§ corpus sweep · daily bars · 2026-08-08

We ran every public Pine script we could find.

12,424 scripts scraped from public GitHub repositories, classified by AST into four grading tracks, and measured — each track by the only question it can honestly answer. The 3,797 that place orders are backtested below; signal indicators are traded through a published harness; plot-only indicators are scored on out-of-sample predictive power against canonical baselines; visual tools are audited for repaint surface. Every guard result is a column, never a pass/fail stamp.

3,797 A · strategies4,058 B · signal indicators2,787 C · continuous indicators1,470 D · visual tools312 libraries (not graded)

Scripts swept

3,797

Track A — declares strategy() or places orders

Eligible measurements

5,478

24.0% of 22,782 script-market pairs

Clear all four guards

Across all six markets combined

…with confirmable fills

Both survivors exit on a trailing stop

The two scripts that clear every statistical guard both exit on a trailing stop. A stop fills somewhere inside a bar and daily bars do not record where — so the engine has to assume a path, and on the higher-scoring of the two that assumption is doing the work: a 0.1% trailing offset on candles with a 3–5% range, 700 trades, a $181 maximum drawdown. We publish it, labelled, and rank it separately. It is not a verdict on the script. It is a statement about what a bar dataset can and cannot confirm.

Where the measurements go

Counted in script-market pairs, not scripts — a script can be ineligible on one market and perfectly measurable on another. None of these rules is a quality judgment; each one marks a backtest that did not produce a measurable result at all.

4,755equity never moved
4,666did not compile or crashed
4,014fewer than 20 closed trades
1,694held a position ≥98% of bars
1,363implied leverage above 1.5×
685account went to zero
127held a position <2% of bars
3,654eligible, but fills depend on an assumed intrabar path (66.7% of eligible — labelled, not excluded)

The same sweep at three timeframes

Daily is the table above, and the only one worth ranking on. The intraday sweeps run for completeness — many of these scripts were written for intraday charts and deserve the fair test — but their histories are far shorter outside crypto, so the six markets are no longer being asked the same question.

Timeframe	Eligible	History (bars)	Mean luck hurdle	Clear all guards
1Dranked above	5,478	3,278–4,000	0.90	2
4H	5,667	1,707–19,653	2.09	0
1H	6,214	5,090–78,545	2.68	3

The spread in the history column is the reason to be careful: at 4H and 1H the crypto markets carry nine years and the equity and metals markets carry two or three, which inflates their buy-and-hold benchmark and makes the excess column mean something different in each row. Daily is the only sweep where all six windows are long enough to say the same thing.

By market

Ranked on Sharpe in excess of buy-and-hold over the identical window. Markets are never averaged together: the windows differ, the regimes differ, and a mean across them lets one spectacular cell carry five poor ones.

BTCUSD

2017-08-17 → 2026-08-07 · 3,278 bars · buy-and-hold Sharpe 0.66 · luck hurdle 0.99

715 of 3,797 measurements were eligible; 622 survive the duplicate collapse. None clear all four guards. A search this wide is expected to throw up a 0.99 Sharpe on luck alone.

