Cogmind Beta 16 Stat Summary (Patreon)

Copied from the forums, where members can see this same content with alternative formatting, as well as stat analyses for previous versions.

The leaderboards have been reset, Beta 16 (and 16.1!) is behind us, and it's time to pull some data from those versions to analyze, examining both the general state of the game as well as anything new and noteworthy in the context of stats.

In most cases here we'll be looking at stats for both 16 and 16.1 together, even though technically this affects the results due to some differences between the two. I'll add a note wherever that's especially relevant. Of course there is also some significant content only present in 16.1, in which case that discussion will naturally exclude the earlier version.

1,730 unique players submitted runs in Beta 16, not deviating much from expectations given its duration and comparisons to previous versions. There were 22,861 runs in all, 614 playing special modes, and 2,152 using non-Rogue difficulty settings. This leaves 20,153 regular Rogue runs to reference for the majority of our stats. (I will be making an exception or two below, noted where applicable.)

Wins

219 different players won at least 1 run, with 1,181 reported wins total (5.8% of all runs).

448 wins (38%) were by the 11 players populating the top 10 list by win count:

Alongside tallying win counts for this beta cycle, I included some other interesting stats for these players, highlighting highs and lows as well.

"ae" by far got the highest number of wins, many of them relatively quick flight runs of various types, though scoring on the very low side overall.

C0-2MO did a fair number of pacifist speed runs this time around, but plenty of others as well--overall spending a lot of time playing.

CaptainWinky shares a lot of videos and does a huge variety of runs, tending toward mid-duration highly destructive runs. He's also just really good at Cogmind, earning the highest win rate among the top 10 despite doing crazy stuff, just really good at going with the flow (the way Cogmind was primarily meant to be played :D).

szymekc does longer runs that generally aim for a large number of maps and extended wins, tending to be highly destructive but more towards the end, including some of the cycle's most destructive wins.

Hermelin streams a lot, including directly on Discord, also uploading videos for some runs. This approach no doubt contributes to the longer average length of his runs, but also in particular to this beta Hermelin was our foremost Unchained hunter (which we'll get to later), and the abnormally high sterilization rate is another contributing factor as one of the more reliable Unchained summoning rituals.

Like szymekc, "the" and AngelSpirit both also take longer, more destructive routes.

Despite having the lowest win rate on average in this group, shapeshifter submitted a high total number of average wins, both in duration and score.

chebyrekser plays on the slower side but wins quite a lot.

alice_fexa also does videos (Japanese), also known for a wide variety of run types.

Savant runs came in the slowest on average, but also holds the highest average score in this group and the second-highest win rate. Surprisingly in combination with one of the lowest sterilization rates across runs, which is instead normally seen in low-scoring or speed-focused runs.

Switching gears to look specifically at the individual top 10 runs, we see an almost completely different set of names! Literally szymekc is the only one populating both lists :P (though if we expand this to the top 20 scores, most are also shared with the previous list)

The highest-scoring run was by Darcyisverycute, at 399,170, making heavy use of the Gui. EMP Blaster before it finally got nerfed again in Beta 16.1 (the story behind that is recorded in those release notes if you missed that and are interested). This 33-hour run did 5 sterilizations in all and destroyed 2,445 robots, which is insane, spending nearly half the run at high security, using RIF as a complementary strategy while obliterating Zion, W base, SF, 4 different Unchained (there were only 5 total in the game at the time xD), and of course the other major extended targets. Overall it was a very long run that visited 43 different maps. They have some Cogmind videos here, but this particular run was not recorded, to my knowledge, just reported along the way.

It's not surprising to see SE master Wyvric sitting at #2 with their typical approach.

Pogmind was a RIF-heavy run making significant use of allies and a good mix of prop across the run, cycling through everything before eventually turning to our newer types of legs.

Szymekc breezed through most of their run on flight before switching to a concept-heavy legs build for the final three depths. Szymekc allied with the UFD, as with in fact half of these top 10 runs, a common approach to help take down the optional new powerful target added in Beta 15, though in several cases also taking the opportunity to complete the UFD mission before ducking out to continue on with their own long-term plans. (Notice that the entire top 10 is accompanied by +++ on the leaderboards, all except for the run by Cogchaaaan.)

