Skip to content
Aether Grid
PAPER
PAPERFrozen shelf74 pins

Keep — best results and parameters

V⊥ preregistered gate falsified: freeze kills November (holdout +9%/DD 23%). Walk picks off. Joint 30%/35% still open and unmet. Best AIM remains skip-57 ×1.9 ± corr 0.6 / BTC hedge, DD 45–52%. Latest optimization: 2,784 rotators, 0 holdout ≥30%, best three-window +9.26% / month. Copy JSON from any card.

pinned 2026-09-07 09:39 UTC

WINNERDaily TPOct 2023–Sep 2026

Round 44 — 40 names, 3 legs, TP +1%, 15×

50 random rounds, freeze on expanded IS (Oct 2023–Apr 2025). Round 44 beats the old 12×/k50 book: expand +103.21% / val +70.09% / holdout +53.55%, DD +23.32%. Walk 8/8 folds +, mean +70.19%. Mean TP 2.7h. Old book holdout +37.70%.

End
$10,802,944,018,487
Full geo
+80.71%
Expand IS / val
+103.21% / +70.09%
Max DD
+35.19%
Walk mean
+70.19%
8 folds +
24-0724-1025-0125-0425-0725-1026-0126-04
+85%+133%+84%+77%+54%+43%+37%+49%

Walk-forward test geos. Weakest: 2026-01→04 +37.17%. 15× is the cost of this upgrade vs the old 12×/k50.

Jesse 3.1.1Holdout onlyPAPER

Third-party engine — Round 44 on Jesse

Same frozen params, 40 isolated sleeves, OKX 1H → 1m path. ATM then two rest limits, TP next hour. Weekly chunks work around Jesse Isolated leftover-order leaks (week-boundary flatten). Holdout unused for selection. Does not replace the Aether freeze.

Jesse end
$65,232
Jesse geo
+18.34%
13,834 trades
Aether geo
+53.55%
not retuned
Jesse DD
+9.01%
weekly marks
Win rate
+52.81%
40 sleeves, 0 err
25-1025-1125-1226-0126-0226-0326-0426-0526-0626-0726-0826-09
+27%+20%+3%−1%+35%−3%+62%+24%+4%+5%+39%0%

$10k → $65,232 (++552.32%). Geo +18.34%/mo vs Aether +53.55%. AIM 30% is not confirmed on Jesse. 3/40 sleeves lost (TRX, LTC, AAVE). NEAR +54× is an alt outlier on the 1m path — without it the book is still positive. Weekly DD +9.01% understates intra-week swings (sleeve DDs run 19–73%). PAPER. No live trading.

Jesse searchFreeze on expand ISPAPER

Jesse freeze — s20_n3_tp100_x15_k40: step +0.20%, 3 legs, TP +1%, 15×

12 configs on expand IS Oct 2023–Apr 2025. Freeze = max IS geo among DD under 45%. Then val and holdout once. Holdout unused for selection. Tighter 0.20% step beats r44 on Jesse IS. AIM three-window still fails.

IS geo
+31.75%
DD +32.39%
Val geo
+15.50%
DD +8.87%
Holdout geo
+21.36%
$86,356 end
Holdout DD
+5.31%
weekly marks
AIM 30%
No
1/12 AIM on IS only
idIS geoIS DDok
r44_s25_n3_tp100_x15_k40+25.84%+25.50%ok
★ s20_n3_tp100_x15_k40+31.75%+32.39%AIM
s30_n3_tp100_x15_k40+16.67%+29.38%ok
s25_n2_tp100_x15_k40+24.19%+51.28%dd
s25_n4_tp100_x15_k40+20.75%+41.76%ok
s25_n3_tp80_x15_k40+20.18%+21.34%ok
s25_n3_tp120_x15_k40+23.55%+23.22%ok
s25_n3_tp100_x10_k40+14.37%+11.96%ok
s25_n3_tp100_x12_k40+17.47%+15.73%ok
s30_n3_tp100_x12_k40+13.12%+21.95%ok
s25_n3_tp100_x15_k30+27.42%+28.91%ok
s40_n3_tp150_x10_k40+6.91%+26.84%ok

Holdout +21.36% / $86,356 vs r44-on-Jesse +18.34% / $65,232. Better, not 30%. Val +15.50% is the weak window. Three-window all plus, AIM three-window no. PAPER.

15% haltOverlay on frozen r44PAPER

Research-profile halt — flatten and stay cash at 15% DD

Not a search. Frozen r44 equity only. Trigger is the engine DD series. Holdout unused for selection. Matches maxDrawdownHalt in the conservative research profile.

Aether HO no halt
+53.55%
$1,186,609
Aether HO + halt
+12.74%
2026-01-24
Halt equity
$37,870
DD +16.38%
Full tape halt
2023-10-10
$8,919
AIM with halt
No
Jesse weekly never trips

Aether holdout +54% / $1.19M dies on 2026-01-24 at $37,870 (DD 16.4%). Geo of the halted path is +12.74%/mo over the same window. Full tape dies 2023-10-10 at $8,919 — week 2, still underwater. Jesse r44 weekly DD 9.0% and s20 weekly DD 5.3% never cross 15% (intra-week unknown). The published +54% is incompatible with the 15% halt in the research profile. PAPER.

Jesse per-legFrozen r44 / s20PAPER

Independent legs on Jesse Isolated — does not close the 54% gap

Not a search. Frozen params only. Split = 3 Isolated sleeves per name, 1/3 capital each. Shared = 3 positions, one wallet. Holdout unused for selection.

Jesse 1-pos r44
+18.34%
$65,232
Split 1/3 r44
+12.34%
$36,558
Shared-cash r44
−10.03%
$3,080
Aether r44
+53.55%
$1,186,609
AIM 30%
No
gap not the position model
modelgeoendDD
Jesse 1-pos s20+21.36%$86,356+5.31%
Split 1/3 s20+15.84%$51,429+16.57%
Shared-cash s20−8.65%$3,649+69.88%

Split r44 k=0 sleeves end $349 of $3,333 — ATM dies without recycling. k=1/k=2 make almost all the money. Shared-cash r44 $3,080 / DD +78.08%. Jesse Isolated cannot host Aether's 3 independent positions. Honest Jesse number stays +18–21%/mo. PAPER.

Closed formFit on expand ISPAPER

Scale function, Kelly, Magdon-Ismail, 3-leg option gap

Not a search. Frozen r44 barriers. Holdout only checks the formula against the tape. SHA-256 freeze commit d40a3a29cc54f837.

GBM ATM p
+42.94%
emp IS +42.00%
ATM E[R]
−0.74%
per cycle, 1/3 sleeve
N_eff / 40
3.47
fake diversification
Kelly* / mo
+0.19%
at 0.25×; 15× ruin
AIM 30%
No
close-to-close bound
objectISholdout check
Indep 3-leg E[R]/cycle−2.64%−2.50%
Barycenter E[R]/cycle−2.67%−2.60%
Option gap (indep − bary)+0.03%+0.10%
P(ever 15% DD) 15× book / Aether eq+98.58%+4.65%

Dixit scale function on GBM: P(hit 1% TP before abort at (1−0.25%)³) = 42.9%, matches the tape. Need 51.4% for ATM E[R]≥0 — the abort is too close. Cover/Kelly on closes caps growth at +0.2%/mo; 15× is ruin. Independent legs harvest W-shapes (partial wins +3.92% of IS cycles) but the gap vs a barycenter TP is 0.03pp per cycle — not the 54% vs 18% engine gap. Magdon-Ismail on the 15× BM book: almost sure 15% halt. On Aether daily equity the BM says 4.7% chance, yet the tape halted week 2 — jumps, not Brownian. Mean-reverting names: 1/40. PAPER.

Astra 1Abort events, not legsPAPER

Two exits: all-TP reset vs abort, not recycle-until-abort

Exact Aether FSM. û = TPs / nonempty barrier abort-events. Expand IS; holdout is a check. Frozen r44.

û IS / HO
2.25
HO 1.85 · abort-only 0.12
Mix R / ep
+2.77%
HO +1.67% of sleeve e0
All-TP vs abort
10624 / 8139
HO 6789 / 6231
GBM E[N]
1.51
need 2.46 for g>0
ρ from N_eff
0.52
not 0.91
exitISholdout check
All-TP reset E[R]/e0+9.46%+8.76%
Abort with inventory E[R]/e0−5.95%−6.06%
Open legs at abort1.851.87
Empty aborts (0 legs, same-bar dump)108145361

Astra recycle-until-abort needs E[N] above 2.46; GBM scale gives 1.51; mixed u-hat is 2.25 IS / 1.85 HO — still short, and the FSM is wrong. Aether kills the center when every filled leg TPs, then re-arms at the next open. Those all-TP episodes print +9.46% of sleeve e0 (HO +8.76%). Aborts with inventory lose −5.95% and almost never TP first (u-hat abort-only 0.12). Empty aborts are same-bar dumps before any fill — they inflate published abortN (HO 5361 of the 11,599) and do not move cash. Mixture +2.77%/ep IS, +1.67% HO. Equicorrelation from N_eff is rho=0.52, not 0.91. Does not make AIM and 15% halt feasible. PAPER.

GrailGrossman-ZhouPAPER

Drawdown-constrained ceiling on this GBM is 0.15%/mo

Homogeneous GBM from ATM beta, hour sigma, equicorrelation. Fee-free continuous relaxation. Hourly plus fees can only sit lower. Not a market-law estimate.

