Search in War Thunder for the mode of personal oppression (PPU) using the analysis of win/loss statistics

From 07/12/2019 to 07/19/2019, I carefully collected and tabulated the statistics of my wins and losses in the game War Thunder, in order to verify the presence or absence of RPU by testing statistical hypotheses about the equality of averages. After 2 years, I finally mustered the strength and courage to publish the data obtained and present the results of the analysis for public discussion.

The data obtained allow us to draw a conclusion about the hypothesis of the presence / absence of RPU in the game, lead to very unusual conclusions regarding the object under study, have scientific novelty and have not been expressed in other publications for 2 years. Also in the article I will try to draw some reasonable conclusions about the characteristics and properties of the object under study.

Study Methodology

A method for testing statistical hypotheses was chosen. Having collected data on each of my wins/losses for the period from 07/12/2019 to 07/19/2019, I decided to test the hypothesis about the equality of the means in the two samples. The null hypothesis was that the ratio of wins to losses, without the presence of the RPU regimen, on average, would be the same daily. If you code win as 1 and lose as 0, when comparing the average, there will be no significant differences in the samples by day.

An alternative hypothesis was that in the presence of the RPU mechanics in the game, statistically significant differences in the averages will be observed, which cannot be explained by the difference in the state of the playing biological object of observation, in the number of wins, at the moments when the RPU is turned on, at the moments when the RPU plays against the player, at the moments when the RPU is playing for the player and when the RPU is not active.

hidden text

Looking ahead, I can say that I reliably found such differences, but not at all where I was looking initially.

Observations were completed after reliable detection of differences in mat. expectation of different data samples and give an idea only about the properties of the RPU mechanics during observations. Further study is needed to obtain more general or relevant conclusions about the operation of mechanics.

Description of the dataset

The data is represented by three variables.

win – 0 loss, 1 win.

date – July day from 12 to 19.

timestamps – the hour when the battle ended and the battle report was received. from 1 to 24.

For the correct classification of battles, two additional variables had to be introduced

battletype

Figure 2. sq. Play with or without a squad. Records with type 1 were used for the analysis – playing alone, without a squad.

Depending on the date, the following mean win values were obtained for each day of observation.

Whatever tests I run, all tests show that the null hypothesis is confirmed. Even the seemingly huge difference between the 12th and the 15th is completely random.

T-TEST GROUPS=date(12 15) /MISSING=ANALYSIS /VARIABLES=win /CRITERIA=CI(.95).

Table 2. Results of comparing the average number of wins between July 12th and 15th, 2019. We accept the null hypothesis that the two means are equal.

It seems that everything, RPU does not exist, the conclusion is unambiguous and there is no doubt left that this is a fiction. But there is still a distribution by the hour. Let’s see it. And here a very curious effect opens up. It turns out that there are hours of the day when playing in which the number of wins is statistically different from other hours of the day. Let’s take all the fights that ended at 18:00 and compare all the fights that ended at 22:00.

T-TEST GROUPS=timestamps(22 18) /MISSING=ANALYSIS /VARIABLES=win /CRITERIA=CI(.95).

Table 3. Results of comparing the average number of wins for fights that ended at 18 hours and fights that ended at 22. We reject the null hypothesis about the equality of the two variances and reliably reject the null hypothesis about the equality of the two means. Sig = 0.004

It must be said that more analysis does not show any differences between fights at other times. Even if there is doubt that such a coincidence is coincidental, due to the multitude of comparisons, the data looks convincing to me.

Conclusion

The RPU mechanic in War Thunder undoubtedly exists and is set up so that a person who constantly plays does not notice its presence. To achieve the effect of its absence, the RPU mechanic uses a more subtle algorithm than just “always on” or “always off” to oppress or play along. It’s safe to say that, for me, in July 2019, the mechanics were set up to counteract my efforts from 6pm to 7pm and make gameplay easier from 10pm to 11pm. The rest of the time, the RPU mechanics are either disabled or work in both directions in a balanced way. It is possible that not for everyone the hour of the wolf and the hour of skill are chosen in the same way, and it is possible that not for everyone this is exactly the hour of the day. Answers to these questions require further research and analysis.

Data file used for analysis I attach.

I would be grateful for constructive criticism. Thank you in advance.