Boris ran a mostly-hover build with an early companion evo into SQC domination. Also UFD focused (completed mission and left).

Terminus: Scrap Engine!

Dwarfinator: Beta 16's highest-scoring Borg run, executed almost entirely on wheels. (UFD)

saltine: Scrap Engine!

Cogchaaaan: A UFD flight run transitioned to legs combat for the final two depths.

spider: Another UFD flight build for much of run before gradually transitioning into an 8-weapon scrap tank for the last few maps.

You can see these individual scoresheets (and all others) via the archived Beta 16 leaderboards. (Some of these runs were streamed in discord, but I'm not sure if/which might have also been uploaded elsewhere.)

54.0% of all wins were of the basic type (a typical portion):

Catching Up

Before we get into newer content, there are cases where I didn't yet have the scoresheets recording data useful for reflecting recent features, or it was too early to say much, so I wanted to catch up on a few of those...

As of Beta 16 we finally got SS stats in the form of Scrap Suit Usage (%) and Parts Cannibalized.

Usage percentages accumulated across entire runs aren't necessarily representative of how useful the tech was for a given run, due to the possible shifting of build types across longer runs. For example Terminus, one of the major SS fans, had a 235k-point win containing the highest number of parts cannibalized during the beta (261), but composite SS use only accounted for a 14% of that run by the end.

Even kite, with the highest-recorded SS use by turn count across an entire winning run (49%) accumulated that value entirely on a single map (the final one before they won). (21 other runs reported a higher percentage use overall, though none of those went on to win.)

spider, who had the second-highest number of cannibalized parts (117), also only used the upgraded version after acquiring it from its late-game source.

Only between 1-2% of winning runs used SS to any respectable degree, again many of these focused on the late-game UFD applications, pretty useful for convenience in terms of firepower and build maintenance there. This is essentially how it was designed to begin with--good for certain stretches of a run, excelling at a relatively specific subset of build types, and situational assuming at least the right complementary parts are available.

Beta 16 scoresheets also started recording data relevant to another piece of UFD-sourced tech: CPS Tubes. They're not a listed numerical stat, but now that their targets are reported to the history log I can parse them to pull out those details.

2.8% of runs used at least 1 CPS Tube. Also 43.6% of those runs used only one.

That doesn't mean they're not useful, of course, just situational (being a consumable, after all). Browsing through the list of typical targets, even a lot of these "single uses" were in winning runs, and in many cases reserved for very worthy targets. Even for late-game targets this doesn't usually mean someone reserved the option for more than half a run after the most common source of CPS, rather they more often would have acquired it from a later source (since sourcing them is not especially hard with UFD support, though does require some planning). There will also be another major source in the future--lore-wise CPS tech was not originally developed by the UFD, even though it clearly suits their tactics...

A total of 1,249 different targets were hit by CPS Tubes in Beta 16 (various abbreviations used to avoid too many spoiler names):

(The list above leaves out a couple hundred scattered targets that were clearly just randomly chosen (sometimes even accidental) or don't fall under any notable category.)

Unsurprisingly, Botcubes top the list as 15.1% of CPS targets, not just increasing the salvage return when they fall, but even more importantly the only way to guarantee the original cube drops as well in order to rebuild. It's a good combo. Also the counts here can be inflated by doing this repeatedly on the same Botcube for each rebuild, where possible.

Various UFD bots intended to be recovered as scrap come in a close second at 14.4% of targets, also not surprising since they're a good source of backup parts for SS users while also being a source of CPS to begin with, lowering the opportunity cost of doing so.

The rest of the list is populated mostly by more powerful targets or at least those equipped with one or more specific parts players find especially valuable. "Hero" on this list includes not just Heroes of Zion, but any otherwise named good combat ally.

As for player-wise usage statistics, Hermelin alone hit 190 targets with CPS across 53 different runs... a true salvage madman accounting for 15.2% of all uses in Beta 16. Actually a fair number of other players totaled 30-50 targets across their runs, just no others nearly that high.