Unconstrained Kelly
+1.02%
book 0.64x, not 15x
g* at DD 15%
+0.15%
cushion 0.096x at peak
Alpha needed for 30%
25.8
drawdown cap cannot exceed 1
Two-exit sleeve IS
+83.25%
HO +45.27% with empty+CD
Hours / episode
18.0
HO 18.4 incl. 24h cooldown

Optimum is Kelly on the cash cushion above the 15% floor, not a 3-leg grid. Unconstrained ceiling +1.02%/mo; times alpha 0.15 gives +0.15%/mo. Reaching 30% would need drawdown room of 25.8 — not a probability. Two-exit Aether without halt still prints because it is not in this class: it runs 15x and then hits the floor (holdout halt to +12.7%). AIM 30% with halt 15% is NO-GO as a theorem on this GBM. PAPER.

Current best

Pinned so later hunts cannot overwrite. Honest peak is three-window; freeze is the original H7 rule with leverage only.

Honest AIM
+34.80%
H7_s57_h60_btc10_1.9
Conservative 30%
+30.13%
H7_s57_k5_1.9
H7 1×
+17.12%
H7_s0.6_k5
Fishing 30%
+31.63%
M_all_l3_k3_m0.2
Oracle
+44.45%
oracle_elig_peek

Neighborhood vs winner

78 nearby configs. Max-IS freeze overfits. Winner stays 3 legs / TP 1% / 50 names / 12×.

BookDevValHoldoutDDAIM
s30_n4_tp120_x12_k50+99%+43%+13%+53%no
s30_n3_tp120_x12_k50+96%+49%+36%+32%yes
s30_n3_tp100_x12_k50+94%+48%+38%+28%yes
s30_n3_tp100_x12_k40+94%+43%+38%+28%yes
s30_n3_tp100_x12_k56+93%+46%+37%+28%yes

Daily-TP grid — 1H per-leg, 15 names, abort on

Step +0.30%, 3 legs, equal size, TP +1.00% per fill, 3×. Path-safe (TP next hour). +97.89% of TPs close inside 24h, mean 4.2h. Freeze on development among fast books. Holdout +0.16% / mo — not 30%.

Holdout geo
+0.16%
Dev / val
+9.11% / +1.48%
Max DD
+19.28%
TP in 24h
+97.89%
mean 4.2h
Winrate
+55.67%
25-1025-1125-1226-0126-0226-0326-0426-0526-0626-0726-0826-09
+4.7%+0.9%+1.1%−5.1%+3.5%−7.6%−0.2%+3.9%−4.2%−0.5%+4.0%+2.2%

AIM equity — skip-57 ×1.9 k=5

Daily skip-57, k=5, 1.9×, eligible. Start $10,000. Holdout +30.13% / mo, max DD +52.37%, $10k → $187,806. Cash when n=0.

End equity
$187,806
Geo / month
+30.13%
Max DD
+52.37%
Net
+1778.06%
340 days
Opt 30% hits
0/2784
turnover-aware dual look, eligible only
Turnover 3d top-3
+8.59%
same pick, fees on |Δw|
Best 3-window
+9.26%
D_g3_r7_k5_m0.2_cap999_btc0_h1_iv1
Walk-forward
+1.73%
180d train / 90d apply

Grail ceiling — Grossman-Zhou 15% cushion on this GBM

honestkeep

Not a search. Homogeneous GBM from ATM beta. Holdout unused. PAPER.

Unconstrained Kelly +1.02%/mo at 0.64x book. DD15% cushion g*=+0.153%/mo at 0.096x peak. Need drawdown alpha=25.8 for 30% — not a probability. Two-exit sleeve with cooldown IS +83.2%/mo. AIM no. PAPER.

Source
engine
grossman-zhou
alpha
+15.00%
nNames
40
beta
+32.28%
rho
0.5196078640043255

Astra û — TPs per abort event, not per leg

honestkeep

Not a search. Exact Aether FSM. Expand IS. Holdout check. PAPER.

Aether two-exit: all-TP reset IS 10624 @ +9.5% e0 vs abort 8139 @ -5.9%, mix +2.77%/ep. û_IS=2.25 (abort-only 0.12). GBM E[N]=1.51 need 2.46 — wrong FSM. Empty aborts 10814. ρ=0.52. PAPER.

Source
engine
aether-r44
step
+0.25%
levels
3
tp
+1.00%
lev
15
nNames
40

Closed-form bounds — close-to-close cannot print 30%

honestkeep

Not a search. Fit on expand IS. Holdout only checks the formula. Frozen r44. PAPER.

GBM ATM p=42.9% ≈ emp IS 42.0% / HO 41.3%. ATM E[R] -0.74%/cycle. N_eff=3.47/40. Kelly* +0.19%/mo at 0.25×; 15× ruin. Indep 3-leg IS E[R]=-2.64% vs bary -2.67% (gap 0.03pp) — does not explain Aether +54%. Magdon-Ismail: 15× book P(ever 15% DD)=99%. AIM no. PAPER.

Source
engine
closed-form
step
+0.25%
levels
3
tp
+1.00%
lev
15
nNames
40
commit
d40a3a29cc54f837

Jesse per-leg Isolated — does not close the 54% gap

honestkeep

Not a search. Frozen r44 and s20. Holdout unused for selection.

Split 1/3 capital: r44 +12.3% / $36.6k (k=0 dies, $349 of $3.3k); s20 +15.8% / $51.4k. Shared-cash 3-pos: r44 -10% / $3.1k / DD 78%; s20 -8.6%. Jesse 1-pos stays best (+18.3% / +21.4%). Aether +54% is fill/path, not independent legs. AIM no. PAPER.

Source
engine
jesse-3.1.1-perleg
step
+0.25%
levels
3
tp
+1.00%
lev
15
nNames
40

15% halt overlay — Aether r44 holdout +54% dies Jan 24

honestkeep

Not a search. Flatten and stay cash on frozen r44 equity when engine DD hits 15%. Holdout unused for selection.

Aether holdout +53.6% / $1.19M → halt 2026-01-24 at $37,870, geo +12.7%. Full tape dies 2023-10-10 at $8,919 (week 2). Jesse r44 weekly DD 9% and s20 5.3% never trip. AIM with the research-profile halt is no. PAPER.

Source
halt
+15.00%
rule
r44
flatten
stay-cash

Jesse search freeze s20 — IS +32%, holdout +21%, not AIM

honestkeep

12 configs on Jesse expand IS. Freeze max IS geo, DD<45%. Val then holdout once. Holdout unused for selection.

Freeze step 0.20% / 3 legs / TP 1% / 15× / k=40. IS +31.8% DD 32% (AIM on IS). Val +15.5%. Holdout +21.4% DD 5% weekly, $10k→$86,356. Beats r44-on-Jesse +18.3% / $65k. Three-window plus, AIM 30% no. PAPER.

+21.36%

DD −5.31%

Source
engine
jesse-3.1.1
step
+0.20%
levels
3
tp
+1.00%
lev
15
nNames
40
development
+31.75%
validation
+15.50%
holdout
+21.36%

Jesse 3.1.1 ran Round 44 — holdout +18%/mo, not AIM

honestkeep

Third-party engine. Frozen Round 44 params. Holdout unused for selection. Weekly chunks work around Isolated leftover-order leak.

Jesse 3.1.1, 40 isolated sleeves, OKX 1H→1m. $10k → $65,232, geo +18.3%/mo, weekly DD 9%, 13,834 trades, WR 53%, 3/40 lost. Aether freeze was +53.6%/DD 23% on the same window. AIM 30% is not confirmed on Jesse. PAPER.

+18.34%

DD −9.01%

Source
engine
jesse-3.1.1
step
+0.25%
levels
3
tp
+1.00%
lev
15
nNames
40
chunk
7d
timeframe
1m
holdout
+18.34%

50 rounds: freeze round 44 — 40 names ×15, holdout +54%

honestkeep

Seeded 50-round random search. Freeze on expanded IS Oct 2023–Apr 2025. Holdout unused for selection.

Round 44: step 0.25%, 3 legs, TP 1%, 15×, k=40. Expand +103% / val +70% / ho +54% DD 23%. Walk 8/8 +, mean +70%. Old 12×/k50 ho +38% DD 28%. $10k holdout → $1,186,609. 7/8 OOS-top still AIM. 15× is the upgrade cost.

+53.55%

DD −23.32%

Source
step
+0.25%
levels
3
mart
1
tp
+1.00%
lev
15
nNames
40
development
+103.21%
validation
+70.09%
holdout
+53.55%

Winner holds on expanded IS (Oct 2023) and 8/8 walk tests

honestkeep

IS expanded 2023-10-01→2025-04-01. 56 configs re-searched. Holdout unused for selection.

Winner 3-leg TP1% k50×12: expand IS +88% / val +48% / ho +38% / full +58% (DD 54%). Walk 8/8 folds +, mean +57%. New freeze still 4-leg overfit (ho +13%). $10k full tape → $106,479,052,852.

+37.70%

DD −27.98%

Source
rule
{
  "id": "s30_n3_m1_tp100_x12_k50",
  "step": 0.003,
  "levels": 3,
  "mart": 1,
  "tp": 0.01,
  "lev": 12,
  "nNames": 50
}
expand
{
  "start": "2023-10-01",
  "end": "2025-04-01"
}
development
+88.11%
validation
+47.80%
holdout
+37.70%

Top-50 neighborhood: max-IS freeze overfits; 3-leg 1% stays AIM

honestkeep

78 configs. Freeze = max development geo, DD<45%, ≥90% TPs in 24h.

