Here's more intel, and it's a little frustrating:

izirayd normally plays spiders here, in line behind Godde. Godde can beat my rovers on this map with spiders. izirayd can't - or hasn't yet.

In this match, I picked rovers again because A. It is likely to beat izirayd's spiders and B. He deserves the opportunity to overcome it. There's mild factory RPS involved in that initial pre-plop presumption.

The reality is, izirayd is sick as all hell with badgers and decides to gamble on Cloakbots instead, because he knows 1 key detail: Rovers cannot deal with knight on a map like this. Bonus for him that scorcher is in a shit place vs glaivespam atm.

We make our picks and the match isn't decided, but it's heavily weighted from this point on. It's also why you either psychically pick "the right factory" or try your best to delay an almost inevitable defeat.

Movement at the top of the ladder is barely dictated by player skill anymore. It's heavily influenced by "who made the right factory guesses today". Starting to feel like a merri-go-round waste of time.

Next time we get matched up on this map, I'll be thinking to myself - "do I now have to rotate into jumpbots in order to completely invalidate a cloakbot pick?" thus further spinning the endless wheel of RPS. If I get it right, it'll be izi wasting his time queueing up and picking cloakbots in good faith.

Seriously, RPS needs toning down hard because it's too impactful now.

Knight has too much range. Badger/sling is too strong. Thug is too tanky. Kodachi/pyro are too extreme in outcomes. Spiders don't have enough downsides on favoured maps. Halberds are too cheesy, lances are king in a game of arty rushing. At least amph got smoothed back down to a reasonable state.

I do wonder if it would be a better contest at the high level if both players were forced to pick the same factory to plop - either through a vote or randomly chosen. Balancing the factory RPS sounds next to impossible.

Even if we had all-factory starts or an omni factory, you would see the same shit over and over. It looks like this: Quickscout with flea, safe expansion with convict/constable/welder, safe defense with bandits/bolas/snitch/imp/rogue/venom/archer. The game begins to get to bogged down and you'll see a few slings-badgers-firewalkers into lances guarded by outlaw, thug and knights with owls overhead being mandatory. Locusts might be some perpetual round-the-sides annoyance. You might see a few other units for variety depending on the map (crab on this one), but I honestly doubt the game would stray from this meta too far. There are too few options that shut guarded arty down for it be any different. The only way to subvert it is to tweak the units in good faith.

P.s. Someone please add a "# of factory tokens" game modifier so I can put the above prediction to the test.

Dregs I feel like you are putting to much thought into this.
Do you want to know how i descide which factory i pick each game?
I just choose what i feel like playing at that moment.
If the RPS is bad for me, so be it. Thats just a bigger challange then...

Try out some non-optimal textbook playstyles too ;)

Create a non-ELO ladder; call it an ongoing tournament if you like. Players agree on a map, agree on each other's starting factories, and agree on a wager for ladder points at some odds.

Put a meta-meta-game on top of the meta-game. Give the top tier players a chance to get matches that they feel are fair, or at least where they are compensated for unbalanced matches via risk-reward for potential gains in their ladder positions. The highest-ranked players on this ladder will be those who are best able to accurately assess the degree of advantage that any one factory provides over another factory on a given map.

Forbid mirror matches.

Dregs, if you actually have a specific series of high level factory counters you are concerned about, I would be quite interested to hear it, especially about the specific units that make that true, and the ones that definitely aren't a problem there.

Its very plausible to me that a series of relatively small tweaks could probably help here pretty substancially.

I feel like I have historically seen a quite disproportionate number of complaints about factory RPS on Ravaged replays. Changing ZK may only be able to do so much for this map in particular.

Dregs correct me if I'm wrong, but you seem to be talking about a more specific form of pregame RPS than people usually use. The basic form of pregame RPS has one factory have a general inherent advantage over another. You seem to be talking specifically about particular maps, or possibly even particular pairs of players.

