War Robots WR Data and Analysis

By the Numbers: The View from the Bottom

Jay's taken Gepard Diary somewhere it's never been before: to Android!

“Hey Dredd,” Aurora Excelsis Chapter Leader SprintingGoat had messaged me the other day, “what’s your take on Pixo’s motivation for structuring the Gold chests the way they have? It’s actually the first time I’ve been genuinely concerned about where they might be taking the game, where the state of gameplay is running a very, very distant second behind revenue raising. It just seems a surefire way to flood the middle and upper low tiers with Ancilots… essentially bringing ‘stale’ gameplay upstream.”

I told him that I wasn’t especially worried. After all, my own calculations had shown that the Gold chests “only” offered +20% EV. To get that, you’d need to invest 1,000Au, which most players at the lower level probably didn’t have to spare. And since EV only normalizes across an average, one-off Gold chests had a 2/3rds chance to be a losing proposition. How flooded could the market be? In that regard, it’s sort of like the stock market- the ones making a lot of money on the stocks tend to be the ones with a lot of money to put into the stocks to begin with.

“Yeah,” he replied, “I agree on the whole, but I have to think there’d be a lot of folks out there saving for their fifth slot or first Lancelot, and given how the spin cycle tends to work, as more people tend to get roped into gambling their savings, that a really decent chunk will come out with a garage full of stuff. I’m not convinced either way, but it certainly has me thinking there’ll be a big difference in the ‘average hangar.”

I still wasn’t convinced. “Another thought I had was that this may be a result of so many people upset about the past events. If you look at it long-term, like the last year, this may do nothing more than break the community even when compared to past losses.”

Although I didn’t realize it at the time, turned out SprintingGoat and I were- somewhat ironically- talking past one another.


Back when the new matchmaker had rolled out, I downloaded War Robots onto an Android tablet I have but seldom use, and went through the tutorial and some early-level play. I thought it might be interesting to see what the new-player experience was from a data and analysis perspective, but as things continued to evolve and I found myself busy on iOS, the idea fell by the wayside.

Unlike past events, there’s a lot of positive feeling about the Anniversary one- unsurprising given the +EV chests. One veteran player remarked that if they were coming into the game right now, they’d buy $100 of Gold and run it all through the lottery without a second thought. “What better way,” the idea went, “to stock up your hangar early with things you know you’ll eventually need anyway?”

It was a compelling thought. The kind of compelling thought that only grew more compelling the more I ignored it. And the more I ignored it, the closer it got. The challenge to put my money where my mouth was became the tipping point, and after charging up my tablet, I logged in to my old level 6 Android account. $100 later, I headed to the casino.

When I walked out, I had a five-slot hangar that was ready to rumble with a Geppo (Gekko Gepard), Orkan Galahad, Tulu Doc, Orkan Rogatka, and Ancile/Zeus Fury.

Pretty good for Private 1!

Level: 15

Cups: 3

Victories: 25

On to the numbers! But first, a quick note about sample size. My standard sample is 110 data points (the data from 10 matches). That’s not possible at these lower levels, as I’m finding myself promoted before I’ve played ten matches. So my sample size for Private 1 is 88, and for Bronze 3 is 66 (should have been 77 but lost a game from a technical issue).


Average Player Level (Private 1): 16.36

Average Player Level (Bronze 3): 18.94

While this stat has lost some of its relevance in my iOS reporting, since I’m at level 30 there and so is the bulk of my opposition, it’s certainly relevant here. Clearly in each case I’ve been lower than the average level, but since the matchmaker cares only for performance, I can hardly say I’m surprised. My five-slot beast-mode hangar is one of the strongest I’ve seen in the lower levels, so I’m punching well above my weight if you look at my level.

Average Victories (Private 1): 198.51

Average Victories (Bronze 3): 234.58

Holy Maloney, now those are some experienced pilots! Your average player in Private one is a seasoned veteran of (going by the 50% win ratio rule of thumb) around 400 matches! Now I see why Android players often say their platform is much harder than iOS, where-

…wait, what’s that you say?

