Jump to content

I Scraped Everyone's* Unit Kill Ratios And Here's What I Found (sponsored by DNN)


hidude45454
 Share

Recommended Posts

@hidude45454 I’d be interested to see what the data looks like before you took the logarithm of the y axis.

It looks like it might be logistic (your line of best fit isn’t linear) although I don’t imagine the google sheets graphs you’re using has support for such a regression.  Would recommend checking out some python packages like numpy, sklearn, and scipi if you’re interested in the data manipulation and learning more from it.  R and matlab are also useful options too if you have access to them!

Link to comment
Share on other sites

40 minutes ago, Mr. Berrycloth said:

Very cool! But what was the point of having/using the log kill ratios? (Besides making the graph look pretty, doesn’t that mess with the data?) Also what was the base of the log you used?

I think it could be be somewhat misleading reading a log graph, but to my understanding I don't think it messes with the data since you can always unlog it. (And it's with a base of 10)

31 minutes ago, Cooper_ said:

@hidude45454 I’d be interested to see what the data looks like before you took the logarithm of the y axis.

It looks like it might be logistic (your line of best fit isn’t linear) although I don’t imagine the google sheets graphs you’re using has support for such a regression.  Would recommend checking out some python packages like numpy, sklearn, and scipi if you’re interested in the data manipulation and learning more from it.  R and matlab are also useful options too if you have access to them!

Here is the original data, I cleaned up a few data points in order to make it look kind of prettier (I think)

image.thumb.png.b7b80b5c0b19e780a08e4b13b5da8f8e.png

Here the best line of fit is kind of linear, which I suspected was the case if a line of fit on a log graph was logarithmic (I can manually choose which type of line I want.) I've used numpy/sk-learn/scipy a little bit before for class stuff so agree that it would probably be good practice, but figured Google Sheets would probably be a bit more accessible and was also feeling a bit lazy at the moment I guess 😜

Link to comment
Share on other sites

1 hour ago, hidude45454 said:

I think it could be be somewhat misleading reading a log graph, but to my understanding I don't think it messes with the data since you can always unlog it. (And it's with a base of 10)

Here is the original data, I cleaned up a few data points in order to make it look kind of prettier (I think)

image.thumb.png.b7b80b5c0b19e780a08e4b13b5da8f8e.png

Here the best line of fit is kind of linear, which I suspected was the case if a line of fit on a log graph was logarithmic (I can manually choose which type of line I want.) I've used numpy/sk-learn/scipy a little bit before for class stuff so agree that it would probably be good practice, but figured Google Sheets would probably be a bit more accessible and was also feeling a bit lazy at the moment I guess 😜

Thanks! I guess it makes sense that the data is linear, wars at higher tiers are fought with more troops.

Link to comment
Share on other sites

Oh hey I actually made an appearance... interestingly my tank ratio is not that good, only 1.7 -- and then 2.8 for planes, 2.5 for ships, so that ain't half bad. Most likely my soldier kill ratio is that high because of how many raiders I countered when I was lower tier. Very interesting data! 

Edited by Cassia
Link to comment
Share on other sites

13 hours ago, hidude45454 said:

*Everyone that's currently active or in an alliance, idgaf about the rest lmao!

Sponsored by DNN: https://discord.gg/MDYdBkzFh4

The full sheet, if anyone wants to check it out: https://docs.google.com/spreadsheets/d/1n0w9tpZIKI_dAdSm2p6PHjDIEAQvxGkmjspNG_Ebs44/edit#gid=914942025

In a moment of boredom I scraped the soldiers/tanks/planes/ships kills/losses of everyone in the game and compiled a bunch of ratios and statistics for your enjoyment. The sheet is pretty self-explanatatory but I screenshotted all the stats anyways just to go into some analysis/explanation of what they mean.

Kill Loss Ratio Histograms:

image.thumb.png.226cea5e7bfaab4f0a28beb0083c22e8.png

Perhaps surprisingly, these histograms peak at about a 0.7-ish kill to loss ratio rather than at an exact 1:1 ratio. This suggests that there is a smaller group of better fighters in the game that can take advantage of a larger pool of worser fighters that skews the graphs. Note that the final peak in all graphs is from all the people who are in the range from the cutoff point to the maximum ratio in that unit (presumably mostly from the raiders in the C1-9 range).

Kill Loss Ratio Medians:

image.thumb.png.d871fac4edd49b82a290c3b7c60e7060.png

The overall median and tier-based medians more or less seems to confirm this. Theory: The median of tank, plane, and ship kill ratios increases the higher tiers we get because of 1) Generally more experienced members in higher ranges, and 2) More downdeclaring potential to kill more units.

Kill Loss Ratio Percentages:

image.thumb.png.cb5781bdd266e5887f98975bedefb14d.png

How to read this graph: for ex. if you are a C15 with a 1.2 tank kill loss ratio, that is saying your ratio is better than about 70% of people in that range (not bad!)

Kills Per City Trends

image.thumb.png.e1eb2d889ad149b1cfd009e44d937279.png

I used the log of unit kills per city rather than the unit kills per city itself because without it the graphs looked super ugly, lmk if that wasn't the right approach. Here I was curious if lower city nations could pound for pound compete with higher city nations based on kills per city, and from looking at the highest points at each city count the answer appears to be yes, but on the overall higher city nations generally trend higher anyways, probably mostly because they have played the game longer and therefore had more time to kill units.

Best and Worst Ratios in Each Tier and Overall

image.thumb.png.b53542d1d0a123ec91127f0850e9d18f.png

image.thumb.png.5ad1c30b8a0d3b76c78ec6bb82d00aa8.png

image.thumb.png.5203aed21569f64832828f0c80176a6f.png

image.thumb.png.b0bff507000070c24b652138137343fb.png

image.thumb.png.87e3a1c971ba39a8eb916087a148f0d7.png

image.thumb.png.0d3913cca4123c5ccad5ea7c1cd2f1d9.png

Fairly self-explanatory, feel free to tag yourself or your friends who made these lists! 😉 For the overalls, I wanted it to actually look kinda like a legit list so I filtered it only to people with 5m+ soldier losses, or 100k+ tank losses, or 7k plane losses, or 1k+ ship losses (thanks to Ramona for the thresholds.)

Outliers

image.thumb.png.4b90c043fcd7be2e39eb70678896ba7e.png

Some people might be wondering how tf some people's ratios are so high. As two examples, MIBHG Country (the first linked nation) has built less than 200k soldiers in 2500+ days, and Dusty (the second linked nation) has built literally zero tanks since rerolling (lmao). Obviously neither of these people are particularly good at fighting but have good ratios because of these choices regardless. This is the case with quite a few people, especially the lower you go with raiders and stuff who also primarily target lots of people who don't really know how to fight back. So, just a note that ratios don't mean everything (and I care about total unit kills anyways, but PnW already has a leaderboard for that 😉)

 

Note: I have no experience with statistics classes, so pls no kill if I did something wrong (but please point it out!) Feel free to suggest any additional fun data points to scrape and I will do my best to fill all interesting suggestions.

Not particularly good? that hurts me deep🥺, I like my tankless memes

 

Edited by Dusty
  • Haha 1
Link to comment
Share on other sites

23 hours ago, hidude45454 said:

I think it could be be somewhat misleading reading a log graph, but to my understanding I don't think it messes with the data since you can always unlog it. (And it's with a base of 10)

Here is the original data, I cleaned up a few data points in order to make it look kind of prettier (I think)

image.thumb.png.b7b80b5c0b19e780a08e4b13b5da8f8e.png

Here the best line of fit is kind of linear, which I suspected was the case if a line of fit on a log graph was logarithmic (I can manually choose which type of line I want.) I've used numpy/sk-learn/scipy a little bit before for class stuff so agree that it would probably be good practice, but figured Google Sheets would probably be a bit more accessible and was also feeling a bit lazy at the moment I guess 😜

I don’t know how much linear algebra experience you have but a line of best fit will always be linear.  Other fits will require a different regression.  You use the log-linear plot to show data that spans multiple orders of magnitude easily, I.e an exponential function would look linear.  I take back the logistic argument though, seeing the full distribution, this looks pretty linear itself just with a bit of variance from the outliers. 

Link to comment
Share on other sites

Second part with infra ratios, from a suggestion from Cassia:

The sheet: https://docs.google.com/spreadsheets/d/1n0w9tpZIKI_dAdSm2p6PHjDIEAQvxGkmjspNG_Ebs44/edit#gid=731748600

Destroyed Lost Ratio Histogram:

image.png.9accad7941d91aecf5d399f4b05438e7.png

Already here you can see a pretty significant change vs. units, where the peak settles much closer around the 1:1 ratio.

Destroyed Lost Ratio Medians:

image.png.8a797dfc857dfa8b89be2273bd1cf1a1.png

This pretty much confirms the above graph. I think there is a pretty strong theory as to why these numbers are so high -- simply because I only recorded data from active players in alliances. Active players in alliances raid inactive players not in alliances and get easy infra boosts as a result. This is even more true in the lower tiers where raiding statistics factor in much heavier than war statistics.

Destroyed Lost Ratio Percentages:

image.thumb.png.26580bc19f4fe6858bafc874b41fe66c.png

Adding to this theory even more, the higher up you go, the more wars factor into that and the lower most people's ratios get, whereas in the C1-9 range most of the raiding actives are well past the 2.5 infra destroyed/lost ratio.

Kills Per City Trends

image.png.ad6dc99e8f9c703e1e146b364b7ef87d.png

Looks pretty much the same as the unit graphs.

Best and Worst Ratios in Each Tier and Overall

image.png.c46c0a314d24384c769249379435bd4c.png

image.png.9347c64f5021c20d3858a7a757c0654a.png

image.png.d677c876f9d26bf37dcf4084c53c4e35.png

image.png.de894af61dda986354cbd6127a4f1d27.png

image.png.f0f471832b605cfc4692e459acd62a5b.png

image.png.ec9eb34299f6b2de0be0b6540478b957.png

Here some additional, albeit fairly intuitive patterns start emerging -- pirates with not much infra to begin with generally top losts, while crappy farmers who don't know how to fight settle at the bottom. The overall list is filtered only to people with 100k+ infra lost. Shoutout to this dude https://politicsandwar.com/nation/id=159527 for being one of the worst fighters I've ever seen and killing 3.5k infra in 1000+ days 😂

As always, let me know if you have any more interesting suggestions to pull!

  • Upvote 3
Link to comment
Share on other sites

Nice work on the graph work! It's a different way to show who really is good at fighting and others to well ... just Git Gud mate. 

 

676912258_galaxybrain.webp.5c1275f9a627f0a3b84e6f73d163a6c5.webp

My opinion may not reflect those of my alliance or its affiliates. Please read at your own discretion.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
 Share

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use and the Guidelines of the game and community.