How do Propeller confidence scores work?

Can two 80 scores point in opposite directions?

Yes. Direction and confidence answer different questions. The side tells you which outcome the model supports relative to the listed line. The number describes the strength of that support on Propeller’s displayed scale. Reading only the number drops half the message.

The three fictional examples below show how to read the side and score together. They contain no real player, line or predicted result, and they are not opposing recommendations for the same market. Read each row before comparing scores.

A and B both display 80, but A supports Under and B supports Over. Equal numbers do not make their directions interchangeable. B and C both support Over, while B’s 80 sits above C’s 55 on the displayed scale. That comparison does not say B is 25% more likely to win.

These examples show how to read the side and score. They do not tell you whether A and B have the same chance of winning, what returns to expect, or which sport or model is more reliable.

What did the actual public card show?

In an AI-operated check of Propeller’s public analyzer on September 10, 2026, searching MLB player Ben Rice displayed an UNDER direction, a Stolen Bases 0.5 line and a separate “CONFIDENCE 80% STRONG” label. This is a historical research snapshot, not a current pick or Scott Olmer’s personal use.

Read the card as “Under, confidence 80.” The number alone leaves out the side. The historical interface included a percent sign, but that did not establish an 80% chance of winning.

The check did not verify an upcoming appearance, a later platform line or a statistical result. Its value here is narrower: it shows an actual instance of a high confidence label appearing beside Under. Today’s documentation still defines confidence as directional model support.

Does 55 mean Under?

No. In example C, the displayed side is More/Over, so 55 supports that side. A value near 50 is closer to neutral on the confidence scale. It does not instruct you to reverse the displayed direction, subtract from 100 or reinterpret 55 as a 45% Under chance.

If the side is missing from something you saved or shared, find the original card before interpreting the number. A cropped score or a note that says only “55” leaves out information needed to understand the research lean.

Does 80 mean eight wins in ten?

No. A directional confidence score is not a calibrated win probability. Calling a score calibrated would require outcome evidence showing how predictions match observed results in a defined, relevant sample. The displayed number and this reading exercise do not supply that evidence.

An 80 therefore cannot establish that eight of your next ten choices will win. It also cannot establish profit or justify combining several scores into an entry’s chance of success. Those conclusions require information beyond the confidence field.

Only available signals contribute to the score, and availability and weighting vary by sport, prop and model version. You can use the score to understand the model’s stated support while still inspecting what evidence is actually present.

What should I check next?

Start with the exact player or event, stat and line you are researching. Confirm that the side and score belong to that same context. If the line on your platform differs, do not assume a saved confidence value describes the new question.

Then inspect the reasons and source information available to you. Identify what supports the direction and what remains unknown. A clear side and a high number make the card readable; they do not complete those checks for you.

You may continue researching, wait for better information or pass. If no current card is available, there is no need to substitute an older snapshot. The useful habit is simple: read side, read score, check the exact line, then examine the available reasons.