Граф коммитов

bfc1da9de1 fix: the sequence models were reading the window backwards main AnimateDread 2026-08-07 18:28:44 -04:00
da54639996 feat: expectancy stop - halt when the measured result says the strategy loses AnimateDread 2026-08-07 14:20:00 -04:00
b3b7e7bceb fix: excursion window must not depend on the barrier it sizes AnimateDread 2026-08-07 13:57:23 -04:00
32ffeb99f3 fix: normalise the asymmetry target - the raw one is confounded by volatility AnimateDread 2026-08-07 13:04:13 -04:00
a7701f032b feat: derive the ATR multiples from measured excursions - no hardcoded geometry AnimateDread 2026-08-07 12:06:25 -04:00
2c78f3b90d diag: is "optimal SL/TP" learnable? Score the features against excursions AnimateDread 2026-08-07 10:22:41 -04:00
3482b6c238 feat: entry/SL/TP stop being inputs - the barrier geometry is measured AnimateDread 2026-08-07 09:39:30 -04:00
9e1c72aacc fix: make the indicator tuner actually measure, and gate what it installs AnimateDread 2026-08-07 09:31:06 -04:00
e5ceed6466 fix: MI diagnostics never ran on a resumed model - the stated intent was never achieved AnimateDread 2026-08-07 09:08:01 -04:00
0d5892357b diag: report the indicator tuner's candidate spread - "no improvement" is ambiguous AnimateDread 2026-08-06 21:59:10 -04:00
cccf94f9ca fix: correct the lag profile across lags too - it contradicted itself AnimateDread 2026-08-06 16:28:57 -04:00
04ee2e113a fix: gate the barrier-geometry winner on a family-wise null, not its own AnimateDread 2026-08-06 14:04:16 -04:00
3271f1ea93 diag: MI feature-lag profile - close the blind spot in every MI verdict so far AnimateDread 2026-08-06 13:51:39 -04:00
8c5ea639ee feat: extend ADWyckoffEventStream with new range-lifecycle parameters and update related features AnimateDread 2026-08-02 17:08:48 -04:00
8ccbddb051 Add new research scripts for trading strategy analysis AnimateDread 2026-08-02 12:25:20 -04:00
f1b7dcf7f3 fix: correct MI sample alignment and improve BN weight diagnostic report AnimateDread 2026-08-02 08:12:47 -04:00
5f647ba5db fix: improve error messages and suppress false sharing-violation logs AnimateDread 2026-08-02 01:09:18 -04:00
3459db887b infra: bid/ask M1 bars and an order engine that makes the fill bug unavailable AnimateDread 2026-08-02 00:08:30 -04:00
82cdbb40df research: Wyckoff context REPLICATES on a second, independent trigger AnimateDread 2026-08-01 23:57:30 -04:00
c998d655ee research: Wyckoff CONTEXT has real predictive content - 8/8 positive slopes AnimateDread 2026-08-01 23:56:05 -04:00
9a4ae635e0 research: the two canonical Wyckoff trades tested whole - both negative AnimateDread 2026-08-01 23:34:35 -04:00
79f108ab84 research: RETRACTION - the retail-fade edge was a fill-timing artifact AnimateDread 2026-08-01 23:17:55 -04:00
821f16df8b research: model-selected retail fades hold up out of sample (after killing a big leak) AnimateDread 2026-08-01 23:15:00 -04:00
c9b489e5d8 research: the retail fade DOES clear cost on EURUSD - correcting the earlier verdict AnimateDread 2026-08-01 22:33:11 -04:00
c645c98f31 chore: keep the reference PDFs out of the repository AnimateDread 2026-08-01 22:18:18 -04:00
c2dd9eb6aa research: retail setups ARE anti-predictive - and the edge dies with the cost AnimateDread 2026-08-01 22:18:01 -04:00
8e6525d819 research: Wyckoff's law of cause and effect is real but SUBLINEAR AnimateDread 2026-08-01 21:36:14 -04:00
6f7aa4f1e8 research: the stop-run edge fails its own mechanism test - disconfirmed AnimateDread 2026-08-01 21:11:33 -04:00
9ca0ffcc5c research: stop-run/liquidity sweep - the first setup to survive everything AnimateDread 2026-08-01 21:05:26 -04:00
05c6a5c484 research: four more hypothesis families - drift is real, timing still is not AnimateDread 2026-08-01 20:51:02 -04:00
0c2b025c16 fix(research): calendar recorder - separate LIVE from BACKFILL, fix seen-set key AnimateDread 2026-08-01 20:22:12 -04:00
2aa5bfbe28 feat(research): write-once live calendar recorder AnimateDread 2026-08-01 20:14:48 -04:00