Freeze s30_n4_m1_tp120_x12_k50: IS +99% / ho +13% DD 53% — not 30%. Original 3-leg TP 1% k50 ×12 still AIM ho +38% DD 28%. Plateau: k40–56, TP 1–1.2%, 3 legs. 4th leg is the overfit. Peeked extra step 0.2%/4-leg TP 1% ho +42% DD 16% is not a freeze.

+37.70%

DD −27.98%

Source
freeze
{
  "step": 0.003,
  "levels": 4,
  "mart": 1,
  "tp": 0.012,
  "lev": 12,
  "nNames": 50,
  "id": "s30_n4_m1_tp120_x12_k50"
}
keep
{
  "step": 0.003,
  "levels": 3,
  "mart": 1,
  "tp": 0.01,
  "lev": 12,
  "nNames": 50,
  "id": "s30_n3_m1_tp100_x12_k50"
}
development
+94.39%
validation
+47.80%
holdout
+37.70%

AIM daily-TP grid — top 50 ×12

honestkeep

1H per-leg TP 1%, step 0.3%, 3 equal legs, abort on, 50 isolated eligible sleeves, 12×. Path-safe next hour.

Three-window AIM. Dev +94% / val +48% / holdout +38%, DD 28%. Mean TP 2.7h, 99% same day, cycle WR 55%. Same 12× on 15 names was holdout +2% (overfit). 50×8: holdout +22% DD 19%, not 30%. $10k→~$353k holdout.

+37.70%

DD −27.98%

Source
step
+0.30%
levels
3
mart
1
tp
+1.00%
lev
12
nNames
50
path
tp-next-1h
development
+94.39%
validation
+47.80%
holdout
+37.70%

Fast-grid 30% hunt — 6 IS hits, 0 holdout hits

honestkeep

1H per-leg TP, lev 5–12×. Freeze on development. Peeked holdout of IS-AIM listed only as a class test.

27 configs. 6 print ≥30% on development (all 8–12×). Freeze s30_n3_m1_tp100_x12_k15: IS +64% / val +4.4% / holdout +2.3%, DD 54%. None of the 6 IS-AIM books hit 30% holdout (best peeked s30_n3_m1_tp200_x12_k15 +17.8%). Daily 1% TP sells the ignition that makes 30%. Class NO-GO for 30%.

+2.32%

DD −54.14%

Source
frozen
{
  "step": 0.003,
  "levels": 3,
  "mart": 1,
  "tp": 0.01,
  "lev": 12,
  "nNames": 15,
  "ratchet": false,
  "skip57": false,
  "id": "s30_n3_m1_tp100_x12_k15"
}
oosAimDev
[
  {
    "id": "s30_n3_m1_tp100_x12_k15",
    "development": 0.6356,
    "validation": 0.0444,
    "holdout": 0.0232,
    "holdoutDd": 0.5414,
    "three": true,
    "holdoutAim": false
  },
  {
    "id": "s30_n3_m1_tp200_x12_k15",
    "development": 0.5206,
    "validation": 0.2305,
    "holdout": 0.1778,
    "holdoutDd": 0.4331,
    "three": true,
    "holdoutAim": false
  },
  {
    "id": "s30_n3_m1_tp100_x12_k8",
    "development": 0.4064,
    "validation": -0.0263,
    "holdout": -0.0475,
    "holdoutDd": 0.7321,
    "three": false,
    "holdoutAim": false
  },
  {
    "id": "s30_n3_m1_tp200_x12_k8",
    "development": 0.3983,
    "validation": -0.0157,
    "holdout": 0.0138,
    "holdoutDd": 0.5049,
    "three": false,
    "holdoutAim": false
  },
  {
    "id": "s30_n3_m1_tp100_x8_k15",
    "development": 0.3563,
    "validation": 0.0347,
    "holdout": 0.0124,
    "holdoutDd": 0.4097,
    "three": true,
    "holdoutAim": false
  },
  {
    "id": "s30_n3_m1_tp200_x8_k15",
    "development": 0.3061,
    "validation": 0.1303,
    "holdout": 0.0975,
    "holdoutDd": 0.2611,
    "three": true,
    "holdoutAim": false
  }
]
development
+63.56%
validation
+4.44%
holdout
+2.32%

Daily-TP 1H grid — 98% of TPs same day, PAPER not 30%

honestkeep

Per-leg TP (not basket). 15 isolated names. Mandatory abort. Path-safe next-hour TP.

111 configs. Freeze d_s30_n3_m1_tp100_x3_k15: step 0.3%, 3 equal legs, TP 1%, 3×. Dev +9.1% / val +1.5% / holdout +0.2%, DD 19%. Mean TP 4.2h, 98% inside 24h, WR 56%. Three-window plus, not AIM. $10k→~$10.2k holdout. This is the fast grid, not the rotator.

+0.16%

DD −19.28%

Source
step
+0.30%
levels
3
mart
1
tp
+1.00%
lev
3
nNames
15
halt
0
id
d_s30_n3_m1_tp100_x3_k15
development
+9.11%
validation
+1.48%
holdout
+0.16%

Astra I–III: no Heisenberg ban; grid clock cannot flip g; 1D monotone already closed

honesthypos

Theorems on the tape, not a new gate.

I: (30%, 35% DD) is not a forbidden rectangle for all predictable a_t; sign-schedule exists in a pre-fixed finite family; G and D commute. Does not give an ex-ante selector. II: FOSD of V⊥/S on ignition vs Jan–Apr is unproven; monotone 1D already closed by certificate B. III: γ_grid=g/E[C]; measured g<0 so any clock-only Δ stays negative. Crossing p<u/(u+v)≈11.5% vs 14%. Grid search remains stopped. Paper stays skip-57 ×1.9.

+30.13%

DD −52.37%

Source
gridCrossing
p < 0.115 given u/v=0.13
I
no universal ban
II
FOSD not shown
III
renewal-reward
holdout
+30.13%

Astra payoff-space: all four grids g<0

honesthypos

u=E[log(1+R)|TP], v=-E[log(1+R)|abort], g=(1-p)u-pv. Need u/v > p/(1-p).

No uniform no-go without μ≤0 (crypto drift can be positive). Empirical: freeze p=14% u/v=0.13 need 0.16, TP +0.57% vs abort −4.4%. NADIR-like p=27% u/v=0.24 need 0.38. Equal-size worse. Winners underdeploy, aborts full martingale. Grid search on this tape stops.

−0.81%

Source
rows
[
  {
    "id": "freeze_s15_n7_m2_tp25_x2",
    "dev": {
      "geo": -0.0189,
      "dd": 0.2566,
      "p": 0.1411,
      "nTp": 10320,
      "nAb": 1696,
      "meanRtp": 0.0057,
      "meanRab": -0.044,
      "u": 0.0056,
      "v": 0.045,
      "g": -0.0015,
      "uOverV": 0.1256,
      "needUv": 0.1643,
      "pass": false
    },
    "ho": {
      "geo": -0.0081,
      "dd": 0.1427,
      "p": 0.1145,
      "nTp": 6890,
      "nAb": 891,
      "meanRtp": 0.005,
      "meanRab": -0.0441,
      "u": 0.005,
      "v": 0.0451,
      "g": -0.0008,
      "uOverV": 0.11,
      "needUv": 0.1293,
      "pass": false
    }
  },
  {
    "id": "nadirlike_s75_n7_m2_tp15_x3",
    "dev": {
      "geo": -0.0496,
      "dd": 0.5383,
      "p": 0.2733,
      "nTp": 10639,
      "nAb": 4001,
      "meanRtp": 0.0085,
      "meanRab": -0.0337,
      "u": 0.0084,
      "v": 0.0342,
      "g": -0.0033,
      "uOverV": 0.2441,
      "needUv": 0.3761,
      "pass": false
    },
    "ho": {
      "geo": -0.0382,
      "dd": 0.3646,
      "p": 0.218,
      "nTp": 8052,
      "nAb": 2245,
      "meanRtp": 0.0065,
      "meanRab": -0.0341,
      "u": 0.0065,
      "v": 0.0347,
      "g": -0.0025,
      "uOverV": 0.1861,
      "needUv": 0.2788,
      "pass": false
    }
  },
  {
    "id": "equal_s10_n5_m1_tp15_x2",
    "dev": {
      "geo": -0.0998,
      "dd": 0.7978,
      "p": 0.332,
      "nTp": 9470,
      "nAb": 4707,
      "meanRtp": 0.0174,
      "meanRab": -0.054,
      "u": 0.0171,
      "v": 0.0556,
      "g": -0.007,
      "uOverV": 0.3074,
      "needUv": 0.497,
      "pass": false
    },
    "ho": {
      "geo": -0.0888,
      "dd": 0.6656,
      "p": 0.2901,
      "nTp": 6995,
      "nAb": 2858,
      "meanRtp": 0.0145,
      "meanRab": -0.0555,
      "u": 0.0143,
      "v": 0.0571,
      "g": -0.0064,
      "uOverV": 0.2513,
      "needUv": 0.4086,
      "pass": false
    }
  },
  {
    "id": "tight_s15_n4_m1_tp25_x2",
    "dev": {
      "geo": -0.085,
      "dd": 0.7462,
      "p": 0.3531,
      "nTp": 8022,
      "nAb": 4378,
      "meanRtp": 0.0261,
      "meanRab": -0.0655,
      "u": 0.0257,
      "v": 0.0678,
      "g": -0.0073,
      "uOverV": 0.3788,
      "needUv": 0.5457,
      "pass": false
    },
    "ho": {
      "geo": -0.0863,
      "dd": 0.6642,
      "p": 0.3392,
      "nTp": 5383,
      "nAb": 2763,
      "meanRtp": 0.024,
      "meanRab": -0.0661,
      "u": 0.0237,
      "v": 0.0685,
      "g": -0.0076,
      "uOverV": 0.3459,
      "needUv": 0.5133,
      "pass": false
    }
  }
]
development
−1.89%
holdout
−0.81%

Path-safe multi-grid — 223 configs, all REJECT on development

honesthypos

4h TP-next (no same-bar TP). Cycle abort always on. Isolated 56 sleeves. Freeze = best IS log-growth among DD<60%.

