ExitBand Pro Stop and Target Probabilities
- 指标
-
Dario Cadeddu
Quantitative Trading Systems Developer focused on risk management, market regime analysis and structured trading workflows for MetaTrader 5. - 版本: 1.100
- 激活: 10
ExitBand Pro answers one practical question before you place a trade:
"With this stop loss and this take profit, how likely is price to hit my TP first, my SL first, or neither of them within the next few hours?"
It shows these three probabilities for 9 common stop/target combinations, updated at every closed bar. The probabilities are calibrated: when ExitBand says 40%, the event happened about 40% of the time across many years of out-of-sample data. ExitBand does not predict direction and does not give entry signals. It tells you what your stop and target are realistically likely to do under current conditions.
WHAT YOU SEE ON THE CHART
A compact panel shows:
- Expected range: how far price is expected to move in the next hour (M15, M30) or the next 4 hours (H1). This distance is the unit "u".
- A table of stop loss sizes (0.5u, 1u, 1.5u, also shown in pips) against take profit sizes (1x, 2x, 3x the stop).
- For each cell, three numbers: % TP first / % SL first / % neither within the horizon.
Example (EURUSD M15): SL 1u = 8.0 pips, TP 2x = 16.0 pips: 11 / 39 / 51
This means: 11% the target is reached first, 39% the stop is reached first, 51% neither is reached within 4 hours.
The horizon is 4 hours on M15 and M30, and 16 hours on H1.
WHY IT IS USEFUL
- Size stops and targets on facts, not habits. A 2R target that is reached first only 11% of the time within 4 hours is a different trade from one that is reached 30% of the time.
- See when a target is unrealistic for the time you plan to stay in the trade. A high "neither" percentage means price is likely to drift without reaching either level.
- Adapt to conditions. The same pip distances behave very differently in the Asian session, at the London open or on a quiet Friday. ExitBand takes volatility and time of day into account.
- Compare cells before entering, instead of adjusting the stop after you are in the trade.
HOW IT WORKS
- The expected range comes from the VolBand Pro engine, which is built into ExitBand. No other product is required.
- For each of the 9 cells, a regularized multinomial logistic model estimates the three probabilities from volatility and time-of-day features. It contains no directional inputs.
- The model is re-estimated every 4 weeks on a rolling window of about 2 years. It uses only data available at the time, so there is no look-ahead and no repainting.
- Long and short trades are pooled. The probabilities apply to either direction.
VALIDATION
All tests were pre-registered: hypotheses, thresholds and code were fixed and hashed before each test was run, and each test was run once.
- Calibration (expected calibration error at most 2 percentage points, extreme deciles within 4 points) passed on all instruments in every test:
- M15: 18 major and cross pairs, 2017 to 2019, out of sample.
- M15: 10 exotic pairs never used in development (USDSEK, USDNOK, EURSEK, EURNOK, USDPLN, EURPLN, USDCZK, USDMXN, USDZAR, USDSGD), 2017 to 2026, confirmed on a second data feed.
- M30 and H1: same 28 pairs, same thresholds.
- The MetaTrader 5 code reproduces the reference model to within 1e-12, bar by bar, on M15, M30 and H1.
- A forward confirmation on new data runs from November 2026 to February 2027. Its results will be published in the product comments.
REQUIREMENTS
- MetaTrader 5, timeframes M15, M30 or H1. On other timeframes the panel says the timeframe is not supported.
- About 2.5 years of price history on the chart's timeframe, so the first model can be estimated. With less history the panel shows "Not enough history".
- Designed and validated on forex pairs.
- The first calculation takes about 20 to 30 seconds. After that, each new bar is processed instantly.
INPUTS
- Days of history with computed probabilities (default 250): how far back the probability buffers are filled.
- Show panel, Panel corner, X/Y offset, Font size, colors.
- Publish values (global variables): publishes the latest probabilities as terminal global variables.
- Export CSV (verification only): writes all values to MQL5\Files for independent checking.
FOR EXPERT ADVISOR DEVELOPERS
All values are available through iCustom and CopyBuffer, on closed bars:
- Buffer 0: expected range u (price units).
- Buffers 1 to 3: model features (volatility surprise, forward session activity, current activity).
- Buffers 4 to 30: probabilities. Buffer = 4 + 3 x cell + outcome.
- Cell = 3 x (stop index) + (target index). Stop index: 0 = 0.5u, 1 = 1u, 2 = 1.5u. Target index: 0 = 1x, 1 = 2x, 2 = 3x.
- Outcome: 0 = SL first, 1 = TP first, 2 = neither.
- Example: SL 1u, TP 2x is cell 4, so buffers 16 (SL), 17 (TP), 18 (neither).
- Empty value: EMPTY_VALUE when the model is not available yet.
LIMITATIONS (PLEASE READ)
- ExitBand is not a trading system and has no directional edge. It describes how stops and targets are likely to resolve. It does not tell you where price will go.
- The probabilities are computed on bid prices, without spread or commission. On wide-spread symbols the real chance of reaching a target is lower.
- A stop and target hit inside the same bar count as a stop, which is conservative.
- On H1 the 16-hour horizon crosses the daily rollover and the overnight session.
- The quality of the information varies by symbol. On some pairs (for example NZD crosses on H1) the probabilities stay calibrated but differ less from a fixed table.
- Past calibration does not guarantee future results. Trading involves risk of loss.
SUPPORT
Questions and feedback are welcome through the comments section or private messages on MQL5.
