Rules frozen before their future, judged every month on real candles
A backtest can always be touched up afterwards. Not here: each rule below was frozen, with its fingerprint, before the first candle that judges it. What you get: what well-known rules are worth on a future nobody had seen, and a method you can replay at home.
Where things stand
No verdict yet, and that is expected: a verdict needs candles that did not exist on the day of the freeze. Each entry says when its first judgement becomes readable. Until then, the monthly reading shows the numbers as they come, as information.
The entries
200-day trend on bitcoin
BTCUSDT 1d · frozen 17 November 2026 22:14 UTC · judged from 18 November 2026
Not enough candles yet (0 since the freeze). First readable judgement on 15 August 2029 (week 143).
Rule written in advance (Brock, Lakonishok and LeBaron, 1992), no tuning. On 24 years of unseen US market it cut the maximum drawdown by 2.4 without beating buy-and-hold (16 Nov 2026, inconclusive). Here it meets a future nobody has seen.
What this follow-up can show. Here a “holds” is out of reach, by a rule sealed at the freeze: deflated Sharpe of the history 70.3% (95% required). What stays useful: whether the rule protects on a real future as it did on the past (maximum drawdown), and how far it stands from buy-and-hold.
| History before the freeze | After the freeze | |
|---|---|---|
| Period | 21 February 2024 → 16 November 2026 (1000 candles) | none yet |
| Rule | +14.8 % | — |
| Buy-and-hold | +19.6 % | — |
| Max drawdown of the rule | 31.7 % | — |
| Trades | 12 | — |
| Deflated Sharpe (history) | 70.3 % (1 trial(s) counted) |
Checkpoints fixed at the freeze (free reading): 15 August 2029 (week 143), 12 May 2032 (week 286). Certified reading (30 independent events): first point 5 November 2036.
The frozen rule, its fingerprints and its files
strategy "Tendance 200 jours" version 1.0
asset: BTCUSDT
timeframe: 1d
capital: 10000
entry:
condition: close > SMA(200)
size: 100%
exit:
condition: close < SMA(200)
Fees 10 bp per fill · tv_exact_v1 · earlier trials declared: 0 · CC BY-NC-SA 4.0
rule sha256 933c77bb3017868c298eddd63889bd02dc7a333b3144356ad66499a5c5aa074a
Lab engagement SAF-60f84bf2141c · commitment sha256 60f84bf2141ce0d65b4235f6ee73fa747fef44a00c161c272ce9168987076fce
history candles sha256 44dd59c40d9c975afca862985108f1714f3c080563e5808910f135927379af67
journal line 1
- BTCUSDT_1d.csv 44dd59c40d9c975a…
- commitment.json 60f84bf2141ce0d6…
- regle.arenascript 933c77bb3017868c…
- warmup.json e5d933e0ba106cc7…
Last reading : 17 November 2026, Lab 0.1.0-266 (20261117-266-n1), journal line 2.
200-day trend on ether
ETHUSDT 1d · frozen 17 November 2026 22:14 UTC · judged from 18 November 2026
Not enough candles yet (0 since the freeze). First readable judgement on 1 January 2031 (week 215).
The same rule, unchanged, on a second market.
What this follow-up can show. Here a “holds” is out of reach, by a rule sealed at the freeze: deflated Sharpe of the history 43.3% (95% required); 8 trades on the history (10 required). What stays useful: whether the rule protects on a real future as it did on the past (maximum drawdown), and how far it stands from buy-and-hold.
| History before the freeze | After the freeze | |
|---|---|---|
| Period | 21 February 2024 → 16 November 2026 (1000 candles) | none yet |
| Rule | +3.4 % | — |
| Buy-and-hold | -20.1 % | — |
| Max drawdown of the rule | 41.8 % | — |
| Trades | 8 | — |
| Deflated Sharpe (history) | 43.3 % (2 trial(s) counted) |
Checkpoints fixed at the freeze (free reading): 1 January 2031 (week 215), 14 February 2035 (week 430). Certified reading (30 independent events): first point 5 November 2036.