#	Script	Sharpe	vs B&H	Deflated	Out-of-sample	Breakeven	Trades
1	f350bbc9469dunverified fills	1.23	+0.57	0.83	1.05	130bp	429
2	9df83dc1035d	0.98	+0.33	0.48	0.76	350bp	79
3	481eadba9429unverified fills	0.93	+0.27	0.40	0.20	249bp	71
4	3d10a14e6366unverified fills+1 copy	0.91	+0.25	0.36	0.08	690bp	70
5	79ee75d8a4f1unverified fills	0.90	+0.24	0.37	0.67	376bp	104
6	de2b3ad147b6unverified fills	0.90	+0.24	0.36	1.05	173bp	191
7	6d7745c56b7e	0.88	+0.23	0.34	0.43	712bp	42
8	e75dba25276aunverified fills	0.87	+0.22	0.31	0.92	387bp	86
9	3fa2a8943294	0.87	+0.21	0.32	0.45	59bp	408
10	bb5ea915a6f2unverified fills	0.87	+0.21	0.32	0.94	398bp	42
11	bcef98bb3554unverified fills	0.86	+0.20	0.30	0.63	239bp	126
12	3f2d12f10ac4	0.85	+0.20	0.30	0.66	850bp	45
13	b7dbba45076eunverified fills	0.84	+0.18	0.27	0.08	655bp	71
14	10a77dd7801funverified fills	0.84	+0.18	0.26	1.11	216bp	38
15	0824a9731deaunverified fills	0.83	+0.18	0.25	0.56	93bp	194
16	a8412e9e7194unverified fills	0.83	+0.18	0.28	0.78	888bp	37
17	18108d19313f	0.83	+0.17	0.27	0.50	199bp	116
18	8c2d382277ceunverified fills	0.83	+0.17	0.27	0.61	471bp	84
19	82bdbd84e6cf	0.83	+0.17	0.27	0.68	212bp	173
20	00b46c35e453unverified fills	0.82	+0.16	0.24	0.95	157bp	68
21	d9106db72344unverified fills	0.81	+0.16	0.25	0.27	45bp	735
22	97bb41d55176unverified fills	0.80	+0.14	0.23	0.61	71bp	282
23	e03fb2f48bd9	0.79	+0.14	0.23	0.76	1673bp	67
24	51cb4a80f2fd+1 copy	0.78	+0.13	0.22	0.73	487bp	64
25	27e332631bd1	0.78	+0.13	0.19	0.68	307bp	22

ETHUSD

2017-08-17 → 2026-08-07 · 3,278 bars · buy-and-hold Sharpe 0.56 · luck hurdle 0.94

863 of 3,797 measurements were eligible; 717 survive the duplicate collapse. 2 clear all four guards, and none of those has fills we can confirm. A search this wide is expected to throw up a 0.94 Sharpe on luck alone.

#	Script	Sharpe	vs B&H	Deflated	Out-of-sample	Breakeven	Trades
1	8b0ccbe43bfdunverified fills	2.50	+1.94	1.00	2.38	162bp	700
2	f350bbc9469dunverified fills	1.49	+0.93	0.99	1.67	251bp	427
3	e8ce11a33670unverified fills	1.03	+0.47	0.69	1.54	103bp	90
4	320a76d7cda6unverified fills	1.00	+0.43	0.58	1.96	56bp	282
5	e0f64d88f71funverified fills	0.95	+0.39	0.52	1.05	357bp	92
6	92ef1156b121	0.90	+0.33	0.44	0.78	570bp	52
7	786f08429a15unverified fills	0.86	+0.30	0.39	1.15	206bp	83
8	7e0bd15865f4unverified fills	0.84	+0.28	0.37	0.49	355bp	126
9	0906c0321dd3unverified fills	0.82	+0.26	0.34	0.45	366bp	95
10	9df83dc1035d	0.82	+0.26	0.34	0.45	454bp	72
11	6995914c771dunverified fills+1 copy	0.78	+0.21	0.28	0.66	502bp	87
12	bb5ea915a6f2unverified fills	0.76	+0.20	0.26	0.31	507bp	36
13	f814fdd4baa4unverified fills	0.76	+0.20	0.26	0.40	885bp	36
14	f6451e46d9b0unverified fills	0.75	+0.19	0.24	1.37	95bp	164
15	641bd91b80b0unverified fills	0.74	+0.18	0.24	0.64	193bp	176
16	47ca82903670unverified fills	0.74	+0.18	0.23	0.73	347bp	46
17	d2146a1673e9unverified fills	0.73	+0.17	0.21	0.48	186bp	424
18	6b6d612edc1eunverified fills	0.71	+0.15	0.17	0.74	130bp	47
19	51cb4a80f2fd+1 copy	0.71	+0.15	0.21	0.45	617bp	58
20	2c4833f9ea17unverified fills	0.71	+0.15	0.21	0.23	227bp	143
21	3770e8fe4b4c+1 copy	0.71	+0.14	0.20	0.49	277bp	103
22	8086936c0135unverified fills	0.70	+0.14	0.20	0.30	546bp	71
23	8c2d382277ceunverified fills	0.69	+0.13	0.19	0.77	468bp	81
24	bcef98bb3554unverified fills	0.69	+0.13	0.18	0.49	224bp	122
25	0ce2a5c82945unverified fills	0.68	+0.12	0.18	0.45	567bp	77

SPX500

2010-09-10 → 2026-08-06 · 4,000 bars · buy-and-hold Sharpe 0.80 · luck hurdle 1.01

1,002 of 3,797 measurements were eligible; 845 survive the duplicate collapse. None clear all four guards. A search this wide is expected to throw up a 1.01 Sharpe on luck alone.

