Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
133 commits
Select commit Hold shift + click to select a range
ec81fda
research: open hyperliquid-lower-vol track with NB01 baseline from th…
miohtama Sep 9, 2026
dc64e81
research: add NB02 lower-vol baseline with standard chart formatting
miohtama Sep 9, 2026
3810803
research: add equity-curve smoothing experiment plan for the lower-vo…
miohtama Sep 9, 2026
ae7e647
research: add implementation snippets to the smoothing experiment plan
miohtama Sep 9, 2026
91c462a
research: NB03a data quality, attribution and power for the smoothing…
miohtama Sep 9, 2026
b2aa5d1
research: NB03b decision-aligned feature screen for the smoothing track
miohtama Sep 9, 2026
696234d
research: NB04 portfolio vol-target sweep (REJECT on development window)
miohtama Sep 9, 2026
ece7e8a
research: NB05 pool-cap sweep (REJECT, tightening the cap hurts both …
miohtama Sep 9, 2026
5369d4c
research: NB06 breadth/concentration sweep (REJECT; 25% concentration…
miohtama Sep 9, 2026
0e78cfb
research: NB07 drawdown-based sizing sweep (REJECT)
miohtama Sep 9, 2026
13cbf31
research: NB09 consistency-selection sweep (REJECT) plus a strong NB0…
miohtama Sep 9, 2026
1e8f999
research: NB11 close-out - plateau check fails, hold-out confirms the…
miohtama Sep 9, 2026
5fbd926
research: mark the smoothing experiment plan as executed
miohtama Sep 9, 2026
3d31101
research: add anchor-parity verification for the smoothing track
miohtama Sep 9, 2026
aaf5a19
research: fix two harness defects and re-run NB03a-NB11
miohtama Sep 9, 2026
bcc9c43
research: verify NB03b feature screen alignment
miohtama Sep 9, 2026
b036b53
research: apply the review fix list and re-run NB03a-NB11
miohtama Sep 9, 2026
686d95e
research: NB12 all variants side by side
miohtama Sep 9, 2026
cf10079
research: move the whole track to the full 2026-01-01..2026-09-08 window
miohtama Sep 9, 2026
1aee045
research: NB13 audit of the age barrier in the inclusion rules
miohtama Sep 9, 2026
7f8c770
research: NB14-NB19 evidence-weighted allocation plan (Draft 2)
miohtama Sep 9, 2026
7f456f7
research: NB14 decision-aligned screen of the evidence scores
miohtama Sep 9, 2026
ef9d7d2
research: NB15 placebo frontier control
miohtama Sep 9, 2026
6dd3c96
research: NB16 evidence-weighted selection - REJECT, comprehensively
miohtama Sep 9, 2026
13ebe15
research: NB17 evidence-weighted sizing - REJECT
miohtama Sep 9, 2026
3a5bfb3
research: NB18 core/satellite sleeves - REJECT (with the plan's best …
miohtama Sep 9, 2026
ede4938
research: NB19 close-out - NOTHING ADOPTED, plan complete
miohtama Sep 9, 2026
c95be78
research: independent Codex reviews of NB13-NB19
miohtama Sep 12, 2026
53d3f44
research: act on the NB13-NB19 reviews, part 1 (NB13 re-run, NB14, pl…
miohtama Sep 12, 2026
433d47d
research: act on the NB13-NB19 reviews, part 2 (NB15-NB17 re-run, hea…
miohtama Sep 12, 2026
913b661
research: act on the NB13-NB19 reviews, part 3 (NB18-NB19 re-run, hea…
miohtama Sep 12, 2026
643be99
research: stability-leads plan (NB20-NB23), draft 2 after gpt-6-astra…
miohtama Sep 13, 2026
b902ccd
research: stability-leads plan draft 3 and the shared modules it needs
miohtama Sep 13, 2026
88de369
research: NB20 reporting-versus-trading inactivity - DIAGNOSTIC
miohtama Sep 13, 2026
f582c94
research: fix two bugs the splice verification caught before any note…
miohtama Sep 13, 2026
c36c7a8
research: NB21 the vol-matched family as a candidate - REJECT on the …
miohtama Sep 13, 2026
0a9fbed
research: NB22 complementary downside selection - REJECT, and the pro…
miohtama Sep 13, 2026
708f497
research: NB23 Sortino leg swap - REJECT, with the forward-fill defec…
miohtama Sep 13, 2026
690d9b0
research: the executed stability-splice verification notebook
miohtama Sep 13, 2026
79234f9
research: rebuild and re-run NB20 so it is reproducible from the curr…
miohtama Sep 13, 2026
9f52f1f
research: NB24 close-out - NOTHING ADOPTED, stability-leads plan comp…
miohtama Sep 13, 2026
9a46460
research: stability-leads plan EXECUTED - NOTHING ADOPTED
miohtama Sep 13, 2026
e9cfbd7
research: reviewer tooling - de-noised notebook extractor and shared …
miohtama Sep 13, 2026
1069129
research: NB21 review - independent Codex review, 7 of 8 findings app…
miohtama Sep 13, 2026
66b7c3c
research: NB20 review - independent Codex review, one code fix and ei…
miohtama Sep 13, 2026
fb8ca42
