With the implementation done, experiments can now be conducted with the simulation to study the various aspects of emotions and behaviour.
(Team 0 with emotions and team 1 without emotions)
In this experiment, the aim is to find out how a team with emotions will perform against a team without emotions and why does it perform in this manner. 100 of such experiments, running 5000 cycles each time, were carried out.
The following are extracts of the parameter file used. The random seed is change at each run.
Parameter File for Simulation - Experiment 1
================================
Random Seed
[Seed] [1]
Pause time
[Pause] [0]
No_of_cycles/turn for current simulation
[Cycle] [5000]
Size of World
[World_Size] [200] [200]
Number of Emotional Variable
[Em_No] [3]
[Team0]
[Em_Stat] [1] [1] [1]
[Em_Thres] [75.0] [25.0] [75.0] [25.0] [75.0] [25.0]
[Team1]
[Em_Stat] [0] [0] [0] Note team 1 has emotions switch off.
[Em_Thres] [75.0] [25.0] [75.0] [25.0] [75.0] [25.0]
[Rule_No] [26] No of rules
[Defensive] [8]
[Offensive] [16]
[Others] [2]
Defensive
[0.5] [1] [1] [2] All the strengths of the rules are set to 0.5 and the effects of
[0.5] [2] [1] [2] the emotion on the rules follows that as specify in chapter 4
.
.
.
[Distance_Def] [5]
[Near] [0.05] Defination of 'near'
[Mid_Near] [0.25] Defination of 'Mid_Near'
[Middle] [0.45] Defination of 'Middle'
[Mid_Far] [0.60] Defination of 'Mid_Far'
[Far] [1.00] Defination of 'Far'
[Team_Name] [STAR WARS] Team Name
[Agent_No] [6] Number of Agents Six agents per team
[ID] [1] Both teams have similar "type" of players.
Player capabilities
[Sight] [150.0] Sight is set to 3/4 of the field.
[Speed] [5.0] Speed is slightly faster for "offensive" players.
[Tackling] [0.3] Player 1 - 3 have better offensive abilities, ie
[Throwing] [0.7] catching and throwing and player 4 - 6 has better
[Catching] [0.7] defensive ability, tackling.
Player current emotional state
[Joy] [50.0] Emotional state is set at 50.0 for all.
[Pride] [50.0]
[Appreciation] [50.0]
.
.
.
[ID] [4]
Player capabilities
[Sight] [150.0]
[Speed] [3.0]
[Tackling] [0.70]
[Throwing] [0.30]
[Catching] [0.30]
.
.
Next team....
The results collected were,
After 100 runs, the following statistics about final scores were obtained.
| No. of Wins | |
| Team 0 (with emotions) wins | 76 |
| Team 1 (without emotions) wins | 11 |
| Draws | 13 |
Clearly, it shows that the team with emotions performs better. A closer look into the emotion states of the team with emotion was necessary to investigate why this was so.
Below is the extracts of a run in which the team who emotions fought back a deficit and won 24 - 18.
| Output (Cycle no, score, Value of Joy, Pride and Appreciation) |
Explanation. (Note that team with emotions' score is the first value.) |
| Cycle: 100 0 - 0 49.5 49.2 51.4 |
At cycle 100, score was 0 - 0. Emotion is constant for all |
| Cycle: 650 0 - 1 39.6 41.4 40.6 |
Team 1 takes the lead. Joy in the team drops, Pride towards own's team decreases and aggression against opposition increase. |
| Cycle: 1100 0 - 3 26.4 28.2 5.34 |
At cycle 1100, team 1 led by 0 - 3 and all variables dipped tremedously. Appreciation variable dipped below threshold and a slightly more aggressive play is adopted. |
| Cycle: 1150 1 - 3 31.1 45.6 19.1 |
Team 0 finally scored it's first goal and emotions variables are raised slightly only as they are still trailing 1 - 3. |
| Cycle: 2900 9 - 12 29.8 75.5 28.2 |
At cycle 2900, team with emotions still trails 9 - 12. The joy variable is low but the pride variable has crossed the threshold. A closer look into the action revealed a series of successful action which results in the high pride value. |
| Cycle: 3650 14 - 14 78.6 69.6 33.4 |
Here, the score is tie. The reason for this fightback was due to the high pride in its own team-mates which results in more throws and runs leading to goals. Joy has raised considerably and crossed the threshold resulting in somewhat more complacent play with more failure in actions and a dipping pride. |
| Cycle: 5000 24 - 18 92.3 80.4 69.4 |
Runs ends with the team 0 pulling away with a strong win. Emotion values are move upwards. |
The next sample run shows team 0 leading until a series of failed action which results in their loss.