#	Script	Sharpe	vs B&H	Deflated	Out-of-sample	Breakeven	Trades
1	8b0ccbe43bfdunverified fills	1.44	+0.64	0.96	1.97	16bp	889
2	10e5023f17bfunverified fills	0.87	+0.07	0.27	0.74	135bp	67
3	933c5c63a17d	0.87	+0.07	0.28	0.67	44bp	279
4	9ed647d79f1a	0.86	+0.06	0.27	1.00	19bp	245
5	350f6fea94d8unverified fills	0.86	+0.06	0.25	0.58	19bp	113
6	2954f8fe430bunverified fills	0.85	+0.05	0.26	1.17	188bp	53
7	0f5405895f90unverified fills	0.84	+0.04	0.24	0.55	120bp	79
8	e75dba25276aunverified fills	0.81	+0.00	0.20	0.76	770bp	21
9	c4abe310bb5b	0.80	+0.00	0.19	0.87	114bp	109
10	09112c0ed304unverified fills+1 copy	0.80	-0.00	0.19	1.15	10bp	248
11	c82f32e839bd	0.79	-0.01	0.19	0.70	157bp	53
12	c32ee9a09a20unverified fills	0.78	-0.02	0.18	0.74	44bp	217
13	47ca82903670unverified fills	0.78	-0.02	0.18	0.71	78bp	65
14	ffcca3b65c2aunverified fills	0.77	-0.03	0.17	1.02	52bp	153
15	b10b53e4a43b	0.77	-0.03	0.17	0.81	161bp	120
16	a5ef13bf4849unverified fills+1 copy	0.76	-0.04	0.15	0.97	85bp	62
17	c85c954ef05cunverified fills	0.76	-0.04	0.16	1.46	24bp	150
18	9df83dc1035d	0.75	-0.05	0.15	0.90	45bp	106
19	9e794798c0fd	0.72	-0.08	0.12	0.61	94bp	80
20	81db64760c31unverified fills	0.72	-0.08	0.12	0.48	66bp	110
21	ef7ac175c108unverified fills	0.71	-0.09	0.12	0.98	28bp	207
22	c330af749e92	0.71	-0.09	0.11	1.02	123bp	69
23	394796be9266unverified fills	0.71	-0.09	0.12	0.37	44bp	186
24	6d7745c56b7e	0.71	-0.09	0.12	0.81	63bp	66
25	b0c066e1b159	0.71	-0.09	0.11	1.08	99bp	67

QQQ

2010-09-10 → 2026-08-06 · 4,000 bars · buy-and-hold Sharpe 0.94 · luck hurdle 1.06

1,011 of 3,797 measurements were eligible; 844 survive the duplicate collapse. None clear all four guards. A search this wide is expected to throw up a 1.06 Sharpe on luck alone.

#	Script	Sharpe	vs B&H	Deflated	Out-of-sample	Breakeven	Trades
1	8b0ccbe43bfdunverified fills	1.17	+0.23	0.66	1.65	20bp	806
2	10e5023f17bfunverified fills	1.00	+0.06	0.40	0.94	194bp	70
3	cafc7bd8d462unverified fills	0.96	+0.02	0.35	0.63	22bp	136
4	ef7815e76f1cunverified fills	0.94	+0.01	0.32	1.28	105bp	229
5	b21aaf4506e8unverified fills	0.91	-0.02	0.28	0.40	351bp	39
6	41a59c37b715unverified fills	0.88	-0.06	0.21	0.89	63bp	136
7	0f5405895f90unverified fills	0.87	-0.07	0.23	0.78	173bp	73
8	394796be9266unverified fills	0.86	-0.07	0.22	0.74	52bp	229
9	933c5c63a17d	0.85	-0.08	0.21	0.71	41bp	366
10	8880640d00d1unverified fills+1 copy	0.85	-0.09	0.21	0.51	362bp	20
11	9ed647d79f1a	0.84	-0.09	0.20	0.83	26bp	228
12	90dc8c511bfbunverified fills	0.84	-0.09	0.19	0.81	185bp	50
13	c32ee9a09a20unverified fills	0.84	-0.10	0.20	0.61	61bp	204
14	b91cba1b6688unverified fills	0.83	-0.10	0.19	0.57	498bp	21
15	c82f32e839bd	0.83	-0.11	0.18	0.61	298bp	40
16	c8c53ca8f96cunverified fills	0.83	-0.11	0.18	0.26	137bp	41
17	fa4858c800aaunverified fills	0.83	-0.11	0.18	0.57	26bp	77
18	a8412e9e7194unverified fills	0.82	-0.12	0.17	0.56	1248bp	20
19	5374ee946141unverified fills	0.81	-0.13	0.17	0.96	185bp	46
20	e75dba25276aunverified fills	0.81	-0.13	0.16	0.70	1156bp	22
21	c330af749e92	0.80	-0.13	0.15	0.85	127bp	93
22	9df83dc1035d	0.80	-0.14	0.15	0.74	64bp	104
23	282448ec3ee7unverified fills	0.79	-0.15	0.15	0.71	1130bp	24
24	e96912a44343unverified fills	0.79	-0.15	0.14	0.91	59bp	266
25	1a005e6b1836unverified fills+8 copies	0.78	-0.15	0.13	0.70	211bp	49