research: NB22 review - independent Codex review, 4 findings applied,…
miohtama Sep 13, 2026
1bdd5a4
research: NB23 review - independent Codex audit, verified and applied
miohtama Sep 13, 2026
57f61a5
research: NB24 review - independent Codex review, verified and applied
miohtama Sep 13, 2026
328c8de
research: align the bootstrap Sharpe with the adoption statistic, and…
miohtama Sep 13, 2026
eadfac1
research: NB21 re-run against the ddof=1 bootstrap fix
miohtama Sep 13, 2026
f82bb84
research: NB20 re-run against the fixed shared modules
miohtama Sep 13, 2026
c14df05
research: NB23 re-run on the ddof=1 bootstrap fix
miohtama Sep 13, 2026
d22e832
research: fold the five independent reviews into the plan
miohtama Sep 13, 2026
5f8c013
research: NB22 re-run against the aligned bootstrap and the NaN guard
miohtama Sep 13, 2026
d174c8c
research: NB24 re-run against the aligned bootstrap and the NaN guard
miohtama Sep 13, 2026
d278c5d
research: update the plan's family-wise figures after the ddof realig…
miohtama Sep 13, 2026
d09595b
research: NB25 all stability-leads variants side by side - DIAGNOSTIC
miohtama Sep 13, 2026
0288bf9
research: NB26 was the drop a fluke - both, in different halves
miohtama Sep 13, 2026
1b6a73f
research: fold NB26's decomposition into the plan's lead 1 result
miohtama Sep 13, 2026
efaf18f
research: NB25 review - 8 of 9 Codex findings confirmed, plus 3 more
miohtama Sep 13, 2026
8814941
research: NB26 review - the random null was one draw, and the sparse …
miohtama Sep 13, 2026
c988bba
research: correct the plan's NB26 summary after NB26's independent re…
miohtama Sep 13, 2026
d40e06d
research: record the first review round's outcomes in the shared revi…
miohtama Sep 13, 2026
2a96501
research: new research rules - maximise Sharpe subject to a 20% CAGR …
miohtama Sep 14, 2026
c285075
research: drop the CAGR floor, and record that Sharpe is return in th…
miohtama Sep 14, 2026
63b63c6
research: objective now constrains the mechanism, not just the outcome
miohtama Sep 14, 2026
8b1235a
research: add diversification to the objective, and state how it reso…
miohtama Sep 14, 2026
1eff907
research: stable-selection plan (NB28-NB31), and the superseded event…
miohtama Sep 14, 2026
33c5558
research: stable-selection plan draft 2 - all 11 blocking review find…
miohtama Sep 14, 2026
5162218
research: stable-selection shared modules, verified
miohtama Sep 14, 2026
8c093b4
research: NB28-NB31 stable-selection batch - NOTHING SHORTLISTED
miohtama Sep 14, 2026
02ccaae
research: NB32 return floor + stability rank - stability rankers REJE…
miohtama Sep 14, 2026
efaaa80
research: NB32 revised after Codex review - fee/cap decomposition, ra…
miohtama Sep 14, 2026
e2a9b02
research: NB32 third run after second Codex review - cap genuinely re…
miohtama Sep 14, 2026
d287be8
research: NB32 fourth run after third Codex review - common-cycle inv…
miohtama Sep 14, 2026
3e04105
research: NB28-NB31 revised after Codex review and re-run; headings g…
miohtama Sep 15, 2026
048d888
research: NB28-NB31 second review round applied and re-run
miohtama Sep 15, 2026
4c6873e
research: NB28-NB31 third review round applied and re-run
miohtama Sep 15, 2026
e64ea59
research: NB28-NB31 fourth review round applied; fee discrepancy is o…
miohtama Sep 15, 2026
b1b0af1
research: NB33 incumbent vs leads on the incumbent's window and the f…
miohtama Sep 15, 2026
4c8f108
research: plan 34 Draft 2 after Codex review; six rule amendments; NB…
miohtama Sep 15, 2026
3d45a52
research: NB34 calm-score screen - gate 5 fails on the return clause'…
miohtama Sep 15, 2026
322dc56
research: NB34 round 2 - reachability oracles added (the vol oracle p…
miohtama Sep 15, 2026
b94decc
research: NB35 - REJECT both centres; random exclusion beats measured…
miohtama Sep 15, 2026
7fc3790
research: NB35 review applied (null wording; churn bias direction); N…
miohtama Sep 15, 2026
7d1f725
research: NB36 close-out - NOTHING SHORTLISTED; 51/51 reproduce at 1e…
miohtama Sep 16, 2026
e295e19
research: NB36 review applied - mirror verified in-kernel, fail-close…
miohtama Sep 16, 2026
cf7ba94
research: idiot-gate audit - gate 5 return clause and gate 9 beat-all…
miohtama Sep 16, 2026
84b8615
research: NB37 threshold crash filter - EXPLORATORY; 1.0 knee costs 1…
miohtama Sep 16, 2026
9709774
research: NB37 review applied - UNEVALUATED verdicts for unrun gates,…
miohtama Sep 16, 2026