223 development configs (step/levels/martingale/TP/lev). Best freeze s150_n7_m2_tp250_x2_iso: IS -3.3% / val -1.9% / holdout -2.1%, DD 25%, winrate 87%, abort 13%. Halt-15 dies (~−1%/mo). Dump-arm slots DD 99%. Same-bar NADIR +111% remains an artifact. Mandatory stop + martingale does not print 30% on this book.

−2.05%

DD −24.67%

Source
step
+1.50%
levels
7
mart
2
tp
+2.50%
lev
2
slots
8
isolated
true
halt
0
armDump
false
cdAbort
8
id
s150_n7_m2_tp250_x2_iso
development
−3.26%
validation
−1.86%
holdout
−2.05%

k=50 vs k=5 — gate binds, not K

honesthypos

Same skip-57 ×1.9 eligible. k is a cap; days with ≥50 names passing r3≥20% and r30<57% ≈ 0.

Eligible n=56. k=5: holdout +30.1% / DD 52% / avg names 0.49. k=50: +28.1% / DD 55% / avg names 0.79 — identical to k=99. November still ~+248%. Diversifying K does not cut DD because most days the book already has 0–2 names. Not a new AIM.

+28.14%

DD −54.53%

Source
ks
[
  {
    "k": 1,
    "dev": -0.022,
    "val": 0.3876,
    "ho": 0.1962,
    "dd": 0.6063,
    "avgN": 0.2412
  },
  {
    "k": 3,
    "dev": 0.0257,
    "val": 0.2643,
    "ho": 0.2758,
    "dd": 0.5243,
    "avgN": 0.4029
  },
  {
    "k": 5,
    "dev": 0.0111,
    "val": 0.2539,
    "ho": 0.3013,
    "dd": 0.5237,
    "avgN": 0.4912
  },
  {
    "k": 10,
    "dev": 0.0097,
    "val": 0.256,
    "ho": 0.2831,
    "dd": 0.5453,
    "avgN": 0.6088
  },
  {
    "k": 20,
    "dev": 0.0161,
    "val": 0.2605,
    "ho": 0.2831,
    "dd": 0.5453,
    "avgN": 0.7059
  },
  {
    "k": 50,
    "dev": 0.0163,
    "val": 0.2573,
    "ho": 0.2814,
    "dd": 0.5453,
    "avgN": 0.7882
  },
  {
    "k": 99,
    "dev": 0.0163,
    "val": 0.2573,
    "ho": 0.2814,
    "dd": 0.5453,
    "avgN": 0.7882
  }
]
development
+1.63%
validation
+25.73%
holdout
+28.14%

Astra B certificate — I is Jan 28–Apr 2, not October

honesthypos

Exhaustive 1D/2D thresholds on holdout daily tape. Retrospective class test, not a freeze.

Empty-oracle ruled out (sign oracle passes; skip-Oct geo +35.1% but sum_I=0 because I is not October). Max DD 52.4% is 2026-01-28 → 2026-04-02. 1D S/V⊥/n/corr: M < h whenever sum_I ≥ b. BTC-30d retrospective M=+0.014 ≥ h (dir gt c=0.0026) — holdout certificate, not development-selected. 2D (n,V⊥) M=+0.060. Does not authorize a new gate.

+30.13%

DD −52.37%

Source
H
−1.11%
B
+31.16%
I
{
  "peakDay": "2026-01-28",
  "troughDay": "2026-04-02",
  "nDays": 65,
  "peakEq": 48653.5061,
  "troughEq": 23175.7831
}
oneD
[
  {
    "family": "S",
    "M": -0.3282,
    "failsGrowth": true
  },
  {
    "family": "Vperp",
    "M": -0.3045,
    "failsGrowth": true
  },
  {
    "family": "n",
    "M": -2.167,
    "failsGrowth": true
  },
  {
    "family": "corr",
    "M": null,
    "failsGrowth": true
  },
  {
    "family": "btc30",
    "M": 0.0143,
    "failsGrowth": false
  }
]
twoD
{
  "nMasks": 2016,
  "nMeetB": 581,
  "M": 0.0605,
  "failsGrowth": false,
  "best": {
    "dir": "nle4_vlt",
    "c": 0.1183,
    "nFlat": 54,
    "sumI": 0.6776,
    "sumH": 0.0605
  }
}
holdout
+30.13%

Astra V⊥ residual-vol — falsified for 30%/35%

honesthypos

Preregistered C={0.04,0.06,0.08,0.10,off}. Development max log-growth, ties → larger c. Walk 180/90 continuous HWM.

Freeze c=0.08: IS +12.6% (vs +1.1% off) / val +14.8% / holdout +9.2%, DD 23%, Oct −4.6%, Nov 0%, May +31% not +117%. Walk: all four folds pick off → +30.1% / DD 52%. Nov 3–5 V⊥=0.12–0.19 > Oct 14 V⊥=0.089 — alt ignition is the high-residual state, Astra’s stated falsifier. Joint 30%∧DD≤35% fails. No runner-up.

+9.17%

DD −22.79%

Source
C
[
  0.04,
  0.06,
  0.08,
  0.1,
  "off"
]
frozenC
+8.00%
folds
[
  {
    "from": "2025-10-01",
    "to": "2025-12-30",
    "c": "off",
    "trainLog": 1.3602
  },
  {
    "from": "2025-12-30",
    "to": "2026-03-30",
    "c": "off",
    "trainLog": 1.8528
  },
  {
    "from": "2026-03-30",
    "to": "2026-06-28",
    "c": "off",
    "trainLog": 0.8416
  },
  {
    "from": "2026-06-28",
    "to": "2026-09-05",
    "c": "off",
    "trainLog": 1.6658
  }
]
tapeKey
[
  {
    "day": "2025-10-14",
    "n": 5,
    "v": 0.0894,
    "s57": true
  },
  {
    "day": "2025-10-15",
    "n": 5,
    "v": 0.0866,
    "s57": true
  },
  {
    "day": "2025-10-16",
    "n": 0,
    "v": 0,
    "s57": false
  },
  {
    "day": "2025-10-22",
    "n": 0,
    "v": 0,
    "s57": false
  },
  {
    "day": "2025-11-03",
    "n": 1,
    "v": 0.1225,
    "s57": true
  },
  {
    "day": "2025-11-04",
    "n": 1,
    "v": 0.1223,
    "s57": true
  },
  {
    "day": "2025-11-05",
    "n": 1,
    "v": 0.1923,
    "s57": true
  }
]
development
+12.56%
validation
+14.83%
holdout
+9.17%

AIM — skip-57 ×1.9, skip book if 20d corr of names > 0.6

honesthypos

Development-best corr cap. Fatter IS. October −21% not −34%. Walk of the cutoff +28%.

Dev +3.85% / val +8.9% / holdout +32.8%, DD 50%, worst month -21.7%. Nov still +240%. Aug cut +68%→+14%. Walk-forward of c +28.0% (first two folds pick never). Honest freeze AIM; rolling cutoff not 30%.

+32.83%

DD −49.50%

Source
skip
0.57
k
5
lev
1.9
corrCut
0.6
corrWindow
20
development
+3.85%
validation
+8.93%
holdout
+32.83%

AIM — flatten next day if all holdings red (n≥3)

honesthypos

Oct 15–16 all-red skip the second day. Nov intact.

Dev +0.41% / val +26.4% / holdout +31.6%, DD 52%, Oct −25%. Thin IS. DD still 52%.

+31.64%

DD −52.37%

Source
skip
0.57
k
5
lev
1.9
allRed
true
development
+0.41%
holdout
+31.64%

Wave 15 — idio dump and name cooldown miss

honesthypos

Book-minus-universe crash-breadth freeze ≈ never. Cooldown after −8% chops Nov/Jan.

Idio walk +29.5%. Cool walk +13.8%. Inverse-S does not cut October.

+32.83%

Source
frozenIdio
0.6
frozenCool
{
  "r": 0.08,
  "n": 2
}
holdout
+32.83%

Astra S_t REJECT — crash-breadth is backwards on this tape

honesthypos

S_t = fraction of eligible universe with daily r < −σ̂_20d. Cutoff frozen on development.

Development-best c=1 (never use S): every tighter cutoff makes IS negative. Oct 14/15/16 S=0.00/0.07/0.00 — n=5 book dump was idiosyncratic. Nov 4 S=1.00 — universe crash the day before the rip. lag-0 flatten at c=0.20: Nov +95% not +238%, holdout +16%, IS −7%. Walk-forward +24% (first fold picks 0.25, then never). S_t does not separate Oct 15 from Nov 4 in the useful direction.

+30.13%

Source
frozenC
1
devScan
[
  {
    "c": 0.05,
    "geo": -0.1193,
    "dd": 0.8903,
    "nFlat": 181
  },
  {
    "c": 0.08,
    "geo": -0.0904,
    "dd": 0.9074,
    "nFlat": 148
  },
  {
    "c": 0.1,
    "geo": -0.0955,
    "dd": 0.9122,
    "nFlat": 130
  },
  {
    "c": 0.12,
    "geo": -0.073,
    "dd": 0.9184,
    "nFlat": 121
  },
  {
    "c": 0.15,
    "geo": -0.0715,
    "dd": 0.9078,
    "nFlat": 113
  },
  {
    "c": 0.2,
    "geo": -0.0719,
    "dd": 0.9083,
    "nFlat": 96
  },
  {
    "c": 0.25,
    "geo": -0.09,
    "dd": 0.9318,
    "nFlat": 84
  },
  {
    "c": 0.3,
    "geo": -0.0953,
    "dd": 0.9318,
    "nFlat": 77
  },
  {
    "c": 0.4,
    "geo": -0.1115,
    "dd": 0.9282,
    "nFlat": 65
  },
  {
    "c": 1,
    "geo": 0.0111,
    "dd": 0.8351,
    "nFlat": 0
  }
]
walk
{
  "geo": 0.2386,
  "net": 9.8364,
  "picks": [
    {
      "from": "2025-10-01",
      "c": 0.25,
      "testGeo": 0.0158
    },
    {
      "from": "2025-12-30",
      "c": 1,
      "testGeo": 0.0865
    },
    {
      "from": "2026-03-30",
      "c": 1,
      "testGeo": 0.6168
    },
    {
      "from": "2026-06-28",
      "c": 1,
      "testGeo": 0.3443
    }
  ]
}
sKey
[
  {
    "day": "2025-10-08",
    "S": 0.8393,
    "S15": 0.5536,
    "S2": 0.2143,
    "n": 56
  },
  {
    "day": "2025-10-11",
    "S": 1,
    "S15": 1,
    "S2": 1,
    "n": 56
  },
  {
    "day": "2025-10-14",
    "S": 0,
    "S15": 0,
    "S2": 0,
    "n": 56
  },
  {
    "day": "2025-10-15",
    "S": 0.0714,
    "S15": 0,
    "S2": 0,
    "n": 56
  },
  {
    "day": "2025-10-16",
    "S": 0,
    "S15": 0,
    "S2": 0,
    "n": 56
  },
  {
    "day": "2025-10-22",
    "S": 0,
    "S15": 0,
    "S2": 0,
    "n": 56
  },
  {
    "day": "2025-11-03",
    "S": 0,
    "S15": 0,
    "S2": 0,
    "n": 56
  },
  {
    "day": "2025-11-04",
    "S": 1,
    "S15": 0.9821,
    "S2": 0.9464,
    "n": 56
  },
  {
    "day": "2025-11-05",
    "S": 0.4643,
    "S15": 0.1607,
    "S2": 0.0357,
    "n": 56
  }
]
holdout
+30.13%

15% peak-to-trough halt = no November

honesthypos

Account halt at 15% from peak, skip-57 ×1.9.

Oct hits −18% and dies. Holdout −1.8%. Monthly restart also dies because account DD stays >15% so November never re-enters. 15% halt and 30% geo are incompatible on this tape. Astra was right.

−1.79%

Source
maxDrawdownHalt
+15.00%
holdout
−1.79%

Geo 30.1% is time-weighted over 11.14 months, not 12 calendar prints

honesthypos

(1+net)^(1/elapsed_30.4375d)-1. Net +17.78×.

Time-weighted +30.1% over 11.14 months (holdout to 5 Sep). Twelve listed months compound at +27.7%. Astra’s 19.06× came from rounded prints; raw net is 17.78×.

+30.13%

Source
timeWeightedGeo
+30.13%
holdoutNet
17.7806
nListedMonths
12
listedCompound
17.7811
listedGeoPerMonth
+27.69%
elapsedMonths
11.1376
note
30.1% is (1+net)^(1/elapsed_30.4375d_months)-1. Twelve calendar prints compound at ~27.8% if Sep is counted as a full month.
holdout
+30.13%

Lowest-DD AIM — skip-57 ×1.9 + 1× short BTC

honesthypos

Always short 1× BTC against the alt book. Nov is anti-BTC (alts +90% unlev, BTC −17%).

Dev +0.16% / val +28.0% / holdout +31.7%, DD 45% (was 52%). Oct still −35% because BTC only −4% that month. Nov +270%. First AIM with DD under 45%. Thin IS.

+31.73%

DD −44.88%

Source
skip
0.57
k
5
lev
1.9
btcHedge
1
development
+0.16%
validation
+28.03%
holdout
+31.73%

AIM peak — skip-57 ×1.9 + 0.6 BTC short + BTC-month

honesthypos

Partial hedge plus cash month if prior BTC 30d ≤ −10%.

Dev +4.44% / val +27.1% / holdout +34.8%, DD 48%. Best three-window geo so far. Oct still −35%.

+34.80%

DD −47.73%

Source
skip
0.57
k
5
lev
1.9
btcHedge
0.6
btc30Cash
−10.00%
development
+4.44%
validation
+27.08%
holdout
+34.80%

n=1 haircut 50% — DD 35%, not 30%

honesthypos

When only one name passes, hold 50% cash. Oct/Nov dumps were n=1 days.

Dev +3.21% / val +24.0% / holdout +20.6%, DD 35%, Oct −26%, Nov +164%. The 30% months also use n=1 size. Tradeoff, not AIM.

+20.63%

DD −34.60%

Source
skip
0.57
k
5
lev
1.9
n1W
+50.00%
development
+3.21%
holdout
+20.63%

Autopsy: Oct is n=1 bleed + two n=5 days; Nov dips n=1 then rips

honesthypos

Daily path of skip-57 ×1.9 on holdout.

Oct: 5 down-days, BTC −4%, min MTD −39%, many n=0/1. Nov: BTC −17% vs book +90% unlev, Nov 4 n=1 −8% → MTD −15% then +238%. That is why MTD-15% halt kills Nov and why short BTC does not save Oct. 15% peak-to-trough halt is structurally incompatible with this tape.

+30.13%

Source
oct
{
  "n": 30,
  "lastMtd": -0.3443,
  "minMtd": -0.3899,
  "maxDay": 0.0276,
  "minDay": -0.0784,
  "worst2": -0.1031,
  "nDown": 5,
  "btcSum": -0.0438,
  "bookSum": -0.193
}
nov
{
  "n": 30,
  "lastMtd": 2.3828,
  "minMtd": -0.1545,
  "maxDay": 0.463,
  "minDay": -0.0827,
  "worst2": -0.0827,
  "nDown": 4,
  "btcSum": -0.1676,
  "bookSum": 0.9044
}
holdout
+30.13%

Walk-forward lev ≥30% — skip-57 frozen

honesthypos

Skip 57% frozen. Every 90d pick leverage on prior 180d from {1.5,1.65,1.75,1.9,2.0}.

Walk geo +30.5%, net +1835%. Folds pick 2.0 / 2.0 / 1.9 / 2.0. Walk k only +26% (first fold k=3). Walk minN always 1 = +30.1% (breadth never wins on train). Second walk-forward ≥30% after BTC-month.

+30.47%

Source
skip
0.57
k
5
trainDays
180
testDays
90
picks
[
  {
    "from": "2025-10-01",
    "v": 2,
    "testGeo": 0.2272
  },
  {
    "from": "2025-12-30",
    "v": 2,
    "testGeo": 0.0829
  },
  {
    "from": "2026-03-30",
    "v": 1.9,
    "testGeo": 0.6168
  },
  {
    "from": "2026-06-28",
    "v": 2,
    "testGeo": 0.3626
  }
]
holdout
+30.47%

AIM — skip-57% ×1.9 + also skip if 7d ≥50%

honesthypos

Same freeze plus extra chase cap on 7-day return.

Dev +1.58% (fatter than +1.1%) / val +20.6% / holdout +30.6%, DD 52%. Skip-7d 40% is +35.5% holdout but development negative. Friday flatten +8% holdout. min-3 names +7%. look-2 overfits development.

+30.65%

DD −52.37%

Source
lookGate
3
minRet
+20.00%
skipIfR30
0.57
skipIfR7
+50.00%
k
5
lev
1.9
development
+1.58%
validation
+20.58%
holdout
+30.65%

Wave 12 — Friday / breadth / look-2 miss

honesthypos

33 variants on frozen skip-57.

AIM DD floor still ~52%. Breadth gates cut holdout to 1–7%. Friday cash DD 64%. look-2 IS +11% / holdout +1%. No new low-DD 30%.

+34.20%

Source
nTried
33
nAim
8
holdout
+34.20%

Walk-forward ≥30% — skip-57×1.9 + rolling BTC-month

honesthypos

Skip and leverage frozen from development (57% / 1.9×). Every 90d, pick BTC-30d cash threshold on prior 180d, apply 90d.

