OU GSC Turn limit proposal

[Jorgen]

Listen. It need not necessarily be the case that GSC matches end in stall wars, but the reality is that they happen. However, calling ties in obvious situations isn't self-policing because we've seen several times (particularly in SPL) that people will play them out for hundreds of turns, thinking they have a slight advantage. And they very well might have a tiny theoretical advantage, but not one with which they can win in a reasonable amount of turns. In the end, we just end up wasting everyone's time by extending matches way beyond the point where reasonable people would agree to a draw.

Because this isn't something where we can rely on self-policing, it's necessary to craft an objective turn limit rule for GSC (similar in spirit to the 50-move rule used for chess endgames), after which a tie is automatically enforced/decreed by TDs and hosts. What should follow a tie is up to TDs and hosts to decide on a case-by-case basis, but that's not the issue here. What I'm talking about is just deciding when a draw should be called in the first place.

I'd say the following 2 criteria should be used to determine a tie in GSC:
1) On the turn a Pokemon begins using Struggle, if a KO on either side is not achieved within 50 turns, a draw is called.
2) If a total of 5 straight turns has occurred where neither side has used PP, had a Pokemon spend a turn fast asleep or frozen, nor had a Pokemon take any damage (e.g., from Toxic or Spikes), a draw is called.

With these criteria, I've tried to encapsulate the perpetual-struggle and perpetual-switch tie conditions that have conservative enough turn limits and have specific enough criteria to avoid calling false draws following legitimate tactics. Is there disagreement on either the philosophy or specifics of the proposed turn limit, or are we cool with making this happen ASAP to avoid another Bedshibaer vs. Tiba?

EDIT: A hard turn limit (say, 300) is probably better for simplicity's sake. I don't know why I didn't just start off with proposing that. I guess maybe I wanted to be extra conservative with the whole "call a game" thing, but 300 is plenty of turns, and the second criterion of the complex solution above isn't likely to do what I intend.

 

[Picollo]

^ agree with that.

 

[Lavos]

I agree with both criteria. This is a reasonable and practical approach to GSC which is desperately needed considering the recent game.

 

[Expulso]

I feel like 5 straight turns of the latter condition may be a bit low, but this seems like a good policy overall. :]

 

[Mr.E]

I won't disagree too hard if we put in fairly loose, liberal rules for determining a tie but no, I don't agree. Unless the situation is a true tie, e.g. ADV Wobb mirror (yes I know it's banned but this is exactly why it's banned yo), I want all games to be played out and a true, actual winner determined if at all possible. We can avoid another Bed vs. Tiba by not allowing Tiba to abuse simulator bugs to force a tie out of a losing position, and then having the tournament director make a bad decision to justify his pussy-shit play. (Obviously if the position is 100% lost, then a winner can just be declared on the spot.)

 

[Jorgen]

Yeah, I thought 5 might be low too, but with the wording to exclude sleep & freeze turns from triggering the rule, that just prevents 5x double-switches, which basically never happen. I want the limit to be low to prevent players from trolling the limit. I have to second-guess myself on this though, because players could just space all PP usage by 4 turns in such a situation, which still extends the game a ton.

EDIT: but wait, that would require collusion to extend the game, rather than the current situation where collusion is required to end the game. Thus, such a limit in this case would result in the Prisoner's Dilemma optimum of triggering the draw. Because of the requirement of collusion to keep the game going, the limit miiight be relaxed to something like 10 turns for criterion 2. In my experience though, drawish endgames often result in suboptimal PP usage that would trigger this condition way too late to even matter.

Perhaps a hard turn limit is the better way to go than these complicated conditions. I'd say a 300 hard-turn limit is a pretty fair bound. Speaking from experience, the longest GSC games that end tend to go for roughly 200 turns, whereas these drawish games are known to be such roughly 200 turns in. A limit of 300 gives a 100-turn buffer to even these longest of GSC games, so we shouldn't end up with cutoffs for legitimately-played PP stall wins (e.g., Curse Skarm vs. CurseLax) that would end at turn 220 or something.

