Trav'ling Canuck
Deity
- Joined
- Feb 7, 2018
- Messages
- 2,959
I want to better understand:
Benchmark Objectives / Methodology
To benchmark AI victory times, I will not pursue any victory conditions myself (modified post game 2: I will pursue a Science victory up until the last Space Race project but not complete it). I will not actively attempt to hamper an AI victory. I'll record AI victory time, type of victory, and who was leading in other victory conditions at the time.
To benchmark AI aggressiveness impact, I'll maintain a modest military that should be smaller than neighbouring AI forces and I will not try to cultivate good relations (but I will accept overtures of friendship from the AI). Post-Game 2 I also take steps to make it look like I am closing in on a victory by building a Spaceport and running early Space Race projects. I'll record in general terms the aggressive actions taken by the AI, and then summarize each on the following Impact scale, where each successive level has roughly twice the impact on the player's pursuit of victory:
None (0): AI's action had no impact on the player
Very Low (1): AI's damage was limited to plundering trade routes or pillaging improvements or buildings or a post Classical era Builder or forced a partial shift to a military economy or neutralizing a governor
Low (2): AI killed military unit(s) or a Classical era Builder but did not require a general shift to a military economy for more than 5 turns to deal with or Spy stole gold or a Great Work
Modest (4): AI did not take any cities but pillaged a non-Spaceport district(s) or an Ancient era Builder or post Classical Settler or forced a shift to a general military economy for more than 5 turns
High (8): AI took one or more cities that were subsequently recovered or captured a Classical era Settler that could not be recovered or pillaged a Spaceport district
Very High (16): AI took one or more cities that could not be recovered before accepting peace or captured an Ancient era Settler that could not be recovered
Test Game "Rules"
Subject to some flexibility in game to better achieve the overall benchmarking objectives in the context of the specific game, I'll follow the following rules during the test games:
Test Game Summary Results
All test games posted are played using R&F and all DLCs with the latest patch as of the date of the post. Beginning Test Game 12, this includes the July 2018 patch. These games combined with prior game results as the July 2018 patch purports to deal only with cross-platform support, with no indication than any AI changes were made.
Results as of test game 15:
AI Victory Times
Earliest Observed Victory: 280
Avg. Observed Victory Turn: 318
% Science Victories: 93%
% Cultural Victories: 7%
AI Aggressiveness Impact
Player Eliminated %: 7%
Avg. Aggression Impact: 13.5
Max Aggression Impact: 22
Avg. Aggression Events: 6
Max Aggression Events: 17
If Wars, Avg. Events: 2.5
Max Wars: 4
If Aggressive Spies, Avg. Events: 5.4
Max Aggressive Spy Events: 15
Emergencies
Avg Emergencies Participated in by AI: 3.8
% that Succeeded: 32%
City States
Avg T of 1st death: 21
Avg Surviving at T50 8.6
Avg Surviving at T100 7
Avg Surviving Game 4
Wonders
Wonder // Earliest Observed Build Turn // Avg Observed Build Turn
Alhambra 118 137
Amundsen-Scott Station never built n/a
Angkor Wat 146 213
Apadana 54 102
Big Ben 161 179
Bolshoi Theatre 266 307 [note: only 3 observed constructions]
Broadway 192 230
Casa de Contratacion 146 192
Chichen Itza 93 109
Colosseum 76 125
Colossus 76 112
Cristo Redentor 175 222
Eiffel Tower 236 269
Estadio do Maracana 224 268
Forbidden City 151 174
Great Library 59 83
Great Lighthouse 63 106
Great Zimbabwe 140 171
Hagia Sophia 130 157
Hanging Gardens 23 38
Hermitage 202 216
Huey Teocalli 115 148
Jebel Barkal 53 83
Kilwa Kisiwani 113 156
Kotoku-In 101 112
Mahabodhi Temple 73 86
Mausoleum Halicarnassus 105 148
Mont St. Michel 132 173
Oracle 57 67
Oxford University 152 175
Petra 76 111
Potala Palace 129 147
Pyramids 39 68
Ruhr Valley 148 190
Statue of Liberty 158 185
St. Basil's Cathedral 125 167
Stonehenge 24 31
Sydney Opera House 222 271
Taj Mahal 136 180
Temple of Artemis 35 60
Terracotta Army 121 145
Venetian Arsenal 189 230
- How quickly the AI can win; and,
- How willing / effective the AI can be at interfering with the human player's plans if the player does not put some effort into diplomacy.
Benchmark Objectives / Methodology
To benchmark AI victory times, I will not pursue any victory conditions myself (modified post game 2: I will pursue a Science victory up until the last Space Race project but not complete it). I will not actively attempt to hamper an AI victory. I'll record AI victory time, type of victory, and who was leading in other victory conditions at the time.
To benchmark AI aggressiveness impact, I'll maintain a modest military that should be smaller than neighbouring AI forces and I will not try to cultivate good relations (but I will accept overtures of friendship from the AI). Post-Game 2 I also take steps to make it look like I am closing in on a victory by building a Spaceport and running early Space Race projects. I'll record in general terms the aggressive actions taken by the AI, and then summarize each on the following Impact scale, where each successive level has roughly twice the impact on the player's pursuit of victory:
None (0): AI's action had no impact on the player
Very Low (1): AI's damage was limited to plundering trade routes or pillaging improvements or buildings or a post Classical era Builder or forced a partial shift to a military economy or neutralizing a governor
Low (2): AI killed military unit(s) or a Classical era Builder but did not require a general shift to a military economy for more than 5 turns to deal with or Spy stole gold or a Great Work
Modest (4): AI did not take any cities but pillaged a non-Spaceport district(s) or an Ancient era Builder or post Classical Settler or forced a shift to a general military economy for more than 5 turns
High (8): AI took one or more cities that were subsequently recovered or captured a Classical era Settler that could not be recovered or pillaged a Spaceport district
Very High (16): AI took one or more cities that could not be recovered before accepting peace or captured an Ancient era Settler that could not be recovered
Test Game "Rules"
Subject to some flexibility in game to better achieve the overall benchmarking objectives in the context of the specific game, I'll follow the following rules during the test games:
Spoiler :
- Standard speed, Continents, Standard size, normal number of Civs and City States, all standard settings, Deity level.
- No declaring war. No capturing enemy cites except to liberate your own or liberate city state cities after a war initiated by the AI.
- Expand to between 6 to 10 cities, depending on the available space.
- Maximum military: 2 land units per city (to a minimum of 8, excludes Scouts), 2 naval units per coastal city (or city with Harbour), 1 air unit per city.
- Make no diplomatic overtures, but accept delegations, offers of friendship, etc.
- Make no special effort to satisfy agendas.
- Attempt to maintain Normal Ages (no Dark or Golden Ages if possible).
- District builds: Campus then Commercial Hub or Harbour then Theatre, plus slot in as appropriate:
- x1 city: Encampment
- x1 city: Government Plaza
- x1 city: Industrial Zone
- x2 cities: Aerodrome
- x1 city: Space Port (added effective Game 3)
- if low on amenities: Entertainment Districts/Waterparks
- if lots of amenities: Aquaducts/Neighbourhoods - City build order:
- Ancient Walls (later Walls if undre threat or nothing else to build)
- replace Military units to maximums above
- Traders
- Builders until all worked tiles are improved
- Wonders if useful can be attempted
- Tier 1 and Tier 2 buildings and useful city centre buildings
- [Spies: beginning Test Game 5 I'll stop building Spies]
- Space Projects (but not the last one)
- Archeologists
- Tier 3 buildings if useful in the circumstances
- projects only if nothing else to build, then to boost gold only - Normal starting city build order (i.e. Builder/Scout, Settler, military when barbarians/AI encountered).
- Do not get a Religion. Take whatever pantheon is most useful in the circumstances.
- Do not buy Great Persons with gold or faith.
- [Use Spies defensively only. EDIT: Beginning Test Game 5 I'll stop building Spies entirely, to measure how much damage the AI can/will do if your build no Spy defences ]
- Try to avoid generating Tourists (no Seaside Resorts, National Parks, etc.)
- Use other game systems (Governments, Policy Cards, Traders, Envoys, etc.) however is most useful.
- Play as a civ with no useful bonuses under the above rules (i.e. Tamar or Philip).
- No start rerolls. Play the start you get to observe AI under both fast and slow player starts.
Test Game Summary Results
All test games posted are played using R&F and all DLCs with the latest patch as of the date of the post. Beginning Test Game 12, this includes the July 2018 patch. These games combined with prior game results as the July 2018 patch purports to deal only with cross-platform support, with no indication than any AI changes were made.
Results as of test game 15:
AI Victory Times
Earliest Observed Victory: 280
Avg. Observed Victory Turn: 318
% Science Victories: 93%
% Cultural Victories: 7%
AI Aggressiveness Impact
Player Eliminated %: 7%
Avg. Aggression Impact: 13.5
Max Aggression Impact: 22
Avg. Aggression Events: 6
Max Aggression Events: 17
If Wars, Avg. Events: 2.5
Max Wars: 4
If Aggressive Spies, Avg. Events: 5.4
Max Aggressive Spy Events: 15
Emergencies
Avg Emergencies Participated in by AI: 3.8
% that Succeeded: 32%
City States
Avg T of 1st death: 21
Avg Surviving at T50 8.6
Avg Surviving at T100 7
Avg Surviving Game 4
Wonders
Spoiler :
Wonder // Earliest Observed Build Turn // Avg Observed Build Turn
Alhambra 118 137
Amundsen-Scott Station never built n/a
Angkor Wat 146 213
Apadana 54 102
Big Ben 161 179
Bolshoi Theatre 266 307 [note: only 3 observed constructions]
Broadway 192 230
Casa de Contratacion 146 192
Chichen Itza 93 109
Colosseum 76 125
Colossus 76 112
Cristo Redentor 175 222
Eiffel Tower 236 269
Estadio do Maracana 224 268
Forbidden City 151 174
Great Library 59 83
Great Lighthouse 63 106
Great Zimbabwe 140 171
Hagia Sophia 130 157
Hanging Gardens 23 38
Hermitage 202 216
Huey Teocalli 115 148
Jebel Barkal 53 83
Kilwa Kisiwani 113 156
Kotoku-In 101 112
Mahabodhi Temple 73 86
Mausoleum Halicarnassus 105 148
Mont St. Michel 132 173
Oracle 57 67
Oxford University 152 175
Petra 76 111
Potala Palace 129 147
Pyramids 39 68
Ruhr Valley 148 190
Statue of Liberty 158 185
St. Basil's Cathedral 125 167
Stonehenge 24 31
Sydney Opera House 222 271
Taj Mahal 136 180
Temple of Artemis 35 60
Terracotta Army 121 145
Venetian Arsenal 189 230
Last edited: