fortnite
Claire Morrison
11/1/2022
fn <- read_sheet('https://docs.google.com/spreadsheets/d/1wFb31JqI0QDXmijN82p_oJt3Km28_WsI0Fa8VfFjkp0/edit#gid=0')
head(fn)## # A tibble: 6 × 14
## place elim accuracy damage_to_p distance_m mats_used damage_ta…¹ hits player
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <chr>
## 1 27 0 0 NA 778 110 205 0 claire
## 2 10 6 13 NA 3000 370 430 48 brend…
## 3 8 1 9 NA 2000 210 201 10 claire
## 4 25 1 9 NA 1000 40 313 8 brend…
## 5 49 3 26 NA 473 20 247 19 brend…
## 6 70 0 0 NA 103 0 114 0 brend…
## # … with 5 more variables: date <dttm>, account <chr>, game <chr>, build <chr>,
## # ...14 <dbl>, and abbreviated variable name ¹​damage_takennrow(fn)## [1] 210str(fn)## tibble [210 × 14] (S3: tbl_df/tbl/data.frame)
## $ place : num [1:210] 27 10 8 25 49 70 15 29 99 99 ...
## $ elim : num [1:210] 0 6 1 1 3 0 1 4 1 0 ...
## $ accuracy : num [1:210] 0 13 9 9 26 0 23 9 0 13 ...
## $ damage_to_p : num [1:210] NA NA NA NA NA NA 303 433 0 NA ...
## $ distance_m : num [1:210] 778 3000 2000 1000 473 103 2000 2000 39 10 ...
## $ mats_used : num [1:210] 110 370 210 40 20 0 100 39 0 0 ...
## $ damage_taken: num [1:210] 205 430 201 313 247 114 341 226 102 105 ...
## $ hits : num [1:210] 0 48 10 8 19 0 18 18 0 1 ...
## $ player : chr [1:210] "claire" "brendan" "claire" "brendan" ...
## $ date : POSIXct[1:210], format: "2022-11-01" "2022-11-01" ...
## $ account : chr [1:210] "grapedaddyz" "grapedaddyz" "grapedaddyz" "grapedaddyz" ...
## $ game : chr [1:210] "solos" "solos" "solos" "solos" ...
## $ build : chr [1:210] NA NA NA NA ...
## $ ...14 : num [1:210] NA NA NA NA NA NA NA NA NA NA ...plot all correlations
ggpairs(fn[,c(1:9)])## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
damage dealt
ggplot(fn, aes(x=player, y=damage_to_p, fill=player)) + geom_violin() + theme_bw()## Warning: Removed 14 rows containing non-finite values (`stat_ydensity()`).
kill to death ratio
the standard way to measure success in video games.
here - KDT= eliminations over damage taken (because once you die you are out of the game)
and DDDT= damage dealt over damage taken
fn$KDT<- (fn$elim/fn$damage_taken)*100
fn$DDDT<- (fn$damage_to_p/fn$damage_taken)ggplot(fn, aes(x=player, y=KDT, fill=player)) + geom_violin() + theme_bw()## Warning: Removed 3 rows containing non-finite values (`stat_ydensity()`).
ggplot(fn, aes(x=player, y=DDDT, fill=player)) + geom_violin() + theme_bw()## Warning: Removed 16 rows containing non-finite values (`stat_ydensity()`).
efficiency
what is the average amount of damage dealt per each hit?
for only those games with nonzero hits
# fn[fn=="Inf"]<- NA
#fn[fn=="NaN"]<- NA
fn$effic<- ifelse(fn$hits>0, fn$damage_to_p/fn$hits, 0)
ggplot(fn, aes(x=player, y=effic, fill=player)) + geom_violin() + theme_bw()## Warning: Removed 13 rows containing non-finite values (`stat_ydensity()`).
accuracy
ggplot(fn, aes(x=player, y=accuracy, fill=player)) + geom_violin() + theme_bw()## Warning: Removed 1 rows containing non-finite values (`stat_ydensity()`).
creating a variable that is accuracy divided by place for games with nonzero hits. this variable attempts to capture accuracy over time in a game. there may be games where you have very high accuracy, but are eliminated almost immediately (e.g. 80% acc/98th place = .816). this is not as impressive as having very high accuracy on a higher (lower) place game (e.g. 80% acc/ 5th place = 16). it does not make sense to include games with 0 hits and therefore 0% accuracy in this statistic.
fn$acc_dur<- ifelse(fn$hits>0, (fn$accuracy/fn$place), NA)ggplot(fn, aes(x=player, y=acc_dur, fill=player)) + geom_violin() + theme_bw()## Warning: Removed 21 rows containing non-finite values (`stat_ydensity()`).