I'm asking clarification due to how you're framing pregame RPS. You said that Godde can beat Rovers with Spiders on this map, while izirayd can't and likes to pick Spiders, so you picked Rover. This seems very player specific, and could be read as saying that the pregame RPS exists with izirayd but not with Godde, as the latter knows how to play against Rover. I also expected to see more Fencer since it is generally pretty good vs Cloaky.

On map based factory RPS, I looked at B1000119 2 on Scaryland v1.02, another Rover vs. Cloaky game, and it seems to be a good example. The players are different and the map is different, but it shows a Knight-heavy lineup do poorly against Fencer, Scorcher and a few Badgers. I know the Knights and Reavers were on hold fire at the end, but the army value and attrition graphs were already turning against the Cloaky side at that point. If B1000119 2 on Scaryland v1.02 and B1003291 2 on Ravaged_v2 are both representative games, then we are going to have trouble making a change that fixes the matchup on both maps. Sure, Knight seems pretty powerful, but units are, in part, as powerful as required by their factory, not at a level of power that would be balanced with compared to all other units.

Dealing with factory RPS on a map-by-map basis is very difficult. Part of this is a lack of data. There were only three Rover vs. Cloaky games on Ravaged with loser ELO>2400 in the last 18 months. They were all Cloaky wins, but such data has little relevance with only three matches across many balance patches. I'm not saying that non-general RPS is irrelevant, I just want people to be more aware of when they are reporting factory RPS that is heavily influenced by the map and disambiguate their feedback when they do.

In the end, every 1v1 is on a particular map with a particular pair of players, so even player-based factory RPS matters. If all the factories were perfectly balanced, hypothetically speaking, yet the skill profile of the players was such that every game felt like factory RPS, then there would still be a problem.

A little bit about the lack of fencer use on my part, and decision not to expand on the cliffs: Map chokeyness/rover horizontal dominance.

In scaryland, desert rumble, jurassic sands or say cobalt dream, you are totally free to spam fencers and you may even get success against knights using them like randy would. That's because there's enough open space to juke-fire-juke-fire with them on these maps. In ravaged, there's limited space to do so, and so you can't get the required most out of them for unit to unit combat in this way. The same limitation doesn't affect badger, so it's the better choice here.

Meanwhile, because of rover's relatively horizontal play, my strategy was to prioritize taking the middle 2 expansions where my units would have more space to move around and wouldn't struggle engaging in combat uphill. Very deliberate, definitely at a cost.

The RPS comes in here for a couple of reasons - I'm forced to use rippers to secure expansions, which are expensive and slow compared to the granularity of glaives. I am forced because scorcher is simply too conditional of a unit and frequently suffers against glaives/bandits in adverse conditions. Also, producing scorchers vs cloakbots is a factory RPS nono. Reavers, Knights, Snitches all invalidate the use of scorcher - so why even go deep on them? I won't even get into how easily spread glaives can toast rippers. RPS disadvantage #1.

Next up, I've spoken about why badger and not fencer. Badger does have the luxury of being good vs slings, so the incentive is there. The RPS comes in when knight presents. Obviously we know that a protractor of fencers is good at dealing with them in open spaces, but outside of these circumstances rovers have no unit composition capable of dealing with knights. Dominatrix should be the go-to option here but knights insane capture time, easy cloakability and "highly likely to stun you whilst you turn" range makes it mostly a bad idea. That's ignoring the fact that ronin/glaive/sniper/sling can make domi's life hell. RPS disadvantage #2 - only map related in that fencer couldn't easily come to the rescue.

With the above two combined, this forces an already disadvantaged facswitch.

So what's the core of the issue - Is it the map design, is it the players, is it knights, is it fencers? Hard to get to the bottom of.

I'd say the core of the issue is lack of ability to actively remedy RPS in a game. Facswitches are expensive. Commander upgrades create vulnerability. Turrets are limited (and mostly liable to arty in today's meta). I really believe that cheaper factories or a cheapish factory-morph command need exploring a little more.

[Spoiler]

I haven't watched that game yet, but jumpbots having wonky matchups would not surprise me even a little bit.