(note: I'm posting this a bit earlier than I planned in case it's early enough to implement for Slam in a month, if desired)
As most people are probably aware, Smogon's trophy tours are only seeded in playoffs, using performance in that tournament's qualifying phase for data; this applies to every tournament except OST (and for some reason OLT, but I think I think that's a matter for a separate thread). This works fine for Smogon Tour, in which you have up to 18 lives and so you're well-insulated from bad bracket luck, and it's mostly fine for Masters and OSDT, whose Swiss formats already limit a lot of the bad luck of a randomized bracket (you have three lives and you can't run into a string of undefeated opponents). However, for the single-elim brackets used in OST and for the qualifying tournaments in Slam and Classic, your bracket performance is heavily dependent on the luck of the bracket randomizer.
My proposal is to seed the top 64 for OST, and the top 32 for each Slam Open and each Classic Cup. Reducing that number to top 32 for OST and top 16 for the Opens/Cups works too, but the important point is that I do think that you get diminishing returns from going any deeper. I will list a few different options for how to seed at the end of this post, but I propose to do it according to previous year circuit points in that tier. To my knowledge, the last time that this question was in 2020 in this thread; I'll tag McMeghan in case he has anything else to add from that conversation. As far as I can tell, that thread never had a formal resolution, it just trailed off. I'll add my main arguments here:
I'll say one more thing. The question of whether seeding should be implemented on the basis that it improves these tournaments and the question of how seeding should be determined are different and should be evaluated in turn. If the answer to the former is yes, seeding would improve these tournaments and should be implemented, then I think the goal should be to implement the best version of that. The idea that seeding just shouldn't be implemented if there's no universally agreeable method even though the best option would be an upgrade on the status quo doesn't make sense. It's not nobler to leave tournaments worse than they could or should be through inaction rather than action. I'd love to hear thoughts here; it's been a while since this was discussed formally and I'm sure people will have strong thoughts on the matter.
As most people are probably aware, Smogon's trophy tours are only seeded in playoffs, using performance in that tournament's qualifying phase for data; this applies to every tournament except OST (and for some reason OLT, but I think I think that's a matter for a separate thread). This works fine for Smogon Tour, in which you have up to 18 lives and so you're well-insulated from bad bracket luck, and it's mostly fine for Masters and OSDT, whose Swiss formats already limit a lot of the bad luck of a randomized bracket (you have three lives and you can't run into a string of undefeated opponents). However, for the single-elim brackets used in OST and for the qualifying tournaments in Slam and Classic, your bracket performance is heavily dependent on the luck of the bracket randomizer.
My proposal is to seed the top 64 for OST, and the top 32 for each Slam Open and each Classic Cup. Reducing that number to top 32 for OST and top 16 for the Opens/Cups works too, but the important point is that I do think that you get diminishing returns from going any deeper. I will list a few different options for how to seed at the end of this post, but I propose to do it according to previous year circuit points in that tier. To my knowledge, the last time that this question was in 2020 in this thread; I'll tag McMeghan in case he has anything else to add from that conversation. As far as I can tell, that thread never had a formal resolution, it just trailed off. I'll add my main arguments here:
- Seeding improves the level of talent in playoffs. Seeding would ensure that the best players (or some approximation of it) avoid matchups against other elite players until deep in the bracket. I've seen criticism of both recent OSTs and Classics that their playoffs don't have the level of talent we'd expect from top tournaments; that can be addressed. It's true that this comes at the expense of the occasional exciting matchup in R2-R4, but I'd argue that at least from the spectator's perspective, it's more compelling to have stronger matchups in the later stages (by which point they've outlasted much of the volatility of the tournament and have meaningful odds to win the whole thing), and having to follow the entirety of qualifiers to catch the best matchups is frankly exhausting.
- Seeding builds in a reward for unseeded players that make upsets. Right now, a rookie that signs up for the tournament and takes out a top player early in their bracket can easily be matched up against another top player right after (for instance, this happened to Jojo8868 this OST). I'd argue that they should be rewarded for that!