When a win is not possible or a situation leads to high likelihood of a loss, the optimal play is then to extend the game, and that's exactly what Tiba did. The travesty was more refusing to accept the draw in a game both players knew would go on forever but didn't necessarily see 400+ turns ahead to know for sure who had the tiniest sliver of an advantage, rather than making the game-prolonging plays in themselves.

The issue I have with TDs making judgement calls is probably the same issue they'd have with it; there's too much shit to be thrown at them. Not to mention that, without a clear objective rule to decide a tie, not only is the shit-flinging going to be made worse for it, but situations like tonight where a TD or host isn't available means people waste their time on drawn matches because they need the decree to know for sure that it's a tie.

 

[Mr.E]

The "optimal" play in this instance was only optimal insofar as a simulator-only bug (not a legitimate game mechanic, as far as anyone knows) and a questionable tournament director decision was made regarding said playing into the bug. As far as only the game is concerned, either stalling indefinitely or killing ASAP would've led to the exact same endgame except the latter reaches it hundreds of turns sooner: Tiba is at a clear disadvantage but not a strictly losing position. Since he's not in a strictly losing position he can't (shouldn't) be simply be given the loss, however given that the endgame is identical in both cases he really should have been forced to play out the endgame in a timely fashion to avoid the simulator bug. Consider it an extension of our existing sportsmanship rule.

(If a win is strictly not possible, it is trivial to award that player the loss. The optimal play isn't to extend the game and get it annulled by throwing a virtual temper-tantrum to the rules lawyers, it's to be a reasonable human being and accept you fucking lost, bro.)

I don't begrudge Bed at all for refusing to accept a tie in a situation he is advantaged. The game would not have gone on "forever," just long enough for Tiba to basically abuse the rulebook and hope he got help from the refs (TDs) on a technical limitation outside the game. The travesty is both that Tiba did what he did and that his underhanded, unsportsmanlike actions were actually supported and upheld by the head TD. There's nothing tragic about a game being long. It's tragic Bed got screwed and it directly affects the Raiders' record because it was the difference between a loss and a tie for the week. (It would still be tragic if that didn't matter, since battle difference is a tiebreaker, but it would certainly be less meaningful.)

 

[Jorgen]

Any reasonable cost function will give weight to avoiding disadvantaged odds. If descent is inevitable (assuming optimal play by your opponent), your optimal play is therefore to avoid it as long as possible. A cold, unfeeling GSC bot incapable of being explicitly conniving would have done the same thing, regardless of any artificial turn cap.

100% ties are also super rare. After all, last-Poke Blissey vs. Blissey, the "classic" GSC pure draw, can be ended with ~10 consecutive Struggle crits (assuming max attack; min attack Blissey attempting to minimize Confusion damage cannot possibly win). Same applies to ADV Wobb: critical hit Struggles barely do more than Leftovers, so if you get, like, 100 in a row (even after accounting for the other Wobb's recoil), you win. The only 100% ties are when Pokemon like Wobb and Bliss explicitly make efforts to lower their Attack and boost their Defense + HP, all at once.

We're just talking past each other at this point. I'm going to maintain that if it's not feasible to play the requisite number of turns to force a win, it really shouldn't be considered decided. It's too much to expect that players should play through such a large number of repetitive turns to achieve a definite winner (which they might not even be able to do). The notion that "theoretical" wins be awarded in lieu of finishing matches is also a little silly to me; in this particular case, we could likely put forward a rigorous proof of who would have won in the absence of time constraints, but it took a while to finally figure that out. I really distrust our ability to call a winner before the game does.

 

[Mr.E]

It still feels like just trying to rules lawyer out of a loss they've otherwise earned. Don't be that guy.

A perfect-play bot may have dragged the game out but it would also take zero time to make its move, so playing those turns out wouldn't actually take very long anyway. Besides which we're choosing between two equally optimal options here, as the end result of either branch of play is 100% identical and 100% certain: Struggle can't miss and non-Struggle attacks (particularly all recovery moves in this case) don't have infinite PP. There is neither an advantage to dragging the game out, nor a disadvantage to taking the quick KO and moving to the endgame 53479534785382 turns sooner. Since we are common-sense human beings here rather than bots, it's strictly logical to take the latter action whether or not it's "optimal" from some nebulous game theory standpoint that doesn't actually apply here (slowrolling the "inevitable descent" literally does nothing but waste everyone's time).