The highest usage count by any single player was, surprise, Hermelin with a record 23 in a single run, not especially partial to any one target type, but distributed across Botcubes, Borg, and a variety of other eventual victims throughout an entire run. spider actually came in close with 20 in one run, though less varied overall--mostly targeting UFD bots, and almost entirely during a single lengthy late-game event.

While extracting this CPS data I noticed we technically also have the custom Botcube names set by players, assuming the name change occurred prior to being hit!

This allows us to check out what cool, funny, or telling names folks have been coming up with...

Some themes that become apparent from the list (basically reinforcing what we already know about them):

Botcubes get build and rebuilt, upgraded or downgraded as conditions permit (I removed some cases of this from the shared list to avoid too much repetition)

Botcubes are often created in large part from the loot of a singular enemy or ally, usually but not always a powerful one

Based on the data 20.0% of Botcubes were named (115 out of 575 total), or at least named prior to CPS. A minority 38.8% of CPS'd Botcubes were unnamed, so even ignoring the potential for late namers, most people CPS'ing their Botcubes are also nice enough to bestow a name upon them :)

There are more new stats that fall under the umbrella of "catching up," but I'll cover those at the end due to spoilers, in case some readers want to nope out before that.

New Features

Part studies were added in Beta 16, and have naturally become fairly popular even beyond just folks testing it out because it's one of the latest new mechanics, with many runs choosing to rely on the tech for at least some of their build needs.

42.3% of runs acquired at least one part study, averaging 4.3 per run. By comparison, 100% of winning runs acquired at least one study, averaging 7.7 overall.

The greatest number of studies in a single run was 31, by C0-2MO. 33,663 parts were studied in all.

For this version I also added full studied parts lists, akin to the fab/repair/schematics lists, so we can also examine which parts are popular study targets. Note however that for 3.7% of study-containing runs we don't actually know any of the specific ones obtained, because they were later lost due to an in-game action (I should probably keep that in the scoresheet but mark them as lost, so we still have a record in such cases...)

Below are some summaries of studied parts (no big spoilers by default, since you can't study uniques, prototypes, or non-0b10 parts anyway :P):

Of course the most studied parts are all common ones you can find right at the beginning, which makes sense both due to ease of access but also just a way to eke out some extra benefits when slot count is low, builds are weaker, and you don't have any better targets anyway.

As you can see, three of the top ten are wheels, in particular a great study because these are technically available and useful for the entire duration of a run.

Lgt. Treads is also a good one since it generally takes a little while to reach better treads, but studying them removes one of the bigger reasons one might want better treads in the first place--to avoid being too slow due to going overweight.

Rocket Launchers can also use the accuracy boost, while also being one of the two main early launcher options.

I explored the complete study breakdown by part subtype to see if there were any outliers in terms of frequency, which might suggest preferences, and interestingly most of them fall somewhere between 2,500-3,500 or so, the primary outlier being just guns at around 5,000.

Another telling way to look at the data is to examine which higher-rated parts are studied at the greatest frequency, proving these otherwise relatively easily obtainable parts are still useful in the late game at a time when it's already likely one has sufficient access to prototypes and uniques:

Four of the ten are flight units, studies being a significant boost for those (especially at high slot counts), plus flight builds tend to be more stable in the long run as well as mobile and hacky, so obtaining schematics is even less trouble for them.

Four more are power sources, those often used by some late-game hostiles and therefore a good target even for combat builds that might need to salvage one or more replacements over time, or are using those to begin with (not uncommon).

The remaining two items are among the more popular late-game weapons that happen to not be prototypes and get a lot of general use (HERF more often with QC, Nova often in stacks, but also sometimes just one of the best reliable thermal cannons to keep around).

One of the other big Beta 16 features was temporary free access to the Assimilator, by way of the Subcaves lab. So of course some folks were doing that, as it's essentially yet another whole new play style wrapped around a single part, albeit more challenging. I streamed one myself which was pretty fun.

26.9% of runs visited Subcaves to begin with, compared to 27.8% in Beta 15, so not much difference there (a harder area, after all, and Subcaves were already a newer desirable destination for other reasons in the previous version).