9b99581
research: NB37 second review applied - hysteresis logged per candidat…
miohtama Sep 16, 2026
3786880
research: NB38 trimmed-return screen - DIAGNOSTIC; trimming turns the…
miohtama Sep 16, 2026
88b6c8e
research: NB38 review applied - observed-mark eligibility, reachabili…
miohtama Sep 16, 2026
92cc510
research: NB38 second review applied - stopping-rule framing, persist…
miohtama Sep 16, 2026
37fac04
research: NB39 full-archive event-time screen - 15/16 signals clear t…
miohtama Sep 16, 2026
6ce695c
research: NB39 review applied - non-wrapping 60/90-day blocks, regime…
miohtama Sep 16, 2026
7394787
research: NB39 second review applied - 180-day block sensitivity (15/…
miohtama Sep 16, 2026
3c72b08
research: NB39 third review applied - headline caveat corrected, p la…
miohtama Sep 16, 2026
b49f15a
research: NB39 - restore executed outputs (rebuild had stripped them)…
miohtama Sep 16, 2026
2feb901
research: NB39 fourth review applied - tiled block bootstrap, true va…
miohtama Sep 17, 2026
2c99e89
research: NB39 heading wording
miohtama Sep 17, 2026
8b5c328
research: NB39 fifth review applied - untruncated tiled bootstrap wit…
miohtama Sep 17, 2026
cf58745
research: NB39 sixth review applied - matched paired comparisons on j…
miohtama Sep 17, 2026
e2e0b11
research: NB39 seventh review applied - effective-weight coverage ass…
miohtama Sep 17, 2026
513fcdf
research: NB39 eighth Codex review - no findings
miohtama Sep 18, 2026
4364a20
research: NB40 lead forensics and the cash-sleeve rescue - one positi…
miohtama Sep 18, 2026
031578f
research: NB40 first Codex review applied - disputed-cycle band attri…
miohtama Sep 18, 2026
5aff4a6
research: NB40 second Codex review applied - exploratory framing for …
miohtama Sep 18, 2026
0ece7f2
research: NB40 third Codex review applied - unlimited capped book des…
miohtama Sep 18, 2026
6bb3de6
research: NB40 heading - overlap claim scoped to the runs measured
miohtama Sep 18, 2026
1596395
research: NB40 fourth Codex review applied - mask diagnostics separat…
miohtama Sep 18, 2026
4d57597
research: NB40 fifth Codex review applied - remaining causal and iden…
miohtama Sep 18, 2026
4b6f6ed
research: NB40 sixth Codex review applied - coverage and exit-thresho…
miohtama Sep 18, 2026
d0199d4
research: NB40 seventh Codex review - one wording fix; no material is…
miohtama Sep 18, 2026
124cd56
research: NB40 heading grammar
miohtama Sep 18, 2026
5b6337c
research: NB41 trade forensics - named vaults, entry/exit reasons, va…
miohtama Sep 18, 2026
8afd5cf
research: plan 42 draft 1 - crash-cluster exit on the one-day clock, …
miohtama Sep 18, 2026
45e952e
research: plan 42 draft 1 - 4-hour cadence arm added after the archiv…
miohtama Sep 18, 2026
7d0d854
research: plan 42 draft 2 after Grok review - cluster search cut to a…
miohtama Sep 18, 2026
2f7f401
research: plan 42 draft 3 after the second Grok review - a catch is d…
miohtama Sep 19, 2026
3200518
research: plan 42 draft 3 - two stale sentences in the evidence table
miohtama Sep 19, 2026
012b32e
research: plan 42 draft 4 after the third Grok review - the catch pre…
miohtama Sep 19, 2026
3a5c16c
research: plan 42 draft 5, final for build, after the fourth Grok rev…
miohtama Sep 19, 2026
6f5114c
research: NB42 crash exit - the incumbent's collapse sells were value…
miohtama Sep 19, 2026
689bfce
research: NB42 first Codex review applied - pool-keyed ranking cache,…
miohtama Sep 19, 2026
7679ecb
research: NB42 second Codex review applied - the 19 Aug gate sell pas…
miohtama Sep 19, 2026
8dbe119
research: NB42 third Codex review applied - window A through the run …
miohtama Sep 19, 2026
4a68527
research: plan 42 executed - outcome recorded
miohtama Sep 19, 2026
a50214f
research: NB42 Grok review of the result applied - first four-hour bu…
miohtama Sep 19, 2026
886b387
research: NB43 live hyper-ai vault beside the anchor backtest - same …
miohtama Sep 19, 2026
3b16819
research: plan 44 live-parity backtest, Draft 2 after Grok review - o…
miohtama Sep 19, 2026
ec0d0dc
research: plan 44 F4 corrected - the anchor runs at the same $150k an…
miohtama Sep 25, 2026
8812b59
research: add Hyperliquid IC notebooks without build artefacts
miohtama Sep 29, 2026
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
9 changes: 9 additions & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,15 @@ tmp/
# FinRL model policy files
*.pt