Well, there's no difference in game context anyway. Obviously there are real-world consequences beyond the game here. In that sense, you might say Tiba was indeed correct to drag the game out, in that the simulator bug poses a roadblock to the "inevitability" if Tiba stubbornly refuses to cooperate. Nonetheless it's extremely unsportsmanlike at best, even if the simulator bug wasn't present and he was just doing it to be an asshole (maybe hope Bed disconnects), and exactly where the TD should step in to make everything right... He didn't. Tiba effectively employed an extralegal tactic, then was rewarded by being gift-wrapped a draw from a losing position. It's not really Tiba's fault, he did exactly what he "should" have done even though it makes him the biggest douchenozzle this side of Hazerider. It's the fault of the powers-that-be for letting him get away with it. Kinda like how it's not my former employer's fault I was denied unemployment benefits, it's the State of Ohio's fault for believing their incredibly transparent lies about why they fired me...

Blissey (and Wobb) duels are a true 50/50 at least, so I don't think too many players are going to refuse an offer for a draw nor is anybody getting up in arms if a TD comes in and just declares it so to avoid playing out an average of over a trillion turns (I did the math!) to determine a coin flip. Personally, I would accept a literal coin flip if we really want to avoid ties... but I assume most people wouldn't and there's no good reason to quash ties. Ties should not be decreed if either player has an actual odds advantage unless the players themselves agree and mutually quit battling.

 

[Royal Flush]

Well I can give you that blissey x blissey tie is pretty ok, but in yesterday's match, Bedschibaer HAD the better odds. I'd say even more than I thought since Beds posted that his Lax had EQ so Ttar would have even less chances to drop SDef before dying. It doesn't matter if it would take 1000 or 2000 turns, but in a situation A where Tiba chose to not kill Zapdos, he had 0% chance to win the game since Struggle dealt more damage than lefties recovery. In a situation B where he chose to kill Zapdos and face Lax, he would have 20/25% odds of winning at the very best. Of course people usually don't have the will to play for this long, but we can 100% assure that Tiba had no odds of winning by merely stalling so I don't see how an official draw system would be fair.


So in the end of the story, if we actually apply for instance the 300 turn limit to call an official draw, one player that realized that he could never beat lastmon curselax could exploit a draw here (idk if there's a similar situation, vap/p2 I guess?). It could be a situation like the player with the Curselax realized that this was HIS win condition and then he started to sac all his mons. Then what happens is that the player who supposedly couldn't beat curselax anymore could potentially stall the game over to the 300th turn and get a draw instead of a high odds loss (unless of course the other pokemon couldn't deal enough Struggle damage against lefties, that would be an ok draw). Tiba did that, it worked guys.

 

[Jorgen]

There's always the potential for exploits with a turn limit. The point is to make it high enough where it's no longer a big deal. Personally, I actually don't have a problem with "exploiting" a draw at 300 turns. I figure if you have a winning position, you have to be able to force it within a reasonable time constraint, and 300 turns is well above the limit where a game is considered "too long". The alternatives of granting theoretical wins and mandating specific moves seem way more repugnant to me than imposing pragmatic draws, and the current situation where a game could go for 600+ turns before a TD or host finally comes in and calls it a draw anyway clearly needs to change.

 

[Oglemi]

idk to me a game should be played out to its logical end regardless of # of turns, unless the situation is an absolute draw

I can understand the want to cut extremely long games short, but if a true winner can be decided through the full playing of a match, it should be played to that end and a winner declared.

essentially, i really dislike ties and management-driven decisions when they technically aren't needed, so i'm against a total turn cap

 

[Oglemi]

1) On the turn a Pokemon begins using Struggle, if a KO on either side is not achieved within 50 turns, a draw is called.

fwiw though, i agree with the first^ of the two definitions declaring an absolute draw, the second one i'd rather have be more than 5 turns, how many is enough idk

 

[Crystal_]

I think the only scenario that can be considered a draw is when the best option of both player consists on switching endlessly to avoid wasting PP. IMO criterion 1 is not very reasonable for this reason. If a Pokemon begins to Struggle it's because the game is approaching to the end in some way. I think the only exception to this would be the scenario where both players are in a last-poke struggle war where leftovers is effectively negating all the namage, meaning it's impossible for either player to win. All the other scenarios in which struggle is involved that I would call a draw should already be covered by criterion 2.