4.2% of Subcaves visitors later created at least one Borg (so only 1.1% of all runs), though this number will be lower than the number who actually acquired 1B-0RG's Assimilator to begin with, since it's not possible to use it immediately, and often isn't even something that can or should be used until the mid-game at the earliest due its slots requirements.

About 1 in 20 players created Borg in at least one Beta 16 run. SunShine214's persistence is of particular note, alone accounting for 27 different Borg runs and many more attempts than that via Subcaves over weeks of playing several quick runs per day. SunShine214's most Borg-heavy run created 61 in all, peaking at a simultaneous collective size of 5.

The top Borg leaders created many more in some of their runs:

Only 3 of those runs went on to win, however. A lower 16.5% of Borg runs won in all, though note many Borg runs do not necessarily use the Assimilator for the entire run, anyway, so there are other significant factors involved. On that note, only 29 runs ended with the Assimilator attached, winning or not, this out of 230 total Borg runs. And only 11 of the runs ending with the Assimilator attached were wins (including mine! :D). Again it's the kind of thing players might intentionally ditch towards the end, or even earlier and not by choice if it's lost to bad situations (also Beta 16.1 did not yet make it immune to becoming broken, which I had forgotten to patch in at the time...).

19 different runs (8.3% of Borg runs) reached the maximum collective size (8), almost entirely by different players since it's not usually an effective use of resources to go that large, but it is powerful if you can manage it, and tied to a challenge achievement for that reason :). About half of the runs that reached the maximum collective size did eventually win, reflecting the caliber of player managing to hit that goal in the first place. In my only Borg run I only ever reached a more reasonable size 5 collective :P

Beta 17 swaps out the Assimilator for the Transmogrifier, another build-defining part, so we'll get to look at that in a future stat review.

Unchained

The star of the show has arrived! Obviously one of the main features of Beta 16 was the titular Unchained, so it's time we analyze all things UC.

Despite being a major feature, given their origins and aim of course not everyone experienced them directly, for better or worse. Some folks really wanted to see any fight them, but it's not always that easy, and others would hope to avoid them but get smacked down or forced to flee in terror instead.

Some important notes before we get started here:

While technically 9 different Unchained were added during Beta 16, CY-PHR is excluded from this section of the analysis since both triggering and interacting with them is quite different from the others (we'll cover CY-PHR separate further below).

Of the other 8 UC, five came in Beta 16 and three in 16.1, but except where necessary we'll just discuss them all together.

One of the first things I wanted to confirm is that Unchained are only messing with experienced players, and even those less so on non-Rogue difficulty modes. So taking a look at our 2,094 regular Adventurer/Explorer runs, I see that only 12 of those (0.6%) included Unchained dispatches. 4 of those went on to win, and of the others that did not win, only 2 lost directly or indirectly due to the UC encounter (DW-4LL in both cases :P). In each case only a single Unchained was dispatched across the run, and in 2 cases Cogmind destroyed them (DW-4LL and CL-0N3). In examining all the relevant scoresheets to determine whether player loss could possibly be attributed to the Unchained, I saw that in the two runs where that was indeed the case, one was a RIF build that ducked into a Garrison not long after DW-4LL showed up then died inside, and the other was DW-4LL chasing them to the surface exit in Access, but they had no way through the doors.

From that POV, it does indeed look like this subset of players did not see much impact from Unchained overall, as expected. Further confirmation that new players are safe lies in the metadata, which shows the average cumulative play time for Adventurers/Explorers encountering UC stood at 224 hours.

It's worth noting that due to various trigger/behavior changes since these 12 runs were played (both in Beta 17 and even in Beta 16.1 where some originated in 16.0 data), most of these runs would not have even gotten their one encounter to begin with. You've gotta be really dedicated in those difficulty modes to even have a chance :P

Now though we turn our attention to the bulk of the data, Rogue difficulty, to see where UC stand with the vast majority of players...

3.4% of Rogue runs saw at least one Unchained dispatch, a meaningfully higher value, if still quite low. 52.8% of those runs went on to win, undeterred. That's an even higher rate than the easier Adv/Exp modes, though admittedly the latter group has a small sample size.

But also its worth considering just the type of player both 1) playing Rogue difficulty and 2) getting these dispatches to begin with. Looking at the average play time of those seeing Unchained, it's a whopping 469 hours. Yeah, fairly experienced ;)

The handful of players getting UC at the low end of play time with "only" dozens of hours of experience under their belt were playing Beta 16 at the time and having that happen due to large amounts of hacking, before CY-PHR's arrival could change that in Beta 16.1 for a generally more targeted response.

So all that reconfirms the same trends across even Rogue (yay!).

Before we move on, realize that while Unchained are more likely for veterans, and also usually triggered as a response to more extreme play styles or events, our data will also be affected by those aiming for challenges. For example during Beta 16.1 we had a bunch of folks trying for the new "I'm Coming For You" achievement, and without branches combined with gearing up for a big fight, more than one such player was surprised to actually meet an Unchained, something that wouldn't have happened under normal circumstances in a "normal" run.

Community trends always have a way of working themselves into the data with every version, this one was just of interesting relevance here.

Let's now refocus on other details particular to Unchained in Rogue.

Twice as many wins completely avoided Unchained vs those that reported at least one dispatch. A much closer number of non-w0 wins (those with additional requirements, e.g. extended areas) was achieved between UC and non-UC runs, suggesting that they may in fact derail plans to attempt harder endings, if encountered. This is expected in some cases, and also backed up by anecdotal evidence, but it makes sense to see the effect more broadly reflected in the data.

Of course one of the the main questions on everyone's mind at the start was: Just how are they triggered? What makes Unchained come after you in the first place? Some reasons are pretty obvious, and even used by players to forcefully summon UC in an attempt to hunt them for their parts, or bragging rights. Others a bit less so, but also less common in general, which makes sense since again it's more about adding further challenge in the wake of extreme behaviors.

Over time I did start more explicitly sharing some of the triggers, and continued with more detailed coverage with patrons as I was keeping a close eye on how these mechanics were performing in real runs, while further tweaking values as well as planning for the bigger 16.1 (and later Beta 17) updates.

Beta 16 was the first time I'd ever actively watched stats throughout its months of play, having added a way to see exactly when, where, and how quickly players were engaging in behavior that could lead to Unchained, since these are rarely one-time instant triggers, but instead gradually accumulated values that can cool off as well. Thus I could also determine how close a given run came to triggering Unchained, and at what points, even if it didn't actually happen, or even if a trigger occurred but there was no dispatch for a number of other reasons (they can be blocked or postponed by events, or being in certain irrelevant locations, among other factors).

In examining the trigger data for our purposes here, we'll look at all runs containing active triggers, even if no dispatch actually followed. Although only 693 runs announced at least one Unchained dispatch (the 3.4% figure mentioned earlier), 1,121 triggered Unchained. So that's 5.6% of runs that triggered at least once, a little less than twice as many than actually saw them. There was no meaningful differences between the portion of runs triggering Unchained between 16 and 16.1, though there are other meaningful differences in trigger distribution, as we'll see below.

Here's a breakdown of the trigger types in Beta 16, followed by the same data for 16.1:

The biggest take-away from the first graph is that machine hacking accounted for nearly half (43.8%) of the triggers. At the time we didn't yet have a hacking-specific Unchained response (CY-PHR), so most of the runs containing that trigger never would have actually seen UC at all had that been added already, further cutting down on the threat that might pose some unsuspecting builds. Good because that's a very... asymmetric response :P

I always thought it would be ideal if and when we had enough Unchained to at least somewhat calibrate responses depending on what players did, or their build type, but at the very least it would be necessary to address extreme machine hacking in a different manner, hence 16.1 getting on that right away.

This had two results, one of them clearly visible in the second graph: Hacking triggers dropped as a portion of total UC triggers.

