In this final Splatfest Analysis, we will be taking a closer look at the distributions of the 9 games I had with the Splattershot to determine whether its consistent place in the top 3 was due to me consistently playing well with the weapon, or whether the averages were skewed by particularly high leverage outliers.

The points scored with the Splattershot were relatively consistent around 800-900, but a particularly good game where I scored 1180 points has skewed the mean positively, but it wasn’t so dramatic a swing to cause the mean to drift far from the median; it was near where most of the points were clustered.

The number of kills I got each game with the Splattershot were also consistently around 5-6 kills each game. A slightly low game with 3 kills did cause some skew negatively to the mean, but again it stayed close to the median and where most points were clustered.

The number of deaths each game also showed very little variation, with the most frequent number of deaths being 0. It ranged from 0-4 with a fairly even spread, meaning no individual point had a particularly large influence on the mean of 1.67.

In conclusion, it does seem that I was consistently good with the Splattershot – meaning it was well deserving of its top 3 positions despite its lower sample size.

And with that this utterly riveting multi-part analysis comes to a close! My main take-aways from a data visualisation and analysis stand-point when carrying out these analyses were that:

• Box and Whisker Plots in Excel 2016 are fickle and I’ll likely avoid them in future!
• A more heavily annotated set of graphics may have been more suitable, instead of writing up everything in these report-style blog posts. This would allow the analyses to better live as separate entities from this blog, which is especially important when doing analyses of sensitive data to make sure the data can’t easily be divorced from the analysts’ findings and any caveats to the data.
• Using Excel to do ‘proper’ data analysis with the AnalysisToolPak is very easy. Until starting this I was unaware Excel could do things like linear regression (almost) natively, usually I would move my data over to a more specialised analytical software package like SPSS (and I have a particularly large disdain for SPSS!).

Thanks to those of you who took the time to read my ramblings and please don’t be afraid to leave any feedback you have in the comments section.

Here is the Excel file I used to create all of the graphics for this analysis: Link.

Up next in this multi-part analysis is a look at my performance with each weapon I used during the Splatfest. I’ll take a closer look at my Win to Loss ratio by each weapon and how well I did overall with them regardless of outcome of the game.

To get things started we’ll take a look at the number of games I used each weapon – this is necessary to help contextualise the rest of this analysis. Those weapons that have a high number of uses are much more likely to be reliable when making assumptions about how good or bad I was with the weapon as they are much less likely to be influenced by random chance. For example with the Krak-On Splat Roller, I can’t be 100% sure whether any of my results are definitely because I’m bad with the weapon, because with only 2 uses my performance could be as a result of a number of factors like: my team composition (what weapons my team mates were using); the enemies team composition (did I just get unlucky and fight against those weapon types that counter Roller weapons?); or whether I was just having a bad/off game. Weapons like the Sploosh-o-Matic are more likely to have the effects of those previously mentioned factors smoothed out simply by having a larger sample size – I’m much less likely to have uncomplimentary team composition every game for 21 games than I am for 2 games as an example.

Now let’s look at my Win to Loss ratio with each weapon:

Ketchup vs Mayo, indicating my best (ketchup-coloured) and worst (mayo-coloured) Win:Loss ratios, yum.

Those weapons I did worst with (Blaster, Tentatek Splattershot and .52 Gal) are all weapons I used less than 5 times, so it’s hard to draw any concrete conclusions about whether I would have averaged out higher/lower with more games – but I can tell you that I did purposefully stop using these weapons because I felt completely useless with them (when we look at the next couple of charts we’ll see whether I actually was useless with them). The Tri-Slosher is the clear winner out of these – there are rumblings within the Splatoon 2 community that this weapon is overpowered so I’m not massively surprised by this outcome. Its large ink spread, ability to 2-hit kill from a decent range and lack of ‘RNG’ (it will always have the same ink spread every shot unlike other guns) mean it’s much easier to consistently do well with, which is most likely why I did so much better with it than other weapons over the course of the Splatfest. Of the weapons I used a lot of times (> 5 uses), the lowest win to loss ratio was with the Kelp Splat Charger, which as discussed in my previous post I got significantly (statistically speaking!) worse with throughout the Splatfest, overall losing a game for every game I won.

KDA is calculated using Kills plus Assists all divided by Deaths.

Mayo coloured bars are those weapons I did particularly poor with… Or you could think of it as the colour for the Krak-On Splat Roller, either works.

There’s one clear conclusion we can draw from these… I suck with the Krak-On Splat Roller. As you can see I definitely was not doing well with this weapon with the lowest average points, KDA and average number of kills (and highest average number of deaths). I stopped using that weapon with good reason! Even accounting for potential bad team compositions or enemy counters I don’t think I would have done any better, I was absolutely terrible with it.

Shooter weapons like the Splattershot and Sploosh-o-matic tended to be at the top of the average points overall, and as these weapons are generally recommended for Turf War it makes sense.

Despite my high win to loss ratio with the Tri-Slosher it ended up fairly middle of the pack for average points scored, being at the bottom of the group of weapons that I used frequently. However, it topped the average number of kills and was very close to the front of the pack with regards to KDA, potentially being why I won more games with it.

The other weapons I used less than 5 times also don’t seem to have performed particularly awfully. The .52 Gal was consistently at the bottom of the pack but was a country mile better than my performance with the Krak-On Splat Roller, and wasn’t too far from the rest of the pack, while the others were fairly spread out, with the Blaster even being the weapon with the 2^nd most kills and 3^rd highest KDA. The Blaster’s low point scores were its major downside, meaning I may be good with it, but it’s not particularly suited for Turf War so I may have been correct in swapping out from it.

The Splattershot maintained a top 3 position in each of these metrics, which is probably why it averaged out to a positive win to loss ratio too. Being the all-rounder weapon it fits in well here – especially as it was the first weapon I used a good amount in both Splatoon 1 and during the first week after Splatoon 2 was released. The lower sample size (9) makes it hard to properly compare it with my other most used weapons, so I may take a look at its distributions for each of these metrics in a future part to see if any high leverage points may have skewed the average heavily for this weapon.

So in conclusion, I should never ever use the Krak-On Splat Roller (without a good amount of practice) and I was right to only use it twice. I tended to do best with the Tri-Slosher, with high amounts of kills and a decent KDA. My most consistently good weapon seems to be the Splattershot, keeping top 3 positions across all the metrics I looked at (although its lower sample size means it’s less easily comparable with my other most used weapons).

That’s it for this part! I’ll be taking a deeper dive into some of the specifics I brought up within this part of the analysis and looking at how I performed when winning and losing in the next exciting (… ahem) installment of this analysis.

Splatoon 2 came out on the Nintendo Switch a couple weeks back, and they’ve had their first ‘Splatfest’ since the game was released. ‘Splatfests’ are worldwide events that pit two similar things against each other in an opinion poll (for this Splatfest it was Mayo vs Ketchup). After voting for your favourite, you’ll then be playing the game mode ‘Turf War’ against members of the opposite team, to determine which side is the best at Splatoon. I picked Ketchup, and I battled valiantly to prove that Mayo is bad and everyone who likes it deserves to lose at Splatoon.

During this Splatfest, I used Nintendo’s phone app to collect the data of my Splatfest matches in order to analyse my performance over the 24-hour event. I’ll be using this data for a multi-part analysis, covering a variety of sections like which weapons was I best with, what stages I was better at and more!

Today’s post will be covering whether I improved at the game throughout the Splatfest, mainly focused on correlation analysis.

Did I improve my performance during Splatoon 2’s ‘Splatfest’ event?

I wanted to check whether I had improved at all throughout the Splatfest – I did this by looking at my points scored, number of kills and number of deaths, and seeing whether any improved as I played more.

Looking at the correlation matrix and the P-Values for each of the options, most of my performance stayed relatively similar and the main reason my performance tended to vary was down to random chance, not me showing marked improvements or deterioration; apart from two things:

1. I actually got worse at covering turf the more I played with the Charger weapon type (shown by the highly negative correlation coefficient of -0.63). This wasn’t simply due to random chance (and was therefore significant), as its P-Value was below the Alpha value of 0.05.
2. I tended to die less the more I played. I had slightly negative correlation (with a coefficient of -0.27) between me playing more games in the Splatfest and the number of deaths I had. This was also not due to random chance, as the P-Value again was less than the Alpha value.

I could tell that both things were happening as I played. When I first started playing with the ‘Kelp Splat Charger’ I was having absolutely amazing games, often being top of my team’s scoreboard, but as I played with it more, and when playing during the second day of the Splatfest I could definitely tell my aim was getting worse and I was generally not playing as tightly as when I started. As the Splatfest went on I could also tell that I was slowly becoming more conservative in my play style; I felt that my team mates were always going in and dying constantly and I thought it would be best to generally stay back and make sure my team could super-jump to me (it rarely ever happened though… my team mates were playing like they had never played the game before, despite most being a similar level to me).

I also did a quick check to see whether points scored, number of kills and number of deaths all correlated, and it wasn’t really a surprise to see that as I scored more points I tended to get more kills and die less, but it’s nice to confirm it!