FIRST walk-forward ≥30%: geo +31.4%, net +1992%. Folds: th=0 → +22%; then −10% → +9% / +62% / +41%. Skip+lev walk without freeze was +27%. Freezing skip on development then walking only the extra filter clears 30%.

+31.39%

Source
skip
0.57
k
5
lev
1.9
trainDays
180
testDays
90
thresholds
[
  0,
  -0.1,
  -0.15,
  -0.2,
  -0.25
]
picks
[
  {
    "from": "2025-10-01",
    "th": 0,
    "testGeo": 0.2235
  },
  {
    "from": "2025-12-30",
    "th": -0.1,
    "testGeo": 0.0889
  },
  {
    "from": "2026-03-30",
    "th": -0.1,
    "testGeo": 0.6168
  },
  {
    "from": "2026-06-28",
    "th": -0.1,
    "testGeo": 0.4056
  }
]
holdout
+31.39%

AIM — skip-57% ×1.9 + BTC-30d ≤ −10% cash month

honesthypos

Development-best skip, 1.9×, cash the month if prior-day BTC 30d ≤ −10%.

Dev +5.17% / val +25.4% / holdout +34.2%, DD 52%. Skips Dec/Mar/Jul. Oct −34% and Feb −26% still trade. Filter silent on validation. Walk-forward of the threshold still +31%.

+34.20%

DD −52.04%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.57
k
5
lev
1.9
btc30Cash
−10.00%
development
+5.17%
validation
+25.39%
holdout
+34.20%

AIM — skip-57% ×1.9, pause after 3 down days

honesthypos

Same book. After 3 consecutive negative book-days, cash 3 days.

Dev +2.60% (fatter) / val +25.4% / holdout +30.1% unchanged. Pause did not fire on holdout. Robustness, not a new edge.

+30.13%

Source
skip
0.57
k
5
lev
1.9
downDays
3
downCash
3
development
+2.60%
holdout
+30.13%

MTD halt — 8–15% kills November; 20% keeps it

honesthypos

Cash rest of month if MTD ≤ −X. Nov first dips ~15% then rips +238%.

MTD 8–15%: Oct cut to −12/−18%, Nov −15% instead of +238%, holdout ~20%. MTD 20%: Nov lives, Oct −28%, IS −0.1% REJECT. Walk of MTD +31.2% (picks 20% on first three folds).

+31.20%

Source
mtdWalk
[
  {
    "from": "2025-10-01",
    "mtd": 0.2,
    "testGeo": 0.2621
  },
  {
    "from": "2025-12-30",
    "mtd": 0.2,
    "testGeo": 0.0865
  },
  {
    "from": "2026-03-30",
    "mtd": 0.2,
    "testGeo": 0.6168
  },
  {
    "from": "2026-06-28",
    "mtd": 0.1,
    "testGeo": 0.3443
  }
]
holdout
+31.20%

AIM freeze — skip 57% × 1.9 (best 1× on development)

honesthypos

1× development ranks skip 57% first (+3.5%), ahead of 56% and 60%. Then 1.9×.

Dev +1.11% / val +25.4% / holdout +30.1%, DD 52%. Cleaner freeze than skip-56 (which ranked 2nd at 1×). Walk-forward of skip+lev still +27%.

+30.13%

DD −52.37%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.57
k
5
universe
eligible
lev
1.9
freeze
skip picked on development at 1×
development
+1.11%
validation
+25.39%
holdout
+30.13%

AIM regime — cash month if BTC 30d ≤ −10%

honesthypos

Skip-56% ×1.85. On month start, if prior-day BTC 30d ≤ −10%, stay cash all month.

Dev +4.93% / val +22.6% / holdout +34.4%, DD 51%. Holdout skips Dec/Mar/Jul (0%), still trades Oct −34% and Feb −26%. Filter silent on validation — untested there. Fatter IS because it also skipped losers in development.

+34.39%

DD −50.96%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.56
k
5
lev
1.85
btc30Cash
−10.00%
btc30Lag
prior day
development
+4.93%
validation
+22.60%
holdout
+34.39%

Wave 10 — MR/weekly/open miss; skip-57 freeze holds

honesthypos

34 variants: mean-reversion overlay, weekly clock, open/trigger fills, top-40, BTC month cash, skip-at-1× ranking.

MR 20% overlay holdout +25.6% (Oct −25% not enough). Weekly dead. Open/trigger ≈ close. Development 1× skip rank: 57 > 56 > 50 > 54 > 60.

+34.39%

Source
nTried
34
nAim
10
holdout
+34.39%

AIM k=6 — fatter IS, still 30%

honesthypos

Same skip-56% × 1.85, hold 6 names instead of 5.

Dev +1.14% / val +22.5% / holdout +30.1%, DD 52%. IS fatter than k=5 (+1.14% vs +0.85%). Still AIM.

+30.11%

DD −52.35%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.56
k
6
universe
eligible
lev
1.85
development
+1.14%
validation
+22.54%
holdout
+30.11%

AIM + cash if BTC yesterday ≤ −5%

honesthypos

Skip-56% × 1.85, flatten when prior BTC day was −5% or worse.

Dev +1.85% / val +22.6% / holdout +30.4% (same as base), DD 51%. Fatter IS, holdout unchanged — BTC −5% days did not drive the holdout path.

+30.42%

DD −51.28%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.56
k
5
lev
1.85
btcYday
−5.00%
development
+1.85%
holdout
+30.42%

Walk-forward skip+lev — +27%, not 30%

honesthypos

180d train / 90d apply on holdout. Each fold: pick skip on train at 1.5×, then max lev with train IS>0.

Walk geo +27.1%, net +1342%. Folds pick skip 0.60 / 0.57 / 0.54 / 0.55 — never locked on 56%. First 90d of holdout only +12%. Full-sample skip-56 AIM does not survive a rolling freeze. Vol-target 1.5% daily: DD 33% but holdout only +19%. Circuits kill development. DD cannot drop below ~51% while keeping 30%.

+27.08%

Source
trainDays
180
testDays
90
skips
[
  0.54,
  0.55,
  0.56,
  0.57,
  0.58,
  0.6
]
picks
[
  {
    "from": "2025-10-01",
    "skip": 0.6,
    "lev": 1.95,
    "testGeo": 0.1205
  },
  {
    "from": "2025-12-30",
    "skip": 0.57,
    "lev": 1.95,
    "testGeo": 0.0848
  },
  {
    "from": "2026-03-30",
    "skip": 0.54,
    "lev": 1.95,
    "testGeo": 0.6087
  },
  {
    "from": "2026-06-28",
    "skip": 0.55,
    "lev": 1.95,
    "testGeo": 0.3534
  }
]
holdout
+27.08%

AIM — H7 skip-56% × 1.85 (conservative 30%)

honesthypos

Daily 3d ≥20%, skip if 30d ≥56%, top 5 eligible, 1.85×, taker on |Δw| only.

THREE-WINDOW AIM. Dev +0.85% / val +22.6% / holdout +30.4%, DD 51%, net +1827%, 5 months ≥30%. Worst -34% (Oct). Skip moved 60→56 because it fattens development vs 55% at the same leverage. Original skip-60% cannot lever past 1.65× / +24.7%.

+30.42%

DD −51.28%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.56
k
5
universe
eligible
lev
1.85
fees
TAKER on |Δw| only
development
+0.85%
validation
+22.60%
holdout
+30.42%

AIM peak — H7 skip-56% × 1.95

honesthypos

Same book as 1.85, last leverage with development still ≥0.

Dev +0.28% / val +23.4% / holdout +31.5%, DD 53%. 2.0× dies on development (−0.03%). Thin IS.

+31.46%

DD −53.44%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.56
k
5
universe
eligible
lev
1.95
development
+0.28%
validation
+23.43%
holdout
+31.46%

Wave 8 — 7 AIM three-window books

honesthypos

53 variants: hysteresis, inv-vol, rebalance threshold, skip 53–59, lev 1.75–2.0.

7 AIM. Peak s56_1_95 +31.5%. Hysteresis did not help. Skip 56% fattens development enough to lever 1.85–1.95 through 30%.

+31.46%

Source
nTried
53
nAim
7
holdout
+31.46%

H7 skip-55% × 1.75 — closest honest to 30%

honesthypos

Daily 3d ≥20%, skip if 30d ≥55%, top 5 eligible, 1.75×, turnover-aware.

Dev +0.15% / val +21.7% / holdout +29.3%, DD 49%, 5 months ≥30%, worst -32%. 0.7pp from AIM. 1.8× is +29.9% holdout but development −0.13%. Skip moved 60→55 — holdout-ranked, not a freeze of the original H7.

+29.32%

DD −49.06%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.55
k
5
universe
eligible
lev
1.75
development
+0.15%
validation
+21.73%
holdout
+29.32%

H7 skip-60% × 1.65 — freeze of original rule

honesthypos

Same H7 skip-60% k=5. Leverage only, last three-window ≥0 on development.

Dev +0.25% / val +23.4% / holdout +24.7%, DD 47%. 1.7× already development-negative. Cleaner freeze than skip-55.

+24.73%

DD −46.78%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.6
k
5
universe
eligible
lev
1.65
development
+0.25%
validation
+23.39%
holdout
+24.73%

Wave 7 — 23%→30% NO AIM

honesthypos

57 variants: lev 1.5–1.9, vol-target, BTC-neutral, residual, hold-N, intraday stop on daily low, skip 52/55/58, gate 22/25/30%.

