Ramshutu
02-02-2012, 02:50 PM
Okay, i have been investigating loss mechanics, i found it quite an interesting itch to scratch and felt this may need its own thread.
I've been running some numbers and comparing them to what I am seeing in game. I am using some broad raid info I've gained from two accounts, a level 12 and level 35 to give me some confidence that what I see isn't down to a fluke of unit makeup or level.
Essentially, we all know that when you attack, the unit list is scanned one by one in some fashion to determine what units are lost. How this list is scanned, how many and why you lose the units you do I shall attempt to explain.
1. using the 'consume_percentage' fields of the database (link below) I have calculated that if there was a straight loss calculations (running through each unit and rolling a dice) the odds that of surviving a raid with 0 losses is <0.2%
2. On my low level account, I am seeing 0-2 losses, with around 75% of attacks and 30% of raids resulting in no loss.
3. On my higher (lol almost said high there) account, I am seeing 1-3 losses, with <5% of attacks resulting in no loss and <2% of raids resulting in no losses.
4. Due to 1, I can say for almost certain that the number of units you loose does not depend on the units you have. While I am thinking of starting a 3rd account to check, I suspect that having 100 high loss units will loose the same number as 100 equivalent low loss units. If there were a simply a fixed loss cap and straight roll on all units, I would expect to see more losses than I am seeing on both accounts, the low level account has far too many no losses. This means that the loss cap is random.
5. With the above, we have two options. Firstly, the loss count could be random, in that the game decides how many units you will loose, and then rolls to choose which you loose. Alternatively, the loss cap could be random (the game decides how many units, at most, you will lose), this is difficult to test; you would have to have a consistent set of units and work out what the chances of not losing any versus how many you do lose ( for example, having 200 units with 1% consume would yield a 13% chance of keeping all units, with an average consume of 2% this drops to about 2% chance there's very little differences to compare otherwise)
6. Programmers are lazy. I am one, and can speak from experience. With this in mind, I suspect (but cannot yet proove) that there is a random loss cap rather than random loss count, based on either level, number of units, or allies (these are the only real differences between the two accounts, but will be able to tell which at some point soon....) if the dice were rolled until you made up the losses, the calculation may take time.
7. What I am having trouble with, is the unit choice. the units are searched in some order, a dice roll is conducted for each. If this fails, then the unit dies. If the cap is reached or you have no more units to search, the loss calculation ends, and that's it. The choices for the order are a) high to low loss, b) alphabetical c) inventory order d) reverse attack list order and d) completely randomised.
8 I am pretty sure from implementation and from the statistics that it isn't d) or c) I loose too many rangers, and too few destroyers compared to the number I have for either to be true. A) almost fits the stats, but I have too many blips loosing units low down the list for it to match the stats. B) seems to fit a little better... Avengers are quite high up the list, but I would expect this implementation to be convoluted.
So, if anyone has more detailed info or their own unit losses compared to overall unit breakdown (how often do you loose unit x, how many do you have, and what number of higher loss units do you have),
In summary, however, I think that regardless of whether there is a random loss count or loss cap per fight, the statistics mean that there isn't much of a difference in the result, with lots of units at a very low loss, your still pretty likely to loose them regularly.
bulking up on meat shields is the only way of avoiding bad losses. For example, having 50 rangers in your raid will almost certainly eat up the losses you would otherwise have on more expensive ones (it would mean a very small chance of making it through the rangers without loss). For example, Having 20 harriers as the lowest loss rate units would mean that you would loose at lest one harrier 1/3rd of raids.
I will be doing some investigating into the infirmary now, I suspect that this will lower the loss cap randomiser which means that you will loose the same units, but less (or none of them)
However, the data I have is all circumstantial, without befriending a dev with mountain dew and cheetos, then teasing out industrial secrets with my wit and charm, I could be horribly, horribly wrong and so would welcome anyone else's data.