CY-PHR's targeted approach pretty directly messes with hackers by interfering with some hack results, leading some players to reduce overall hacking use by cutting down on only semi-helpful hacks or those that can be replaced by other forms of intel gathering, though other players don't mind and just power through the interference (as intended, if you're a really good/dedicated hacker it shouldn't matter too much since they only mess with a minority of hacks anyway). We'll probably see a bit of a rebound in willingness to deal with CY-PHR in Beta 17 since two of CY-PHR's main annoyances were significantly nerfed.

Also worth considering alongside the second graph is that most of those HACK_MACHINES triggers were no longer dispatching actual Unchained, but attracting CY-PHR instead. In fact, comparing the two versions, more than two times as many runs in Beta 16 had actual dispatches compared to 16.1, despite each version having a similar total trigger count. So had CY-PHR been around earlier in the first batch of UC, many of those early runs wouldn't have actually gotten attacked by Unchained :P

Not only were many of the Unchained in Beta 16 triggered by faster hack-flight style players, but those particular builds were also generally not as worried about the results of such a dispatch, since they could probably flee most UC threats anyway. Can't escape CY-PHR results so easily, if at all once triggered, so that changes the calculus.

One of the other big changes in trigger results is the HACK_ROBOTS frequency, which dropped significantly to almost nothing in 16.1 (only one reported instance in all runs, compared to 12 in Beta 16), this because RIF-heavy runs had too many associated triggers by default, and allowing them to accumulate in parallel could mean two or even three Unchained encounters across a run, which was a bit excessive :P. Prior to that I had already made it possible for certain triggers to reset others if they took effect, to avoid results like this, but didn't apply the same behavior to the several possible triggers big RIF users might encounter even during otherwise mostly normal play. After changing that major bothackers were at least no longer getting repeatedly harasses for like half the game... though it's hard to avoid attracting at least one this way. While the portion of winning RIF runs encountering at least one Unchained remained relatively consistent between Beta 16 and 16.1, the ratio of such wins encountering more than one was cut in half (29.6% to 15.0%).

Our final trigger-related graph should come as no surprise, though serves to reinforce the idea that triggers are cumulative in nature, showing here that a greater number of them occurred further into a run:

Anyone triggering them down at the lowest depths is doing it on purpose :P

Also noteworthy is that -1/Access triggers often result in no Unchained, as players may have either set off a trigger only later into exploring that map, or there is not enough time to actually dispatch one before Cogmind has left for the surface or somewhere else.

Taking a closer look at dispatch data, the total number of maps under threat of UC per run averaged 2.0. In winning runs the average was a higher 2.25, somewhat increased by mainly those players proactively hunting them, or using strategies known to attract many (e.g. RIF before it was updated).

As for type, an average 1.28 unique Unchained were dispatched in runs that had any at all. Most interesting to look at among this data is of course the high end--what runs (and by who) are seeing many different Unchained?

Again, most runs see only one at most, with only a small number getting more than that and any runs at the highest end are out to actively hunt them, so here's the top 10 list for unique Unchained attracted, by individual run:

As you can see there are a lot of ties given the smaller numbers, while it's also apparent that a couple players in particular are really interested in hunting these foes down :P. A majority of UC in these runs were destroyed.

Looking at the cumulative number of UC kills across all runs by each player, some of the same names pop up, especially H3-MLN and C0-2MO (the former being famously bent on this goal and check out that insane number!):

Though also interestingly there are quite a few different names as well, including for example szymekc there near the top, where most of these other folks were doing numerous runs and just destroying the one Unchained that got in their way, if any.

Notably a few skilled players still never or rarely got Unchained, playing in "non-extreme" ways, while others were getting them almost every run, though again a good portion of this in particular for Beta 16 (not 16.1) was due to hacking, which 1) was later replaced by CY-PHR and 2) it should be noted that CY-PHR does not count in any of this dispatch data! We'll get to them in a bit, but their dispatching and response is completely unrelated to the normal UC mechanics, and they can even be active simultaneously with other Unchained.

Unchained hunting will likely fall in Beta 17 if only because they're not new anymore, destroying them may be less reliable given some have semi-effective new defenses against traps (originally one of the more reliable ways to safely take them down), and X0-1V1 is the next big thing, which tends to push runs along in its own way anyway and blocks all other Unchained from interfering.