# Generated backtesting and research artefacts outside the research notebooks.
**/_artifacts*/

# Temporary review working files and research manifests.
scratchpad/hyperliquid-ic/_review/
scratchpad/hyperliquid-ic/**/*.jsonl
scratchpad/hyperliquid-ic/**/*.err
scratchpad/hyperliquid-ic/_build/manifest_*.json

# Dotenv
# May contain secrets
.env
Expand Down
13 changes: 11 additions & 2 deletions .vscode/settings.json
Original file line number Diff line number Diff line change
Expand Up @@ -46,12 +46,21 @@
"statusBar.background": "#006666",
"statusBar.foreground": "#e7e7e7",
"statusBarItem.hoverBackground": "#009999",
"statusBarItem.remoteBackground": "#006666",
"statusBarItem.remoteBackground": "#ff00ff",
"statusBarItem.remoteForeground": "#e7e7e7",
"titleBar.activeBackground": "#006666",
"titleBar.activeForeground": "#e7e7e7",
"titleBar.inactiveBackground": "#00666699",
"titleBar.inactiveForeground": "#e7e7e799"
"titleBar.inactiveForeground": "#e7e7e799",
"activityBarTop.activeBackground": "#009999",
"activityBarTop.background": "#009999",
"activityBarTop.foreground": "#e7e7e7",
"activityBarTop.inactiveForeground": "#e7e7e799",
"commandCenter.foreground": "#e7e7e7",
"statusBar.debuggingBackground": "#006666",
"statusBar.debuggingForeground": "#e7e7e7",
"activityBar.activeBorder": "#e7e7e7",
"activityBarTop.activeBorder": "#e7e7e7"
},
"peacock.color": "#006666"
}
8 changes: 7 additions & 1 deletion CLAUDE.md
Original file line number Diff line number Diff line change
Expand Up @@ -129,14 +129,20 @@ Don't format code.

- We check for performance, management, entry/exit fees but vaults do not usually have these. Do not assume any slippage on vaults: deposits and redemptions fill at NAV (share price), and performance fees are typically internalised in the share price series.

## Backtesting build artefacts

- Saved cell outputs inside research notebooks may be committed with the notebooks.
- Never commit generated backtesting files outside notebooks, including intermediate datasets, caches, plots, logs, reports, and run manifests. Add a matching `.gitignore` rule whenever a workflow creates them, before staging or committing files.
- Keep reusable scripts, notebooks, and written research conclusions outside generated artefact directories so they remain available for review. If an external generated file is already tracked, remove it from the Git index without deleting the local file.

## Editing files

- When you need to modify files the editable Python packages live under ~/code/trade-executor
- Never edit files under ~/code/docs

## Indicator cache

If the notebook crashes because of what looks like indicator cache issue: we have added or edited indicators and the data is not correctly recalculated, you can clear the indicator cache wtih `clear-backtesting-cache` skill.
If the notebook crashes because of what looks like indicator cache issue: we have added or edited indicators and the data is not correctly recalculated, you can clear the indicator cache wtih `clear-backtesting-cache` skill.

This skill should not be used unless the notebook crashes because of indicator data problems.

Expand Down
Loading
Loading