GOLD

2010-09-10 → 2026-08-06 · 4,000 bars · buy-and-hold Sharpe 0.52 · luck hurdle 0.79

1,014 of 3,797 measurements were eligible; 846 survive the duplicate collapse. None clear all four guards. A search this wide is expected to throw up a 0.79 Sharpe on luck alone.

#	Script	Sharpe	vs B&H	Deflated	Out-of-sample	Breakeven	Trades
1	8b0ccbe43bfdunverified fills	1.18	+0.65	0.93	0.96	16bp	817
2	e75dba25276aunverified fills	0.80	+0.28	0.53	1.64	892bp	21
3	ba3d466c57f7unverified fills	0.77	+0.24	0.47	1.58	731bp	27
4	bafac95c4f45	0.71	+0.19	0.37	0.89	16bp	239
5	f5c79326a9dbunverified fills	0.70	+0.18	0.36	0.80	61bp	33
6	47ca82903670unverified fills	0.69	+0.17	0.35	1.21	58bp	79
7	54a337f2ed35	0.68	+0.15	0.31	0.65	22bp	118
8	110c8ca80323unverified fills	0.67	+0.15	0.32	1.23	150bp	20
9	129af442b78aunverified fills+1 copy	0.65	+0.13	0.29	0.86	57bp	1042
10	815bc5ac8ef1unverified fills	0.65	+0.13	0.29	0.84	153bp	35
11	a6baae6cc1f2	0.65	+0.12	0.29	0.98	107bp	29
12	984f25ff4a3eunverified fills	0.64	+0.12	0.28	0.97	98bp	43
13	46f51aa7fa77unverified fills+2 copies	0.63	+0.11	0.25	1.27	22bp	107
14	e18ef3854449unverified fills	0.63	+0.10	0.26	1.16	195bp	31
15	97e2ab193cd4	0.62	+0.10	0.27	1.12	252bp	45
16	08675d7e8462	0.62	+0.09	0.25	1.09	81bp	84
17	dfce8ddf6c28unverified fills	0.61	+0.09	0.24	1.12	27bp	169
18	e65cb47877d6unverified fills	0.60	+0.07	0.22	1.08	13bp	542
19	850d0973bf7dunverified fills	0.59	+0.07	0.24	1.07	61bp	252
20	ef7815e76f1cunverified fills	0.58	+0.06	0.21	1.43	61bp	193
21	63e93f09d32c	0.58	+0.06	0.22	1.07	56bp	126
22	e34b5f5c4058unverified fills	0.58	+0.05	0.19	0.96	34bp	152
23	10e5023f17bfunverified fills	0.58	+0.05	0.20	0.93	62bp	70
24	c101398d3287	0.58	+0.05	0.21	0.89	93bp	54
25	c8c53ca8f96cunverified fills	0.57	+0.05	0.20	0.92	102bp	36

EURUSD

2011-03-27 → 2026-08-07 · 4,000 bars · buy-and-hold Sharpe -0.11 · luck hurdle 0.63

873 of 3,797 measurements were eligible; 721 survive the duplicate collapse. None clear all four guards. A search this wide is expected to throw up a 0.63 Sharpe on luck alone.

