animatedread/Warrior_EA
Watch
5
8
Fork
You've already forked Warrior_EA
 11
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Commit graph

  • 47a5ef338b Refactor Warrior EA: Integrate custom signal modules, enhance voting mechanism, and improve management features main AnimateDread 2026-09-13 14:32:40 -04:00
  • 5f8a2b1df8 Remove obsolete log and data files: deleted cpu_directml.log, opencl.log, and profiling.csv to clean up the repository. AnimateDread 2026-09-13 14:32:28 -04:00
  • ac57a720e0 wip: snapshot before the KISS restructure AnimateDread 2026-09-11 06:16:08 -04:00
  • e37f15edb3 build: tag books-d1-census-1 for the first deploy of the swing-book and census fixes AnimateDread 2026-09-10 07:16:06 -04:00
  • e80993f10b feat(book): daily and weekly charts are swing books - no session window, no weekly flat AnimateDread 2026-09-09 13:17:00 -04:00
  • dbe28d5478 fix(census): the symbol census does not run in the tester AnimateDread 2026-09-09 08:59:31 -04:00
  • 0ddc33dd1e fix(journal): a virtual firing fills like the real order, and the label pays the spread at its fill AnimateDread 2026-09-08 21:48:13 -04:00
  • 649ca53fa9 fix(book): the plan line applies the management policy before it prints AnimateDread 2026-09-08 21:14:04 -04:00
  • 6376720fe0 feat(book): Book_Management input - as the book states, or stop/target/time stop only AnimateDread 2026-09-08 21:12:04 -04:00
  • 034301bbc0 fix(book): R is the original risk - breakeven was ending management on the first move AnimateDread 2026-09-08 20:17:29 -04:00
  • 757383221a fix(meta): sort rows chronologically and average the walk-forward over three seeds AnimateDread 2026-09-08 20:15:15 -04:00
  • d6d14484d1 fix(meta): reload only when the file changes, compress the forest, choose on a lower bound AnimateDread 2026-09-08 19:17:37 -04:00
  • fa0d0fccc3 perf(journal): in the tester, read the journal once at deinit, not daily AnimateDread 2026-09-08 18:48:56 -04:00
  • 10463eba2f feat(meta): the meta-labeller - a random forest in pure MQL5, learned from the shadow journal AnimateDread 2026-09-08 18:41:21 -04:00
  • af3db90353 feat(journal): the shadow journal - every firing, independently, next to the trades taken AnimateDread 2026-09-08 18:32:38 -04:00
  • ad327735db feat(journal): the tester fills the database, and the journal can skip a categorical loser AnimateDread 2026-09-08 17:36:02 -04:00
  • d91bb3a0cb feat(tester): tag the entry with every setup armed on its side AnimateDread 2026-09-08 12:59:21 -04:00
  • ff4833615f feat(tester): tag every entry with the setup that shaped it AnimateDread 2026-09-08 12:36:10 -04:00
  • 8097ccd9a0 fix(book): a short on a netting account was invisible to its own time stop AnimateDread 2026-09-08 12:11:15 -04:00
  • 3ee32784f3 fix(tester): the EA was long-only because every run inherited tradingdirection=LONG_ONLY AnimateDread 2026-09-08 11:45:35 -04:00
  • 9ab41d3752 diag(arming): count the sides separately - the EA is long-only AnimateDread 2026-09-08 11:06:41 -04:00
  • 4da7bf322d feat(book): don't buy into resistance - the headroom rule, measured AnimateDread 2026-09-08 10:15:44 -04:00
  • bce0ae742f feat(agreement): four is the threshold once the spread is actually charged AnimateDread 2026-09-08 10:09:18 -04:00
  • 19bef6dbd4 test(verdict): a decade says the spread is the whole gap AnimateDread 2026-09-08 07:01:34 -04:00