 

[Jorgen]

For what it's worth, I'd suggested 10 turns for the second criterion after I realized the necessity of collusion to "troll" the limit (i.e., switch back and forth 9 times, then use a PP, thereby maximally extending the game). Still not sure this is a great idea though, as realistically, people playing through drawish games will still use PP, perhaps on the off-chance that it catches the opponent making a mistake/misclicking/whatever.

I understand not wanting mgmt to step in but it's been shown to be inevitable for games like this. If it's inevitable, it's better there be a clear, objective reason for it than needing to make ad hoc decisions all the time.

Also a "tie" need not necessarily end in "tie". Depending on the situation, if a tie isn't an acceptable outcome, the host can force a rematch (overtime) or, if if that isn't possible, flip a coin (shootout). Determining exact odds can be tedious and relies on the assumption of optimal play, so trying to calculate those for a more accurate "coin flip" is both not feasible and not particularly in the spirit of the human nature of the game, where suboptimal play is part of it all.

As for wanting games to only be ties that are theoretical ties: those are actually incredibly rare. If 2 Blisseys are in a struggle war, a long string of crits (which is inevitable, but will in all likelihood take forever to realize) will decide the match. The only exception is if they both drop their Attack DV, in which case yeah, it's a theoretical tie.

But what if one dropped its Attack DV and the other didn't? Is the appropriate course of action for mgmt to call it a win for the side who didn't drop the DV, as they theoretically win if given infinite time to play? I don't think that's a great idea, as allowing mgmt declaring W/L decisions is worse on the face of it than just giving it the power to declare noncommittal ties/rematches/whatever. Triple-scald-crit-burn is the precedented exception, but that's a different situation where, instead of explicitly streamlining the "true" ending, the operating principle is preventing internet/server problems from unfairly influencing the game (although for what it's worth I don't feel strongly that that should supercede mgmt non-committal-ness).

How about a situation where one side can safely switch forever to conserve PP, and the other must react to the opponent? Say, (Curse) Skarmory and Blissey vs. (Curse) Snorlax and Zapdos. In this case, there's clearly a winner in the long run (Lax & Zapdos), but the game still takes forever to play out, and the "optimal plays" to be asssumed are way harder to determine. Do we consider such epistemically likely wins (i.e., we know the vastly favored side, but it's on the certainty that the optimal play for Skarm and Bliss is to use PP more often than Snorlax and Zapdos) the same as ontologically likely wins (i.e., we know the favored side because optimal play is deterministic and we can therefore straightforwardly calculate the odds)? Would it be fair to flatten the odds for one but not the other when terminating a too-long game?

Also before somebody comes in with this point, this solution is for tournaments & such where time constraints are way more relaxed than a real-time tournament (i.e., Tour). In that situation, you do sort of have to bite the bullet and call a winner.

 

[Expulso]

It could also be a good idea to make the first condition "once one Pokemon runs out of PP...", or they could just switch in and out of said Pokemon without incurring the penalty for using Struggle. Then again, if it were like this, one player could just keep all theirPokemon with 1 PP on a move and switch around as much as they want. :/

 

[ohgeedubs]

2) If a total of 5 straight turns has occurred where neither side has used PP, had a Pokemon spend a turn fast asleep or frozen, nor had a Pokemon take any damage (e.g., from Toxic or Spikes), a draw is called.

Instead of this, why not borrow threefold repetition from chess? I'm not familiar with GSC, but there might be some marginal cases similar to chess where in order to get the desired position you need, you have to play very careful, positional (but relatively passive) moves in order to get into a certain position (which may take more than 5 turns?). Both players making constant switches might be drawish, but if one player can eventually "force" the opponent into a threatening position, things can happen. But if you have threefold repetition, where you reach the exact same one v. one scenario (same sleep turns, same pp, same pokes on the field, same hp), well now it's starting to look like a cycle where both players' best plays are to stay within the cycle.