#	Script	Sharpe	vs B&H	Deflated	Out-of-sample	Breakeven	Trades
1	8b0ccbe43bfdunverified fills	1.71	+1.82	1.00	1.27	9bp	863
2	d2146a1673e9unverified fills	0.62	+0.73	0.49	0.35	66bp	178
3	abc8ce376cb5unverified fills	0.48	+0.60	0.28	0.10	12bp	353
4	3c7af8b6c7aa	0.47	+0.58	0.26	0.67	25bp	24
5	9960631a4fe0unverified fills	0.45	+0.56	0.23	0.62	67bp	87
6	524cd1698d6eunverified fills	0.43	+0.54	0.19	0.33	16bp	32
7	017e4bf7251bunverified fills	0.42	+0.53	0.18	-0.45	3bp	592
8	30a49af03fffunverified fills	0.41	+0.53	0.20	0.66	44bp	22
9	46f9f3639fbf	0.40	+0.51	0.18	0.38	39bp	27
10	af6605f124bcunverified fills	0.39	+0.50	0.16	-0.11	280bp	2162
11	8ecad2a97c87unverified fills	0.39	+0.50	0.17	0.71	18bp	360
12	c49c494ee52dunverified fills	0.37	+0.49	0.16	0.92	19bp	115
13	77884fa81d6funverified fills	0.37	+0.48	0.16	0.42	13bp	270
14	97756a7feea9	0.36	+0.47	0.14	0.42	9bp	166
15	c2684d73ea76	0.36	+0.47	0.14	0.22	8bp	108
16	1ae3ee549758	0.35	+0.46	0.14	0.30	12bp	86
17	1ab9d8c982ebunverified fills	0.35	+0.46	0.13	-0.15	22bp	78
18	fedbb19d402bunverified fills	0.34	+0.45	0.12	-0.04	16bp	56
19	0dd0fffbcfe8unverified fills	0.33	+0.44	0.12	-0.34	8bp	70
20	70887d4d62bd	0.33	+0.44	0.12	0.10	10bp	94
21	337fa4fc672aunverified fills	0.32	+0.44	0.11	0.15	10bp	70
22	116ae7ac5736unverified fills+16 copies	0.31	+0.42	0.10	0.73	4bp	328
23	0f5bde3b2524unverified fills	0.31	+0.42	0.10	0.10	19bp	86
24	bafac95c4f45	0.30	+0.42	0.09	-0.20	3bp	185
25	952cdcbe6346unverified fills	0.30	+0.41	0.09	-0.14	2bp	372

Scripts that show up more than once

63 scripts reach the published rows of at least two markets, 4 of four or more. Each is reported by its worst market, never its best — that is what makes the number mean something.

Script	Markets	Worst vs B&H	Worst out-of-sample	Lowest deflated	Fills
8b0ccbe43bfd	ETHUSD · SPX500 · QQQ · GOLD · EURUSD	+0.23	0.96	0.66	assumed intrabar
47ca82903670	BTCUSD · ETHUSD · SPX500 · QQQ · GOLD	-0.24	0.62	0.08	assumed intrabar
e75dba25276a	BTCUSD · SPX500 · QQQ · GOLD	-0.13	0.70	0.16	assumed intrabar
9df83dc1035d	BTCUSD · ETHUSD · SPX500 · QQQ	-0.14	0.45	0.15	confirmable
08675d7e8462	BTCUSD · ETHUSD · GOLD	+0.07	0.60	0.13	confirmable
10e5023f17bf	SPX500 · QQQ · GOLD	+0.05	0.74	0.20	assumed intrabar
79ee75d8a4f1	BTCUSD · ETHUSD · EURUSD	+0.05	0.06	0.08	assumed intrabar
933c5c63a17d	SPX500 · QQQ · GOLD	-0.08	0.67	0.15	confirmable
394796be9266	SPX500 · QQQ · GOLD	-0.09	0.37	0.12	assumed intrabar
9ed647d79f1a	SPX500 · QQQ · GOLD	-0.09	0.83	0.17	confirmable
b21aaf4506e8	ETHUSD · SPX500 · QQQ	-0.10	0.35	0.11	assumed intrabar
f814fdd4baa4	ETHUSD · SPX500 · QQQ	-0.20	0.40	0.11	assumed intrabar

§ track B · signal indicators · same rulebook

2,658 indicators that say “Buy” — so we bought.