Type-wise Unchained kill totals across all runs, show a distribution that's entirely within expectations:

You have a lot more total kills for those already available in our first batch than when all were present in 16.1, and the most aggressive Unchained lead their respective batches.

While it would be cool to look at which Unchained were responsible for destroying players, from that end the best we can do is look at direct kills (or indirect kills by obvious Unchained-related combat bots), and even then the results are pretty inaccurate since Cogmind is fairly resilient and able to escape for a time, if only to be taken out by something else later, or even at the same time when ganged up on by multiple hostiles plus the Unchained themself. Overall not really worth the huge effort of compiling that data :/

I'll close out this section on Unchained stats by reiterating comments from the latest Beta 17 release announcement, that there have been even further adjustments to Unchained based on player feedback and stat analysis, generally making them slower to dispatch, chase for fewer maps in some cases, and somewhat more lenient in other ways.

I'll also be de-emphasizing their development going forward--despite having designed all the others, as we can see they only really matter to a pretty small minority of players, even taking into consideration the portion of players who are trying to go after them, so are not a great use of dev resources. Maybe we'll still see some of their tech later, since it's cool stuff, but the existing Unchained do a decent enough job on their own (about half having been added now, 10 of the 19 planned), and we'll also be getting some new foes one step down from Unchained anyway ;)

CY-PHR

Cypher! The Unchained that's sort-of-not-really an Unchained, at least in practice since they aren't mutually exclusive with other dispatches, and have a unique and very specific triggering mechanism that gives ongoing progress updates as you approach the trigger point.

Most runs will never engage with Cypher, it's mainly the hacking-heavy runs, essentially very heavy, with hacks of certain types more likely to trigger this over the long run. Some players don't care about triggering Cypher, some want to see/find Cypher anyway (good loot, or to keep the full benefits of extreme hacking throughout a run by proactively taking Cypher out of the picture), and yet others adjusted their play style to minimize the chance of having to deal with Cypher while retaining the most desired portion of their hacking potential (for example not doing as many low-benefit early-game hacks, or certain other hacks that can be replaced by infowar or other means of intel).

For Cypher stats we want to examine two main states: Having attracted initial interest from Cypher, or fully triggering whereby they'll start to take effect (and therefore interfere with a portion of otherwise successful hacks). Of course these stats will only consider Beta 16.1, because Cypher wasn't present in Beta 16.

9.5% of all runs attracted initial interest from Cypher, although this isn't especially telling since many runs don't reach the end anyway, and Cypher starts taking notice at only half of their trigger threshold. More representative would be a look at Cypher interest across only winning runs, which stands at a much higher 62.2%. That's just the first stage of course, so looking further we see that 29.1% of those actually triggered Cypher (or 18.1% of all winning runs).

The above graph excludes runs completely ignored by CY-PHR, but also breaks out the other result of a fully-triggered Cypher: Boredom. A bored CY-PHR (determined randomly) ends up instead just passing your info along to some random Unchained that is dispatched later to dispose of you instead. It's not common, but it mixes things up.

As for depth, as expected Cypher will generally be triggered later in a run. Some players, especially UC hunters of course, might try their best to trigger Cypher as early as possible to improve one's chances of going after them, since it's a somewhat longer process and might affect routing decisions.

Tracking down CY-PHR isn't easy, but 10 runs not only did that but also managed to destroy them (that's 4.7% of triggered runs). Unfortunately I don't have data for whatever likely greater number of runs found Cypher but did not get the kill, instead just at least spooking them to keep Cypher from being a nuisance anymore by proving Cogmind is clearly capable of getting to them :). For Beta 17.1 I'll be adding a record of that to the history log, and also awarding bonus score for it. (One of many good reasons to go through all these stats each version is to get an idea of what interesting factors we might be missing from the scoresheet, so I can add those in!)

Meta

Before moving on to spoiler territory, I wanted to revisit one of the meta stats we looked at last time. We don't need to cover all that stuff again, input methods and platforms etc, but especially relevant to Beta 16 is global achievement rates!

