Dancing Experiments - Results 120806
On the Friday (27 July 2012) of the second weekend of the Dance Festival V we did several experiments to look at how different dance HUDs and dance control systems are affected by various techniques used to control dance lag.
Prior to the experiments I outlined what we were going to try to do in a blog post. Shadow Tarber and Jariah Yuhara volunteered to help and thanks go to them.
This note describes what happened.
At the scheduled time for the first experiments I invited DANCE QUEENS members to the sim to participate in the experiment. During all the experiments the number of avatars on the two sims (stage and audience) ranged from 17 to 30. Unfortunately not all the avatars on the sims were participating in the experiments since some were preparing for their dance shows.
The time available required that we change the order and number of experiments. Two sets of experiments were actually run. The first set was aimed at seeing the impact on dance lag of controlling avatar script level and avatar movement using a split sim. Six experiments were performed using the Barre, Huddles and Fleursoft HUDs. Data collected included measuring total audience sim script time (from the region controls), measuring number of avatars on the two sims, measuring performer judgment of lag, and measuring audience judgment of dance lag.
The second set of experiments was aimed at seeing the impact on dance lag of controlling avatar script level and avatar movement using a single sim. Four experiments were performed using the Barre abd Huddles HUDs. Data collected included measuring total audience sim script time (from the region controls), measuring number of avatars on the two sims, measuring performer judgment of lag, and measuring audience judgment of dance lag.
For you experimental design experts, no experiments were done to look at the confounding effects of the variables.
The results are here.
The experiments and the session for performing the experiments were done hastily and the experimental session showed it. The desired number of avatars (40+) were not on the sims, not all the avatars on the sims participated in the experiments, those who did participate were not trained on what to do prior to the experiment, and the performers were not trained prior to the experiment. The result was a relatively chaotic mess. Although many avatars did their part in terms of script levels and seating, and performers did their part, the control of the experimental conditions was very poor. The audience members and performers also were not trained on how to judge lag. Measurement of the region script time however is a good indicator of server demand from the entire audience sim and does give some insight into the server demands.
The results of the experiment are not meaningful.
Repeat the experiments with good controls by:
- Recruiting 40 avatars for a scheduled testing session
- Recruiting three avatars for a scheduled testing session
- Limiting access to the sims to performers and audience members only
- Training audience members on how to identify and quantify dance lag, and how to control avatar script levels
- Training performers on how to identify and quantify dance lag
- Planning each experiment
All I want to do is dance. Sometimes, I wonder if all this effort is worthwhile, but, I want to dance without lag. So, onward I trudge.