
vol_ratio	Volatility	0.00121	2	-0.8699
obv_slope	Volume	0.00096	3
zscore_20d	Mean reversion	0.00064	3
rsi_14	Mean reversion	0.00059	6
volume_z	Volume	0.00054	4

IV measures how much a feature separates risers from non-risers. Because WoE is ln(non-event/event), a negative coefficient means the feature predicts upward moves. Features below the IV floor were dropped before fitting.

The scorecard — points by feature band base score 561, 20 points to double the odds

Feature	Group	Band	Rows	Rose	WoE	Points
vol_63d	Volatility	<= 0.1695	10,575	33.02%	0.702	0
vol_63d	Volatility	(0.1695, 0.1978]	10,575	45.46%	0.1771	10
vol_63d	Volatility	(0.1978, 0.2224]	10,574	46.98%	0.1156	11
vol_63d	Volatility	(0.2224, 0.2464]	10,575	48.66%	0.0484	13
vol_63d	Volatility	(0.2464, 0.2728]	10,574	51.27%	-0.0559	15
vol_63d	Volatility	(0.2728, 0.309]	10,575	53.27%	-0.1361	16
vol_63d	Volatility	> 0.309	42,298	55.01%	-0.2063	17
beta_63d	Relative strength	<= 0.5357	21,150	43.71%	0.2479	0
beta_63d	Relative strength	(0.5357, 0.685]	10,574	45.61%	0.1708	1
beta_63d	Relative strength	(0.685, 0.9217]	21,149	46.26%	0.1446	2
beta_63d	Relative strength	(0.9217, 1.043]	10,575	48.42%	0.0582	4
beta_63d	Relative strength	(1.043, 1.187]	10,574	55.28%	-0.2171	9
beta_63d	Relative strength	> 1.187	31,724	56.49%	-0.2661	10
vol_21d	Volatility	<= 0.1504	11,195	38.04%	0.4824	0
vol_21d	Volatility	(0.1504, 0.1806]	11,195	44.22%	0.2271	0
vol_21d	Volatility	(0.1806, 0.2062]	11,194	46.79%	0.1231	0
vol_21d	Volatility	(0.2062, 0.2321]	11,195	49.28%	0.0234	0
vol_21d	Volatility	(0.2321, 0.2639]	11,194	49.71%	0.0065	0
vol_21d	Volatility	(0.2639, 0.3019]	11,195	52.48%	-0.1046	0
vol_21d	Volatility	(0.3019, 0.5198]	33,583	54.29%	-0.1773	0
vol_21d	Volatility	> 0.5198	11,195	55.28%	-0.2175	0
ret_252d	Momentum	<= 0.04096	39,058	45.85%	0.1621	0
ret_252d	Momentum	(0.04096, 0.0983]	7,810	47.9%	0.0798	2
ret_252d	Momentum	(0.0983, 0.1683]	7,811	49.2%	0.0277	2
ret_252d	Momentum	(0.1683, 0.2715]	7,811	55.46%	-0.2236	7
ret_252d	Momentum	> 0.2715	15,623	58.56%	-0.3501	9
atr_pct	Volatility	<= 0.01655	22,626	41.93%	0.3203	4
atr_pct	Volatility	(0.01655, 0.01853]	11,313	46.07%	0.1521	3
atr_pct	Volatility	(0.01853, 0.02041]	11,314	49.41%	0.0183	2
atr_pct	Volatility	(0.02041, 0.02254]	11,312	50.95%	-0.0436	1
atr_pct	Volatility	(0.02254, 0.02509]	11,313	51.95%	-0.0834	1
atr_pct	Volatility	(0.02509, 0.02851]	11,313	53.18%	-0.1327	1
atr_pct	Volatility	(0.02851, 0.03312]	11,313	53.43%	-0.1426	1
atr_pct	Volatility	(0.03312, 0.04054]	11,313	54.23%	-0.175	0
atr_pct	Volatility	> 0.04054	11,313	55.58%	-0.2296	0
ma50_vs_ma200	Trend	<= -0.007352	34,333	46.26%	0.1431	0
ma50_vs_ma200	Trend	(-0.007352, 0.01494]	8,583	46.83%	0.1205	0
ma50_vs_ma200	Trend	(0.01494, 0.0596]	17,166	48.95%	0.0353	2
ma50_vs_ma200	Trend	(0.0596, 0.08936]	8,583	52.57%	-0.1095	5
ma50_vs_ma200	Trend	> 0.08936	17,166	57.99%	-0.3292	10
px_vs_ma200	Trend	<= 0.05005	51,499	47.53%	0.0923	0
px_vs_ma200	Trend	(0.05005, 0.08115]	8,583	48.68%	0.0462	0
px_vs_ma200	Trend	(0.08115, 0.1226]	8,583	50.34%	-0.0204	0
px_vs_ma200	Trend	> 0.1226	17,166	57.07%	-0.2915	0
ret_126d	Momentum	<= 0.06648	57,893	47.98%	0.0722	3
ret_126d	Momentum	(0.06648, 0.1157]	9,649	48.11%	0.0671	3
ret_126d	Momentum	(0.1157, 0.1809]	9,649	50.34%	-0.0221	3
ret_126d	Momentum	(0.1809, 0.2972]	9,649	54.77%	-0.2001	1
ret_126d	Momentum	> 0.2972	9,649	56.75%	-0.2803	0