Ahh, right, tankers.

So this is the point in today’s programme where we point out that mean (“average”) is an occasional liar and frequent thief. In each league I ran into a pilot which had in excess of 10,000 wins under their belt- and not the same pilot! Others “only” had a few thousand. These were hardly the norm, but their numbers are so great that they yank the average up and present a patently false picture.

Instead, to get a better idea of what I’m actually facing, we’re going to employ a value called median, which is the number in the center if you take all the values and sort them highest to lowest. That means the person who has 10,000 has the same weight as someone who has 10.

Average Median Victories (Private 1): 35.50

Average Median Victories (Bronze 3): 48.00

Ahh, that’s more like it! Our revised statistic here underlines the inexperience of the player pool I”m facing. And that’s no slight on them- we all got started somewhere. But maybe we didn’t quite expect to have to face some guy with 10,000 wins- or with $100 worth of chests.

This naturally begs the question, just how prevalent are tankers? Longtime readers will remember the post here that took the notion of “rampant clubbing” to task, concluding (from my own experience) that clubbing was rare in frequency, but highly disruptive (and thus likely to be remembered and/or overstated) in experience. While I can’t necessarily pinpoint a tanker given the range of playstyles the practice affords (from hard tanking, to beacon-only, to “fading” or giving well below your usual effort), I can at least use career victories as a handy proxy for experience. Where you draw the line is up to you.

Fig. 1: Win distribution

Average Trophies (Private 1): 28.03

Average Trophies (Bronze 3): 43.14

Again, no surprises here. With the cups statistic now reflecting seven days’ worth of activity instead of ten, these numbers will be smaller than we may have previously been accustomed to. Nevertheless, I’ve used this value as a proxy for activity, and this is in line with what I’d expect to see for new players building a body of experience. It’s worth noting too that the value is derived from damage dealt, and most players at this level aren’t dishing out a ton (yet).

A 50% jump in activity seems about right for moving to the next level. That wouldn’t fly in the later stages of progression, but dealing with the modest numbers here it’s not too onerous a task at all.

Player Divisions

Here’s the distribution of what leagues the matchmaker pulled from in order to populate my games.

Fig. 2: League distributions

Although given the smaller than accustomed sample sizes, I do find the matchmaker choices interesting in how far- and how often- it needs to pull in for players. Note that the most frequent opponent when I was Private 1 was by far Bronze 3, which is a likely indicator of pool size and player distribution within the game itself. A small number of new and veteran players on each end of the spectrum, and a vast body in the middle.

In other words, a bell curve!


Here comes the fun part. Since the event has given lower-spectrum players the opportunity to field higher-spectrum hangars, this has the potential to be all over the place.

Average Hangar Slots (Private 1): 3.02

Average Hangar Slots (Bronze 3): 3.14

A modest uptick between Leagues, little surprise there. At this stage of the game, both 1000Au and 5000Au for hangar expansion can be cost-prohibitive. For those still occasionally wondering at the value of a fifth slot, this provides ample illustration.

Average Robot Level (Private 1): 2.98

Average Robot Level (Bronze 3): 3.89

Average Highest Weapon Level (Private 1): 4.40

Average Highest Weapon Level (Bronze 3): 4.54

A big jump in bots not mirrored by a reasonably comparable jump in weapons either indicates that players at this level are prioritizing one over the other, or that something else is at play here. While there’s certainly no discounting the former, given how important health and speed are to bots, it’s much likelier that as you begin to see more Leos and Natashas enter the player pool, the more the bot level is going to experience an immediate spike. This is because those bots are already level 6 when you purchase them, so they impact the average straightaway without having to be leveled up to significance first.

Not much of an increase in highest weapon levels here, but looking back to the league distribution only about 18% of the Bronze 3 player pool were in Leagues that were too high for my Private 1 matchmaker to draw from. In other words, I was dealing with a lot of the same level of players. And since weapon upgrading is on a fixed axis of time (versus victories, cups, and other participation-dependent values), it potentially would advance more slowly than pilots themselves. Later in the game it becomes much harder to advance up the ladder, but down here it doesn’t seem to take nearly as long.

What bots are people running? 

This is where we get to really look under the hood at this engine of progress we’ve been given. One of the things the player level system did was restrict the ability of players to play certain bots and weapons until they’d “unlocked” them through sufficient leveling.

Want to run a Lancelot? No problem, but you’re gonna need to be level 15 in addition to having 5,000Au. A wallet will let you bypass one of these on the spot, but can only speed the other one along (through premium, if you go that route). With the chest prizes ignoring level altogether, you could be level 5 (the minimum to participate) and win a Lancelot, and begin playing it right away.

And this gets to the heart of SprintingGoat’s concern, as discussed in the introduction of the article. By allowing players to shortcut the system, has harm been done to the lower levels of play? I didn’t initially think so, but I’m a bit more mixed now. Let’s pause here for some data.

Fig. 3 Bot distributions

Because of the smaller sample sizes, you don’t want to go too deep into the analysis, but it is interesting to see some trends starting to take form here. We know from my other research that Griffins comprise a huge part of the leater-league metagame, around 25%, so seeing them start moving boldly in that direction is perfectly consistent. Since they unlock only at level 16, the low Private 1 league prevalence is largely attributable to unavailability, not lack of interest.

The Natasha, which unlocks at level 18, also begins to move upwards, as she’s a very common early bot for many. Note too that the Patton appears to slip here, which is also consistent with later findings. It’s not a terrible bot, but the more common loadouts that make it effective- Aphids and (debatably) Gekkos– are both Au weapons. Without those, it’s slower speed and tendency to lose weapons with damage make it less desireable.

Here’s the same data, resorted by prevalence in Bronze 3.

Fig. 4 Bot distribution by prevalence

Again, few surprises. Early-game bots like the Destrier, Cossack, and Vityaz going down. Later-game bots like the Leo and Natasha going up. But take a moment to note the results in the Type column. The overwhelming majority of bots I encountered at the lowest leagues were Silver. The proportion of Au/WSP/Event bots were miniscule.

This does help put some of the concern about the lower leagues in perspective. We may face hundreds of bots- but you can bet you’re gonna remember that one Butch a lot more than you will your thirtieth Golem.


As I mentioned above, I’m a bit mixed on this. It does seem to be a case of the problem being a little overstated, but my concern in handing a bunch of strong bots out to lower-level players is that while the pilots should see themselves up the ladder through winning, it will be at a lower pace because they simply lack the Silver for mass upgrading. My Level 15, five-slot monster has only 2.5m Ag to its name, so I have to upgrade carefully and deliberately.

This- potentially- could have an effect similar to a slow-moving air bubble traveling up a hose. I’m not going to say anything conclusive on this with early, small-sample data, but this will be a focus of my attention as I move up the ladder with my own ridiculously overpowered hangar. I’ll be looking to take a sample with each move up the ladder, so expect more Android data in the days ahead.

The other point worth noting is the delicious irony on display. I began this blog in part looking at the clubbing phenomenon, and have been fairly clear and consistent on my opinion of that sort of activity. And yet, by doing nothing more than simply enjoying Pixonic’s latest event, I have become- undeniably- a seal clubber.

There are few joys for me in games like this, where I ran around the battlefield nuking Destriers and Golems with my virtually untouchable Galahad. It was nearly a one-bot game until I just stopped caring about self-preservation, and when I lost the Galahad I hopped in the Geppo and finished off the last few remaining opponents.


So yeah, definitely looking forward to working my way up the ladder. Adversity is the whetstone of the mind!

Thanks for reading.


2 comments on “By the Numbers: The View from the Bottom

  1. Pingback: By the Numbers: At the Gates of Hell – Gepard Diary

  2. Pingback: By the Numbers: Apple Seeds – Mech*Spectrum

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: