Project Bathyscaphe- the name I’m using to refer to my use of a Pixo-given iOS account to collect data rising up through the leagues from scratch- has been a treasure trove of ‘lessons learned.’ Some of these lessons are obvious and clear-cut, like the value of shelving the high-level bots and weapons for awhile and working with an army of Lights, relying on my own player skill and five-slot hangar to get me through quickly.
In a sense, though, another of the recurring problems I faced was the opposite issue. I was so focused on trying to get through the lower ranks as quickly as possible, in the hopes of minimizing my footprint, that I actually compromised the mission a bit.
As any good statistician will tell you, sample size is everything. If you doubt that, try running a poll on a political issue and asking only one person for their answer. Obviously, you need a lot of data points in a sample to make it representative, and the number I’ve settled on over time with these analysis articles is 110: the data of ten matches.
The problem I ran into was that the ascent in the lower leagues is fast. Like, really fast. If you show a streak of aptitude, Pixonic is very keen to get you up to better competition straightaway, leaving less-skilled adversaries in the churn below to grind a skillset off of one another.
As a result, I experienced the data collection equivalent of the bends, ascending faster than was healthy, and so some of my sample sizes are lower than I’d like. Sixty-six. Eighty-eight. So take the results with an extra grain of salt.
Now that said, they don’t seem that far off from expectation, so if there’s an abundance of variation distorting the result, it’s not much of a distortion. Imagine it’s like the difference in a few percentage points of interest in your savings account- but your savings account only has twenty dollars in it.
So with the preamble out of the way, howabout some data? Today we’ll be adding in the data from Bronze 3 to Bronze 1.
Here’s a look at the opposition as I coasted through iOS Bronze. Since these graphs will be cumulative for reference and convenience, I’ve tried highlighting the new bits, let me know if that works or not?
Again, remember that I don’t have any data for early Android, so that part’s blank. But overall, both iOS and Android seem to be tracking fairly consistently in terms of player level and median victories. Median victories are used as a proxy for experience (the longer you play, the more wins you collect), while trophies are a proxy for activity level.
There we see a big difference- a jump of almost fifty points- but I would certainly hesitate to conclude that iOS players at this level are just that much more active than their Android brethren. Note also the slight dip in player level in iOS Bronze 2 compared to Bronze 3 and Bronze 1. This isn’t a straight-line process, just a trending one.
Figures 2 and 3 above represent player populations, by which it’s meant what level opposition am I facing at each level? One of the things we often hear in explanation for one thing or another is that the iOS population of War Robots players is significantly smaller than the Android one.
This data bears that out. Note that my matches in iOS needed to pull from a much broader range to find enough people to fire. My Bronze 1 matches had to pull as high as Silver 1, which is somewhat startling given the potential power imbalance represented by those two designations.
You know what else jumps out at me? The stunning lack of parity in iOS matchups. Bronze 3 Android was a bit off in that regard, with the most players being pulled from Bronze 2 (39.39%), but Bronze 3 still had a significant presence (31.82%). Both Bronze 2 and Bronze 1 had solid parity, with the highest pull being from players in the same league as myself.
But iOS? It’s all over the board. Most of my fellow players in Bronze 3 came from Bronze 1. An overwhelming number of my Bronze 2 cohort came out of Silver 3 (more than the rest combined, if you leave out the Unassigned), and fully half of my Bronze 1 matchups pulled from Silver 3– and with an equal amount coming from Silver 2 as Bronze 1!
I have to wonder with a touch of concern what the impact will be of Pixonic’s stated goal of introducing a more rigid player pool system, where you should be playing only within your league. If iOS has to forage this far afield to fire my matches, how much longer will my wait times be?
To my friends at Pixonic: longer queue times are far less welcome than player population variance.
See, while I find the numbers interesting in what they describe, I cannot say that I’m having bad matches- on any of my accounts on either platform. Put another way, I’m not myself experiencing the problem that a more monogamous player pull will fix, and if it means I might have to wait longer for matches, that’s an undesirable change.
Now that we’ve looked at who is playing, let’s turn to what they’re playing. The first graph is one I forgot to include in my last stats article, so we’ve got a little ground to cover.
Hangar slots- to no-one’s surprise- happily tracks upward with each advance in league. We’re not far above three-slot territory here and won’t be for awhile, but the values in this table are ones that should quite steadily trend upwards. Between Private 2 and Bronze 1, for instance, note that average bot level effectively doubled (2.24 to 4.60).
Another takeaway is that players continue to prioritize leveling up their weapons over their bots. In the pre-matchmaker days, this was directly attributable to the fact that the tier system seemed to weigh bots more heavily than weapons, so the maxim of “keep your bot two levels below your weapons” was the order of the day. While the difference now is much closer to one level than two, either old habits die hard, or perhaps that’s simply the correct play.
Finally, in terms of the actual bots themselves?
Right, that’s a wall of numbers. Let’s see what sense we can make of it.
Griffin: Look, should we be surprised that a bot claiming a 25% share of the higher-end metagame is finding use down here? No, we should not. And the sale offerings Pixonic has implemented only driven some of the better bots downward, since that circumvents the level restrictions that had formerly allowed for a more graduated pace of meta shift in the lower leagues.
Leo: Less popular than the Griffin, the Leo’s high degree of survivability thanks to a deep health pool will always ensure it has its adherents at this stage of the game.
Natasha: Another popular early Heavy bot, the Natasha is the bot of choice for those who (putting it charitably) feel that their best work is done on the periphery of the battlefield. Like the Leo and Griffin, recent sales promotions from Pixonic have led to the bot’s greater prevalence in lower-league play than noted previously.
Cossack: The little bot that could has just about reached the end of his sell-by date as we approach Silver, going from above 30% prevalence to less than 5%. Outside of beacon-hungry Event quest chains, there’s little reason to fly the friendly skies in this rig. With any luck, Beacon Rush or other future game modes will make the Lights much more viable, because they’re terrific bots that history has sort of left behind.
Destrier: Our hapless “starter bot” suffers from all the limitations of the Cossack, with none of that bot’s advantages. Outside of novelty (or, perhaps, boredom), there’s little reason anyone would have to level this thing up to 12. It, too, is on the cusp of extinction at the end of the Bronze Age.
Schutze: As above, so below. Once a feared clubber bot, the Schutze is now an afterthought. #rhymetime
TLDR: Silver Heavies are going up, Silver Lights are going down. Gold, WSP, and Event bots continue to be marginal.
That’s it for Bronze. Next time, we’ll jump right into Silver, which is arguably the point of entry for the general “War Robots experience” that carries us up into Diamond.
Thanks for reading!