does efficiency predict damage dealt to damage taken ratio?
ggplot(fn, aes(x=effic, y=DDDT, shape=player,color=player)) +
geom_point()+
geom_smooth(method=lm)## `geom_smooth()` using formula = 'y ~ x'## Warning: Removed 16 rows containing non-finite values (`stat_smooth()`).## Warning: Removed 16 rows containing missing values (`geom_point()`).
do both efficiency and accuracy predict finish?
m2<-(lmer(place~acc_dur+effic+(1|player), data=fn))## boundary (singular) fit: see help('isSingular')summary(m2)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: place ~ acc_dur + effic + (1 | player)
## Data: fn
##
## REML criterion at convergence: 1604.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -1.4167 -0.7764 -0.2964 0.6272 2.8450
##
## Random effects:
## Groups Name Variance Std.Dev.
## player (Intercept) 0.0 0.00
## Residual 490.7 22.15
## Number of obs: 178, groups: player, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 41.2378 3.5744 175.0000 11.537 < 2e-16 ***
## acc_dur -2.7394 0.4259 175.0000 -6.432 1.16e-09 ***
## effic -0.2129 0.1268 175.0000 -1.679 0.0949 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) acc_dr
## acc_dur -0.216
## effic -0.848 -0.046
## optimizer (nloptwrap) convergence code: 0 (OK)
## boundary (singular) fit: see help('isSingular')only accuracy (accuracy/place) significantly predicts place, controlling for each player as they might have different playing tendencies.
randoms <- ranef(m2)
dotplot(randoms)## $player
do efficiency and accuracy predict damage dealt to damage taken?
m1<-(lmer(DDDT~acc_dur+effic+(1|player), data=fn))
summary(m1)## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: DDDT ~ acc_dur + effic + (1 | player)
## Data: fn
##
## REML criterion at convergence: 588.8
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -2.6679 -0.6623 -0.1495 0.4529 4.3226
##
## Random effects:
## Groups Name Variance Std.Dev.
## player (Intercept) 0.2021 0.4495
## Residual 1.4484 1.2035
## Number of obs: 178, groups: player, 3
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 7.164e-01 3.281e-01 3.892e+00 2.184 0.0962 .
## acc_dur 1.479e-01 2.328e-02 1.736e+02 6.352 1.80e-09 ***
## effic 2.927e-02 6.904e-03 1.734e+02 4.240 3.63e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) acc_dr
## acc_dur -0.112
## effic -0.499 -0.048both accuracy (accuracy/place) and efficiency significantly predict damage dealt to damage taken, controlling for each player.
randoms <- ranef(m1)
dotplot(randoms)## $player
how does distance relate to place?
fn$place_cat<- ifelse(fn$place<6, "top 5", ifelse(fn$place<21, "top 20", ifelse(fn$place<51, "top 50", ifelse(fn$place<76, "top 75", "top 100"))))
ggplot(fn, aes(x=distance_m, color=place_cat)) + geom_density(alpha=.3)## Warning: Removed 2 rows containing non-finite values (`stat_density()`).
the games with the highest finishes (top 5) clearly involve the most running!
do the three players differ significantly on accuracy?
tapply(fn$acc_dur, fn$player, mean, na.rm=T)## brendan claire mike
## 2.826538 1.537823 1.740210fn$BvAll<- (fn$player=="brendan")*-2/3 + (fn$player=="claire")*1/3 + (fn$player=="mike")*1/3
fn$MvC<- (fn$player=="brendan")*0 + (fn$player=="claire")*-1/2 + (fn$player=="mike")*1/2
mcSummary(lm(acc_dur~BvAll+MvC, data=fn))## lm(formula = acc_dur ~ BvAll + MvC, data = fn)
##
## Omnibus ANOVA
## SS df MS EtaSq F p
## Model 66.661 2 33.331 0.012 1.16 0.316
## Error 5346.319 186 28.744
## Corr Total 5412.980 188 28.792
##
## RMSE AdjEtaSq
## 5.361 0.002
##
## Coefficients
## Est StErr t SSR(3) EtaSq tol CI_2.5 CI_97.5 p
## (Intercept) 2.035 0.478 4.261 521.799 0.089 NA 1.093 2.977 0.000
## BvAll -1.188 0.834 -1.424 58.248 0.011 0.932 -2.833 0.458 0.156
## MvC 0.202 1.345 0.150 0.651 0.000 0.932 -2.451 2.856 0.881no, there are no significant differences in accuracy among players.
what about efficiency?
