Stage 1 Battle Mechanic Test Results For all tests the following stats are applicable: Attacker Stats Hero: 197 Attack Rune: 10 Healing Potion: 10 Blood Lust/Frenzy: OFF Defender Stats (Level 9 Goblin Camp) Hero: 0 (Assumption) Attack Rune: 9 Healing Potion: 9 OVERVIEW The idea of these simulations is to gather enough data to make progress in determining an accurate model of the battle mechanic. No single formula is ever going to accurately capture the battle mechanisms. Instead a series of formulas will most likely be formed over time in order to try and best represent the battle mechanism. A level 9 goblin camp was chosen as it contains only Artillery troop types. This allows only other Artillery troop types to be sent as there should be no bonus or penalty when Artillery troop types fight each other. Each simulation is run five times in order to determine a reasonably accurate average. ATTACKER RIDUCOULSY OUTNUMBERED To simulate attacking a player defending with 25.6 million might 100 archers were sent against a level 9 goblin camp which had 1000 archers, 600 ballistae and 500 catapults. Multiplying everything by 1000 results in simulating 100k archers being sent against a player defending with a huge army. In pure troop numbers the attacker is outnumbered 21 to 1 whilst in terms of might they are out gunned by 64 to 1. The attacker lost all troops on each attack whilst an average of 81.8 troops were lost on each attack by the defender. There was a positive deviation of 3.8% and a negative deviation of 3.4%. ATTACKER OUTNUMBERED To simulate attacking a player defending with 2.56 million might 1000 archers were sent against a level 9 goblin camp which had 1000 archers, 600 ballistae and 500 catapults. Multiplying everything by 100 results in simulating 100k archers being sent against a player defending with a large army. In pure troop numbers the attacker is outnumbered 2.1 to 1 whilst in terms of might they are out gunned by 6.4 to 1. The attacker lost all troops on each attack whilst an average of 822.4 troops were lost on each attack by the defender. There was a positive deviation of 2.0% and a negative deviation of 2.4%. DEFENDER OUTNUMBERED To simulate attacking a player defending with 256k might 10000 archers were sent against a level 9 goblin camp which had 1000 archers, 600 ballistae and 500 catapults. Multiplying everything by 10 results in simulating 100k archers being sent against a player defending with a medium army. In pure troop numbers the defender is outnumbered 4.8 to 1 whilst in terms of might they are out gunned by 1.6 to 1. The defender lost all troops on each attack whilst an average of 1319.8 troops were lost on each attack by the attacker. There was a positive deviation of 3.0% and a negative deviation of 2.4%. DEFENDER RIDUCOULSY OUTNUMBERED To simulate attacking a player defending with 25.6k might 100000 archers were sent against a level 9 goblin camp which had 1000 archers, 600 ballistae and 500 catapults. Multiplying everything by 1 results in simulating 100k archers being sent against a player defending with a small army. In pure troop numbers the defender is outnumbered 48 to 1 whilst in terms of might they are out gunned by 16 to 1. The defender lost all troops on each attack whilst an average of 1310.2 troops were lost on each attack by the attacker. There was a positive deviation of 2.7% and a negative deviation of 1.6%. ASSUMPTIONS It is assumed that might is not a true indicator of troop strength. This seems to be supported by the large discrepencies seen in past battles. In game the various Tiers are given a star value to indicate their attack and life. Those star values have been converted into numerical values and a modifier applied based upon several factors. The following formula has been used to calculate the Modifier: Code: Modifier = 1 + (Hero / 200) + (Research / 20) + (Item / 5) Where: • Hero: The level of Hero used in the the attack. • Research: The level of Research, ie Attack Rune for attack modifier and Healing Potion for life modifier. • Item: Whether or not Item buffs were uses, attack modifier and life modifier. Where 1 indicates used and 0 indicates not used. By placing in the correct values the Attack Modifer and Life Modifer for both the attacker and defender can be determined. Therefore the following modifiers will need to be calculated: 1. 2. 3. 4. Winner Attack Modifier Winner Life Modifier Loser Attack Modifier Loser Life Modifier LIMITATIONS All attacks consisted of only Tier 1 troops. This limits any conclusions to single Tier attacks only. Test data for multi-tiered attacks will have to be gathered to expand on battle mechanic model. CONCLUSION The very first thing to note is there is definitely a random factor of between 1.6% and 3.8% involved. It is not a huge deviation and for the purposes of determining the battle mechanic model it can be ignored. As long as calculated values fall within the deviation then they can be considered as accurate. Another consistant observation is the deviation between calculated results and expected results when one side is considerably outnumbered. From the above tests it was shown that approximately a 50% bonus was applied to the winner. That is the winner had their losses reduced by 50%. Additional tests were run to confirm or deny this and a new pattern emerged. Attacks were performed at a range of percentages above the calculated strength of the defender. What emerged was an inverse curve relationship between the bonus applied and the strength of the army sent in comparison to the defender. The greater the difference the greater the bonus but it seemed to be maxed out at around 50%. The following formula was extropalated from the data gathered: Code: Winners Bonus = 5 * SQRT(Difference) Where: • Winners Bonus: The bonus applied to the winner, to determine actual troop losses, in percentage. • Difference: The percentage by which the winner overpowers the loser. ie Difference = ((Winners Attack / Losers Attack ) * 100) - 100 Note: 1. Winners Bonus has a ceiling of 50% The Winners Bonus formula is by no means perfect but is proving to be relatively consistent with the data gathered. Occassionaly a result will appear that is outside of the expected Random Deviation. That would suggest one of the following possibilities: 1. 2. 3. 4. The Winners Bonus formula is wrong The Random Deviation is greater than expected The method of calculating Actual Strength is wrong Some other unknown factor is having an effect An initial formula can be concluded from the test data. It will need to be updated as more data is gathered but could be considered as a general rule of thumb for now. Code: Winners Losses = Losers Attack / Winners Life Modifier - ((Losers Attack / Winners Life Modifier) * (Winners Bonus / 100)) A great many more simulations are needed to start drawing an accurate conclusion. This is just the first in many many steps. Input from others is greatly appreciated but please give detailed results otherwise the data is useless.