Weighting Mystery Revealed!!!

[dorme84]

i think this is simply a rough estimate.

I'm not sure troop weight counts.
I also think upgrading your TH adds weight even if you don't upgrade anything else.

[Krodas0616]

Noticed that totally maxed 9 except for one level on the air sweeper weighed 67k while another 9 in a different clan weighed 73k but with only 21/20 heroes and barely any lava walls done.

[random2]

From my observations the contribution for each defense towards the war loot is nonlinear. You can easily check this by looking at engineered bases like defenseless th10s or Th10s with th8 level defenses. Each new defense adds much more for bases worth 40k than bases worth 100k.

The reason the numbers work for some is because most people have the same defenses. When you have so many variables that are often correlated (those with L7 wiz towers will have L7 ADs etc) it's easy to come up values that fit the data and many people with close to identical bases will report good results.

Also, for TH9s (the most interesting use of war weights, imo) the difference in weight between a max TH9 and a TH8 is 20k. The difference between a max th9 and a a new TH9 with new defenses is something like 12k. The weights don't have to be very accurate to get a close result.

I suspect that each defense/wall etc probably has a value associated to it because it's the easiest to program. However, that war weight might be transformed (log?) to get to the war loot.

[cjRandom]

From my observations the contribution for each defense towards the war loot is nonlinear. You can easily check this by looking at engineered bases like defenseless th10s or Th10s with th8 level defenses. Each new defense adds much more for bases worth 40k than bases worth 100k.

The reason the numbers work for some is because most people have the same defenses. When you have so many variables that are often correlated (those with L7 wiz towers will have L7 ADs etc) it's easy to come up values that fit the data and many people with close to identical bases will report good results.

Also, for TH9s (the most interesting use of war weights, imo) the difference in weight between a max TH9 and a TH8 is 20k. The difference between a max th9 and a a new TH9 with new defenses is something like 12k. The weights don't have to be very accurate to get a close result.

I suspect that each defense/wall etc probably has a value associated to it because it's the easiest to program. However, that war weight might be transformed (log?) to get to the war loot.

Hey random2,

can you explain to me where do you see a wight in-game?

[K0n3s]

And those who are saying it wrong or an estimate. These weights weren't written from in game observations. They are completely from Supercell files. The fact they are right or so close to being right for so many means this isn't just a coincidence.

Moskri missed offence weights on Hero's in his sheet which adds an extra 1.9k total.

Whether these numbers are right. I believe they are. However I don't know how they translate. Eg. Small bombs. There are other possible ways to calculate like each value represents a block of bombs unlocked at each town hall. This would bring down the total considerably.
This would then cause the question is this relevant to all values. Walls. Xbows. Infernos?
If we start adding different weight groupings on each group to fit our preconceived max loot values we may be finding what we want to be the values not the actual values.
This is especially true when we don't know these values are supposed to fit directly with loot value.

And @AngryMartians they weren't stolen from your site they are available to everyone with coc. If however you did have these values earlier thanks for sharing with the community...
If you do have these values and have studied them for a bit longer do you mind weighing in and sharing some more advanced findings?

[random2]

And those who are saying it wrong or an estimate. These weights weren't written from in game observations. They are completely from Supercell files. The fact they are right or so close to being right for so many means this isn't just a coincidence.

Moskri missed offence weights on Hero's in his sheet which adds an extra 1.9k total.

Whether these numbers are right. I believe they are. However I don't know how they translate. Eg. Small bombs. There are other possible ways to calculate like each value represents a block of bombs unlocked at each town hall. This would bring down the total considerably.
This would then cause the question is this relevant to all values. Walls. Xbows. Infernos?
If we start adding different weight groupings on each group to fit our preconceived max loot values we may be finding what we want to be the values not the actual values.
This is especially true when we don't know these values are supposed to fit directly with loot value.