  • 7abad65c80 feat(book): port the seven missing triggers, and the classic crowding veto AnimateDread 2026-09-07 23:18:40 -04:00
  • 9e4b03ad44 feat(audit): report what the quorum actually did, not only what it could AnimateDread 2026-09-07 22:51:00 -04:00
  • c961f5c8d7 feat(audit): the EA states its resolved trading plan at init AnimateDread 2026-09-07 22:42:31 -04:00
  • 0c58e55173 feat(adapt): build the writer half of the live-learning loop AnimateDread 2026-09-07 22:37:54 -04:00
  • 6c375ee462 fix(forex): drop the Tokyo window, and close the "open design question" AnimateDread 2026-09-07 21:13:16 -04:00
  • 5418a5e326 feat(book): port the session findings research had and the EA did not AnimateDread 2026-09-07 21:06:51 -04:00
  • 534d622ae5 feat(book): give crypto a measured clock, and say when a quorum is impossible AnimateDread 2026-09-07 20:45:58 -04:00
  • d685a3f952 fix(book): the entry price was wiped by the setup's own other side AnimateDread 2026-09-07 19:57:56 -04:00
  • 5f1c459043 refactor(inputs): commission out of the EA, 63 inputs down to 22 AnimateDread 2026-09-07 19:57:39 -04:00
  • cd3f14febf docs(refactor): slices 5 and 6 are done - the books are in the EA AnimateDread 2026-09-07 17:56:10 -04:00
  • 338e2358be feat(book): the agreement count - the only measured edge, now enforced in the EA AnimateDread 2026-09-07 17:55:42 -04:00
  • f7b12413a4 feat(book): fourteen book setups as signal modules, each owning its own management AnimateDread 2026-09-07 17:53:40 -04:00
  • 213ba83e97 feat(book): CBookSetupSignal - a setup owns its entry, its exit AND its management AnimateDread 2026-09-07 17:46:40 -04:00
  • d7b9013a85 docs(refactor): sections 0 and 3 were still describing the pre-deletion state AnimateDread 2026-09-07 17:27:53 -04:00
  • 6edadfd801 docs(refactor): the plan, updated to what the deletion actually did AnimateDread 2026-09-07 17:27:15 -04:00
  • c6f70d8ee1 refactor(inputs): remove the Neural Networks section - seven dialog switches that did nothing AnimateDread 2026-09-07 17:26:47 -04:00
  • 4e5e6fcae3 refactor(ea): delete on-chart training - the EA stops learning and starts executing AnimateDread 2026-09-07 17:23:36 -04:00
  • 5ffab1a84a docs(refactor): the plan, measured against the code and a real tester log AnimateDread 2026-09-07 15:50:13 -04:00
  • 0686410b42 refactor(logs): two verbosity levels, and a throttle that works in the tester AnimateDread 2026-09-07 15:44:31 -04:00
  • f4c36eebce fix(fleet): stop the sweep that printed on every bar of every tester run AnimateDread 2026-09-07 14:11:15 -04:00
  • 6df9132274 fix(gate+label): certify the trade the EA places - traded population, real fills, no survivorship AnimateDread 2026-09-07 09:20:16 -04:00
  • 84d0af6e91 fix(vote): the classic modules were the live book - inputs, not voters; certify with the live stop AnimateDread 2026-09-06 23:07:53 -04:00
  • a765e0f61f fix(news): weight by scheduled importance, not the post-release verdict AnimateDread 2026-09-06 19:27:33 -04:00
  • 50fad051d4 fix(news): the calendar was not missing, it was still synchronising AnimateDread 2026-09-06 17:03:56 -04:00
  • dc6daa2ff9 fix(news): probe the calendar call instead of declaring the feed absent AnimateDread 2026-09-06 16:31:06 -04:00
  • 15a1e18a5c feat(news): standalone calendar dump, for a terminal that actually has one AnimateDread 2026-09-06 16:22:11 -04:00
  • 9cfcb5f42d fix(news): export the calendar on the upkeep tick, not only at attach AnimateDread 2026-09-06 16:14:26 -04:00
