The PatternX Score: how setups are scored

Every setup that survives pattern detection is given a PatternX Score before anything else happens to it. The score is a weighted sum of 13 factors, each rated from 0 to 100, with weights that add up to 1.00. No AI is involved in computing it, and the same inputs always produce the same score, in Classic and AI mode alike.

The 13 factors

What the weights tell you

Grouped by the question they answer, the 13 weights look like this:

• How good is the pattern? Pattern quality (its geometry) and the fixed reliability prior: 0.24.
• Is the breakout convincing? Breakout strength and volume: 0.22.
• Is the trend behind it? Trend alignment and the higher timeframe: 0.17.
• Can it be traded cleanly? Liquidity, spread, order book and volatility: 0.24.
• What does positioning say? Funding and open interest: 0.05.
• How has PatternX done on this market? The historical win rate: 0.08.

So geometry and the prior carry 24% of the weight, and the remaining 76% measures the breakout and the market around it. The same pattern can therefore score very differently on two markets, or on one market an hour apart.

Two caps are worth knowing. Pattern quality can add at most 13.6 points (85 × 0.16), and the reliability prior adds between 4.72 and 5.44 points. So no setup can reach 100, and the gap between the highest and lowest pattern priors is worth less than one point of score.

A worked example

For example, take a hypothetical long setup with these factor scores: pattern quality 80, breakout strength 70, volume 75, trend alignment 50, multi-timeframe 50, liquidity 70, spread 80, order book 55, volatility 90, funding 50, open interest 50, historical win rate 50 (fewer than 10 closed trades, so neutral) and a reliability prior of 62.

Multiplying each score by its weight and adding up gives 12.8 + 7.7 + 8.25 + 5.0 + 3.5 + 5.6 + 4.0 + 3.3 + 4.5 + 1.5 + 1.0 + 4.0 + 4.96 = 66.11. The setup scores 66: High, and just above the default minimum of 65. Had the volume factor been 45 instead of 75, the total would drop by 3.3 to about 63, and with default settings the setup would not reach you.

Why a fixed formula

A fixed formula has a practical advantage: every score can be traced back to its inputs, as the example above shows, and the same setup is never scored two different ways. The cost is that nothing tunes itself. If a factor matters more on one market than on another, the formula does not notice, and the weights stay the same everywhere.

Categories, confidence and thresholds

The total puts each setup in one of four categories: Elite at 80 and above, High from 65, Medium from 50 and Weak below 50. The score then becomes the signal's confidence, and two thresholds apply in order:

1. The engine floor of 50. Weak setups are discarded before any of your settings are checked.
2. Your minimum confidence. It is 65 by default and adjustable from 50 to 95. At the default, only High and Elite setups reach you. Raising it means fewer signals with higher scores; lowering it to 50 lets Medium setups through.

In AI mode, the context layer can veto a setup or re-weight its confidence before your minimum is applied. The PatternX Score itself is calculated identically in both modes.

What the score is not

• The reliability priors are fixed values, not measured statistics. Each pattern has a set prior between 59 and 68, and the priors do not change as results come in.
• The historical factor is not a backtest. Despite the word "historical", it is PatternX's live win rate on that exchange and symbol, measured per market rather than per pattern, and it only counts once 10 trades there have closed.
• The score is not a probability. A score of 70 does not mean a 70% chance of reaching the target. The "probability" PatternX displays is a heuristic derived from confidence and kept between 25 and 90, not a measured win probability.
• The score is not used to rank signals. PatternX does not order signals by score or choose the highest one. Each setup is judged on its own against the thresholds.
• Backtest scores are not live scores. Backtests run with neutral market context and no higher-timeframe or live win-rate input, so the same setup can score differently there.

Common mistakes

• Setting the minimum near the top of the range. With geometry capped at 85 and the prior at 68, the total cannot exceed about 95 even if every other factor scored 100, so a minimum close to 95 will rarely, if ever, be met.
• Reading the prior as a track record. A higher prior is not evidence that a pattern works better; it is a fixed number.
• Expecting a bigger move from a bigger score. The score says nothing about how far price will travel. Targets come from the pattern's geometry, capped by your take-profit setting.
• Looking only at the total. Two setups scoring 70 can be strong in different ways: one on volume and trend, another on a tight spread and deep liquidity.
• Treating the minimum as a profit dial. A higher minimum gives fewer signals, which is not the same thing as better results.
• Comparing scores across very different markets. Liquidity and spread readings differ systematically between a heavily traded crypto pair and a thinly traded instrument.

Limitations

The weights are fixed. They do not adapt to a market, a timeframe or recent results, and they may suit some conditions better than others. Context factors are snapshots taken at scan time and can change within seconds. Funding and open interest only carry information on markets that have them. And a high score describes how well a setup matches the rules, not what price will do next. See what PatternX can't do, and PatternX Classic for how the score fits into the product.