The analysis has been continued in this post later in the thread (#58), which has a lot more relevant information...
Before voting, please feel free to consider the following information.
I have compared both Drill IV and Combat IV in a few different areas. For the moment I have ignored the side benefits on these promotions, like the healing for Combat IV, and the lower collateral damage on drill promotions. If you believe these are too significant to ignore mention them in your post. Of course, not all units can choose between the two promotion lines, but for the units that can, it can be a tricky decision for many players.
In this experiment, we treat the strength ratio of the attacker to the defender (A/D) as the independent variable, and examine a few other variables. Namely...
1) Probability of attacker winning combat
2) Probability of attacker winning combat unscratched
3) Average hitpoints attacker has after combat
4) Average xp earned by attacker.
For each graph, there is a control where the attacker has no promotions. There are no first strikes (other than on the Drill IV attacker) and no retreat odds involved. The Drill IV attacker has Drill I, II, III and IV. The Combat IV attacker has Combat I, II, III, and IV.
By considering different values for the ratio A/D, we effectively include many different combinations of tile defenses, unit damage and unit strength. [EDIT... It has been brought to my attention by DanF5771 in this post that unit health does in fact have a unique role, since it affects firepower in an unusual way. So here we are technically only comparing undamaged units. I will later on post some more findings with unit health as a variable as well.]
Note: Though I have called this an experiment, it is in fact deterministic. No simulations are involved - these are exact computations (up to machine accuracy anyway).
1) Probability of attacker winning combat
2) Probability of attacker winning combat unscratched (i.e. taking no damage)
3) Average hitpoints attacker has after combat
4) Average xp earned by attacker.
If you are confused about the meaning of the "average" quantities, consider this example.
Suppose there is an oversimplified battle where the attacker will come out of the battle with either 100HP, 50HP or 0HP. Suppose 90% of the time he will win the battle, and 10% of the time the defender will win the battle. Further, suppose he has 50% chance of having 50HP and 40% of having 100HP (50+40 is 90). Then his average hitpoints after battle would be 10% * 0 + 50% * 50 + 40% * 100 = 65HP.
If he would win 3xp if he won the battle, then the average xp earned by him would be 10% * 0 (when he loses the battle) + 90% * 3xp (when he wins the battle) = 2.7xp.
Please don't hesitate to ask anything about any of the data presented here.
In fact, if you think there is another important variable or experiment I should try then please ask. I'll be more than happy to run more tests.
Before voting, please feel free to consider the following information.
I have compared both Drill IV and Combat IV in a few different areas. For the moment I have ignored the side benefits on these promotions, like the healing for Combat IV, and the lower collateral damage on drill promotions. If you believe these are too significant to ignore mention them in your post. Of course, not all units can choose between the two promotion lines, but for the units that can, it can be a tricky decision for many players.
In this experiment, we treat the strength ratio of the attacker to the defender (A/D) as the independent variable, and examine a few other variables. Namely...
1) Probability of attacker winning combat
2) Probability of attacker winning combat unscratched
3) Average hitpoints attacker has after combat
4) Average xp earned by attacker.
For each graph, there is a control where the attacker has no promotions. There are no first strikes (other than on the Drill IV attacker) and no retreat odds involved. The Drill IV attacker has Drill I, II, III and IV. The Combat IV attacker has Combat I, II, III, and IV.
By considering different values for the ratio A/D, we effectively include many different combinations of tile defenses, unit damage and unit strength. [EDIT... It has been brought to my attention by DanF5771 in this post that unit health does in fact have a unique role, since it affects firepower in an unusual way. So here we are technically only comparing undamaged units. I will later on post some more findings with unit health as a variable as well.]
Note: Though I have called this an experiment, it is in fact deterministic. No simulations are involved - these are exact computations (up to machine accuracy anyway).
1) Probability of attacker winning combat
Spoiler :
2) Probability of attacker winning combat unscratched (i.e. taking no damage)
Spoiler :
3) Average hitpoints attacker has after combat
Spoiler :
4) Average xp earned by attacker.
Spoiler :
If you are confused about the meaning of the "average" quantities, consider this example.
Suppose there is an oversimplified battle where the attacker will come out of the battle with either 100HP, 50HP or 0HP. Suppose 90% of the time he will win the battle, and 10% of the time the defender will win the battle. Further, suppose he has 50% chance of having 50HP and 40% of having 100HP (50+40 is 90). Then his average hitpoints after battle would be 10% * 0 + 50% * 50 + 40% * 100 = 65HP.
If he would win 3xp if he won the battle, then the average xp earned by him would be 10% * 0 (when he loses the battle) + 90% * 3xp (when he wins the battle) = 2.7xp.
Please don't hesitate to ask anything about any of the data presented here.
In fact, if you think there is another important variable or experiment I should try then please ask. I'll be more than happy to run more tests.