HIgher Skilled Players Like Artillery Less - Confirmed, But Why?
A note to moderation: Again, I see nothing wrong with this post. I'm not calling out any individual (the chart shows ambiguous data points). I'm not jumping to hyperbole, nor attempting to incite unrest. I would like to, since they took the time to participate in the poll, discuss with the community what they think the results mean. Who knows, it may be instructive for Wargaming as well given I'm unaware of any such study being done previously. I'd love it if the time and work put into this were kept in mind and moderation were kept to removing those posts that attempt to derail or repurpose the discussion onto unproductive subjects. I also hope those replying self-moderate and realize this is not an artillery whine and worthy of more than the usual pithy replies.
Process
I ran a poll on the official forums asking players to give their opinions on artillery (found here). The poll votes were made public, allowing me to have record of each person's individual vote. I then went to WoTLabs and gathered:
- Recent WN8
- Recent Winrate
- Overall WN8
- Overall Winrate
- Number of Battles Fought
- Average Tier
For each player that voted. Because the vote options were in English, I created a numeric scale that corresponds to each vote. This scale was made public in the opening post for the poll. It's as follows:
10 = I strongly dislike artillery. Remove them or the puppy gets it.
9 = Remove artillery
8 = Overhaul, but remove artillery if that's not enough
6 = Overhaul, but keep artillery
3 = Modify artillery, but no overhaul needed
2 = Artillery is fine
1 = I love artillery
The middle option, overhaul but keep, was given a value of 6 instead of 5 as any overhaul would leave the class of vehicles unrecognizable to their current form. This did not feel like it was appropriately represented by a perfectly middle numeric value. Otherwise, the options are perfectly balanced against one another. 3 options for satisfaction with artillery in varying degrees, two of which require no changes. 3 options for disatisfaction with artillery, two of which believe removal should happen.
With these two sets of data, I choose to run a simple bivariate regression analysis. This is a test against data sets, one for each of two related variables, that provides a mathematical confirmation of a correlation between the two related values. In short, it looks at these two sets of data and determines whether or not one influences or predicts the outcome of the other. If you believe my personal artillery opinions make me an unreliable source for this summary, you can head over to wikipedia and get started looking into the process by clicking here.
I have two data sets to work with, from (other than small word differences that account for the audience/moderation staff at each location) this location and another popular English speaking WoT forum known as a haven of players of high skill. However, I ran the results separately for this poll to avoid any accusation of influence as well as one with the two combined. Because the official result was so clear, I've decided not to post the combined results (despite it showing even more correlation and a much higher % calling for the removal of artillery). For this first post, I choose to use Recent WN8 as the skill representation per player.
WN8 is a numerical rating that usesper tankvalues as a basis for determining a mean from which it can compare each player and determine their skill in that tank. Perfect? No, but it's the best numeric rating we have at hand currently. Most of its issues are also exclusive to the very extreme ratings, greater than 3000. It's very reliable in the 0-3000 range. If you have questions about WN8, you can view in depth details by clicking here.
As for data cleanliness, I opted to only ignore votes from players with less than 100 battles. If they lacked a recent rating, I used their overall rating. Otherwise, everyone's vote counted.
Results
At the time of recording for the official forum poll, there were 193 votes. The breakdown by percentage was as follows:
I strongly dislike artillery. Remove them or the puppy gets it: 32 votes, 16.58%
Remove artillery: 14 votes, 7.25%
Overhaul, but remove artillery if that's not enough: 23 votes, 11.92%
Overhaul, but keep artillery: 23 votes, 11.92%
Modify artillery, but no overhaul needed:18, 9.33%
Artillery is fine, or I don't care: 43, 22.28%
I love artillery: 40 votes, 20.73%
For those that are interested, even though they were not the point of the poll, the three "Artillery are in a pretty good spot" options represented 52.34% of the results. The three "Artillery should probably be removed" options represented 35.75% of the results. The "Needs major changes, but the class is necessary" option was 11.92%. (Note that rounding differences will make these not slightly add up, but they're recorded as they were displayed by the forum software). If you wish to divide among "They're mostly fine" and "Needs major work", the community here (at least those that responded)is nearly split with "artillery is fine" havinga ~4% edge.
However, this was not the goal of the poll. The goal of the poll was to run the regression analysis against the two. Below you'll find a chart of a linear plot from that analysis.
To the left is a scale from 1-10. You can relate that back to the votes based on the table I provided at the top of this post, where 10 is the strongest dislike for artillery and 1 is the strongest like for artillery. On the bottom, running left to right, is the axis representing each player's recent WN8, from 0-5500 (I went to 5500 because I wanted the graphs to align between both only the official forum results and the combined results)
A key result from the regression analysisto look out to determine whether or not two related variables are highly correlated is the p-value. A p-value of < 0.05 is considered statistically significant, meaning, it's highly unlikely the relationship is random. The closer you get to 0 the more certain you can be of the relationship between the two. Our p-value for this test was .000000000799. Meaning, it's impossible to believe that the relationship between the two is random.
The conclusion is that, the better you are at WoT, as measured by WN8, the more likely you are to strongly dislike artillery. The orange line of dots shows you what the analysis would predict your opinion is based on your recent WN8. The orange line is what the analysis predicts your vote would be at each recent WN8 point along the way based on the total influence of all votes and recent WN8's. Not how far below 1000 the lowest point is and how well above 3000 the highest points are. Essentially, it predicts that unicums are likely to want artillery removed, blues want an overhaul and greens and below are either happy with it or want much smaller changes.
I'll run the same for winrate again soon, but for now I have to get back to other things. I will post the data tables used on google docs today and update this post with that location. Note that people's WN8 ratings are always shifting slightly as they play, so you may find numbers slightly different from mine, but not by much. I gathered the data in one huge block so it's as accurate as it can be relative to itself.
Conclusion
This shows a relationship, an undeniable relationship that shows the more skill you have at WoT the less likely you are to be satisfied with artillery. What it doesn't show or explain iswhy. That is what is worth discussion. What might the reason be? I have my own thoughts as to why this is the case which I've shared on this forum previously. However, when I did so, some actually objected that this relationship existed to begin with. With that objection removed, I'm curious what their explanations for this relationship would be.
Why is it that the less skilled you are at World of Tanks, the morelikely you are to be satisfied with artillery?