FYI the database of loss rates can be found here:
https://docs.google.com/spreadsheet/lv?key=0AgjEOMAMom92dFhOellvbTAwQ0dvZGJfYk9EQjMxRn c
Thanks to JackRakan for the database information
I've been running some numbers and comparing them to what I am seeing in game. I am using some broad raid info I've gained from two accounts, a level 12 and level 35 to give me some confidence that what I see isn't down to a fluke of unit makeup or level.
Essentially, we all know that when you attack, the unit list is scanned one by one in some fashion to determine what units are lost. How this list is scanned, how many and why you lose the units you do I shall attempt to explain.
1. using the 'consume_percentage' fields of the database (link below) I have calculated that if there was a straight loss calculations (running through each unit and rolling a dice) the odds that of surviving a raid with 0 losses is <0.2%
2. On my low level account, I am seeing 0-2 losses, with around 75% of attacks and 30% of raids resulting in no loss.
3. On my higher (lol almost said high there) account, I am seeing 1-3 losses, with <5% of attacks resulting in no loss and <2% of raids resulting in no losses.
4. Due to 1, I can say for almost certain that the number of units you loose does not depend on the units you have. While I am thinking of starting a 3rd account to check, I suspect that having 100 high loss units will loose the same number as 100 equivalent low loss units. If there were a simply a fixed loss cap and straight roll on all units, I would expect to see more losses than I am seeing on both accounts, the low level account has far too many no losses. This means that the loss cap is random.
5. With the above, we have two options. Firstly, the loss count could be random, in that the game decides how many units you will loose, and then rolls to choose which you loose. Alternatively, the loss cap could be random (the game decides how many units, at most, you will lose), this is difficult to test; you would have to have a consistent set of units and work out what the chances of not losing any versus how many you do lose ( for example, having 200 units with 1% consume would yield a 13% chance of keeping all units, with an average consume of 2% this drops to about 2% chance there's very little differences to compare otherwise)
6. Programmers are lazy. I am one, and can speak from experience. With this in mind, I suspect (but cannot yet proove) that there is a random loss cap rather than random loss count, based on either level, number of units, or allies (these are the only real differences between the two accounts, but will be able to tell which at some point soon....) if the dice were rolled until you made up the losses, the calculation may take time.
7. What I am having trouble with, is the unit choice. the units are searched in some order, a dice roll is conducted for each. If this fails, then the unit dies. If the cap is reached or you have no more units to search, the loss calculation ends, and that's it. The choices for the order are a) high to low loss, b) alphabetical c) inventory order d) reverse attack list order and d) completely randomised.
8 I am pretty sure from implementation and from the statistics that it isn't d) or c) I loose too many rangers, and too few destroyers compared to the number I have for either to be true. A) almost fits the stats, but I have too many blips loosing units low down the list for it to match the stats. B) seems to fit a little better... Avengers are quite high up the list, but I would expect this implementation to be convoluted.
So, if anyone has more detailed info or their own unit losses compared to overall unit breakdown (how often do you loose unit x, how many do you have, and what number of higher loss units do you have),
In summary, however, I think that regardless of whether there is a random loss count or loss cap per fight, the statistics mean that there isn't much of a difference in the result, with lots of units at a very low loss, your still pretty likely to loose them regularly.
bulking up on meat shields is the only way of avoiding bad losses. For example, having 50 rangers in your raid will almost certainly eat up the losses you would otherwise have on more expensive ones (it would mean a very small chance of making it through the rangers without loss). For example, Having 20 harriers as the lowest loss rate units would mean that you would loose at lest one harrier 1/3rd of raids.
I will be doing some investigating into the infirmary now, I suspect that this will lower the loss cap randomiser which means that you will loose the same units, but less (or none of them)
However, the data I have is all circumstantial, without befriending a dev with mountain dew and cheetos, then teasing out industrial secrets with my wit and charm, I could be horribly, horribly wrong and so would welcome anyone else's data.
FYI the database of loss rates can be found here:
https://docs.google.com/spreadsheet/lv?key=0AgjEOMAMom92dFhOellvbTAwQ0dvZGJfYk9EQjMxRn c
Thanks to JackRakan for the database information