tapply(fn$effic, fn$player, mean, na.rm=T)## brendan claire mike
## 22.56149 21.56334 19.57958mcSummary(lm(effic~BvAll+MvC, data=fn))## lm(formula = effic ~ BvAll + MvC, data = fn)
##
## Omnibus ANOVA
## SS df MS EtaSq F p
## Model 197.953 2 98.977 0.005 0.474 0.623
## Error 40482.361 194 208.672
## Corr Total 40680.314 196 207.553
##
## RMSE AdjEtaSq
## 14.445 -0.005
##
## Coefficients
## Est StErr t SSR(3) EtaSq tol CI_2.5 CI_97.5 p
## (Intercept) 21.235 1.238 17.153 61399.642 0.603 NA 18.793 23.676 0.000
## BvAll -1.990 2.182 -0.912 173.493 0.004 0.93 -6.294 2.314 0.363
## MvC -1.984 3.470 -0.572 68.212 0.002 0.93 -8.827 4.859 0.568no, there are no significant differences between players on efficiency.
what about DDDT?
tapply(fn$DDDT, fn$player, mean, na.rm=T)## brendan claire mike
## 2.022930 1.013206 1.579484mcSummary(lm(DDDT~BvAll+MvC, data=fn))## lm(formula = DDDT ~ BvAll + MvC, data = fn)
##
## Omnibus ANOVA
## SS df MS EtaSq F p
## Model 36.265 2 18.133 0.086 9.031 0
## Error 383.484 191 2.008
## Corr Total 419.749 193 2.175
##
## RMSE AdjEtaSq
## 1.417 0.077
##
## Coefficients
## Est StErr t SSR(3) EtaSq tol CI_2.5 CI_97.5 p
## (Intercept) 1.539 0.122 12.591 318.273 0.454 NA 1.298 1.780 0.000
## BvAll -0.727 0.215 -3.375 22.864 0.056 0.937 -1.151 -0.302 0.001
## MvC 0.566 0.343 1.653 5.485 0.014 0.937 -0.109 1.242 0.100there are, however, significant differences in damage dealt to damage taken between players. namely, brendan has a significantly higher DDDT ratio than mike and claire. mike and claire do not significantly differ.
or place?
tapply(fn$place, fn$player, mean, na.rm=T)## brendan claire mike
## 32.71654 32.03571 43.96296mcSummary(lm(place~BvAll+MvC, data=fn))## lm(formula = place ~ BvAll + MvC, data = fn)
##
## Omnibus ANOVA
## SS df MS EtaSq F p
## Model 3105.237 2 1552.618 0.019 2.03 0.134
## Error 158306.687 207 764.767
## Corr Total 161411.924 209 772.306
##
## RMSE AdjEtaSq
## 27.654 0.01
##
## Coefficients
## Est StErr t SSR(3) EtaSq tol CI_2.5 CI_97.5 p
## (Intercept) 36.238 2.309 15.691 188295.957 0.543 NA 31.685 40.792 0.000
## BvAll 5.283 4.064 1.300 1292.183 0.008 0.922 -2.730 13.295 0.195
## MvC 11.927 6.479 1.841 2591.518 0.016 0.922 -0.847 24.701 0.067there do not seem to be significant differences in finish between the group either.
averages by day
fn$date2 <- as.Date(as.character(fn$date), format= "%m%d%y")
fn$date2<- mdy(fn$date)## Warning: All formats failed to parse. No formats found.day<- fn %>% group_by(player, date) %>%
summarise_at(vars("accuracy", "place", "effic"), mean, na.rm=T)
day## # A tibble: 45 × 5
## # Groups: player [3]
## player date accuracy place effic
## <chr> <dttm> <dbl> <dbl> <dbl>
## 1 brendan 2022-11-01 00:00:00 18.1 51.1 8.02
## 2 brendan 2022-11-03 00:00:00 27.2 25 17.7
## 3 brendan 2022-11-06 00:00:00 24.1 19.4 30.9
## 4 brendan 2022-11-07 00:00:00 34.5 25.8 32.0
## 5 brendan 2022-11-08 00:00:00 15.6 29 26.3
## 6 brendan 2022-11-10 00:00:00 14.1 47.2 15.3
## 7 brendan 2022-11-13 00:00:00 19.1 16.4 27.4
## 8 brendan 2022-11-15 00:00:00 22.7 31.2 20.1
## 9 brendan 2022-12-01 00:00:00 23.5 29.9 24.0
## 10 brendan 2022-12-06 00:00:00 32.5 15 28.7
## # … with 35 more rowsggplot(day, aes(x=date, y=place, colour=player)) +
facet_wrap(~player)+
geom_line() +
xlab("")+
theme(axis.text.x = element_text(angle = 75, hjust=1))
ggplot(day, aes(x=date, y=accuracy, colour=player)) +
facet_wrap(~player)+
geom_line() +
xlab("")+
theme(axis.text.x = element_text(angle = 75, hjust=1))
ggplot(day, aes(x=date, y=effic, colour=player)) +
facet_wrap(~player)+
geom_line() +
xlab("")+
theme(axis.text.x = element_text(angle = 75, hjust=1))
are there any differences in xbox gamertags?
