Endless Battle Clause in GSC (and RBY/RSE)

[Earthworm]

GSC doesn't have endless battles unless both players don't want to risk losing and agree to a tie or neither player's Pokemon can outdamage Leftovers with Struggle. This is because there is no Regenerator and Struggle only deals a percentage of the damage dealt to the user. It can be a viable strategy to sweep with Struggle and the logic behind adding the rule in the first place doesn't apply to GSC (or RBY/RSE). It is called "Endless Battle" clause for a reason.

I would like to request the clause's formal removal in these generations or at least GSC where the strategy is particularly viable. It was removed on someone's request a while ago but awareness of this was apparently confined to some GSC players and not widely publicised.

edit: Apparently it may not ever have been enforced in RBY/GSC/RSE. If that is the case then this thread at least serves to make that knowledge public.

 

[pre]

You can have endless battle in RBY via Ghost vs. Ghost Struggle, and in GSC/RSE with Leftovers recovering more than Struggle damage (as you mentioned), so the clause still is completely valid and reasonably named for these gens. However, nothing inflicts staleness in these generations (per https://dex.pokemonshowdown.com/articles/battlerules#endlessbattleclause), and thus "Endless Battle Clause" is really just "battles can only last 1000 turns before ending in a tie". 1000 was an arbitrary choice IIRC from a previous PR thread, though having some upper bound is useful for bounding the amount of time a player needs to sink into a game during ladder play.

I don't think the argument that "endless battles can't happen in these gens so we should remove the clause" is valid, the argument instead seems to be "endless battles are incredibly rare compared to later gens where Regenerator and Leppa etc exist". In later gens, using Leppa or Regenerator Pokemon is also a completely valid strategy and we have the EBC causing artificial win conditions there as well, so I don't really see how Struggle being a valid strategy in GSC or other past gens is different.

If you read one of the 5-10 other threads on EBC you'll find Zarel was opposed to relying on "offer tie" as a crutch, and even though the command has now been programmed into the sim, it still doesn't change the argument.

 

[Earthworm]

I don't think the argument that "endless battles can't happen in these gens so we should remove the clause" is valid, the argument instead seems to be "endless battles are incredibly rare compared to later gens where Regenerator and Leppa etc exist". In later gens, using Leppa or Regenerator Pokemon is also a completely valid strategy and we have the EBC causing artificial win conditions there as well, so I don't really see how Struggle being a valid strategy in GSC or other past gens is different.

The difference between these is that Regenerator and Leppa Berry cause actual endless battles to occur. Correct me if I'm wrong. Rarity does not come into it at all.

To clarify my argument, I am saying that there are legitimate strategies involving Struggle that can reasonably exceed 1000 turns while making semi-consistent progress. Umbreon's Struggle does 7 or 8% to targets with Rest and they only gain 6.25% per turn, however by attacking it they can force it to retreat until it has recovered its own health with Leftovers before attacking again. Games that come down to Struggle in GSC are extremely likely to end within a somewhat reasonable timeframe unless the Pokemon involved deal less than 6.25% to Zapdos or Snorlax.

That said, if ladder battles are a concern, then increasing the cap to 2k would most likely be enough to exhaust realistic win conditions in GSC.

Someone did raise the fact that the browser tab may not be able to handle so much text or whatever. That is a valid point and is beyond my expertise, so I will have to rely on dev advice for whether that issue can be handled to enable the proper conclusions to these longer games to be reached.

 

[Quite Quiet]

Being extremely optimistic with 10 seconds per turn (in reality this is going to be higher) it would take two players 5.5 hours to play 2000 turns to get into a tie. The 1000 turn limit already gets close to a 3 hour game. There is a point where "possible" has to contend with reasonable expectations of players. Is it reasonable to expect players to set aside 3-6 hours of their time to schedule a single game of Pokemon?

As an aside, this post onwards also contribute some to this discussion, for those interested.

 

[pre]

The difference between these is that Regenerator and Leppa Berry cause actual endless battles to occur. Correct me if I'm wrong. Rarity does not come into it at all.

If both sides only have Ghosts Struggling in RBY the battle will not terminate, period. This is exactly as endless as you can get with Leppa or Regenerator. Similarly with Struggle doing less damage on both sides than Leftovers recovers in GSC/RSE. Not endless in the sense that "it takes forever to terminate", but 100% endless as in "it will not terminate, period".

To clarify my argument, I am saying that there are legitimate strategies involving Struggle that can reasonably exceed 1000 turns while making semi-consistent progress.

I agree, but the scenarios I mention above to not make any progress at all. There are also legitimate strategies involving Leppa, Regenerator and the other moves outlined in the existing EBC - the clause seeks to preserve legitimate tactics as much as possible while still conceding that in edge cases/through malicious play they can lead to endless outcomes.

Cards on the table - I hate the Endless Battle Clause because it is inelegant and artificial (and a pain to implement). I only spent so much time rewriting it to eliminate as many false positives as possible (like it triggering if your Pokemon couldn't move for several turns in a row due to paralysis or freeze). However, I believe it to be a necessary evil/best of the bad options available given the mechanics of the game. I don't really care if the limit is changed to be greater than 1000 turns, but I think for consistency the limit should be the same for all generations, and think Quite Quiet's math above checks out and points to the practical reasons as to why 1000 turns was chosen.

 

[Earthworm]

Being extremely optimistic with 10 seconds per turn (in reality this is going to be higher) it would take two players 5.5 hours to play 2000 turns to get into a tie. The 1000 turn limit already gets close to a 3 hour game. There is a point where "possible" has to contend with reasonable expectations of players. Is it reasonable to expect players to set aside 3-6 hours of their time to schedule a single game of Pokemon?

As an aside, this post onwards also contribute some to this discussion, for those interested.

For what it's worth, FOMG and I finished our 1000 turn game in 2 hours. So I wouldn't necessarily assume these games are going to last to the absurd lengths you mention. Typically when games last this long, decisions become very simple and only a few turns require more than minimal thought. 10 seconds for these kinds of situations is well above what will typically be happening.

I agree that anything over 3 hours is getting too long. No-one wants to sit through a game that long. However, I also place a strong value on allowing games to run through to their eventual conclusion when the situation makes it possible. Players should be able to take full advantage of the strategies available, including "stall" strategies if they so desire.


Also I get what you mean now pre, you are saying that there are exceptions that could cause endless games and so the clause should remain even if these situations are extremely rare.

I guess another alternative that would satisfy the requirement would be to require regular progress (net loss of HP on at least one turn in the last x or use of PP) after 1000 turns have passed. This would force plays but at least leave more actual strategies on the table while dealing with your edge cases adequately. Please let me know if I am mistaken.

 

[cityscapes]

i also agree that letting games reach their conclusion is important and i'm a bit disappointed that the spl game wasn't allowed to continue. that being said, it seems unreasonable to me that players should have to dedicate several hours of their time without any major breaks.

i don't see any benefit in increasing the turn cap on ladder. given how much tournament play is valued in comparison, the appeal of ladder play to me is to be able to play out many games against different players and teams. that being said, i can see how it could be annoying if the players were interested in how a scenario played out, but couldn't do that because of the turn limit ending the game immediately.

here are a couple ideas.

1) allow the option to ignore the turn cap on ladder, but only if both players agree to it. after this agreement play proceeds as normal, with both players needing to agree to a draw (so you can't just draw if you're about to lose a 1300 turn game)
2) alternatively, allow anyone to access and use technology that recreates a game from a specified state in challenges. i believe this was something zarel/the tds talked about a bit, but to my knowledge it hasn't seen any use yet. it would allow for the continuation of >1000 turn battles originating on ladder, and allow them to go beyond this limit in friendly games as well.
3) for tour games: pausing the game at 1000 turns and allowing the players to reschedule to resume it. hard 2000 turn limit. this is a really strange solution and comes with the issue of players potentially having days to figure out the most optimal way to play the situation (and likely allowing teammates to help out, unfortunately), but it solves the issue of players being forced to potentially set aside up to 4 hours.

i think that the game state recreation stuff is really underused here. if you can start it from turn 1 with accurate hp/pp/revealed moves, then that solves the problem of the browser crashing as well. you could just recreate the game every 500 turns or so.

these ideas are pretty flawed and i don't expect them to get implemented, but hopefully they're a good starting point.

 

[pre]

My personal ideals for the Endless Battle Clause (which may not be the community's ideals - please push back here if these don't seem reasonable!):

• the clause should interfere with regular play as little as possible (preferably only triggering in cases of a malicious player intentionally causing an infinite battle). Some corollaries:
  • the existence of the clause should alter a player's strategy as little as possible (ie. avoid introducing artificial win conditions)
  • the clause should seek to prohibit as few strategies as possible (ie. don't just ban Ghosts in RBY or Regenerator or Leppa etc if we don't have to)
• the clause's specification should be possible to describe in plain English, such that the average player can reasonably be expected to be able to reason about when it could trigger so that they avoid it
• the clause's specification should be as generic and concise as possible and consistent between metagames
  • this perhaps follows from above, as it helps makes it more understandable/simpler to implement correctly and reason about

I guess another alternative that would satisfy the requirement would be to require regular progress (net loss of HP on at least one turn in the last x or use of PP) after 1000 turns have passed. This would force plays but at least leave more actual strategies on the table while dealing with your edge cases adequately. Please let me know if I am mistaken.