  • eb2a443582 fix(setup): the entry window is measured in BARS, as every book states - the 60 seconds came from an artefact AnimateDread 2026-09-06 15:59:08 -04:00
  • 0dcad40f38 feat(news): export the terminal's own economic calendar - the two news setups become buildable AnimateDread 2026-09-06 15:52:25 -04:00
  • b647b745ec feat(inputs): Asset_Class - the operator declares which book applies to this chart AnimateDread 2026-09-06 14:40:49 -04:00
  • c5da78dff3 feat(setup): the level is a breakout - if the market is already through it, we PASS and log it AnimateDread 2026-09-06 12:58:41 -04:00
  • f5a3aaa131 chore(scripts): tester_results.py - one pass over a finished tester run (log counts, report, journal rows) AnimateDread 2026-09-06 12:17:42 -04:00
  • 92fa576224 feat(setup): the setup owns its trade - delegation, network confluence, feature port, expiry fallback AnimateDread 2026-09-06 12:16:41 -04:00
  • 4004c48d1b feat(journal+signals): the journal becomes a research journal, the setup becomes a voter, and the exit adapts AnimateDread 2026-09-06 11:23:27 -04:00
  • 358c8b74e0 feat(signals): CSignalInsideBarGap - the first research-validated standalone book AnimateDread 2026-09-06 02:22:13 -04:00
  • 035a4920bc docs(research): pipeline analysis, and the two scripts the research depended on AnimateDread 2026-09-06 01:57:18 -04:00
  • 52baf7cea5 refactor(signals): one predicate, one replay bracket - DRY the classic vote paths AnimateDread 2026-09-05 09:01:03 -04:00
  • 3a925fa1d9 feat(features): the classic vote vector becomes model INPUT - meta-labelling wired end to end AnimateDread 2026-09-05 08:52:25 -04:00
  • 095bd27a67 feat(signals): restore CCI, Stochastic, WPR, RVI and Parabolic SAR AnimateDread 2026-09-05 08:21:57 -04:00
  • c99020abb7 feat(signals): complete the Bill Williams suite - Alligator, Fractals, Gator, BWMFI AnimateDread 2026-09-05 08:15:34 -04:00
  • 3ed053e3a8 feat(signals): restore the classic votes as a meta-labelling PRIMARY, with fixed weights AnimateDread 2026-09-05 08:09:30 -04:00
  • 416f232226 feat(signals): freeze the signal weights after the first self-ranking AnimateDread 2026-09-05 07:38:38 -04:00
  • 57abd6d31d test(fleet): add GBPUSD, USDCAD and DAX40 as HELD-OUT instruments for the width cut AnimateDread 2026-09-05 01:53:47 -04:00
  • 4092a94b23 feat(features): second width cut 372 -> 294 - Ichimoku and Wyckoff-failed-structure off AnimateDread 2026-09-05 01:34:11 -04:00
  • 95adc798b7 feat(features): cut the input width 552 -> 372 on doubly-replicated evidence AnimateDread 2026-09-05 01:20:12 -04:00
  • abda14856f refactor(features): delete the cross-asset panel - 564 lines and 117 touch points of dead code AnimateDread 2026-09-05 01:06:25 -04:00
  • 954020b261 feat(optimizers): RMSprop, Nesterov and AMSGrad across all four backends AnimateDread 2026-09-04 23:47:19 -04:00
  • d8c629c8ba feat(research): buffer-health diagnostic for the AD suite - it found a provable dead branch AnimateDread 2026-09-04 23:11:33 -04:00
  • 2b97bf6573 refactor(tuning): delete the in-EA indicator search, replace it with an offline research pass AnimateDread 2026-09-04 22:52:15 -04:00
  • b31f4623e5 fix(memory): the 43 GB was the indicator tuner leaking full-history buffers - 33.3 GB -> 2.85 GB AnimateDread 2026-09-04 22:05:16 -04:00
  • 0405f1e8a4 feat(training): optimizer is per ARCHITECTURE, and the era cap goes 100 -> 10000 AnimateDread 2026-09-04 20:16:37 -04:00