50e1c7ef84 research: flow effect and spread cost decay together and never cross AnimateDread 2026-08-01 20:12:00 -04:00
f83425aad8 research: tick-flow verdict across 4 instruments - real reversal, untradeable AnimateDread 2026-08-01 20:00:22 -04:00
eeacb609b6 fix(research): standardise flow test against the EMPIRICAL null, not a costless coin AnimateDread 2026-08-01 19:51:07 -04:00
91d67db737 perf(research): parallel tick decode, 12x, plus two correctness fixes AnimateDread 2026-08-01 19:14:38 -04:00
fac9f620b0 research: harness for the M5 order-flow edge test AnimateDread 2026-08-01 18:25:44 -04:00
e7a9fe22a7 research: signed order-flow imbalance instead of bare quote-move counts AnimateDread 2026-08-01 18:23:50 -04:00
0526f066ec research: stream tick files into bars with microstructure features AnimateDread 2026-08-01 18:20:34 -04:00
ecb8638996 chore: keep market data out of the repo AnimateDread 2026-08-01 18:12:54 -04:00
ef126845f8 research: decode StrategyQuant .dat tick files AnimateDread 2026-08-01 17:57:11 -04:00
87ee9c0a43 tools: MQL5 tick exporter for offline microstructure research AnimateDread 2026-08-01 17:49:35 -04:00
ceb6342dfd feat(ai): spread as a volatility-regime feature, and fix a stale-index cache in both new blocks AnimateDread 2026-08-01 17:42:40 -04:00
4881ad22b5 research: measure the spread channel, and separate its real signal from its own cost model AnimateDread 2026-08-01 17:37:33 -04:00
d80d9444a5 feat(ai): widen the volume feature block from 1 value to 4 AnimateDread 2026-08-01 17:34:33 -04:00
b035ea29e5 feat(ai): cross-asset currency strength - the first feature not derived from one price series AnimateDread 2026-08-01 17:16:13 -04:00
8710240cd5 fix(signals): revive a dead MA model, and demote Sanyaku from state to event AnimateDread 2026-08-01 17:14:34 -04:00
ae738f59f0 chore(research): drop committed __pycache__, add .gitignore AnimateDread 2026-08-01 16:59:03 -04:00
bd076ddbad research: test the shipped classic patterns for entry edge - and the lookahead that faked one AnimateDread 2026-08-01 16:58:57 -04:00
b6a067b266 research: offline validation kit, and the answer it produced AnimateDread 2026-08-01 16:21:44 -04:00
a77ff64b13 fix(deinit): a full model write was running ahead of the cheap cleanup AnimateDread 2026-08-01 16:06:40 -04:00
7d038df749 research: export the feature matrix and a raw OHLCV grid for offline work AnimateDread 2026-08-01 15:49:57 -04:00
004f2a04f7 fix(diag): the symbol sweep was measuring its own sampling, not the market AnimateDread 2026-08-01 15:11:40 -04:00
168422ff7a fix(labels): the 128-bar horizon ceiling was truncating the shipped label AnimateDread 2026-08-01 14:54:45 -04:00
40af4a4b5b fix(labels): the geometry scan rewarded the labels it should reject AnimateDread 2026-08-01 14:46:25 -04:00
f97ab9f1d6 feat(labels): measure which barrier is predictable at entry, don't guess AnimateDread 2026-08-01 14:32:04 -04:00
4443ce85c1 fix(diag): the alignment scan cried misalignment at its own arithmetic AnimateDread 2026-08-01 14:25:26 -04:00
87c8656b53 diag(autotune): a positive control, and a scan that separates "no signal" from "signal knocked out of step" AnimateDread 2026-08-01 14:12:10 -04:00
9a5f645dc3 diag(autotune): stop making the feature test cost a trained model AnimateDread 2026-08-01 14:01:32 -04:00
9920754dec diag(autotune): five permutations was still a coin flip - use a real test AnimateDread 2026-08-01 13:45:46 -04:00
12a1fbd133 diag(autotune): one label shuffle cannot settle the no-edge question AnimateDread 2026-08-01 13:32:12 -04:00
89eab14ca8 fix(chart): arrows survived the EA that drew them - persist, then clear AnimateDread 2026-08-01 13:28:34 -04:00
018afb1ba9 fix(autotune): MI scorer read an array nobody filled; add the permutation floor AnimateDread 2026-08-01 13:05:50 -04:00
d7eea325fb refactor(ai): extract Layer.mqh and deduplicate AI config AnimateDread 2026-08-01 11:27:28 -04:00
36a2825087 chore(Network): remove unused optimization methods and tidy whitespace AnimateDread 2026-08-01 11:26:59 -04:00