 

[MoxieInfinite]

just keep spikes up and this wont even happen

fuck zapdos and skarmory

 

[tehy]

To be honest, I feel I have something of a solution.

See, low, restrictive limits run the risk of penalizing players for legitimate play or altering their play to avoid penalties (for example, a theoretical 3-turn limit would seriously limit play; even the proposed 5-turn limit could become awkward in a few situations), but high limits that shoot to avoid this can end up failing to restrict a player who pays attention to them (10 turn limit? 9 switches 1 pp; 10 PP takes 100 turns to get used up.) Any middle ground (7 turns?) ends up with the same problems (10 PP=70 turns, and 7 switches might happen once in a blue moon), and even though that's somewhat helpful, it's not a true solution as such. However, people hate management decisions, perhaps because they feel heavy-handed and out of place. What to do?

1) On the turn a Pokemon begins using Struggle, if a KO on either side is not achieved within 50 turns, a TD decision is called.

As a quick aside: "Play on" is a decision and it'll probably be made most of the time.

Right away, any jank-ass gimmicking of the system is dead as dust. In addition, because a TD decision is required, it (hopefully) comes off as more warranted when they do call a draw; after all, they had to come and make SOME decision.

The altered rule is a great example. Let's say some player is losing, slowly but surely. So he wastes PP on one mon on purpose and BANG! Struggles.

At this point, he just needs to stall 50 turns and he draws. Throughout, he can and should play a completely suboptimal style, just to stall till the end. Additionally, this might become a common mode of play, with people purposely wasting PP on a mon early as insurance. Players might even forego filler moves or go sans PP Ups to give themselves a backdoor- and while I don't believe that would happen (especially the move part), wouldn't it be kind of gross if it did?

And yet, this is basically a good rule. It tries to take care of bad situations in a broad enough way to probably catch most of them without need for an adjudicator. What it needs is just a helping hand.

A TD could look at someone purposely wasting PP, laugh, and "play on" their asses. Hell, we wouldn't even see that situation come up!

Meanwhile, 50 turns after a struggle is usually a good place to stop and take a good hard look at the situation.

One last thing-in any case, any of the switching rules should say "Switch or Struggle", since we could just see switch+struggle spam for Leftovers recovery, and thus a never ending game.

Edit: there are two weaknesses to this suggestion, which I forgot to address. Firstly, this would require the TD to be present at all GSC games. I must confess to ignorance at this juncture; it may be that this is already the case. However, if it is not, that could be a problem.

Also, this would cause more management decisions, and while I feel I covered that somewhat, i'd like to address it more thoroughly here. Firstly, they would mostly occur instead of computer-mandated draws, which is by far the lesser of two evils. In addition, these decisions would occur only at junctures considered reasonable by the community and would be required, and finally, they still would not happen that often. Unfortunately, it's still a weakness of the system, but one I hope we can overcome. /edit

In conclusion, I don't think a game should ever be decided by a limit not imposed by the actual cartridge.

 

[Disaster Area]

Can anyone calculate the turn limit in terms of battery life? Just a curiosity.. but it means that that blissey vs blissey scenario might not be something that could happen on cart due to battery life (are there more likely scenarios where this could be of question too?)

 

[froggy25]

With a power supply this is not an issue.

 

[Jellicent]

Can anyone calculate the turn limit in terms of battery life? Just a curiosity.. but it means that that blissey vs blissey scenario might not be something that could happen on cart due to battery life (are there more likely scenarios where this could be of question too?)

You just need an AC adapter, mate~

 

[Texas Cloverleaf]

Can anyone calculate the turn limit in terms of battery life? Just a curiosity.. but it means that that blissey vs blissey scenario might not be something that could happen on cart due to battery life (are there more likely scenarios where this could be of question too?)

Just as a point of interest using an assumed 15 hour battery life (the peak for the one study I googled) and assuming 4 seconds to view animations and select your turn as an optimum you can get through 13,500 turns. Adjustable based on your turn duration estimate naturally.

 

[Disaster Area]

Does the AC adapter make that irrelevant? Past that you have to consider in a human sense what's the limit that you could actually play before you would need to sleep or go to the bathroom for long enough that the timer runs out for one player?

 

[McMeghan]

This is getting ridiculous (not like it wasn't already).