4,058 scripts never call strategy() but emit discrete signals — alertcondition(), buy/sell shapes, arrows. Those signals are machine-readable trade semantics, so a harness rewrites each script as the same canonical strategy: long on any BUY-classified signal at the next bar open, flat on any SELL, 100% of equity, long-only. Signal titles are classified by a fixed keyword table — no model in the loop, and the exact mapping used for every script is recorded. 2,658 of 4,058 mapped cleanly (2,441 with an exit signal, 217 entry-only); the rest are counted, not guessed at. The harnessed scripts then run through the identical pipeline as Track A — same eligibility rules, same deflated Sharpe, same duplicate collapse. One difference is deliberate: the harness sizes every entry at 100% of current equity, so implied leverage is measured against equity rather than initial capital — against initial capital, compounding itself would read as margin.

Signal indicators harnessed

2,658

of 4,058 Track B scripts — the unmapped are listed by reason

Eligible measurements

6,444

40.4% of 15,948 script-market pairs

Clear all four guards

All on the equity indices — none on crypto, gold or FX

…with confirmable fills

The harness never uses stops, so every fill is bar-verifiable

Market	Eligible	After dedupe	Beats B&H	OOS positive	Cost ≥ 20bp	All guards
BTCUSD	1,065	774	379	728	730	0
ETHUSD	1,056	767	401	596	691	0
SPX500	1,116	840	32	807	660	11
QQQ	1,122	840	18	808	733	17
GOLD	1,130	849	159	831	610	0
EURUSD	955	699	348	432	6	0

The pattern is the mirror image of Track A. On the crypto markets, buy-and-hold set a Sharpe no dip-buying signal overlay could beat; on the equity indices the harness found 28 candidates that clear every guard — deflated Sharpe included — and all of them fill at bar boundaries, so there is no intrabar assumption to hide behind. The tables below rank the markets where that happened.

SPX500 — signal indicators

2010-09-10 → 2026-08-06 · 4,000 bars · buy-and-hold Sharpe 0.80 · luck hurdle 0.42

1,116 of 2,658 measurements were eligible; 840 survive the duplicate collapse. 11 clear all four guards. A search this wide is expected to throw up a 0.42 Sharpe on luck alone.

#	Script	Sharpe	vs B&H	Deflated	Out-of-sample	Breakeven	Trades
1	79ca6f94d7ae	0.93	+0.13	0.98	1.11	110bp	77
2	3741497843a6	0.92	+0.12	0.98	0.80	113bp	74
3	5017c92ab7bd	0.91	+0.10	0.97	0.73	48bp	149
4	51f91f8af289	0.89	+0.09	0.97	0.86	19bp	460
5	c759f9d5e54c	0.88	+0.08	0.96	0.64	256bp	28
6	524a8b03766d	0.88	+0.08	0.96	1.08	66bp	117
7	6e1073fb0e17	0.88	+0.08	0.96	0.93	136bp	54
8	0cfe0dc91192	0.88	+0.08	0.97	0.75	54bp	185
9	86e7c9c892cf	0.88	+0.08	0.96	0.78	410bp	20
10	4468a6b5999f	0.85	+0.05	0.96	0.90	66bp	112
11	eceb5c6ef36b	0.84	+0.04	0.95	0.77	48bp	179
12	15b955ec0ddc	0.84	+0.04	0.95	0.72	248bp	39
13	5dab3671c395	0.84	+0.04	0.95	0.82	98bp	72
14	ea1d1a3a95d3	0.84	+0.04	0.95	0.74	123bp	87
15	e19a44f3a32f+2 copies	0.83	+0.03	0.95	1.07	96bp	74
16	da199e035dce	0.83	+0.03	0.94	0.96	74bp	88
17	088de02be0e4	0.83	+0.03	0.95	0.80	277bp	195
18	529b1f0fd956	0.83	+0.03	0.94	0.92	352bp	20
19	a3199187a8ab	0.83	+0.03	0.95	0.64	46bp	176
20	5be17d173c29	0.82	+0.02	0.94	1.01	87bp	75
21	3e31a39b9ca2+5 copies	0.82	+0.02	0.95	0.43	17bp	428
22	1782b7b6857e	0.82	+0.02	0.94	0.68	38bp	178
23	3261672e1e25	0.82	+0.02	0.94	1.09	191bp	54
24	ca29ed426813	0.81	+0.01	0.94	0.23	64bp	109
25	07bb0fb5bcb5	0.81	+0.01	0.94	0.93	162bp	91

QQQ — signal indicators

2010-09-10 → 2026-08-06 · 4,000 bars · buy-and-hold Sharpe 0.94 · luck hurdle 0.47

1,122 of 2,658 measurements were eligible; 840 survive the duplicate collapse. 17 clear all four guards. A search this wide is expected to throw up a 0.47 Sharpe on luck alone.

