Kyzrati
patreon
Cogmind Beta 15 Stat Summary
🕑 Added 2025-11-04 00:07:18 +0000 UTCCopied from the forums, where you can see this same content with alternative formatting, as well as stat analyses for previous versions. I am adding Patreon as a new outlet this time since it may be more convenient for some.
With Beta 16 released to kick off our future with Unchained in the fleshsteel, time to pore over stats collected from Beta 15, which completed the UFD story arc and added a new ending.
1,400 unique players submitted runs in Beta 15, less than last cycle, but actually a proportional increase considering Beta 14 continued for 9 months compared to less than 6 months for Beta 15. About a third of players were brand new to the game with Beta 15 (can't actually get a precise reading on this)--they've got quite a lot of exploring to do ;)
(As usual, remember we only have data from those that proactively opt in to uploading it, so it's not a complete picture, but no doubt represents the majority of the active player base.)
Despite the shorter cycle, Beta 15 brought in quite a lot of runs, 16,621 in all. 449 (2.7%) of those were using special modes, of which were 57.7% were RPGLIKE runs, the rest fairly evenly distributed among other modes. 1,475 (8.9%) of runs used Adventurer/Explorer difficulty. That leaves us with 14,697 runs in the standard mode on Rogue difficulty for our data summary and comparisons.
Wins
Among such Rogue runs, 166 unique players submitted at least one win, so 14.8% of our total Rogue players this time around (1,124). (26.7% of those (374) who tried out Beta 15 have won at least one run in their history of playing Cogmind so far.) There were 607 wins total, or 4.1% of runs. Not high, but just presented for fun since it isn't meaningful in a sense of the commonly asked question "is Cogmind reliably winnable?" since anyone with the skills to "just win" can do so without much trouble then moves on to other harder challenges, like the many extended win types, or sets their own personal goals.
Regarding win types, 55% of wins were the easiest most straightforward one. The remainder divided up among the other nine types, one of which was new in Beta 15! This particular win type is new and spoilery (and among the more challenging ones) so we'll get to more on it further below, but for basic statistical purposes it accounted for 4.1% of the reported wins (25 times in all), outranking most of other "involved" wins since it was one of the new shiny things :)
Lots of high-scoring wins in Beta 15 (score inflation?! why yes...), so good thing Beta 16 has finally brought us a new higher range of score threshold achievements.
The top 10 runs were spread among 7 players, earning 1x "w0" (basic ending), 3x w1, 3x w2, 3x w6. The pattern generally repeats among the high-score wins, occasionally peppered with other win types. The newest win type (w9), though not itself easy, is relatively lower-scoring by comparison because it occurs long before the others, giving comparatively less time to accumulate score through other challenges on the way. Build-wise, these top runs included 3 of the ever-ready-and-waiting Exo (Cog on my mind, Mediocre*2), 2 heavy tanks (aoemica, Terminus), concept hover combat (Pogmind), a walking executioner (Mediocre), 2 SE constructs (Terminus, Wyvric), and a flight launcher platform (Perseus). Most of these runs sterilized at least one floor, if not more.
Prop
There were a couple big moves in terms of propulsion, in particular treads and legs, so I definitely wanted to examine those.
Siege mode finally got a bit of a nerf in Beta 15, leading one to wonder whether or not the changes would affect preferences for treads. As it turns out, apparently not much. Treads vs legs use in runs that made it to at least -6 was surprisingly very steady going from Beta 14 to 15. Preferences only started to diverge in specifically winning runs, with legs stealing some of treads' thunder there.
As usual we can only read so much into this data since it's also a factor of just what people want to do at the time / what might be trending in the community, but at the same time there is the other simultaneous element to consider: We got a lot of neat new legs in Beta 15! Especially mid/late-game legs with special modes, so these were an attractive new thing to play around with :)
Going back to siege mode for a moment though, there was definitely a hit to usage (as intended). In general siege mode was only used about half as much as before--less frequently, and for less time, regardless of whether looking at the top siege runs or even those with less sieging overall. Among the more siege-focused runs, however, actual total damage mitigation was only down by 27% between each version, pointing to smarter, more effective sieging, which was the main goal of the timing nerf.
Leg special modes have been popular as a new option, where available.
Our top 10 martial mode activators averaged 21 activations, for an average 647 total turns, and 34 martial strikes, on average also deflecting 28 incoming hits throughout their run:
The top 10 shielding runs averaged 20 activations, for an average 337 total turns, and fully deflected on average 48 hits:
Thunder Legs are also quite popular, but we don't have aggregate data reflecting that.
Infowar
We now have good data on infowar! It's a big part of most runs, and there are so many sources of it with more being added all the time, so it's time to record all the details for fun and data analysis :D
This data only includes part-based infowar, when we also have many external sources as well, but this alone is telling enough as it needs to be something you're actively using and have decided to dedicate one or more slots to. (Technically we also always have hack data, another main source, but anyway we're interested in builds here.)
So what are people running...
The types of infowar which are used more often (or mostly ignored) shouldn't be too surprising, but we now have specific data to back up any assumptions. Optics, Sensors, Seismic, and Terrain are the most common four, yup, actually followed pretty closely by Structural scanners! Structural was always pretty decent for the right build, not to mention anyone who manages to get the newer DSS added in Beta 14 is probably going to be using it :P
All others are good for their own purposes, but might be situational, harder to acquire, or just a matter of personal taste.
We can see a similar story when looking at which type of infowar each run used the most out of all options:
While the infowar data above is compiled from all runs, compared to winning runs the only meaningful outliers were Cloaking and Jamming, each of which saw about twice as much use in the latter case (they become gradually even more important and useful later, though you also have more slots to use additional infowar simultaneously anyway).
Shout out to spoofing, which almost no one uses (yeah, it's pretty niche :P).
However some types of infowar clearly lesser used in aggregate do still indeed play a big role in some runs, for some players. Like the Active Sensor Suite, an acquired taste with big pluses but also even drawbacks to compensate, and somewhat more challenging to acquire, features prominently in these runs:
Greed and Smarts
One of the new stat sections I added for Beta 15 was aimed at examining the concept of "backup parts," including slot breakdowns and the average number being carried by players and whether or not they were ever used. I indeed spent a while examining this data, calculating various cross-sections in the search for anything of interest, but it turns out... most of these values ended up being rather boring and unsurprising in the bigger picture xD
Perhaps interestingly, although slower heavier builds tend to carry more backups (duh), almost all builds share a similar ratio of unused to carried backups. I.e. fast builds towards the other end of the spectrum carry fewer backups in all, but expend or replace them at the same average proportional rate over a run. One potential outlier is hover, which seems to be about 15% more likely to leave its spare parts unused.
A slot-wise analysis of backup part types confirmed one logical point: Players are more than twice as likely to drop power sources without ever using them compared to other slots. It's a vital component of many builds and you generally don't want to be without one, but sometimes you feel like you can free up an inventory slot since there'll be easy access to more power later, or because power tends to be a relatively low-attrition part in general, players are acquiring better replacements before using the one(s) they already have.
That's really all I've got xD
Note: For the scoresheet to consider a part a "backup" (spare), it must have been picked up and never attached, but kept in inventory for a minimum number of consecutive actions; then later for it to be considered "unused," either the run ended or it was left behind in an earlier map. Currently this does not consider parts you feed to the Scrap Engine as used, even though it is not uncommon to hold parts in inventory for later feeding, so anyone doing such an SE run will find they have a rather high record for "unused" backup parts :P
ALERT: Alert data incoming.
Another new data-heavy section of our scoresheet, one with far more potential value to players looking at them to understand the dynamics of a run, is the "influence" breakdown. Everything is normally just wrapped up in the alert level system, which while it does give a decent idea of the state of things, and the causes are generally pretty obvious, somewhat more detailed values are not as easy to come by until one does a RIF run (or just get help from others/wiki materials).
In aggregate though, this data is not currently all that valuable, as it's already pretty well known and balanced to within intended parameters from the other side of things. Regardless, here I'll present some breakdowns of the data for your general consumption :)
Surprise! (not really xD) Destroying combat robots earns you the bulk of alert in most cases, even more when you add in the extra for squad leaders, while faster builds may be earning a decent chunk of theirs instead by losing combat pursuers. The rest is a bunch of other random stuff that just adds up.
Notes: Data is from runs reaching at least -7 and up, where alert actually starts to matter, and some sources are too small individually to matter in aggregate so have been lumped into "Other," even though scoresheets break it down further.)
Looking instead at runs for the "single highest factor" contributing to alert among those listed, there are basically just two results: Either destroying combat bots (combat runs), or losing combat pursuers (persistent speeders).
As for contributing factors to dropping alert, again it's just two main factors: Entering a new map, or decay over time.
In the bigger picture even purging alert counts for a small to insignificant portion of most runs. The top purging runs only drop approximately 10-20% of alert gains that way (there are some runs with a much higher purge ratio, but those usually don't make it as far, or drive their alert up much to begin with). That said, it's more about the timing of the purge(s) than the amount, since dropping it below thresholds at crucial points when alert is building dangerously high, or might snowball soon, can make a big difference in survivability.
LU-1G1 of the Beta
While for ages we've had actual cave-ins tallied in the scoresheet, before now we had never recorded potential cave-ins, in the sense of movement through spaces with a chance to cave in. In other words, dangerous maneuvering where the player considers the risk worth taking in lieu of some alternative source of certain danger.
Most runs doing upwards of 50 or even 200 such dangerous moves on average end up with about a 7-8% cave-in rate which can vary somewhat up or down depending on the length of the digs in question.
But as we've seen before, some are just built lucky...
Darcyisverycute is hereby awarded "LU-1G1 of the Beta," for a run that made 73 risky moves, the majority of them in what was likely two very long digs in a single map, but not a single cave in.
3.14 and Naughtyusername also each had a run that engaged in lots of 100% successful dangerous maneuvering, one in the same spoiler map as Darcy and the other in Research, though were a good 15 moves behind.
While we're at it let's also award KTG-V3 Anti-LU-1G1 of the Beta, for getting crushed 25 times over 78 potential cave-in moves, all in the same map, for a whopping 32% cave-in rate in what was apparently a very long and almost unsuccessful dig (they went on to get an extended win anyway, and actually this was a mega dig at the end of their run to seemingly avoid impending doom... honestly, that run looks like it was an impressively huge mess at the end (scoresheet--redacted spoilers)).
From here on we'll start getting into increasingly spoilery content... Scrap Engine, Scraptown, UFD, and finally FRG and the new ending after another break.
Scrap Engine
Scrap Engine has been around for a while now and some folks are good at and/or just love running it, a fact which didn't change in Beta 15 (no nerfs :P).
With Scrap Engine's "full integration" into the new faction as of Beta 14, however, for a time the cost of acquiring it did preclude working with other factions, which was a bit limiting so I went forward with plans to provide a method for avoiding such an outcome. Specifically if you visit the Subcaves, then double back to Materials and visit Storage/Recycling that way, the latter map will contain Subdwellers and the researchers have evacuated their lab, allowing you freedom to steal the tech unknown. SE also gets a popularity boost simply from the fact that it is one of the more effective ways to fight a new late-game optional challenge.
So in Beta 15 we did indeed see a resurgence in SE use, with 12.7% of players using it at least once (back up from 6.3% in Beta 14). 2% of all runs used it, though that value is somewhat less meaningful than looking at the portion of runs visiting Recycling that took it, which is 14.5%. Fairly popular overall. (I was planning on running with it again in the near term, but not streaming or playing as much for now...)
60% of SE runs used the non-violent path of acquisition, so 40% chose violence.
20.1% of SE runs won, so higher than the average, which makes sense given that anyone going for this very out-of-the-way unique part probably has a better idea of what they're doing. Indeed the average number of previous runs by players getting SE at all was... 224 :P
The majority of SE runs tended to reach at least Research, except for one unfortunate experience where someone took -9/Storage into Recycling, stole the Scrap Engine, spent hundreds of turns trying to build constructs from all the junk there, then died to a Sentry while roaming about the floor xD (this after successfully blasting quite a few other defenders).
Scraptown
Scraptown was mostly completed previously in Beta 14, except for the main reason it exists! The big event that occurs there (civil war) and full proper alliance with 0bPrime made it an even bigger attraction (while importantly also enabling the intended balance). So how did people fare...
14.3% of runs visited Scraptown; 39.1% of players visited it at least once.
There are now three routes to reach the location, including a couple different events that might provide a direct shortcut, both to help newer players or for just anyone to get there more safely and quickly if desired and the optional is available. 39.5% of runs took a shortcut, while about a third (36.7%) took a longer safer route via Storage, with the smallest group (23.8%) taking the most dangerous route: Subcaves directly to Recycling, which has you potentially facing more Subdwellers and other out-of-depth enemies.
About that big event, its consequences are fairly significant both in the short and long term, so definitely another aspect we want to analyze a bit!
21.8% of Scraptown visits didn't talk to Optimus at all, in some cases because they just missed him in the council room, or maybe just passed through outer Scraptown with other plans (or to do something especially crazy like attack them, or just accidentally pissed them off somehow).
War broke out for 40.5% of those who talked to Optimus, either because they wanted it (looting opportunities!) or just couldn't make it out quickly enough. (Beta 15 also included a bug, now fixed in Beta 16, that did not properly guarantee the ability to exit before the war if you go directly for the exit after talking to Optimus and hearing him tell you to leave, so the event was a little more common than desired in those cases.)
Of course the full effects of the war will not all be apparent on first encounter, and the fact that it happens at all will be a bit surprising/sudden, so some players will adjust plans in the long term to accommodate, one way or another. I would expect more players to leave before the war begins, but only so much relative to those who stick around, since the latter group will also sometimes grow to include those who want to loot the place, depending on the run type and what's available at the time. (Unfortunately we have no way of accurately determining from scoresheets exactly what was looted in this manner.)
With the UFD story arc/civil war complete, there was/is far less reason to ever want to attack them head on yourself. Still one potential reward in particular requires attacking to acquire, and in some cases others may theoretically be easier to acquire via proactive violent means. Also some evil Cogminds just want to attack them for fun. So it's no surprise that attacks on Scraptown dropped from 19.5% of runs to 5.9%. Their behavior in this case was also rebalanced to increase the difficulty of attacking, at least in the long term--one is more likely to want to steal whatever the target is then get the hell out. This properly reflects the vastness of their local resources, unlike before where it was instead set up as a challenge where you could effectively drain them all and treat the place like an early-game arena with numerous rewards at the end :P
85.4% of attackers did not talk to Optimus first, and 87.8% died without ever making it out (compared to 76.6% in Beta 14). 10 were killed by Optimus personally.
The Companion Sword continues to be a big draw for visiting Scraptown, though unfortunately I don't have enough data on that since I forgot to add a key point in the scoresheet... oops, we'll have that in the future.
Botcubes we do have data on, however, and players have been putting together some fun ones, as usual. 9.4% of Scraptown visits created at least one Botcube, down from last version (12.4%) when they were first introduced. 34 runs created two or more. H3-RMN (Hermelin) was our top Botcube connoisseur, creating 10 in a single run, followed by [former top Beta 14 Botcube connoisseur] f3-rnm (fernman) with 6. (Already seeing Hermlin share and fight alongside some amazing Botcube designs in Beta 15.)
0bPrime
So you're working with Optimus and have made it out of Scraptown... time to get hacking!
This was an incomplete feature in Beta 14 given that you could only use UFD hacks to get reinforcements (and by extension resupplied with Scrap Suits / overtuned parts), but as of Beta 15 was properly balanced for its full functionality.
Time to start getting into the new ending and redacted late-game spoiler territory.
On average, winning UFD-allied runs installed Fedlink 11.8 times, though this value includes non-UFD win types (for example due to insufficient installs, or otherwise just derailed runs). Specifically UFD wins (w9) averaged 21.4 installs, the greatest number from a run by puppuppit (36). 9 runs had more installs than that, all the way up to a high of 58 (!), but either lost or ended up going for a different win type. My own UFD win (streamed that one--my first visit to that map!), had the second-lowest number of Fedlink installs out of all the wins xD (12, "beat" only by MTF at 10...).
One thing to remember about these counts is that they're force hacks, meaning an investigation occurred for each one, and therefore higher enemy density with potentially higher alert and attrition. Fun times.
Scrapoid and Scraphulk summoning are also force hacks, though come at the cost of UFD resources on the whole, so most runs tend to use them sparingly, except in emergencies (I did!) or those where the goal of joining is actually to access said resources rather than necessarily aid the UFD.
Beta 15's most prolific Scrapoid summoner was f3-rnm (fernman), calling them in 9 times all the way up to -1/Access.
UFD winners tended to do little to no Scrapoid summoning, to conserve resources, only ever 1~2 at most. That said, those runs on average spent more resources on Scraphulk summoning instead (but still never more than twice per run). Mmm, deadly Scraphulks... Also it takes some build-up before that option even becomes available, so that makes sense, although some non-UFD wins did summon significantly more hulks, in some cases probably after having built up resources before changing direction for whatever reason, or explicitly wanting to make use of Scraphulk gear. Among them Savant had one run that summoned 10 Scraphulks.
Aside from looking at Fedlinks and summoning, I don't have the actual 0bPrime readiness value upon finally triggering the new ending event, since I forgot to put it in the history log, but later we'll have both that and ECA-like per-map values in the scoresheet to see how it fluctuates over time.
Readiness aside, key to the new ending is bomb installation, and we do have info on that below...
"w9"
119 runs met Optimus in the Protoforge, 21% of those took the win (Optimus himself survived in all but 5--sorry buddy I couldn't save you either, but we did it!).
26% left the map at some point, 45% of those after completing the bombing mission, but before the countdown completed (most of which went on to win some other way). The other 55% noped out before bombing was completed, including runs that went on to achieve some other extended win and probably just invited the UFD in to help loot the Protoforge as one huge distraction :)
Among all Drillbomb mission completions, an average 5.8 were installed by the player, 4.2 by the UFD, so players win... for now! :P
There were actually several deficiencies in the UFD Lugger AI in that version which could cause them to do weird unnecessary stuff, sometimes at great cost, and after improvements for Beta 16 I bet the average will shift a bit. I know their performance has improved after giving them better handling for special situations. Technically though the values are still to some degree dependent on player strategy, since one can choose to play more of an escort role, or speedrun your own installations, or help terraform the surroundings to affect routing, or hold important bottlenecks while the UFD does it's thing, and so on.
One aspect of the new ending event I wanted to examine given anecdotal evidence is that it can be pretty long. This isn't unexpected, and perhaps a fair trade in that it occurs multiple maps prior to the "normal" end of a full extended run. While I expect the improved AI might cut down on that a bit, it's also just the nature of the event, a longer adventure containing more leeway for approaching it in your own way over time, and also numerous opportunities to make comebacks as it plays out. Some players even make plans to completely rebuild from scratch after reaching that point.
Anyway, mainly out of curiosity I wanted to check the event play time, but since we only have that info for the run as a whole, action count could suffice. The average run-wise action% for all Drillbomb completions runs in the Protoforge was 19% (i.e. a fifth of the actions taken in a run were in that map). Values ranged from about 10~25%.
Who are these heroes of 0bPrime...
Overall the UFD ending has gone quite well, both as an optional alternative conclusion to the plot while also offering its own fun and unique challenges. Happy with it!
(In hindsight looking at that chart above, there seems to be a correlation between lower score and direct installs by player... might be able to draw some interesting conclusions from that.)
Superfortress
The elephant in the room!
Yeah most people do their best to ignore and avoid the elephant, opting instead to live a long and prosperous robot life, but some are are drawn to its awesomeness, eager to best it in whatever crafty way they can.
Really, we're in spoilers territory here...
Protoforge's Superfortress exists mainly for lore, though also as a deadly but pretty localized threat and optional extra challenge.
Of 299 runs entering Protoforge, 47 (15.7%) destroyed SF. That's a decent chunk, though of course many come well prepared with that goal in mind (and also include patrons, who were testing this out with prerelease runs beforehand as well).
I don't have a full statistical accounting of the methods used... though we do know a range of options available to players, as recorded on Cog-minder. Telefragging it with UFD support is certainly the most common, and by far the easiest to pull off. Indeed 31 of the 47 kills met Optimus in Protoforge. Simply luring also works, in that case.
Of the 47 SF kills, 19 went on to do the first ever +++ runs. One '+' for one boss, and we've had ++ for ages but finally added a suitable enough target for the third "boss." Not really a boss in the same major NPC sense, but as the most powerful bot in the game, and one that raises the rest of the route to even higher difficulty, it really does deserve the recognition of a '+'.
Note that relevant to this section of the world I also eventually want to look a bit at players' choice of Protoforge vs. Section 7, but we should wait on that since Protoforge is "the new thing" as of Beta 15 which has a greater impact on decision-making. For now players are more likely to go there even if they know they're not ready for it, plus there's the added danger of being less familiar with the area to begin with.
Meta
We haven't taken a look at some global player stats for a while, so where do we stand on those...
66 (4.7%) of our 1400 submitting players were using ASCII mode, a value still settling down into its expected range among an expanded player base always comprised of a portion of core roguelike fans but mostly new players/others.
A much greater 328 (23.4%) use keyboard mode.
Primary movement methods:
46% mouse
31% numpad
15% arrows
8% vi
Vi is still surprisingly high given where we are today in terms of player representation!
UI layout:
64% semi-modal (default)
25% modal
9% non-modal
Platform:
93.2% Steam (non-handheld)
4.5% non-Steam
2.3% SteamDeck
Also I don't recall having looked at achievement rates as a global stat before, but it might make sense to check in on those considering Beta 16 just added a ton more achievements. The average completion rate across all players at the end of Beta 15 was 27%.
9 players had >= 90%:
(We know there are players who've hit 100% before, but they weren't playing this particular version--these figures are for Beta 15 submitters only.)