- Seeding removes a significant RNG component from the tournament qualification process. You would be able to know roughly when you'll have to face top opponents in a given tournament by your seed (and accordingly, by your prior performance). Though people have tried to argue otherwise in the previous thread, it just isn't true that winning the tournament/beating everyone in your path is all that matters. For Slam Opens and Classic Cups, every win matters. Making the sixth round as opposed to the fifth as opposed to the fourth can easily be the difference between qualifying for playoffs and not. Those tournaments are improved by reducing the amount of RNG in the qualification process.
- Depending on which data you use for seeding, you can incentivize attendance at particular other tournaments. If you use prior year cup/open performance to seed each cup, you incentivize playing each year, whether or not winning the whole thing is feasible. If you want to do it based on tier circuit standings, you can reward participating in subforum tours too. If you want to incentivize OLT participation, you could make that an input for OST seeding.
- We already know that seeding makes tournaments better, because STour, Classic, Slam, Masters, and OSDT all seed their playoff fields. They could all randomize matchups within their top 16s if they wanted to, and there's a reason they don't. If you accept that, then this isn't a question of "is seeding these tours a worthwhile goal." We can move on to the logistics.
For this OST, I posted an experimental seeding for the tournament, and have posted regular updates about how the bracket has proceeded relative to that seeding. Part of it is a shitpost (there's no meaningful way to rank over half of that bracket), and part of it is to better inform spectators about which players should be considered noteworthy contenders and which matchups are interesting, but the other part is to produce some data about how bracket luck informs the results of the tournament. The point is not to suggest that my formula specifically should be implemented, or even that it's good. Treat it as a ballpark instead.
Slam Opens and Classic Cups don't subject their participants to quite so much bracket variance just because there are five and six of them respectively. With that said, it's still quite impactful; matching up against ABR in R1 of DPP Cup does leave you with another four tournaments, but if DPP was your best tier then mathematically your expected performance in the tournament shot way down just by the luck of the draw. I didn't try to retroactively put together an example seeding for these tours, but let's look through the most recent iterations to see if anything noteworthy happened:
This is just a cursory glance, if somebody knows of more early-round matches like this in either tour then feel free to mention them. The important part isn't even what has happened, though; what's important is that matchups like these can happen, and it's pure RNG as to whether or not they happen. I don't think that the goal of seeding tournaments needs to be to ensure that players like Vert and lax are guaranteed beginners in the first few rounds. Rather, seeding the top 32/64 in OST (16/32 in Cups and Opens) is perfectly fine for the purpose of ensuring that the marquee matchups happen later in the bracket.
- I gave Vert the #3 seed, and frankly that's probably too low; Vert's arguably the best SV OU player on the site. For his top 128 matchup, he was matched up with OLT finalist and teamtour fixture (and my #10 seed) clean, and right after winning that he got my #9 seed Star, who was fresh off a 7-2 run in the tier in SCL.
