Warrior_EA/Variables/ConfidenceBridge.mqh

//+------------------------------------------------------------------+
//|                                                       Warrior_EA |
//|                                                     AnimateDread |
//|                                                                  |
//+------------------------------------------------------------------+
//--- TELEMETRY ONLY SINCE 2026-08-25 - NOTHING HERE MAY STEER A TRADE.
//---
//--- These globals used to be a control path: OpenParams() refreshed them, and the five "Intelligent"
//--- trade-management modes (entry / SL / TP / trailing / lot size) read them back to size real risk.
//--- All five are gone, and the reason is worth keeping next to the numbers themselves: the model's
//--- confidence is KNOWN TO BE MISCALIBRATED (it over-calls by roughly an order of magnitude against
//--- the label prior), so every one of those modes was scaling money by a quantity whose units were
//--- never established. Uncalibrated is not the same as uninformative - it may well rank trades
//--- correctly even while its absolute level is meaningless - which is exactly why the numbers are
//--- still RECORDED here and journalled per trade (aiConfidence / dbConfidence in
//--- Database\TradeJournalManager.mqh, bucketed against realised outcome in TradeJournalReport.mqh).
//--- That report is the evidence a future confidence-scaled mode would have to produce first.
//---
//--- THE RULE: a write to any global in this file must be observable only in the journal, the panel,
//--- or a log line. If a new reader would change an order's price, size, stop or lifetime, it does
//--- not belong here - it belongs behind a measurement.
double g_AISignedConfidence = 0.0;   // -1..1, sign = direction, magnitude = AI confidence; 0 if no AI filter
double g_DBConfidence = 0.0;         // 0..1, historical time-based win rate of the active pattern set
// Live per-tick signed AI confidence (-1..1, sign = predicted direction, magnitude = confidence),
// refreshed every tick/timer from the active AI signal's SignedAIConfidence() in
// CExpertSignalAIBase::ScheduleTrainingIfNeeded(). Kept because it is the only read available while a
// position is already OPEN (an open position generates no further OpenParams() calls), which is the
// interesting half of any future confidence-vs-outcome study - does conviction decay before the loss?
double g_LiveAISignedConfidence = 0.0;
//--- THE LIVE AI VOTE BOARD, and it exists to keep one signal out of another signal's business.
//---
//--- g_LiveAISignedConfidence above is written once per tick and read by the intelligent trailing stop.
//--- With ONE AI signal that is exactly right. With several (an ensemble) every member wrote it
//--- unconditionally, last writer won, and an LSTM entry's stop was moved by whichever member's OnTick
//--- happened to run last (user-identified 2026-08-17).
//---
//--- The obvious patch - have a member average its siblings - trades a scheduling bug for a coupling bug,
//--- and this EA is deliberately built the other way: every signal runs in its own instance, minds its own
//--- state, and VOTES to the orchestrator, which is the only thing allowed to combine opinions. So the
//--- board below is a publish/aggregate pair with that split enforced by shape:
//---   - a member writes ONLY its own slot (PublishAIVote), never reads another's;
//---   - the orchestrator (CExpertSignalCustom::LiveSignedConfidence) calls AggregateAIVotes() to combine.
//--- Same contract as every other global here: one-way, same-tick, no back-pointer into the wizard chain.
//---
//--- The mean, matching the live open decision (a weighted-average vote, not unanimity) and the ensemble
//--- gate's combined-vote score. A member that abstains publishes 0 and dilutes, exactly as it does there;
//--- a member still training publishes 0, so a half-trained ensemble reads WEAKER rather than louder,
//--- which is the safe direction for something that can close a position.
#define AI_VOTE_BOARD_MAX 8
double g_AIVoteBoard[AI_VOTE_BOARD_MAX];
bool   g_AIVoteBoardUsed[AI_VOTE_BOARD_MAX];
void PublishAIVote(int slot, double signedConfidence)
  {
   if(slot < 0 || slot >= AI_VOTE_BOARD_MAX)
      return;
   g_AIVoteBoard[slot] = signedConfidence;
   g_AIVoteBoardUsed[slot] = true;
  }
double AggregateAIVotes(void)
  {
   double sum = 0.0;
   int n = 0;
   for(int i = 0; i < AI_VOTE_BOARD_MAX; i++)
      if(g_AIVoteBoardUsed[i])
        {
         sum += g_AIVoteBoard[i];
         n++;
        }
   return (n > 0) ? sum / n : 0.0;
  }
//--- CombinedConfidence(source) lived here and collapsed the two numbers above into one 0..1
//--- magnitude per the Confidence_Source input. Removed 2026-08-25 with that input and with every
//--- mode that consumed its result. The journal records the two numbers SEPARATELY and always did,
//--- so nothing is lost: a blend can be computed from the recorded columns whenever there is a
//--- reason to believe in one, which is the correct order for that operation.
//+------------------------------------------------------------------+