9756e2b64f fix(deinit): O(n^2) arrow prune blew the shutdown budget and littered 3 charts AnimateDread 2026-08-01 10:38:36 -04:00
6db0519472 perf(autotune): replace the genetic search with a filter score - hours to seconds AnimateDread 2026-08-01 10:29:31 -04:00
eafc6802d9 fix(ui): panel claimed "no directional calls" while the model was signalling AnimateDread 2026-08-01 10:10:56 -04:00
e83e30344f fix: disable auto-tune indicators by default AnimateDread 2026-08-01 10:03:23 -04:00
3bae2f9254 fix: the imbalance correction never ran during the auto-tune search AnimateDread 2026-08-01 00:46:24 -04:00
8ff3f5d632 refactor(inputs): set default SL to ATRx2 and TP to ATRx6 AnimateDread 2026-08-01 00:34:58 -04:00
25813523d3 fix: refuse invalid SL/TP, fix the unreachable deploy floor, scale the horizon AnimateDread 2026-08-01 00:30:49 -04:00
f48bc93f9b refactor(inputs): 96 -> 70 inputs; remove two untested/unusable filter modules AnimateDread 2026-07-31 21:22:02 -04:00
b4a704d309 feat(ai): triple-barrier labels replace exact-pivot ZigZag targets AnimateDread 2026-07-31 20:39:49 -04:00
fa0455f399 diag: inference-path census, to explain zero-trade backtests AnimateDread 2026-07-31 18:24:32 -04:00
0261c013d0 fix: EMA shadow never received the LSTM weight block AnimateDread 2026-07-31 17:52:41 -04:00
ab84998d35 feat(ai): true multi-bar conv and true sequence LSTM AnimateDread 2026-07-31 17:42:33 -04:00
b4d640d0b4 fix: blend-skip warning fired on layers that correctly own no weights AnimateDread 2026-07-31 16:11:16 -04:00
32880f3728 diag: make BlendWeightsFrom say when it skips a layer AnimateDread 2026-07-31 14:38:09 -04:00
251e9711dd diag: report |dW| alongside d|W| per layer, and drop two stale log claims AnimateDread 2026-07-31 14:33:29 -04:00
397b0eac1f refactor(ai): nine class-imbalance inputs down to two AnimateDread 2026-07-31 11:46:57 -04:00
7eb48f5038 feat(trade): anchor SL and TP to the entry price, not the last swing AnimateDread 2026-07-31 10:11:13 -04:00
5a12ae08e2 revert(ai): restore the 4eae763 front-ends - both my rewrites stopped signaling AnimateDread 2026-07-31 10:04:23 -04:00
18f63c7a52 feat(ai): report per-layer weight movement each era AnimateDread 2026-07-31 07:10:09 -04:00
34d6aa42a4 feat(ai): real conv receptive field + the reference's channel pool AnimateDread 2026-07-30 20:05:37 -04:00
5a7ed73df6 fix(ai): positive forget-gate bias, giving the LSTM its window back AnimateDread 2026-07-30 19:15:29 -04:00
b941977612 2 new files AnimateDread 2026-07-30 18:46:03 -04:00
7a753e54ba feat(ai): sequence-LSTM kernels for the OpenCL tier AnimateDread 2026-07-30 18:40:56 -04:00
7a081979b2 feat(ai): make LSTM/HYBRID actual sequence models over bars AnimateDread 2026-07-30 18:22:24 -04:00
3ea54b4f2b feat(dll): fused sequence-LSTM kernels with real backpropagation-through-time AnimateDread 2026-07-30 18:13:50 -04:00
4eae763849 fix(ai): report the metric actually compared; surface the derived front-end AnimateDread 2026-07-30 15:20:30 -04:00
ff06583680 feat(ui): drop the config tag from the plain-language panels AnimateDread 2026-07-30 13:39:08 -04:00
efdd36d183 fix(ai): stop the shutdown save from resurrecting reset weights; size HYBRID's LSTM to its real fan-in AnimateDread 2026-07-30 13:06:09 -04:00
413ff7e9a2 fix(ai): allocate LSTM cell state on the load path, unpinning HYBRID from Neutral AnimateDread 2026-07-30 12:39:14 -04:00
1039ad936f feat(ai): measure precision per confidence tier; fix stale metric labels AnimateDread 2026-07-30 11:47:15 -04:00
ce90fc74b6 fix(ai): discount selection precision by coverage shortfall AnimateDread 2026-07-30 10:40:36 -04:00
45b35b3d1d feat(nn): derive dense depth, train on all history, pin the shape in .cfg AnimateDread 2026-07-30 10:05:40 -04:00
3bc551b6e1 feat(nn): derive conv filter count and LSTM hidden size from the data AnimateDread 2026-07-30 09:22:11 -04:00
ebf2e73667 fix(ui): unique chart tag, product-grade panel, responsive under load AnimateDread 2026-07-30 09:05:58 -04:00
16632c1c3c feat: make batch normalization mandatory, and record the run-3 results AnimateDread 2026-07-30 08:51:10 -04:00