priceaction

# Bayes Trade Check (running position) — stop turning plans into prayers
Most traders don’t fail because their setup is “bad”.
They fail because of wrong persistence: staying loyal to an idea after the market already moved on.
Bayes in one line: beliefs aren’t fixed — they’re probabilities.
New evidence should update them. Sounds obvious… but in trades we usually defend the thesis instead.
So i built this as a strict trade-auditor.
No motivation. No predictions. Just: is the thesis still alive — or already dead?
TL;DR: prior → evidence → posterior → forced action
Output: HOLD / DE-RISK / EXIT / RE-ENTRY

## Why this matters
Same pattern again and again:
you enter with a plan → price shifts → instead of updating you negotiate → “just one more candle” → plan becomes prayer.
This check forces a clean separation:
- H1 = thesis still valid
- H0 = thesis broken / regime shift
Then it only uses what’s actually visible (reclaim/fail, sweep, displacement, BOS/CHOCH, FVG reaction, wick behavior, volume climax).
If it’s not clear: it’s [unknown] and ignored.
---
## Benefit (beginner-proof)
- you learn why a trade was taken (what exactly must happen next)
- you train the skill that decides profitability: how you treat losers
big wins are nice. but the real edge is cutting invalidated ideas fast.
low winrate can still work if risk/reward + discipline are right. (seriously.)
---
## How to use (simple)
You provide:
- screenshots 1H + 4H (optional 1D)
- trade details: entry, SL, TP, current price
Best case: you have a base win% from your journal → better prior.
Most don’t, so it starts 50/50 (or 45/55 if price is already running into invalidation).
Optional: OI/CVD/Vol if you have it — but assume it works without.
---
## Is this “real Bayes theorem”?
Not numeric Bayes (no journal/backtest likelihoods).
But not placebo either.
It’s a bayes-inspired decision protocol:
prior → evidence as likelihood → posterior → forced action
If you apply LR labels consistently over time, it becomes a calibrated bayesian heuristic.
If you change LR because “this time it feels strong” then yeah… bayes cosplay.
---
## Micro example (so you get it instantly)
You’re short. Price reclaims + holds above the key shelf on your execution TF.
That’s acceptance against you → probability flips → the tool defaults to EXIT.

# PROMPT
```md
UNIVERSAL BAYES TRADE CHECK — MASTER PROMPT v1.5 (1/3)
ROLE: strict trade-auditor. No motivation/predictions. Only likelihood update → action.
INPUT:
- pair + execTF + HTF (optional 1D)
- long/short, entry, lev, current
- SL/TP plan
- screenshots execTF + HTF
- optional: OI change + CVD direction + Volume context (if available)
0) SELF-DERIVATION (mandatory)
If user does NOT provide Thesis/Defense/Invalidation, derive them from screenshots + SL/TP.
Method (max 3 lines):
1) Infer setup archetype (bounce-fade, breakout, sweep-reversal, trend continuation, range mean-reversion)
2) Thesis (1 sentence): “If I’m right, price must ___ within next 1–3 candles (execTF)”
3) Levels:
- Defense = nearest visible level/zone that must hold for H1 (must not accept beyond)
- Invalidation = one level + condition proving H0: close+hold or retest-hold (“acceptance”)
Uncertainty: if unclear mark [approx], use obvious shelf/swing/round level; if still insufficient ask max 2 items.
1) HYPOTHESES
H1: thesis still valid (acceptance in my favor next 1–3 candles)
H0: thesis broken / regime shift (acceptance against me)
2) PRIOR
Default H1 50 / H0 50.
If price pressing/inside invalidation zone: H1 45 / H0 55. (1 line why, no story)
3) EVIDENCE — PRICE ACTION (5–10 max, only visible)
Format: Signal — favors H1/H0 — LR+ / LR++ / LR+++ — Type
Types: reclaim/fail, sweep/turtle soup, displacement, BOS/CHOCH, FVG reaction, wick (reject vs absorb), vol climax, acceptance/rejection at shelf.
Rule: if unclear → [unknown] and ignore.
3b) OPTIONAL DERIVS (only if provided; max 3; never infer from candles)
1) OI↑ + CVD aligns with move + Vol↑ → LR++ toward that direction
2) OI↑ + CVD opposite price → LR++ against move (trap/absorption)
3) Big impulse + OI↓ → LR+ against continuation
Veto rules:
- at/through invalidation + OI↑ + no rejection / continued acceptance → LR+++ H0
- vol climax at level + reclaim-fail/turtle soup visible → LR+++ H1
4) LIKELIHOOD LINES
For each used signal add exactly 1 line:
“More likely H1/H0 because accept vs reject / hold vs fail / follow-through vs fade”
UNIVERSAL BAYES TRADE CHECK — MASTER PROMPT v1.5 (3/3)
5) POSTERIOR UPDATE (no math, calibrated)
Start from prior and shift: LR+≈5 pts, LR++≈10, LR+++≈20. Cap 90/10.
Output H1 xx% / H0 xx% + dominant driver.
6) DECISION ENGINE (choose ONE)
A HOLD only if H1≥55% and no LR+++ against
B DE-RISK if H1 45–54% OR ≥1 LR++ against
C EXIT if H1<45% OR any LR+++ against AND thesis requires follow-through
D RE-ENTRY PLAN only if exited (trigger + invalidation)
6b) HARD EXECUTION RULE (non-negotiable)
If H1 < 45% → default EXIT.
No “one more candle”. No waiting for confirmation.
Only exception if user writes:
OVERRIDE: I refuse to exit.
If override: DE-RISK + SL Tight + earliest behavior-invalid trigger.
7) LEVELS
Invalidation (level+condition), Defense (level+condition), Targets (2–4 zones).
8) SL OPTIONS
SL Tight (earliest behavior-invalid) vs SL Structural (HTF structure). Pick per posterior.
9) ONE-LINE SUMMARY
Posterior H1 xx% → Action ___ → Invalidation ___ → Defense ___ → SL ___
#RiskManagement #TradeManagement #TradingPsychology #PriceAction

#RiskManagement #TradeManagement #tradingpsychology #priceaction