Beta 16 caught up on years of content, increasing the total available achievement count from 256 to 484, so naturally we'll see the achievement rate drop in the near term. Compared to 27% from the end of Beta 15, the new average completion rate across all players at the end of Beta 16 was 20.6%.

The number of players at or above 90% dropped from 9 to 4, led by CaptainWinky who has completed 94% of achievements! (including the new set, of course)

Catching Up (Spoilers Version)

This last section is going to do some more catching up on content that isn't necessarily new, but I'm either newly extracting from the scoresheets or wasn't made available until recently, and importantly you may want to nope out if you are currently trying to avoid spoilers--NPC kills, QCS runs, FRG vs S7, and the UFD mission.

FRG was a new map in Beta 15, so at the time it wasn't as worth looking at how it stacked up against the mutually exclusive Section 7, but now that it's no longer as fresh, we should check how the two compare in terms of desirability. Note that desirability is not necessarily a matter of personal preference, if only because we know that sometimes other factors play into the decision to take one over the other, their entrances being at the same location and players having to take into account their current situation and future plans from that point.

Each plays very differently and serves very different purposes, but it turns out that the two are pretty equal as far as visitation counts! 350 (1.7%) of runs went to/through SEC, whereas a greater 489 (2.4%) of runs visited FRG, though 176 such runs were working with UFD, accounting for the majority of the difference since this latter group has a bigger incentive to enter Protoforge.

As for those players helping the UFD infiltrate Protoforge, the readiness distribution is interesting since it reflects a decreasing portion strictly in line with the highest to lowest readiness--players empowering the UFD with a higher level of readiness are both more capable and at the same time even more incentivized to go there and trigger the final mission.

Digging down into the win/loss ratio among FRG UFD runs, we actually see that across most readiness levels the win ratio consistently hovers around 50%.

The lowest levels unsurprisingly break this trend, and a primary reason Maximal readiness does not have an even higher success rate is that many runs at the high end actually choose to go beyond the Protoforge instead of taking the win, and then maybe die at some later point (because the difficulty goes up as a result).

UFD Lugger AI was significantly improved for Beta 16, so I would expect to see an increase in the average Drillbomb installs by UFD vs players...

As reported before, among all completed missions players averaged a majority of installs in Beta 15, 5.8 vs 4.2 (the total required count being 10), but the roles have indeed reversed as of Beta 16, averaging 6.7 by the UFD and 3.3 by the player. I'll admit that obviously the new Red Leader achievement skews the data here (which would have been that much harder without the much-needed AI improvements xD)--in 24 runs the UFD handled all of the installs. So for a better comparison we should at least remove those before calculating the true average. This still leaves the UFD in the lead with an average 5.3 installs vs 4.3 player installs. There were indeed anecdotal reports of the UFD doing a great job on their own in some layouts, even without much help.

I've been adding some new Companion-related records since that feature first came to the game, so now we can see a little bit more about that favorite item. 12.3% of runs visited Scraptown, and some portion of those found and decided to take the Companion Sword. I actually don't know how many that initial number is, i.e. who had the opportunity to begin with, and who took it in its original form, but we can see who was dedicated enough to transform it, and when.

3.0% of Scraptown visitors both took and eventually evolved The Sword into its next form. Probably no surprises here, but the depths at which each evolution occurred is summarized in this graph:

47% of evolved CS runs won. (It's good, but it's no guarantee.)

One other collection of data, incidentally extracted this time because we wanted to look into Unchained kills, was NPC kills. I'm including this just for fun, and have highlighted a few of the more notable values, for example we see the 8R-AWN and 7R-MNS murderers, and four times as many Z-Imprinter over Imprinter. R17++ being so high is kinda interesting? xD

Stats for The Stat GodX0-1V1

The stats are already piling in for the latest popular release featuring X0-1V1, our new worst enemy and savior, which will no doubt be interesting when Beta 17 is concluded. Going forward, some general community stats for Cogmind will never be the same with X0-1V1 in the mix (it's like Scrap Engine on steroids in that regard?!), so we may have to take that into account in future analyses as well.