drawdown_from_high	Mean reversion	<= -0.3161	10,634	58.41%	-0.3446	8
drawdown_from_high	Mean reversion	(-0.3161, -0.2365]	10,633	50.47%	-0.0241	1
drawdown_from_high	Mean reversion	(-0.2365, -0.1165]	31,905	48.83%	0.0419	0
drawdown_from_high	Mean reversion	> -0.1165	53,162	48.67%	0.048	0
dollar_volume_log	Volume	<= 13.86	11,506	54.24%	-0.1751	8
dollar_volume_log	Volume	(13.86, 15.65]	23,010	53.27%	-0.1362	7
dollar_volume_log	Volume	(15.65, 16.1]	11,506	52.59%	-0.1088	6
dollar_volume_log	Volume	> 16.1	69,032	47.56%	0.0926	0
px_vs_ma50	Trend	<= -0.08385	10,781	56.05%	-0.2481	2
px_vs_ma50	Trend	(-0.08385, -0.0489]	10,780	52.04%	-0.0865	1
px_vs_ma50	Trend	> -0.0489	86,243	48.84%	0.0417	0
ret_21d	Momentum	<= -0.09582	11,195	55.79%	-0.2381	0
ret_21d	Momentum	(-0.09582, -0.05676]	11,195	51.47%	-0.0641	0
ret_21d	Momentum	(-0.05676, -0.03205]	11,194	49.12%	0.0301	1
ret_21d	Momentum	> -0.03205	78,362	48.9%	0.0387	1
trend_slope_21d	Trend	<= -1.209	11,195	55.79%	-0.2381	0
trend_slope_21d	Trend	(-1.209, -0.7012]	11,195	51.47%	-0.0641	0
trend_slope_21d	Trend	(-0.7012, -0.3909]	11,194	49.12%	0.0301	1
trend_slope_21d	Trend	> -0.3909	78,362	48.9%	0.0387	1
px_vs_ma20	Trend	<= -0.05414	11,225	54.92%	-0.2029	1
px_vs_ma20	Trend	(-0.05414, -0.03136]	11,224	52.54%	-0.107	0
px_vs_ma20	Trend	(-0.03136, -0.01765]	11,224	49.8%	0.0028	0
px_vs_ma20	Trend	> -0.01765	78,569	48.77%	0.0438	0
ret_63d	Momentum	<= -0.1463	10,579	55.3%	-0.2179	2
ret_63d	Momentum	(-0.1463, -0.08571]	10,572	51.71%	-0.0737	1
ret_63d	Momentum	> -0.08571	84,595	48.96%	0.0364	0
ret_5d	Momentum	<= -0.04895	11,432	54.69%	-0.1932	2
ret_5d	Momentum	(-0.04895, -0.02816]	11,431	52.45%	-0.1033	1
ret_5d	Momentum	(-0.02816, -0.01583]	11,431	49.6%	0.0108	0
ret_5d	Momentum	> -0.01583	80,020	48.85%	0.0407	0
rel_ret_21d	Relative strength	<= -0.08727	11,195	54.58%	-0.189	2
rel_ret_21d	Relative strength	(-0.08727, -0.05305]	11,195	52.1%	-0.0895	1
rel_ret_21d	Relative strength	(-0.05305, -0.03152]	11,194	50.67%	-0.0321	1
rel_ret_21d	Relative strength	> -0.03152	78,362	48.76%	0.0443	0
rel_ret_63d	Relative strength	<= -0.1435	10,575	53.97%	-0.1642	0
rel_ret_63d	Relative strength	(-0.1435, -0.08735]	10,575	51.62%	-0.0701	1
rel_ret_63d	Relative strength	> -0.08735	84,596	49.14%	0.0292	2
range_position	Mean reversion	<= 0.3185	31,900	48.08%	0.0718	0
range_position	Mean reversion	(0.3185, 0.4343]	10,635	48.75%	0.0448	1
range_position	Mean reversion	(0.4343, 0.9475]	53,165	50.61%	-0.0293	2
range_position	Mean reversion	> 0.9475	10,634	52.71%	-0.1135	4
ret_1d	Momentum	<= -0.02214	11,491	53.76%	-0.1559	0
ret_1d	Momentum	(-0.02214, -0.0125]	11,497	51.1%	-0.0491	0
ret_1d	Momentum	(-0.0125, -0.006941]	11,484	49.97%	-0.0037	0
ret_1d	Momentum	> -0.006941	80,434	49.13%	0.0298	0
volume_trend	Volume	<= -0.4219	10,634	52.95%	-0.1233	1
volume_trend	Volume	(-0.4219, -0.3081]	10,633	51.26%	-0.0553	1
volume_trend	Volume	(-0.3081, -0.1518]	21,267	50.6%	-0.0291	1
volume_trend	Volume	> -0.1518	63,800	48.89%	0.0394	0
vol_ratio	Volatility	<= 1.265	95,165	49.58%	0.0116	0
vol_ratio	Volatility	> 1.265	10,574	52.48%	-0.1042	3

This is the deliverable a credit team would hand over: every feature's bands with the points they earn. Each feature's worst band scores 0, so the total is the base score plus everything the current configuration earns. Model output for research — not financial advice, and a 0.5647 AUC means it is wrong very often.