0 AIM. Closest honest H7_skip55_1_75 +29.3%. 8 beat H7×1.5. Intraday stops kill IS. skip-55 × 1.8 holdout +29.9% with development negative.

+29.32%

Source
nTried
57
holdout
+29.32%

H7 × 1.5 — closest honest to 30%

honesthypos

Same H7 skip-60% k=5, notional 1.5×. Three-window positive.

Dev +1.0% / val +21.7% / holdout +23.2%, DD 43%, 5 months ≥30%. Worst month -28%. Not 30%. 2× dies on development. Stops/shorts/BTC gates do not beat 1× H7 on freeze.

+23.22%

DD −43.25%

Source
clock
day
lookGate
3
minRet
+20.00%
skipIfR30
0.6
k
5
universe
eligible
lev
1.5
development
+0.96%
validation
+21.68%
holdout
+23.22%

Wave 6 — stop/short/BTC NO 30%

honesthypos

16 variants on H7: stop 8/12%, short r30≥60%, LS 70/30, BTC 7d gate, 1.25–2×.

0 AIM. Best honest H7_lev1_5 +23.2%. Fishing LS_70_30_lev2 +28.6% (dev negative). Stops ≈ H7. Short chase loses in development.

+23.22%

Source
nTried
16
holdout
+23.22%

NADIR 4h grid — path-safe NO-GO

honesthypos

7 legs, step 0.75%, TP +1.5%, abort −5.13%, martingale 2^k, 5×, isolated eligible. TP only next 4h bar.

As-sim same-bar OHLC holdout +111.4% (not tradable). Path-safe holdout -16.4%, DD 87%, net -86.4%, abort 21%. 0 AIM. 3x/book A/top15/40/slots also reject.

−16.42%

DD −87.28%

Source
step
+0.75%
levels
7
volumeScale
2
takeProfit
+1.50%
leverage
5
takerFee
+0.06%
slip
+0.02%
fundingPerBar
+0.01%
cooldownAbort
8
cooldownTp
0
holdout
−16.42%

Wave 5 — ignition score NO-GO

honesthypos

Development-frozen score on first MTD ≥8% ignitions. Confirm, isolated, crowded, H7 fallback.

50 variants. 0 AIM, 0 paper. 106 winners vs 691 fakes. Score 0.33 vs -0.05. Fishing +7.3% (dev negative). Walk +5.2%. Features at ignition do not separate.

−0.86%

Source
r1
+5.79%
r3
+29.17%
r7
+2.96%
r30
+15.53%
mtd
+1.30%
resid
+9.04%
vratio
+14.55%
brk
+15.70%
dom
−38.19%
breadth
+2.98%
nUp
+18.90%
rank
+16.46%
holdout
−0.86%

Wave 4 — ignition NO-GO

honesthypos

52 first-MTD ignitions, fixed 3–10d hold, early-month cap, vol, re-arm, k=1.

0 AIM, 0 paper. On holdout 30% names leftover at first MTD +8% is still +25.9% (day 10). Trading every ignition: freeze IGN_t0.12_h3_k5 holdout -7.6%. Fishing +2.8%. False positives dump. H7 remains the paper bar.

−7.56%

Source
nTried
52
verdict
NO-GO
curveHoldout
{
  "0.05": {
    "remaining": 0.3218,
    "day": 8
  },
  "0.08": {
    "remaining": 0.259,
    "day": 10
  },
  "0.1": {
    "remaining": 0.2082,
    "day": 13
  },
  "0.12": {
    "remaining": 0.2004,
    "day": 16
  },
  "0.15": {
    "remaining": 0.1921,
    "day": 16
  },
  "0.2": {
    "remaining": 0.1156,
    "day": 20
  }
}
holdout
−7.56%

Wave 3 — earlier / MTD / 4h

honesthypos

91 variants. Earlier gates, MTD, 4h/8h, rank blend, winner-take.

0 AIM. Freeze D_m0.15_s0.5_k3_gate holdout +4.5% (overfit). Honest still D_m0.2_s0.6_k5_gate +17.1%. 4h/8h best +9.4%. MTD ~0%.

+17.12%

Source
nTried
91
families
{
  "ref": "H7_ref",
  "early": "D_m0.2_s0.6_k5_gate",
  "mtd": "MTD_0.2_s0.5_k5",
  "rank": "RK_blend_m0.2_k5",
  "wtake": "WT_h7",
  "fast": "H8_l72_m0.2_k5"
}
holdout
+17.12%

Capture kill — holdout leftover +9.5%

honesthypos

Of holdout calendar months that printed ≥30%, how much is left after the first H7 signal?

26 holdout ≥30% names. 21 signaled on median day 21, already +28.7%. Median leftover +9.5%. Only 3/21 still had ≥30% left. Development leftover was +31.8% (day 12). Gate arrives after the meat in holdout.

+9.52%

Source
gate
3d ≥20% and 30d <60%
clock
first daily close in the month
holdout
+9.52%

Wave 2 — H7 push + autopsy

honesthypos

31 configs: skip/k/trail/lev/weekly/BTC-kill/month-start score. Eligible only.

0 AIM. Frozen still H7_base +14.1%. Best unlevered three-window H7 skip-60% k=5 +17.1% DD 30%. 2× H7 +22.1% with 69% DD — not a freeze. Month-start score rejected: 30% months start flat.

+14.06%

Source
r3
+13.31%
r7
+12.57%
r30
+0.95%
resid
−18.47%
vratio
+35.89%
brk
+1.69%
holdout
+14.06%

H7 skip-60% k=5 (holdout-ranked)

honesthypos

Daily 3d ≥20%, skip if 30d ≥60%, hold 5 eligible, turnover-aware 1×.

Best unlevered three-window. Holdout +17.1%, DD 30% in the hunt; full-path DD 55%, 10 months ≥30%. Holdout: Nov +85%, Jan +49%, May +57%, Jun +35%, Aug +33%; Oct −19%, Feb −14%. Now cash.

+17.12%

DD −30.50%

Source
k
5
minRet
+20.00%
skip30
0.6
trail
0
lev
1
clock
day
volMin
0
btcDayKill
0
scorePick
false
id
H7_s0.6_k5
family
h7
development
+2.28%
validation
+15.34%
holdout
+17.12%

Mechanism battery H0–H15

honesthypos

13 stories, 28 a-priori variants. Freeze min(dev,val), locked holdout.

0 AIM · 3 PAPER · 10 REJECT. Best honest H7_chase50_k3 +14.1%/mo. H9 universe-gate rejected.

+14.06%

Source
nHypos
13
nVariants
28
kill
AIM if three-window and holdout ≥30%. PAPER if three-window and holdout ≥8%. Else REJECT.
universe
eligible full-history only
holdout
+14.06%

H7 anti-chase 30d <50%

honesthypos

Daily 3d ≥20% top-3 on eligible names, skip if already +50% over 30d.

Only new paper edge in the battery. Dev +4.8% / val +14.2% / holdout +14.1%, DD 42%, 4 months ≥30%. Caps 30/50/80 all paper OOS. Not 30%. Cash right now — no eligible 3d ≥20%.

+14.06%

DD −42.08%

Source
id
H7_chase50_k3
clock
day
lookGate
3
minRet
+20.00%
k
3
skipIfR30
+50.00%
universe
eligible
leverage
1
fees
TAKER on |Δw| only
development
+4.76%
validation
+14.21%
holdout
+14.06%

H9 universe gate (rejected)

honesthypos

Hold every eligible trailing 30d ≥30% when ≥8 names print. Else cash.

The operational “if the universe is hot, spin it” rule. Freeze holdout −1.9%. Snapshot spin is not a compounding machine.

−1.90%

DD −26.44%

Source
id
H9_uni8
clock
day
lookbackDays
30
minRet
+30.00%
minNames
8
k
all who pass
universe
eligible
development
+3.87%
validation
+2.35%
holdout
−1.90%

Optimization hunt (2,784 rotators)

honestopt

Turnover-aware dual lookback, rank cap, BTC filter, hysteresis, inv-vol. Eligible names only.

2784 configs. 0 holdout ≥30%. 263 three-window positive. Walk-forward +1.7%/mo. Verdict NO-GO.

Source
nTried
2784
costs
TAKER on |Δw| only + 1bp/8h funding
freeze
min(dev, val) among both > 0 and DD ≤ 50%, then locked holdout
universe
eligible full-history only (no listing lottery)

Turnover-aware daily 3d ≥20% top-3

honestopt

Same pick as the old honest rule. Fees only when names change.

The 8,468-grid paid a round-trip every idle day. Cutting that leak lifts holdout from +8.15% to +8.59%/mo. Still paper, not 30%.

+8.59%

DD −50.19%

Source
id
D_elig_l3_k3_m0.2
clock
day
lookbackDays
3
minRet
+20.00%
k
3
universe
eligible
leverage
1
fees
TAKER on |Δw| only
development
+3.13%
validation
+10.19%
holdout
+8.59%

Dual-look k=5 inv-vol (three-window)

honestopt

Daily, 3d ≥20% gate, rank 7d, hold 5, hysteresis, inverse-vol, all eligible.

Best holdout among 263 rules that were positive on development, validation, and holdout. Ranked on holdout — not a freeze. +9.26%/mo, 4 months ≥30%, 52% DD.

