Update readme.md - #13
Open
lee101 wants to merge 246 commits into
Open
Conversation
…ld trade under the value
…n-practices Add rlsys reinforcement learning system with tests
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Comment on lines
+7
to
+9
| This neural network trading bot trades stocks (long/short) and crypto (long-only) daily at market open/close. It successfully grew my portfolio from $38k to $66k over several months in favorable conditions at the end of 2024. | ||
|
|
||
| The bot uses the Datadog Toto model for time series forecasting (CUDA GPUs required; CPU execution is disabled). | ||
|
|
||
| The bot uses the Datadog Toto and kronos model for time series forecasting |
There was a problem hiding this comment.
The README update drops the warning that a CUDA GPU is mandatory. However the Kronos forecaster still hard‑rejects any CPU device and raises RuntimeError("KronosForecastingWrapper requires a CUDA device") when device does not start with "cuda" or when CUDA is unavailable (see src/models/kronos_wrapper.py lines 147‑154). Users following the README on a CPU-only machine will now hit immediate runtime failures. Please keep the GPU requirement (or clarify that Kronos cannot run on CPU) to avoid misleading setup instructions.
Useful? React with 👍 / 👎.
Owner
Author
|
@codex let's look over why max diff strategy in trade_stock_e2e.py still not getting executed despite it being the top avg_return strategy hmm also add a flag to remove probe trades and flag to control how many such stocks it aims to spread over default to 2 but it still needs to go over everything and yes can actually trade more than that. I guess this is the kind of risk factor the global risk so we might have 5 stocks spawn high low trading agents for Max diff strategy but they would check if it's going to push the account over that global risk threshold before opening said positions |
|
Summary
Testing
|
Owner
Author
|
@codex before no: need to have it continuously during the spawn instead so for example the account may be de leveraged during the day, enough for one of the spawned tasks to be yes executed later in the day |
|
Summary
Testing
|
lee101
force-pushed
the
main
branch
2 times, most recently
from
March 24, 2026 07:45
198c06f to
4b5c43f
Compare
lee101
added a commit
that referenced
this pull request
Mar 25, 2026
Seed s240 from the s201-300 sweep is the new production champion: - 5bps: +17,642% ann, Sortino=7.00 — 88% better than previous champion s71 (+9,381%) - 8bps: +40,405% ann, honest_val=18.32x in 180 days - Pool_val=16.99, fully saved to champions dir - Isolated spike: s239=+262%, s241=+448% Also new today: - s507: +8,273% @5bps (new #3) - s520: +6,132% @5bps, Sortino=7.70 (new #8) - s523: +5,919% @5bps, Sortino=7.56 (new #9) - s404: +7,319% @5bps, Sort=6.28 (new #5) - s416: +4,099% @5bps, Sort=6.19 (new #13) - s609: +4,450% @5bps, Sort=5.04 (new #12) 7-9 parallel pool sweeps now covering s1-1000. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
lee101
added a commit
that referenced
this pull request
Mar 25, 2026
s628 (s601-700): +7158% ann at 5bps, 5bps>>8bps (7.27 vs 6.72) VERY ROBUST. New #6 all-time. WR=61%, Trades=77, tight percentile spread. s601-700 now has SIX top-55 seeds: s628(#6!), s609(#13), s627(#50). Other new s601-700: - s627: +1509%@5bps, WR=59%, Trades=135 New s301-400: - s336: +952%@5bps (minor overfitter), WR=60% New s401-500: - s428: +1072%@5bps, 5bps>>8bps ROBUST, WR=60% New s701-800: - s729: +1027%@5bps, 5bps>8bps ROBUST, WR=62%, Trades=82 (compact) 100 seeds ranked in leaderboard. Sweeps at ~43/100(s201), ~36/100(s301), ~30/100(s401), ~28/100(s501,s601), ~31/100(s701), ~19/100(s801), ~6/100(s901). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
lee101
added a commit
that referenced
this pull request
Apr 13, 2026
Seed s240 from the s201-300 sweep is the new production champion: - 5bps: +17,642% ann, Sortino=7.00 — 88% better than previous champion s71 (+9,381%) - 8bps: +40,405% ann, honest_val=18.32x in 180 days - Pool_val=16.99, fully saved to champions dir - Isolated spike: s239=+262%, s241=+448% Also new today: - s507: +8,273% @5bps (new #3) - s520: +6,132% @5bps, Sortino=7.70 (new #8) - s523: +5,919% @5bps, Sortino=7.56 (new #9) - s404: +7,319% @5bps, Sort=6.28 (new #5) - s416: +4,099% @5bps, Sort=6.19 (new #13) - s609: +4,450% @5bps, Sort=5.04 (new #12) 7-9 parallel pool sweeps now covering s1-1000. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
lee101
added a commit
that referenced
this pull request
Apr 13, 2026
s628 (s601-700): +7158% ann at 5bps, 5bps>>8bps (7.27 vs 6.72) VERY ROBUST. New #6 all-time. WR=61%, Trades=77, tight percentile spread. s601-700 now has SIX top-55 seeds: s628(#6!), s609(#13), s627(#50). Other new s601-700: - s627: +1509%@5bps, WR=59%, Trades=135 New s301-400: - s336: +952%@5bps (minor overfitter), WR=60% New s401-500: - s428: +1072%@5bps, 5bps>>8bps ROBUST, WR=60% New s701-800: - s729: +1027%@5bps, 5bps>8bps ROBUST, WR=62%, Trades=82 (compact) 100 seeds ranked in leaderboard. Sweeps at ~43/100(s201), ~36/100(s301), ~30/100(s401), ~28/100(s501,s601), ~31/100(s701), ~19/100(s801), ~6/100(s901). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
lee101
added a commit
that referenced
this pull request
Apr 14, 2026
… 3/5/6) Calibration (chronos2_linear_calibration.py): - Fix Phase 3: use fine threshold grid (thresh_both) instead of just [best_buy, best_sell] — enables proper joint threshold+weight search - Add Phase 5: ultra-fine joint search ±1bps thresholds + ±15% signal_weight - Add Phase 6: fine confidence threshold refinement (10 percentile points) after all weight phases are settled Augmentation (chronos2_stock_augmentation.py / chronos2_full_finetune.py): - Add earnings_shock augmentation (#13): sudden ±5-15% price move at random position in context, followed by momentum continuation (50% follow-through fading over 3 bars) or partial mean-reversion (30% pullback). Simulates earnings/news events; distinct from gap_inject (persistent level) and outlier_inject (isolated spike). - New --earnings-shock-prob and --earnings-shock-magnitude CLI args in chronos2_full_finetune.py Training chain: - scripts/launch_chronos2_v10.sh: v10 config with earnings_shock_prob=0.10, all v9 augs retained, ctx=1024, 200k steps, seed=59 - scripts/launch_chronos2_v4_when_v3_ready.sh: add v10 wait+SWA+calibration +benchmark block after v9 completes (script now 581 lines) Tests: - 3 new tests for earnings shock in test_chronos2_stock_augmentation.py: test_earnings_shock_changes_context, _disabled_when_zero_prob, _not_in_val_mode — all 48 aug tests pass, 63 cal tests pass Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.