  • 780bb97fb5 fix(persistence): a saved "complete" model came back FROZEN - migrate it to live-and-learning AnimateDread 2026-09-04 20:03:33 -04:00
  • ca47183c40 fix(labels): the ZigZag readiness test asked the wrong question - the fleet would idle all weekend AnimateDread 2026-09-04 19:46:03 -04:00
  • f78f567804 fix(lifecycle): deploying makes a model LIVE, not FINISHED - let the NN keep learning AnimateDread 2026-09-04 19:26:02 -04:00
  • fd7759596e fleet: 6 -> 5 in the symbol list - six charts left only 4.2 GB of headroom AnimateDread 2026-09-04 17:01:16 -04:00
  • 528178776c feat(fleet): 3 -> 6 instruments - pooled observations are the only remaining lever AnimateDread 2026-09-04 16:47:21 -04:00
  • 8141c4e000 feat(pca): measure the STANDARDISED component count - the one that should set k AnimateDread 2026-09-04 16:38:12 -04:00
  • c3dde7cdba fix(capacity): the budget line claimed the encoder was at its floor - it is not AnimateDread 2026-09-04 16:20:54 -04:00
  • 3988f01820 feat(capacity): state the standard parameter budget - total params <= N_eff / alpha AnimateDread 2026-09-04 15:56:52 -04:00
  • e8c7d275c4 fix(capacity): count both neuron hierarchies, and never print an uncounted zero AnimateDread 2026-09-04 15:36:46 -04:00
  • 0847529707 fix(capacity): the capacity line was counting one layer of a sequence model AnimateDread 2026-09-04 14:53:44 -04:00
  • 3580324573 feat(features): sweep the prune threshold - can correlation pruning alone get the width down AnimateDread 2026-09-04 14:42:01 -04:00
  • 10f90c5091 feat(features): name the redundant columns and run the greedy correlation prune AnimateDread 2026-09-04 14:33:17 -04:00
  • 94628ba7a9 feat(risk): make the tight stop rungs falsifiable - break-even slippage per hit AnimateDread 2026-09-04 14:23:24 -04:00
  • 01c3a1f2d8 feat(risk): measure what the STOP is costing - the exact counterfactual, in R AnimateDread 2026-09-04 13:33:11 -04:00
  • af59b41820 fix(features): a Wyckoff sentinel was 1137x wider than every other column AnimateDread 2026-09-04 13:12:01 -04:00
  • d2574f4c3a feat(label): measure the horizon frontier - and it says the horizon is NOT the lever AnimateDread 2026-09-04 12:58:42 -04:00
  • 2a82b0bd9e fix(ichimoku): read the Senkou spans RAW - stop depending on the wrapper shifted buffers AnimateDread 2026-09-04 12:41:07 -04:00
  • ceef283ead feat(features): mask OFF - stop doing the network job with a univariate screen AnimateDread 2026-09-04 12:23:37 -04:00
  • 9550f94997 fix(ichimoku): the throwaway probe turned out to be the fix - it is now a deliberate primer AnimateDread 2026-09-04 11:06:37 -04:00
  • ddb9fa0897 fix(features): the Ichimoku failure branch is PER BAR - a CopyBuffer there stalled the fleet AnimateDread 2026-09-04 10:40:16 -04:00
  • b724fc58a2 fix(ichimoku): back ON - the stdlib was never the problem, the double-shift was AnimateDread 2026-09-04 10:14:13 -04:00
  • 58982c8651 fix(measure): a zero-cost instrument has no cost axis to split on AnimateDread 2026-09-04 09:06:48 -04:00
  • cf62fc22a8 perf(features): masked out means switched off - 27.8 GB of commit becomes 3.3 GB AnimateDread 2026-09-04 09:01:33 -04:00
  • 0b2cf77606 fix(features): Ichimoku OFF - MT5 will not reliably serve the Senkou buffers AnimateDread 2026-09-04 08:53:37 -04:00
  • 97e55c3879 fix(pool): the reader kept both defects the writer was repaired for AnimateDread 2026-09-04 08:45:31 -04:00