# Success probability

Success probability is a model-estimated percentage attached to a pre-trade verdict, expressing the model’s assessed chance of meeting its defined success criteria, not the chance of earning a return.

## What it means

A percentage beside a pre-trade verdict expresses uncertainty about an assessed outcome; it does not describe a stock’s expected price change or the return on a basket. Its meaning depends on what the model counts as *success* and the conditions under which that outcome is assessed. Without those definitions, a percentage cannot be translated into a specific chance of profit, a fill or any other particular event.

PreTrAIde attaches a model-estimated success probability to its GO, NO-GO or MODIFY verdict for each stock line and for the whole basket. It ranges from 5% to 95%. The percentage is decision support, not a guarantee or a forecast of returns.

## How it is calculated

A general way to describe a probability model is `p_raw = f(x, c)`, where `p_raw` is its initial probability estimate, `f` is the model, `x` is the information it uses, and `c` represents the conditions and success criteria being assessed. A percentage is `100 × p_raw`.

PreTrAIde’s stated output rule is `p_display = min(95%, max(5%, p_raw))`. Here, `p_display` is the probability shown with the verdict, `p_raw` is the model’s estimate before that bound, `max` applies the 5% floor, and `min` applies the 95% ceiling. Both probabilities use percentage-point units in this expression. This rule describes the **bounds**, not the model’s full probability calculation.

**Calibration** is a separate question: when a model assigns a given probability to many comparable cases, how often does its defined success outcome actually occur? For example, estimates near 70% would be well calibrated if roughly 70% of those cases met the defined criteria. Testing that requires recorded predictions, consistent outcome definitions and later observations. The 5%–95% rule alone neither establishes nor measures calibration; it also means the displayed number may differ from an unbounded estimate.

## Worked example

Consider a **hypothetical** stock basket assessed before an order is placed. Suppose a model produces a raw estimate of 98% for its defined success outcome.

1. Apply the upper bound: `min(95%, 98%) = 95%`.
2. Apply the lower bound: `max(5%, 95%) = 95%`.
3. Display 95% beside the verdict—not 98% or a claim of certainty.

If a different hypothetical raw estimate were 3%, the same rule would display 5%. An estimate of 70% would remain 70%. These round numbers demonstrate the bounding rule only; they do not reproduce a PreTrAIde analysis or establish how frequently any outcome occurs.

## Why it matters before you trade

A probability provides a compact way to read a verdict alongside its reasoning, risk assessment and execution conditions. Its precision can be misleading if treated as a measured fact about one future trade. A 70% estimate does not mean that seven of the next ten trades will succeed, and it says nothing by itself about the size of possible gains, losses or trading costs.

To interpret any such number, ask what counts as success, what time horizon and conditions apply, and whether comparable past estimates have been checked against outcomes. Even a well-calibrated model can be wrong on an individual case. Bounds below 100% and above 0% avoid displaying certainty or impossibility, but they do not prove that 95% or 5% is the true probability.

## How PreTrAIde uses it

PreTrAIde performs pre-trade analysis on a user-built basket before any order is placed. Its verdict includes the success probability, a risk assessment, a suggested execution strategy and plain-language reasoning. Its AI agents reason from quantitative scores, news and analyst ratings. A guardrail clamps probabilities to 5%–95%; a GO with a success probability under 50% is downgraded. Guardrails move verdicts only in the conservative direction.

The product facts do not specify a public success-event definition, probability-calibration results or a complete formula for generating the raw estimate. The displayed percentage should not be read as a validated historical success rate. PreTrAIde does not place orders; the user trades through their own broker.

## Common mistakes

- **Reading it as a return forecast:** A 70% probability is not a predicted 70% gain or a stated chance of making money.
- **Treating a bound as calibration:** Clamping an estimate to 5%–95% does not show that past predictions matched observed outcomes.
- **Ignoring the verdict and reasoning:** The percentage is one part of a broader pre-trade assessment, not an instruction to trade.
- **Assuming zero risk at 95%:** The upper bound is still an uncertain model estimate, and an unfavorable outcome remains possible.

---
Source: https://pretraide.com/glossary/success-probability