The frozen rule, its fingerprints and its files
strategy "Tendance 200 jours" version 1.0
asset: ETHUSDT
timeframe: 1d
capital: 10000
entry:
condition: close > SMA(200)
size: 100%
exit:
condition: close < SMA(200)
Fees 10 bp per fill · tv_exact_v1 · earlier trials declared: 0 · CC BY-NC-SA 4.0
rule sha256 e83e89fc3ad79bb430cf6f03fc28a68365ec2749358f211d199d5a1c2929732e
Lab engagement SAF-0d1f281d15a3 · commitment sha256 0d1f281d15a3dd7c57918c946cc6e7f2fdc6abc498612bc6d8cb8cf4abfd7d9f
history candles sha256 781e3e2d0d36305ea824edd7487363bdc9565cc26140e47dc7e3fad256e751ce
journal line 3
- ETHUSDT_1d.csv 781e3e2d0d36305e…
- commitment.json 0d1f281d15a3dd7c…
- regle.arenascript e83e89fc3ad79bb4…
- warmup.json 72ad96253a818503…
Last reading : 17 November 2026, Lab 0.1.0-266 (20261117-266-n1), journal line 4.
Ether above its 56-day average, reviewed every 12 candles
ETHUSDT 1d · frozen 17 November 2026 22:14 UTC · judged from 18 November 2026
Not enough candles yet (0 since the freeze). First readable judgement on 15 November 2028 (week 104).
The closest case to a 'holds' in the 16 Nov 2026 campaign: +67.9% against -45.1% for buy-and-hold over 2.6 set-aside years, but deflated Sharpe 78.7% (95% required). It was picked among 21 settings: those 20 other trials are declared.
What this follow-up can show. Here a “holds” is out of reach, by a rule sealed at the freeze: deflated Sharpe of the history 21.4% (95% required). What stays useful: whether the rule protects on a real future as it did on the past (maximum drawdown), and how far it stands from buy-and-hold.
| History before the freeze | After the freeze | |
|---|---|---|
| Period | 21 February 2024 → 16 November 2026 (1000 candles) | none yet |
| Rule | +84.8 % | — |
| Buy-and-hold | -41.1 % | — |
| Max drawdown of the rule | 39.6 % | — |
| Trades | 40 | — |
| Deflated Sharpe (history) | 21.4 % (23 trial(s) counted) |
Checkpoints fixed at the freeze (free reading): 15 November 2028 (week 104), 13 November 2030 (week 208), 8 November 2034 (week 416). Certified reading (30 independent events): first point 5 November 2036.
The frozen rule, its fingerprints and its files
strategy "Au-dessus de la moyenne 56 jours, revue toutes les 12 bougies" version 1.0
asset: ETHUSDT
timeframe: 1d
capital: 10000
entry:
condition: close > SMA(56)
size: 100%
exit:
timeout: 12 bars
Fees 10 bp per fill · tv_exact_v1 · earlier trials declared: 20 · CC BY-NC-SA 4.0
rule sha256 49077f4adcc146e84d8fc8396a53f5a22be22bca8f91863749cec067b1100f6a
Lab engagement SAF-1fcf2dc67c57 · commitment sha256 1fcf2dc67c57fa7dd1779da87c6aa088b20b804ea5d6cfd4af7dc3a64b55c237
history candles sha256 781e3e2d0d36305ea824edd7487363bdc9565cc26140e47dc7e3fad256e751ce
journal line 5
- ETHUSDT_1d.csv 781e3e2d0d36305e…
- commitment.json 1fcf2dc67c57fa7d…
- regle.arenascript 49077f4adcc146e8…
- warmup.json 72ad96253a818503…
Last reading : 17 November 2026, Lab 0.1.0-266 (20261117-266-n1), journal line 6.
Positive control (synthetic market with a planted edge)
TEMOINPLUS 4h · frozen 17 November 2026 22:14 UTC · judged from 18 November 2026
Not enough candles yet (0 since the freeze). First readable judgement on 10 February 2027 (week 12).
Made-up market, announced as such: after a volume spike, the next 6 candles rise on average. Expected answer: 'holds'. If not, our pipeline is broken, and we will say so.
Expected answer: holds. If the registry says otherwise, the pipeline is at fault and this page will say so.
| History before the freeze | After the freeze | |
|---|---|---|
| Period | 5 July 2025 → 17 November 2026 (3000 candles) | none yet |
| Rule | +938.2 % | — |
| Buy-and-hold | -82.1 % | — |
| Max drawdown of the rule | 8.8 % | — |
| Trades | 118 | — |
| Deflated Sharpe (history) | 100.0 % (1 trial(s) counted) |
Checkpoints fixed at the freeze (free reading): 10 February 2027 (week 12), 5 May 2027 (week 24), 20 October 2027 (week 48). Certified reading (30 independent events): first point 16 June 2027.