| Output (Cycle no, score, Value of Joy, Pride and Appreciation) |
Explanation. (Note that team with emotions' score is the first value.) |
| Cycle: 200 2 - 0 77.5 79.2 69.2 |
At cycle 200, team 0 takes a 2 - 0. Joy and Pride soared above the thresholds. |
| Cycle: 2150 7 - 1 98.6 99.1 91.8 |
Team 1 takes a tremendous 7 - 1 lead. All variables above their thresholds and this results in aggressive play which cause the team with emotions to perform even better. |
| Cycle: 3550 16 - 15 45.1 29.2 50.4 |
A series of failed actions cause a dip in all emotion variables which resulted in the other team gaining on them. |
| Cycle: 4350 18 - 24 6.3 2.46 11.6 |
A check with the results shows that a continuing series of failed
actions led to team 1 taking a led. All emotion variables are now below the lower threshold and a somewhat more aggressive play is again adopted. |
| Cycle: 5000 22 - 25 17.1 18.4 13.3 |
Run ends with Team 0 catching up slightly. Emotion variables up slightly. |
This first sample run shows that a series of successful actions can cause a team with emotion (team 0) to play more aggressively, more throws and run forwards, and cause a change of tide and win the game. A look into the other runs shows that if the emotion variables values are high, it results in more aggressive play and more wins. The second sample run however shows that there is an occasional hitch, as a result of the random number, can cause a series of failed actions and the emotion variables drops and "loses confidence" resulting in the other team without emotions catching up and winning. The random number in this case provided some sort of an element of luck and irrational gameplay by players which exists in games.
However, the second sample run is a rare case where the tide is turned when the team with emotions are leading and the emotion variables values are high. The general situation is the first sample run, where the team with emotions wins when they start lead to the game. This is due the increases aggression when the upper threshold is crossed.
With this experiment comes a question. Does a aggressive team means a "sure win" team? This question led to the second experiment.
(Team 0 (a much weaker team) with emotions and team 1 without
emotions)
This second experiment was a follow-up to the first. The aim is to find out whether a aggressive physically weaker team with emotions will still perform better and why. The situation is the same as previously, team 0 with emotions and team 1 without emotions but this time around, team 0 is a much weaker team with low physical values and high emotion values so it start off aggressively. The upper threshold are lower and the lower thresholds raised so that will be aggressive most of the time. Again 100 of such experiments, running 5000 cycles each time, were carried out.
The following are extracts of the parameter file used. The parameters on the whole are the same with experiment one with changes in thresholds and physical abilities. The random seed is again change at each run.
Parameter File for Simulation - Experiment 2
================================
Random Seed
[Seed] [1]
.
.
.
[Team0]
[Em_Stat] [1] [1] [1] Switch Emotion Variable on or off
Set Threshold for emotions (first the upper, then the lower thresholds)
[Em_Thres] [60.0] [40.0] [60.0] [40.0] [60.0] [40.0]
Emotion Thresholds are lower and raised respectively to increase
aggression
[Team1]
[Em_Stat] [0] [0] [0] Note team 1 has emotions switch off.
.
.
.
[Team_Name] [STAR WARS] Team Name
[Agent_No] [6] Number of Agents Six agents per team
[ID] [1]
Player capabilities
[Sight] [150.0] Sight is set to 3/4 of the field.
[Speed] [5.0] Speed is slightly faster for "offensive" players.
[Tackling] [0.1] Physical abilities are now lower to illustrate a
[Throwing] [0.3] weaker team.
[Catching] [0.3]
Player current emotional state
[Joy] [90.0] Emotional state is set at a high of 90.0 for all to
[Pride] [90.0] illustrate a aggressive team.
[Appreciation] [90.0]
.
.
.
[ID] [4]
Player capabilities
[Sight] [150.0]
[Speed] [3.0]
[Tackling] [0.30]
[Throwing] [0.10]
[Catching] [0.10]
.
.
.
[Team_Name] [STAR TREK] Team Name Team 1 has the parameters as
experiment one.
[Agent_No] [6] Number of Agents
[ID] [1]
Player capabilities
[Sight] [150.0]
[Speed] [5.0]
[Tackling] [0.5]
[Throwing] [0.7]
[Catching] [0.7]
Player current emotional state
[Joy] [50.0]
[Pride] [50.0]
[Appreciation] [50.0]
.
.
.
[ID] [6]
Player capabilities
[Sight] [150.0]
[Speed] [3.0]
[Tackling] [0.70]
[Throwing] [0.50]
[Catching] [0.50]
The results collected were the same as experiment one,
After 100 runs, the following statistics about final scores were obtained.
| No. of Wins | |
| Team 0 (with emotions) wins | 0 |
| Team 1 (without emotions) wins | 100 |
| Draws | 0 |
There were no wins for the weaker team with emotions. In most runs, they beaten by more than five goals.
Below is an extract of a typical run in which the team with emotions but weaker physical abilities are decisively defeated.
| Output (Cycle no, score, Value of Joy, Pride and Appreciation) |
Explanation. (Note that team with emotions' score is the first value.) |
| Cycle: 50 0 - 0 85.2 85.5 85.1 |
At cycle 50, score was 0 - 0. Emotion variables are still high from initial settings. |
| Cycle: 150 0 - 1 68.2 77.1 81.3 |
Team without emotions scores and emotion variables starts to decrease |
| Cycle: 2500 6 - 12 7.39 7.06 8.6 |
Half way through the simulation, emotion variables values are very low and team 1 who have better physical continues to lead. |
| Cycle: 5000 12 - 25 0 0.127 0 |
Simulation ends with all three emotion variable at close to zero or zero values. And the team with emotions but weaker physical is convincing beaten. |
The results clearly shows that an aggressive team cannot replace a physically competent team in this simulation. Emotions have an definite effect on a team's performance. However, it must be matched with adequate physical abilities. The system is therefore not built in such a manner that the aggressive team always wins. There are elements of randomness and physical abilities parameters.
(Both teams with emotions)
Experiment three was aim to study if emotion stabilisation exists in the simulation when both teams have emotion activated and both teams have similar physical attributes.
Again 100 of such experiments, running 5000 cycles each time, were carried out.
The following are extracts of the parameter file used. Two teams with similar parameters are used this time around. The random seed is again change at each run.
Parameter File for Simulation - Experiment 3
================================
Random Seed
[Seed] [1]
.
.
.
[Team0]
[Em_Stat] [1] [1] [1] Switch Emotion Variable on or off
Set Threshold for emotions (first the upper, then the lower thresholds)
[Em_Thres] [75.0] [25.0] [75.0] [25.0] [75.0] [25.0]
[Team1]
[Em_Stat] [1] [1] [1] Note team 1 has emotions switch off.
[Em_Thres] [75.0] [25.0] [75.0] [25.0] [75.0] [25.0]
Both teams have emotion switched on and the same threshold.
.
.
[Team_Name] [STAR WARS] Team Name
[Agent_No] [6] Number of Agents Six agents per team
[ID] [1]
Player capabilities
[Sight] [150.0] Sight is set to 3/4 of the field.
[Speed] [5.0] Speed is slightly faster for "offensive" players.
[Tackling] [0.5] Normal physical abilities.
[Throwing] [0.5]
[Catching] [0.5]
Player current emotional state
[Joy] [50.0] Emotional state is set at 50.0 for all.
[Pride] [50.0]
[Appreciation] [50.0]
.
.
.
[Team_Name] [STAR TREK] Team Name Team 1 has the parameters as Team 0
[Agent_No] [6] Number of Agents
[ID] [1]
Player capabilities
[Sight] [150.0]
[Speed] [5.0]
[Tackling] [0.5]
[Throwing] [0.5]
[Catching] [0.5]
Player current emotional state
[Joy] [50.0]
[Pride] [50.0]
[Appreciation] [50.0]
.
.
.
The results collected were,
After 100 runs, the following statistics about final scores were obtained.
| No. of Wins | |
| Team 0 (with emotions) wins | 36 |
| Team 1 (without emotions) wins | 41 |
| Draws | 23 |
The number of wins for both teams were quite even, with very close score line of less than three goal separating the teams. There are also quite a number of draws at 23%.
Below is an extract of a typical run.
| Output (Cycle no, score, Value of Joy, Pride and Appreciation) |
Explanation. (Note that team with emotions' score is the first value.) |
| Cycle: 200 0 - 0 Joy 46.3 54.1 Pride 48.9 52 Appreciation 48 1 51.9 |
At cycle 200, no score. Emotion remains relatively stable around 50s. |
| Cycle: 250 0 - 1 Joy 19.8 72.2 Pride 36.7 60.8 Appreciation 39.6 65.5 |
Team 1 scored a goal and it's emotions increases. Team 0's emotions decrease due to the goal. |
| Cycle: 300 1 - 1 Joy 47.5 59.9 Pride 44.9 46.2 Appreciation 49.6 55.4 |
Team 0 equalised and emotions are once again stabilised again. |
| Cycle: 1450 5 - 6 Joy 48.9 81.5 Pride 35.1 78.3 Appreciation 53.2 79.8 |
Team 1 scored another goal |
| Cycle: 2000 6 - 6 Joy 59.6 63.1 Pride 61.6 61.1 Appreciation 58.3 50.6 |
Team 0 scores and emotions stabilises again. |
| Cycle: 3850 15 - 11 Joy 75.8 9.17 Pride 93.6 5 21 Appreciation 76.4 6.14 |
Team 0 scored a series of goal with a series of actions in favour of them. Emotions now de-stabilises. |
| Cycle: 5000 18 - 17 Joy 49.5 71.9 Pride 69.8 71.8 Appreciation 69.2 59.2 |
At the end of simulation, team 1 had a series of successful actions resulting in the other team catching up. Emotions levels somewhat moves closer together again. |
The results of experiments three shows that there is some sort of emotion stabilisation when there are no significant events to disrupt it. Significant events like the random running in favour of one team can change the course of events and de-stabilised the values of the variables as shown in cycle 3850 and the end of simulation. It shows that if two team equally match team were to play, the luckier team will win or the team which do not play the rational way will win or lose. The large number of draws were a result of even random number which didn't provide an edge for either teams.
A follow-up experiment to this one could be to narrow the upper and
lower thresholds to simulate agents with more volatile emotions and to see if emotion
stabilisation still exist. However time has run out at this point and this experiment was
not conducted. The next concluding chapter summarises the results obtained.
Copyright © 1995 to Hay Yong Joo