"Regular progress" ends up being quite difficult to get right - this is what the old EBC implementation attempted to do, only it started hitting false positives when a Pokemon couldn't move for N turns in a row thanks to a status condition (among other reasons). Attempting to quantify "regular progress" is quite complex/non-trivial to implement in a non-buggy way given the interactions between all the effects that can occur in a game, and will almost certainly end up breaking given how the EBC would have to be implemented (high surface area and complicated). Only attempting to check for "regular progress" after 1000 turns seems much more desirable than checking it from Turn 1 as we were attempting before, only I'm still not really in love with this idea because it still involves the same inscrutable and complex bandaid as before. Losing a game because of a bug in the simulator's implementation of an artificial termination condition rubs me the wrong way. Usually, "difficult to implement" shouldn't really matter (there are a whole bunch of difficult to implement things Game Freak did that we recreate), only perhaps there is a reason Game Freak just slapped a total game timer on cartridge play instead of attempting to implement such a clause (part of me wonders if determining whether a Pokemon battle will terminate or not is actually the Halting problem and thus literally impossible, but we'll leave that as an exercise for the reader).

tl;dr: IMO, switching how the EBC works in edge case scenarios just complicates things as all of a sudden there are new rules 2-3 hours into the match (and rules a player is likely to be significantly less familiar with given they are applied so rarely by design). 1000 turns (the only part of the current EBC that really matters for this thread) is dead simple to understand and affects a small minority of matches and makes the minimum number of changes compared to cartridge play. If you can't win in 1000 turns and are forced to rematch after a tie, perhaps chose a team that will win quicker (while this is sadly slightly hostile to stall, its hardly the most impactful anti-stall decision the community has made).

 

[Earthworm]

To address the above, I can respect that position and would instead like to suggest again simply increasing the turn limit to 2k in GSC, which I believe meets your ideals pre. I would not be opposed to suggestion 3 from Gurpreet Patel (Sent you a Friend Request)'s post above, i.e. allowing tour players to enforce a break after 1k turns, circumstances permitting.

With a longer turn cap we can still encourage players to go for moves that make progress in games where they should only be taking minimal risks to further their position and allow these longer term win conditions to play out. I think that is both a desirable and necessary thing for strategy and game purity reasons, so I hope the relevant parties will accede to this request.

 

[pre]

consistent between metagames

I'm uncomfortable with us changing the turn limit for one specific metagame, and also don't think 2000 is a good choice for non-GSC metagames.

 

[Earthworm]

I'm uncomfortable with us changing the turn limit for one specific metagame, and also don't think 2000 is a good choice for non-GSC metagames.

It's not ideal I agree, but I think we can't reasonably expect to have the same rules across the board for generations that are so different from one another. I hope we can find compromise on this point as allowing games to play through to their natural conclusions unobstructed by artificially imposed rules seems like it should be prioritised.

 

[obi]

As a data point, here is the end of a battle log from Pokemon NetBattle:

Begin Turn #1022
Mr. Mime used Recycle!
But it failed!
---------------------------------
Porygon2 used Recycle!
But it failed!
---------------------------------
End of turn #1022
Bob Jones's Mr. Mime: 100% HP
e to the pi i's Porygon2: 374 HP
tayottt: for 1021 turns
surfingsuicune has left.
tayottt: jesus christ
L. Sammy(retired): SAVE THE LOG NOW OBI
e to the pi i: I am.

Begin Turn #1023
Mr. Mime used Recycle!
But it failed!
---------------------------------
Porygon2 used Recycle!
But it failed!
---------------------------------
End of turn #1023
Bob Jones's Mr. Mime: 284 HP
e to the pi i's Porygon2: 100% HP
Bob Jones: !
-_- ItsPat -_-: ...
AnthraxTheGreat: rofl
L. Sammy(retired): oh boy
tayottt: bob jones= jazz player
L. Sammy(retired): the biggest turn in pokemon history
L. Sammy(retired): coming up
tayottt: lol
Bob Jones withdrew Mr. Mime!
Bob Jones sent out Smeargle (Lv.100 Smeargle)!
Porygon2 used Assist!
Porygon2 used Recover!
Porygon2's attack missed!
Bob Jones sent out Smeargle (Lv.100 Smeargle)!
Porygon2 used Recycle!
But it failed!

Begin Turn #1024

NetBattle v0.9.6
Log saved 1/26/2006 at 7:27:29 PM

And then the server crashes. So we know this wasn't a frequently encountered issue back then -- the server wasn't regularly crashing from people doing this, and I only knew of one other person who knew that something happened at 1024 turns.

We could look at the battle logs and see how often games of each generation have hit the 1000 turn limit in the past few years. I suspect very rarely.