- Also in the round of 128, my #7 seed lax (who led all entrants in SV OU sheet wins at the time) matched up against Nat, who earned my #4 seed by being one of the most dominant players in SV OU teamtour games all generation long
- Defending champion and my #5 seed Fusien got eliminated in the second round by aim. I don't actually have a problem with that; aim basically hasn't played in the tier all generation. What is unfortunate, however, is that aim was "rewarded" for that win by matching up against leng loi, another SPL veteran and my #28 seed, just two rounds later. And I suppose it's fortunate for leng loi that she got that matchup, because she herself had drawn OLT finalist and my #6 seed Ewin in the second round. If those early-round upsets hadn't happened, it would have been quite funny to imagine a world where we got Vert-clean, Nat-lax, and Fusien-Ewin all in the round of 128
- We can identify other examples of "unseeded player drawing a top player in the early rounds, winning, and then drawing another top player right after;" I don't know if people will have much sympathy because this involved an act win, but while Jojo8868 drawing my #12 seed zS in the second round is something that could easily happen in a seeded tournament, seeding would have prevented them from facing Eternal Spirit, 7-2 in the tier the previous SPL and my #29 seed, in the round of 128
Slam Opens and Classic Cups don't subject their participants to quite so much bracket variance just because there are five and six of them respectively. With that said, it's still quite impactful; matching up against ABR in R1 of DPP Cup does leave you with another four tournaments, but if DPP was your best tier then mathematically your expected performance in the tournament shot way down just by the luck of the draw. I didn't try to retroactively put together an example seeding for these tours, but let's look through the most recent iterations to see if anything noteworthy happened:
- SoulWind losing to Malekith in R1 of BW Cup
- SoulWind losing to ABR in R2 of DPP Cup
- Mako losing to Garay oak in R1 of ADV Cup
- Kate losing to JustFranco in R2 of UU Open (both made playoffs the year before, Kate made UU Open finals, JustFranco was a UU starter in SCL the year before)
- I'm not going to mention a few lesser ones (vani losing to melancholy0 in R2 of GSC Cup or mayo losing to thelinearcurve in R2 of ADV Cup) because even they raised an eyebrow I'm not certain each player would have been seeded
This is just a cursory glance, if somebody knows of more early-round matches like this in either tour then feel free to mention them. The important part isn't even what has happened, though; what's important is that matchups like these can happen, and it's pure RNG as to whether or not they happen. I don't think that the goal of seeding tournaments needs to be to ensure that players like Vert and lax are guaranteed beginners in the first few rounds. Rather, seeding the top 32/64 in OST (16/32 in Cups and Opens) is perfectly fine for the purpose of ensuring that the marquee matchups happen later in the bracket.
Okay, all I've done so far is illustrate the kinds of matchups that can and do happen that seeding would prevent. I'm going to respond to a few of the counterarguments presented when this last came up in 2020.
I have two replies here. First, as someone who did indeed play a top player in their first ever tournament, I object to the idea that getting matched up against a top player right away is necessarily discouraging. It's exciting to play against someone you recognize and know is way better than you! If JimmyLovesPichu matches up against Finchinator in round 1, sure, they might be frustrated that they're overwhelmingly likely to lose, but I think they're also very likely to think "wow, I get to play against someone famous? that's so cool! I could never get this on ladder!" I probably wouldn't have made this point, but when I brought this up on Discord a few weeks ago, someone expressed exactly this! (yes, they replied to Drifting who was saying the exact opposite, but I mentioned my thoughts on that at the outset)
Second, we can be more precise about how many people this actually impacts. Seeding the top 64 would specifically change the matchups involving top 64 seeds, ensuring that none of them match up before the round of 64, ensuring that anyone who beats a top 64 seed wouldn't face another one until the round of 64, and ensuring that the highest seeds don't match up against each other until even deeper into the tournament. If we were to start OST again with seeding (and let's assume I did the seeding; the actual seeding would look slightly different, but it wouldn't really change the overall numbers that much), here are the matchups that wouldn't be possible:
The point is that seeding doesn't meaningfully make life easier or harder for the new player, because it doesn't actually matter who they face all that much unless they face another noob, and seeding the tournament doesn't really change the likelihood of that much. Anyone can beat a top player, and new players are going to be substantial underdogs against any experienced player. The players actually impacted are the top seeds. By my math, the top 64 seeds in this OST have a winrate of around 72% against everyone from the 65th seed onwards; my top 32 seeds have a 76% winrate against the same sample. One obvious takeaway, of course, is "wow, Pokemon has a lot of variance!" But the other crucial takeaway is that top seeds are impacted far more impacted far more by going from "top 32/64 opponent" to "not a top 32/64 opponent" than beginner players are; their odds of winning the match drop from north of 70% to 50% or lower just through bracket luck. It's not about whether top seeds deserve to avoid hard matchups more than beginners (and I'm obviously not making this argument for self-interest, because I would absolutely not be a top seed in any of these tours!), it's about changing a system that currently presents far more risk for top players than it does for beginners, with the ultimate goal of making tournaments more competitive and more entertaining.