differences might occur if a certain username is a higher level, or has a propensity to be in a harder or easier lobby.
ggplot(fn, aes(x = place, y = account, group = account, fill=account)) +
geom_density_ridges() +
facet_wrap(~game)## Picking joint bandwidth of 6.49## Picking joint bandwidth of 11.6
ggplot(fn, aes(x = DDDT, y = account, group = account, fill=account)) +
geom_density_ridges() +
facet_wrap(~game)## Picking joint bandwidth of 0.572## Picking joint bandwidth of 0.463## Warning: Removed 16 rows containing non-finite values (`stat_density_ridges()`).
ggplot(fn, aes(x = elim, y = account, group = account, fill=account)) +
geom_density_ridges() +
facet_wrap(~game)## Picking joint bandwidth of 1.3## Picking joint bandwidth of 0.741
ggplot(fn, aes(x = damage_to_p, y = account, group = account, fill=account)) +
geom_density_ridges() +
facet_wrap(~game)## Picking joint bandwidth of 217## Picking joint bandwidth of 161## Warning: Removed 14 rows containing non-finite values (`stat_density_ridges()`).
sensivity analysis to see if there are differences in game type: solos vs duos
mcSummary(lm(place~game, data=fn))## lm(formula = place ~ game, data = fn)
##
## Omnibus ANOVA
## SS df MS EtaSq F p
## Model 9939.894 1 9939.894 0.062 13.649 0
## Error 151472.029 208 728.231
## Corr Total 161411.924 209 772.306
##
## RMSE AdjEtaSq
## 26.986 0.057
##
## Coefficients
## Est StErr t SSR(3) EtaSq tol CI_2.5 CI_97.5 p
## (Intercept) 16.793 5.011 3.351 8178.241 0.051 NA 6.914 26.672 0.001
## gamesolos 19.942 5.398 3.695 9939.894 0.062 NA 9.301 30.583 0.000ggplot(fn, aes(x=game, y=place, fill=game)) + geom_violin() + theme_bw() +
facet_wrap(~player)## Warning: Groups with fewer than two data points have been dropped.
mcSummary(lm(acc_dur~game, data=fn))## lm(formula = acc_dur ~ game, data = fn)
##
## Omnibus ANOVA
## SS df MS EtaSq F p
## Model 15.77 1 15.770 0.003 0.546 0.461
## Error 5397.21 187 28.862
## Corr Total 5412.98 188 28.792
##
## RMSE AdjEtaSq
## 5.372 -0.002
##
## Coefficients
## Est StErr t SSR(3) EtaSq tol CI_2.5 CI_97.5 p
## (Intercept) 3.091 1.054 2.934 248.472 0.044 NA 1.013 5.170 0.004
## gamesolos -0.839 1.135 -0.739 15.770 0.003 NA -3.077 1.399 0.461ggplot(fn, aes(x=game, y=acc_dur, fill=game)) + geom_violin() + theme_bw() +
facet_wrap(~player)## Warning: Removed 21 rows containing non-finite values (`stat_ydensity()`).## Warning: Groups with fewer than two data points have been dropped.
mcSummary(lm(effic~game, data=fn))## lm(formula = effic ~ game, data = fn)
##
## Omnibus ANOVA
## SS df MS EtaSq F p
## Model 500.681 1 500.681 0.012 2.43 0.121
## Error 40179.634 195 206.049
## Corr Total 40680.314 196 207.553
##
## RMSE AdjEtaSq
## 14.354 0.007
##
## Coefficients
## Est StErr t SSR(3) EtaSq tol CI_2.5 CI_97.5 p
## (Intercept) 25.742 2.666 9.657 19216.214 0.324 NA 20.485 30.999 0.000
## gamesolos -4.499 2.886 -1.559 500.681 0.012 NA -10.192 1.193 0.121ggplot(fn, aes(x=game, y=effic, fill=game)) + geom_violin() + theme_bw() +
facet_wrap(~player)## Warning: Removed 13 rows containing non-finite values (`stat_ydensity()`).## Warning: Groups with fewer than two data points have been dropped.
interestingly, there are significant differences in finish between solos and duos, but not accuracy or efficiency.