And @AngryMartians they weren't stolen from your site they are available to everyone with coc. If however you did have these values earlier thanks for sharing with the community...
If you do have these values and have studied them for a bit longer do you mind weighing in and sharing some more advanced findings?

Just because it works for some people doesn't mean it's right. Most people's upgrades are very correlated and it's not difficult to find parameter values for each defense that will produce the correct weight for a majority of people. Most low th9s have similar defenses to other low th9s. Most mid th9s have similar defenses to other mid th9s.

For example, if you interchange the weight of air defenses, mortars, wiz towers, and teslas, the result will be the same for most bases. This is because most bases have 4 of each and roughly the same level. You can make the same argument with many of the defenses/troops. This is the reason why the formula seems to work for many bases. Just because it works for some bases doesn't mean that the values are even close to right.

The formula becomes wrong when you look at non-standard bases. For example, my th9 is missing a few defenses since th6. The spreadsheet says he is worth 31k. In actuality he is worth 49k. In another example, the formula says 41k while the actual worth is 54k.

Also it seems odd to me that troop levels matter in the formula when there was no indication otherwise. If troops contribute around 10k towards war loot levels, surely someone would have noticed it before.

[K0n3s]

Just because it works for some people doesn't mean it's right. Most people's upgrades are very correlated and it's not difficult to find parameter values for each defense that will produce the correct weight for a majority of people. Most low th9s have similar defenses to other low th9s. Most mid th9s have similar defenses to other mid th9s.

For example, if you interchange the weight of air defenses, mortars, wiz towers, and teslas, the result will be the same for most bases. This is because most bases have 4 of each and roughly the same level. You can make the same argument with many of the defenses/troops. This is the reason why the formula seems to work for many bases. Just because it works for some bases doesn't mean that the values are even close to right.

The formula becomes wrong when you look at non-standard bases. For example, my th9 is missing a few defenses since th6. The spreadsheet says he is worth 31k. In actuality he is worth 49k. In another example, the formula says 41k while the actual worth is 54k.

Also it seems odd to me that troop levels matter in the formula when there was no indication otherwise. If troops contribute around 10k towards war loot levels, surely someone would have noticed it before.

I think you missed my point that these numbers weren't found by looking at any bases. Normal or Abnormal. They are from files that give each building, troop, trap, wall, spell and hero a weight.

[random2]

I think you missed my point that these numbers weren't found by looking at any bases. Normal or Abnormal. They are from files that give each building, troop, trap, wall, spell and hero a weight.

They were supposedly obtained from a file that has, presumably, the name of the defenses, and some value. Doesn't mean it is war weight.

Also, the only way to validate the formula is to compare it to actual bases. My point is that comparing them to normal bases doesn't mean that it is close to correct because of the linear dependence of many of the defenses.

[K0n3s]

They were supposedly obtained from a file that has, presumably, the name of the defenses, and some value. Doesn't mean it is war weight.

The values are actually clearly labelled as such within the encrypted files.
And by majority of users it has confirmed the correlation. But as my first post says the translation of the values to war weight hasn't been completely fine tuned yet. There are a few variables that need to be figured out. Perhaps you can make a sheet of your levels and your current war weight for comparison?

Getting data from non standard bases will be the best way to check and fine tune.
The guys at gaymartians are filled with non standard bases and use the same weighting system as this.
This confirms that it does translate. They have had more time to fine tune based on the non standard bases which give better results as you can check individual defence upgrade results.

[wiped]

I used the calculator spreadsheet to check 3 TH9 accounts and 1 TH7 account. The TH9 numbers are reasonably close, but the TH7 number is way off.

The errors from real values for 3 TH9 accounts: -1.2%, -1.4% and -5.2%.
The errors from real value for the TH7 account: -23%

In general terms, it seems wrong that troop upgrade weights are directly added to the base weight for ranking on the map and loot bonus. I'd also imagine the HP a base has should contribute to its weight somewhat, i.e., buildings that don't attack should have some contribution.