The frozen rule, its fingerprints and its files
strategy "Témoin positif : pic de volume, tenir 6 bougies" version 1.0
asset: TEMOINPLUS
timeframe: 4h
capital: 10000
entry:
condition: volume > SMA(volume, 20)
size: 100%
exit:
timeout: 6
signal_exit: off
Fees 10 bp per fill · tv_exact_v1 · earlier trials declared: 0 · CC BY 4.0
rule sha256 1bf520a950009dc58d8f57511113de36874a2138219a3d4a4f386b8ebe6c84f1
Lab engagement SAF-2e5fb4604d51 · commitment sha256 2e5fb4604d51eb2f155094e1e88daf49a6a2eb564e7320df4d56256f0a5f8ae9
history candles sha256 448055774f414f2d48a048b3a3711072b6451993b1474ba16aa2121a033389ec
journal line 7
- TEMOINPLUS_4h_histoire.csv 448055774f414f2d…
- commitment.json 2e5fb4604d51eb2f…
- regle.arenascript 1bf520a950009dc5…
- warmup.json bd0dbc41d03e3684…
Last reading : 17 November 2026, Lab 0.1.0-266 (20261117-266-n1), journal line 8.
Negative control (synthetic pure-noise market)
TEMOINZERO 4h · frozen 17 November 2026 22:14 UTC · judged from 18 November 2026
Not enough candles yet (0 since the freeze). First readable judgement on 16 December 2026 (week 4).
Made-up market with no edge at all, same rule. Expected answer: never 'holds'.
Expected answer: never “holds”. If the registry says otherwise, the pipeline is at fault and this page will say so. Its history already rules out a “holds”: deflated Sharpe of the history 0.9% (95% required); it checks that this door holds.
| History before the freeze | After the freeze | |
|---|---|---|
| Period | 5 July 2025 → 17 November 2026 (3000 candles) | none yet |
| Rule | -80.3 % | — |
| Buy-and-hold | -48.8 % | — |
| Max drawdown of the rule | 85.6 % | — |
| Trades | 410 | — |
| Deflated Sharpe (history) | 0.9 % (2 trial(s) counted) |
Checkpoints fixed at the freeze (free reading): 16 December 2026 (week 4), 13 January 2027 (week 8), 10 March 2027 (week 16). Certified reading (30 independent events): first point 8 August 2029.
The frozen rule, its fingerprints and its files
strategy "Témoin négatif : pic de volume, tenir 6 bougies" version 1.0
asset: TEMOINZERO
timeframe: 4h
capital: 10000
entry:
condition: volume > SMA(volume, 20)
size: 100%
exit:
timeout: 6
signal_exit: off
Fees 10 bp per fill · tv_exact_v1 · earlier trials declared: 0 · CC BY 4.0
rule sha256 5f06f4028a25a6470328bad1627031dee0b5f8940b34a2b602a09cdcbe22ad45
Lab engagement SAF-604c96505004 · commitment sha256 604c9650500417cb9fb98bf00110373144b05cdb962de655781e00e01b924cd9
history candles sha256 5877949b736e84dae316d5b7ee4dcf130e03662ca57c23195ecaea482b2e4457
journal line 9
- TEMOINZERO_4h_histoire.csv 5877949b736e84da…
- commitment.json 604c9650500417cb…
- regle.arenascript 5f06f4028a25a647…
- warmup.json efe389842192d151…
Last reading : 17 November 2026, Lab 0.1.0-266 (20261117-266-n1), journal line 10.
Mirage control (planted edge that stops at the freeze)
TEMOINMIRAGE 4h · frozen 17 November 2026 22:18 UTC · judged from 18 November 2026
Not enough candles yet (0 since the freeze). First readable judgement on 3 March 2027 (week 15).
Made-up market: the same edge as the positive control exists through the whole history, then vanishes exactly when the judged period starts. A great past, nothing after. Expected answer: never 'holds'.