#	Script	Sharpe	vs B&H	Deflated	Out-of-sample	Breakeven	Trades
1	42f1f172d23d	0.99	+0.06	0.98	0.76	12bp	1003
2	0137de1d5e67	0.98	+0.05	0.98	0.96	307bp	36
3	dba85b1b6369	0.98	+0.04	0.98	0.60	68bp	186
4	524a8b03766d	0.98	+0.04	0.98	0.78	60bp	151
5	2b0ed3671100	0.98	+0.04	0.98	0.93	64bp	152
6	b3a4769d80ea+1 copy	0.97	+0.04	0.98	0.73	295bp	53
7	038808513a6f	0.97	+0.04	0.98	1.02	47bp	217
8	ea1d1a3a95d3	0.97	+0.04	0.98	0.58	171bp	88
9	6028e8d2cdd4	0.97	+0.03	0.98	0.71	62bp	162
10	d62c73561854	0.97	+0.03	0.98	0.93	85bp	115
11	377d8da362c5	0.96	+0.02	0.98	0.98	202bp	52
12	a20d716bc520	0.96	+0.02	0.97	0.87	273bp	39
13	2e353c0677c5	0.95	+0.02	0.97	0.87	38bp	662
14	00ea6298c592	0.95	+0.01	0.97	0.80	115bp	88
15	102043b49f6f	0.95	+0.01	0.97	0.95	110bp	90
16	15b955ec0ddc	0.94	+0.01	0.97	0.92	412bp	34
17	3a89397cd94f+1 copy	0.94	+0.00	0.97	0.64	570bp	20
18	eed9a8509d41	0.94	+0.00	0.97	0.97	435bp	22
19	afe1590d37e8	0.93	-0.00	0.97	1.09	90bp	101
20	21262a431420	0.93	-0.01	0.97	0.64	123bp	142
21	225ddfa5712a	0.93	-0.01	0.96	0.89	48bp	2424
22	695bc710ccc8	0.92	-0.01	0.96	1.10	70bp	152
23	4468a6b5999f	0.92	-0.01	0.96	0.85	87bp	113
24	3261672e1e25	0.92	-0.02	0.96	0.56	311bp	44
25	51f91f8af289	0.92	-0.02	0.96	0.90	25bp	441

§ track C · continuous indicators · predictive power

The median indicator knows less than RSI(14).

2,787 scripts only plot — no orders, no discrete signals — so they are not backtested; their plotted values are tested for whether they know anything about the next day. The statistic is the Spearman rank correlation (information coefficient) between a signal's value and the next day's return. The best plot per script and the direction of the relationship are chosen on the first 70% of history only; everything published comes from the last 30%. Price-scale overlays are measured as their distance from price, and signals that are essentially “time” are rejected as nonstationary.

One more hurdle, and it is the one that matters: most plots collapse to some flavour of price-vs-moving-average, and whatever IC that carries belongs to the market's return structure, not to the script. So every market gets a baseline — the best out-of-sample IC among five canonical signals (SMA20/50 distance, RSI14, 5- and 20-day momentum) — and a script only counts for what it measures above that line.