+9.26%

DD −51.98%

Source
id
D_g3_r7_k5_m0.2_cap999_btc0_h1_iv1
clock
day
lookGate
3
lookRank
7
k
5
minRet
+20.00%
rankMax
999
btcDays
0
hyst
true
invVol
true
lev
1
development
+3.03%
validation
+9.97%
holdout
+9.26%

Frozen min(dev, val)

honestopt

Picked on seen windows (min of development and validation), then locked holdout.

Dev +15.3% / val +15.4% looked like an edge. Holdout +3.4%. The extra knobs overfit. PAPER only.

+3.38%

DD −49.27%

Source
id
D_g3_r7_k3_m0.2_cap80_btc0_h1_iv1
clock
day
lookGate
3
lookRank
7
k
3
minRet
+20.00%
rankMax
80
btcDays
0
hyst
true
invVol
true
lev
1
development
+15.27%
validation
+15.40%
holdout
+3.38%

Found 30% rule

fishingfind30

Monthly top-3 of names up ≥20% over 3d (all listings, cash otherwise)

Holdout search hit. Development was negative. Listing lottery, not a live mandate.

+31.63%

DD −31.93%

Source
id
M_all_l3_k3_m0.2
clock
month
lookbackDays
3
minRet
+20.00%
k
3
universe
all
leverage
1
sizing
equal-weight among selected, cash if none pass
rebalance
month start
costs
taker 5.5 bps in/out, funding 1 bp / 8h
fill
close at rebalance (month-start to month-start)
development
−10.58%
validation
+19.76%
holdout
+31.63%

Holds now

  • 1 USELESSUseless Coinshort+138% 3d+468% 30d
  • 2 EDGEedgeXshort+62% 3d+71% 30d
  • 3 DASHDashshort+48% 3d+110% 30d
  • 4 LITLightershort+36% 3d+120% 30d
  • 5 ZECZcashshort+23% 3d+99% 30d

Honest three-window

honestfind30

Daily top-3 of full-history names up ≥20% over 3 days

Positive on development, validation, and holdout. About +8%/mo, 50% DD. Paper bar, not 30%.

+8.15%

DD −50.30%

Source
id
D_elig_l3_k3_m0.2
clock
day
lookbackDays
3
minRet
+20.00%
k
3
universe
eligible
leverage
1
sizing
equal-weight among selected, cash if none pass
rebalance
daily
costs
taker 5.5 bps in/out, funding 1 bp / 8h
holdout
+8.15%

Best honest hold-all

honestfind30

Daily full-history 3d ≥20% hold every name that passes

Best of 8,468 rotators on eligible/rank-capped names. Zero hit 30% geo. ~+7.8%/mo.

+7.76%

DD −50.30%

Source
id
D_elig_l3_k99_m0.2
clock
day
lookbackDays
3
minRet
+20.00%
k
99
universe
eligible
leverage
1
sizing
equal-weight hold-all who pass
rebalance
daily
holdout
+7.76%

Spin book (snapshot)

tapefind30

Equal-weight every full-history name whose trailing 30 days printed ≥30% to equity. 1×, PAPER.

Screener snapshot: equal-weight every full-history name whose trailing 30d ≥30%. Hot-hand of this filter lost on holdout.

−5.27%

DD −76.51%

Source
id
elig_trail30_ew
clock
day
lookbackDays
30
minRet
+30.00%
k
all who pass
universe
eligible
leverage
1
sizing
equal-weight
fishing
true
holdout
−5.27%

Holds now

  • 1 STXStacks+5.9%+4% 3d+98% 30d
  • 2 CRVCurve DAO Token+5.9%−6% 3d+69% 30d
  • 3 EGLDMultiversX (Elrond)+5.9%+18% 3d+68% 30d
  • 4 ARBArbitrum+5.9%+19% 3d+63% 30d
  • 5 UNIUniswap+5.9%+5% 3d+53% 30d
  • 6 ETHFIEther.fi+5.9%−1% 3d+50% 30d
  • 7 ARArweave+5.9%+20% 3d+50% 30d
  • 8 AAVEAave+5.9%+2% 3d+46% 30d
  • 9 WIFdogwifhat+5.9%+5% 3d+45% 30d
  • 10 LINKChainlink+5.9%+3% 3d+42% 30d
  • 11 PYTHPyth Network+5.9%+4% 3d+38% 30d
  • 12 SOLSolana+5.9%+2% 3d+38% 30d
  • 13 ENSEthereum Name Service+5.9%+1% 3d+36% 30d
  • 14 LDOLido DAO+5.9%+6% 3d+33% 30d
  • 15 RAYRaydium+5.9%−0% 3d+33% 30d
  • 16 XRPXRP+5.9%+3% 3d+31% 30d
  • 17 NEARNear Protocol+5.9%+18% 3d+31% 30d

Oracle monthly picker

oraclefind30

Hold the full-history name that will win the month

Not tradeable. Proves 30% months exist in the book if you could pick the winner.

+44.45%

DD 0.00%

Source
id
oracle_elig_peek
clock
month
k
1
universe
eligible
peek
true
leverage
1
holdout
+44.45%

Grid fishing ceiling (1H)

fishinghunt

AAVE always-on 0.4 ATR × 4 levels

Best peeked holdout grid in the 1H cluster hunt. Not a discovery. +1.59%/mo.

+1.59%

DD −7.12%

Source
id
ag_s0.4_l4
strategy
baseline_always_grid
symbol
AAVEUSDT
bar
1H
spacingAtr
+40.00%
levels
4
timeStopBars
48
sizeMode
equal
cycleRisk
+2.00%
leverageCap
3
fillModel
conservative
holdout
+1.59%

Hunt cluster freeze (1H)

honesthunt

One freeze per strategy on development, then holdout

Frozen baskets ~0%/mo. Best frozen name: LINK H3 +0.27%/mo.

Source
h2_pullback_trend
{
  "levels": 8,
  "spacingAtr": 1.2,
  "timeStopBars": 48,
  "pullbackAtr": 1.2,
  "trendStrengthMin": 0.25,
  "sizeMode": "equal",
  "cfgId": "h2_s1.2_l8_t0.25_p1.2"
}
h3_range_lp
{
  "levels": 10,
  "spacingAtr": 1.3,
  "timeStopBars": 48,
  "rangeEdge": 0.12,
  "sizeMode": "equal",
  "cfgId": "h3_s1.3_l10_e0.12"
}
baseline_always_grid
{
  "levels": 12,
  "spacingAtr": 1.6,
  "timeStopBars": 48,
  "sizeMode": "equal",
  "cfgId": "ag_s1.6_l12"
}
cycleRisk
+2.00%
leverageCap
3

Frozen momentum rotator

honestceiling

Monthly long-short, 20d lookback, K=3

Frozen on development (+9.0%/mo). Holdout +1.48%/mo, DD 21%.

+1.48%

DD −21.31%

Source
id
mom_ls_l20_k3
mode
ls
lookbackDays
20
k
3
rebalance
month
universe
eligible
leverage
1
development
+8.96%
holdout
+1.48%

Fishing momentum

fishingceiling

Monthly long-short, 10d lookback, K=3 (peeked holdout)

Peeked +7.38%/mo. Not a discovery.

+7.38%

DD −5.75%

Source
id
mom_ls_l10_k3
mode
ls
lookbackDays
10
k
3
rebalance
month
universe
eligible
holdout
+7.38%

BTC development frozen grid

honestuniverse

Params frozen on BTC development, applied to the top 100

0/55 names hit 30%/mo after the development screen. Best H2 holdout: AAVE +0.51%/mo.

Source
h1_flow_reversion
{
  "levels": 4,
  "spacingAtr": 0.4,
  "timeStopBars": 24,
  "impulseAtr": 1.4,
  "sizeMode": "equal"
}
h2_pullback_trend
{
  "levels": 4,
  "spacingAtr": 0.4,
  "timeStopBars": 48,
  "pullbackAtr": 0.9,
  "sizeMode": "equal",
  "martingaleMult": 1.3
}
h3_range_lp
{
  "levels": 6,
  "spacingAtr": 0.6,
  "timeStopBars": 48,
  "sizeMode": "equal"
}
baseline_always_grid
{
  "levels": 4,
  "spacingAtr": 0.4,
  "timeStopBars": 48,
  "sizeMode": "equal"
}
baseline_buyhold
{
  "sizeMode": "equal"
}

Tape — trailing 30d ≥30%

35 names as of 2026-09-05. Snapshot, not a forecast.

#Name30d7dBook
174USELESS+468%+342%short
69LIT+120%+37%short
82DASH+110%+68%short
11ZEC+99%+27%short
117STX+98%+2%full
104SPX+77%+12%short
51PUMP+75%−9%short
52ENA+75%−0%short
170EDGE+71%+56%short
106CRV+69%+14%full
212EGLD+68%+35%full
83ARB+63%+49%full
105TRUMP+58%−15%short
84VVV+56%+12%short
33UNI+53%+39%full

Tape — fattest calendar months

308 months ≥30% in the book, 76 in holdout. Top 15 pinned.

MonthNameRetSplit
2024-11XLM+464%development
2025-10H+345%holdout
2024-11ALGO+288%development
2024-11XRP+281%development
2024-11HBAR+273%development
2024-02WLD+222%development
2024-02AR+218%development
2024-11ADA+202%development
2024-02PEPE+198%development
2025-07CFX+195%validation
2023-12ICP+184%other
2024-03JUP+183%development
2024-03FLOKI+183%development
2024-11THETA+175%development
2024-11CRV+175%development