Expected answer: never “holds”. If the registry says otherwise, the pipeline is at fault and this page will say so.
| History before the freeze | After the freeze | |
|---|---|---|
| Period | 5 July 2025 → 17 November 2026 (3000 candles) | none yet |
| Rule | +1950.3 % | — |
| Buy-and-hold | -55.7 % | — |
| Max drawdown of the rule | 8.6 % | — |
| Trades | 127 | — |
| Deflated Sharpe (history) | 100.0 % (3 trial(s) counted) |
Checkpoints fixed at the freeze (free reading): 3 March 2027 (week 15), 16 June 2027 (week 30), 12 January 2028 (week 60). Certified reading (30 independent events): first point 26 May 2027.
The frozen rule, its fingerprints and its files
strategy "Témoin de mirage : pic de volume, tenir 6 bougies" version 1.0
asset: TEMOINMIRAGE
timeframe: 4h
capital: 10000
entry:
condition: volume > SMA(volume, 20)
size: 100%
exit:
timeout: 6
signal_exit: off
Fees 10 bp per fill · tv_exact_v1 · earlier trials declared: 0 · CC BY 4.0
rule sha256 bb79d9cc0cb56a6db40dde5d3ee31139e3449bb5e9e4f2196f998b8b108e8f09
Lab engagement SAF-5504ff42d838 · commitment sha256 5504ff42d838ea51d6086f744f52d0bf90467a25861d2d4cd0df9926d2b0722b
history candles sha256 40bcfba817530f3c007f75745a2799c12e5c1edd45ce596b2ef0df86d61181cb
journal line 11
- TEMOINMIRAGE_4h_histoire.csv 40bcfba817530f3c…
- commitment.json 5504ff42d838ea51…
- regle.arenascript bb79d9cc0cb56a6d…
- warmup.json 1e0aac1e95b46d80…
Last reading : 17 November 2026, Lab 0.1.0-266 (20261117-266-n1), journal line 12.
How a rule is judged
- The judge is the Strategy Arena Lab itself (free desktop app, version 0.1.0-266), with its forward_band tool: at the freeze it seals the script, the setting, the fees, the market, the warm-up candles and the time; the period judged starts at the first candle that opens after the freeze.
- Each month it adds the candles closed since, only if they match the ones already stored, and replays the rule with no position before the freeze.
- 'holds' at a checkpoint: at least 10 trades after the freeze, positive gain above buy-and-hold on the same candles, SPA significant at the checkpoint's threshold, deflated Sharpe of the history >= 95% (sealed at the freeze), at least 10 trades on the history; no past decision changed by later candles. Checkpoints fixed at the freeze: weeks H, 2H, 4H..., threshold halved at each (5% in total, free reading).
- Between two checkpoints, the reading is provisional: the numbers are shown, the verdict waits.
Why three controls
A registry that can only say “no” proves nothing. The three controls are made-up markets, announced as such, whose whole series (past and future up to 2028) is fixed by a published seed and fingerprinted at the freeze: one with a real planted edge (expected: holds), one of pure noise, and one whose edge stops exactly at the freeze, a great past with nothing after (both expected: never holds). If one of them answers otherwise, the fault is ours.
Check it yourself
- The journal: registre.ndjson (12 lines, head 3aef814c2b5ba9b4…). Each line carries the SHA-256 of the previous one: changing or removing a line breaks every following link. verifier_registre.py checks it in one command.
- For AIs and scripts: registre.json.
- To replay an entry without us: download the Lab, import the same candles, and give forward_band the frozen rule; the commitment.json file of each entry lists everything that was sealed.
- External timestamp: the manifest of the freeze (fingerprints of the journal and of every file) is to be anchored with OpenTimestamps; until that is done, the date of the freeze rests on our word and on the Lab’s clock, and this page says “pending”.
What this registry does not say
- It gives no advice and promises no gain. The real rules place no order; they are research on market data.
- The horizons are long: slow rules trade a few times a year, and a verdict needs enough trades. Nobody can make that faster without weakening it.
- Each reading names the Lab version that produced it. A later version may read the same candles; it will be a new line, never a rewrite.
- Real-market candles: Binance, archived daily by strategyarena.io since 16 November 2026; derived files under CC BY-NC-SA 4.0. Controls: CC BY 4.0.
Another follow-up on future candles, older and separate: forward test no. 3 (Dip buyer, sealed 29 September 2026).
Research on past and future market data, never advice. Data generated 17 November 2026.