Scored across ≥4 markets

1,944

of 2,787 Track C scripts

Median excess IC

-0.012

Median script trails the canonical baselines out-of-sample

Beat their baseline

27.3%

530 scripts measured above the line

Clear all four guards

Positive, consistent, robust and above baseline — all at once

#	Script · best signal	Excess IC	OOS IC	Worst mkt	Consistency	Signals tried	Signal quality
1	fbea15d82c25Intraday / Weekly High (vs price)	+0.305	0.339	0.272	6/6	12	100
2	e452484a0aabRSI-Weekly	+0.082	0.113	0.101	6/6	3	100
3	4651171f9885Up Trend 1 (vs price)	+0.049	0.074	0.023	6/6	12	96
4	fae8ceb48187STR Below - ST (vs price)	+0.051	0.083	0.020	6/6	8	95
5	e4ed61638b6bUp Trend (vs price)	+0.027	0.054	0.023	6/6	24	91
6	6c9d16e92142Down Trend (vs price)	+0.021	0.048	0.024	6/6	6	81
7	dd003b2ac3a1Up Trend (vs price)	+0.021	0.048	0.024	6/6	6	81
8	f8022d901e31Down direction (vs price)	+0.021	0.048	0.024	6/6	4	81
9	39319cfec20aVolatility Stop	+0.043	0.074	-0.041	5/6	2	78
10	56790ded32b8Plot 1	+0.043	0.068	-0.047	5/6	13	78
11	71f4156832beVolatility Stop	+0.043	0.076	-0.041	5/6	2	78
12	ba4bbe146a12EMA 20 (W)	+0.042	0.073	-0.063	5/6	4	78
13	d38d6c638ef2EMA 20 (W)	+0.042	0.073	-0.063	5/6	4	78
14	6c5986223a9a405 EMA	+0.039	0.069	-0.052	5/6	23	78
15	e5ccc760188eMonthly trigger	+0.039	0.069	-0.053	5/6	12	78

Baselines per market (signed out-of-sample IC, next-day horizon): BTCUSD mom5 0.020 · ETHUSD mom5 0.025 · SPX500 rsi14 0.033 · QQQ mom5 0.036 · GOLD rsi14 0.026 · EURUSD rsi14 0.042. An excess IC near +0.03 held across four or more markets is a genuinely unusual result at this horizon; most of the corpus sits below zero on that column, which is the honest headline.

§ track D · visual tools · not gradeable, and saying so

1,470 scripts draw. We refuse to invent a score for them.

Support zones, session boxes, dashboards, candle paint — these place no orders, emit no signals and plot no series, so any number attached to them would be an invention. What can be said mechanically is what kind of tool each one is, and whether its display could even be reproduced from historical bars — the repaint surface.

What they are

658levels & zones
315other utility
267tables & dashboards
154labels & annotations
42bar & background coloring
28candle overlays

Repaint surface

473 pull higher-timeframe data with request.security() — 308 of them without an explicit lookahead argument, and 106 with lookahead explicitly on — their historical display shows information that did not exist at the time
24 keep varip state, which updates intrabar and cannot be reproduced from bar data at all
The verdict for the whole track is the same and it is published as such: not gradeable (visual tool) — an honest non-grade, not a low one.

How to read the columns

Sharpe: Annualized from daily equity returns, ×√252. No fee is charged in this column — the breakeven column prices execution instead.
vs buy-and-hold: The same Sharpe minus what simply holding the asset earned over the identical window. On a market that tripled, beating zero is not an achievement.
Deflated: The probability the edge is real rather than the luckiest of a large search. It falls as the search widens, which is the point: with hundreds of candidates in contention, an ordinary Sharpe stops being evidence. Trials are grouped by return correlation first, because a scraped corpus is the same dozen ideas re-uploaded.
Out-of-sample: Sharpe over the last 30% of history, which the author could not have fitted to. The decay against the first 70% is what tells you whether the rules were tuned.
Breakeven: The cost per unit of turnover at which the edge disappears. Spot crypto runs about 10bp all-in, so anything near that is a rounding error rather than an edge.
Guards: Four ticks, left to right: beats buy-and-hold, positive out-of-sample, breakeven ≥ 20bp, deflated Sharpe ≥ 0.95.

What this sweep does not tell you

Every script ran on daily bars with its default inputs. A script written for 15-minute charts, or one whose author expects you to tune its lengths, is being asked a question it was not written to answer. Intraday sweeps are running separately.
Order fills follow TradingView's rule — market orders at the next bar's open. That rule is implemented deliberately, but it has not yet been confirmed against TradingView itself on a reference strategy. Until it is, treat absolute returns as provisional; the ranking depends on relative ordering, which is far less sensitive.
A failing script here is not a worthless script. Most of what this corpus contains was written to be read on a chart by a person, not to be run unattended. Measuring it as an automated strategy is a narrow question, and this page only answers that one.