How The System Scores

LiveJudge uses any of 4 different methodologies to arrive at normalized scores based on the number of contestants and judges within a particular class. Beginning in late January 2026, event admins will be able to choose the method that suits their competition the best.

We further allow users to scale those scores (e.g. 1000 is the best score always) like teachers often "curve" scores for tests. 

Normalization Methodologies

Judges have different “centers” and “spreads.” One judge may hand out mostly 9s while another prefers 6s. Normalization aligns those personal scales so the contestant is rewarded for performance—not for drawing the “easy” judge.

Raw Average

There will be no mathematical adjustment to standardize the scores entered by your judges. All judge scores within a round will be averaged for each contestant.

High/Low Judge Drop

The admin is given the opportunity to drop the top and bottom scores from a chosen number of judges during a given round. There must be at least FIVE assigned judges per round in order to implement this method. Note that this is not the same as round drops, described below.

Z-Score Normalization

Z-score normalization, also known as standardization, transforms a dataset to have a mean of 0 and a standard deviation of 1. This is achieved by subtracting the mean from each data point and then dividing by the standard deviation. The resulting z-scores indicate how many standard deviations a data point is away from the mean. 

Think of each judge’s score sheet like a ruler with its own zero mark and spacing.

1. Find the judge’s average (their personal middle).
2. Measure how far each score sits above or below that middle.
3. Divide by their typical spread so big and small rulers become the same length.

After that quick math:

• 0 = exactly that judge’s average.
• Positive numbers = above their normal.
• Negative numbers = below.

Because every judge’s ruler now shares the same center (0) and spacing (1 unit = one typical jump), generous and harsh judges cancel out and the playing field levels automatically.

Olympic-Z

Considered the gold standard for scoring, this method uses Z-Score normalization on the net raw scores and then drops high/low per your settings.
 

Round Drops

If your event has 3 or more rounds, you may elect to ignore all but the final round to determine the winner (you can set the number of competitors who advance after a given round) or average or accumulate the scores over all rounds. In the latter scenarios, the software gives you the opportunity to drop the lowest X rounds. This option does NOT address biases or tendencies among judges. Rather, it encourages competitors to "go for it" knowing that if they mess up, that round can be disregarded. 

Scaling

Absolute/Standard Scaling

This is a proportional method of fitting all contestants within a given range (e.g. 1-100 or 500-1000). This method awards a final score based on the raw score compared to the total best raw score possible. 

Example:

• Winner gets a 93.5/100 possible points. 
• On a scale of 0-1000, the 93.5 is equal to 93.5% of 1000 = 935
• On a compressed scale of 500 - 1000, 93.5% of 500 (the range) = 467.5. Everyone gets 500 base points, so the winner gets 467.5 + 500 = 967.5

Top Pinning Scaling

Top pinning differs from absolute scaling in that the admin only specifies what the winner's score must be (e.g. 1000). From there, all other competitors are measured as a percentage of the winner's unscaled score multiplied by 1000. 

Example:

•  Using the winner score of 93.5 and a second-place score of 90, we award the winner 1000 by rule.
• The second-place finisher got 90/93.5 = 96.25% of the winner's score. 96.25% of 1,000 = 962.5

Take‑away for Event Hosts

The LiveJudge system allows you to select your preferred normalization, scaling and round-drop rules for each class in your competition class library. This gives you enormous flexibility to run multiple rubrics within and across events. If you always use the same methodology, just set it once within a class and use that class across events.