I have two replies here. First, as someone who did indeed play a top player in their first ever tournament, I object to the idea that getting matched up against a top player right away is necessarily discouraging. It's exciting to play against someone you recognize and know is way better than you! If JimmyLovesPichu matches up against Finchinator in round 1, sure, they might be frustrated that they're overwhelmingly likely to lose, but I think they're also very likely to think "wow, I get to play against someone famous? that's so cool! I could never get this on ladder!" I probably wouldn't have made this point, but when I brought this up on Discord a few weeks ago, someone expressed exactly this! (yes, they replied to Drifting who was saying the exact opposite, but I mentioned my thoughts on that at the outset)
Second, we can be more precise about how many people this actually impacts. Seeding the top 64 would specifically change the matchups involving top 64 seeds, ensuring that none of them match up before the round of 64, ensuring that anyone who beats a top 64 seed wouldn't face another one until the round of 64, and ensuring that the highest seeds don't match up against each other until even deeper into the tournament. If we were to start OST again with seeding (and let's assume I did the seeding; the actual seeding would look slightly different, but it wouldn't really change the overall numbers that much), here are the matchups that wouldn't be possible:
- Fusien - shiloh (R1)
- leng loi - Ewin (R2)
- Antonazz - Eeveeto (R2)
- clean - Ash KetchumGamer (R3)
- Dasmer - Esteb4n (R3) (Esteb4n already played a top 64 seed in Thiago Nunes)
- Gilbert arenas - Azluc On Top (R3) (Azluc On Top already played a top 64 seed in freezai)
- Vert - clean (R4)
- Nat - lax (R4)
- Dasmer - Let's Rumble Shall We (R4)
- aim - leng loi (R4) (aim already played a top 64 seed in Fusien)
- Eternal Spirit - Jojo8868 (R4) (Jojo8868 already played a top 64 seed in zS)
- Vert - Star (R5) (they're both high enough seeds that they can't meet this early)
The point is that seeding doesn't meaningfully make life easier or harder for the new player, because it doesn't actually matter who they face all that much unless they face another noob, and seeding the tournament doesn't really change the likelihood of that much. Anyone can beat a top player, and new players are going to be substantial underdogs against any experienced player. The players actually impacted are the top seeds. By my math, the top 64 seeds in this OST have a winrate of around 72% against everyone from the 65th seed onwards; my top 32 seeds have a 76% winrate against the same sample. One obvious takeaway, of course, is "wow, Pokemon has a lot of variance!" But the other crucial takeaway is that top seeds are impacted far more impacted far more by going from "top 32/64 opponent" to "not a top 32/64 opponent" than beginner players are; their odds of winning the match drop from north of 70% to 50% or lower just through bracket luck. It's not about whether top seeds deserve to avoid hard matchups more than beginners (and I'm obviously not making this argument for self-interest, because I would absolutely not be a top seed in any of these tours!), it's about changing a system that currently presents far more risk for top players than it does for beginners, with the ultimate goal of making tournaments more competitive and more entertaining.
Let's talk about how seeding would actually work. In terms of how the bracket would be determined, I'd just propose that it can be as simple as deciding the top 64 seeds (for OST, as an example) and randomizing every other player. You then generate the bracket so that in the round of 64, it's 1 vs 64, 2 vs 63, and so on. If the 1 seed gets upset, the player that beats them continues along their bracket path, no reseeding involved. The round of 32 would have the 1st seed against the 32nd seed, the round of 16 would have the 1st seed against the 16th seed, and so on until a final in which the 1st seed would be projected to face the 2nd seed (I say all of this because in theory you could partially randomize the seed matchups, so like, the 1st seed plays either the 3rd seed or the 4th seed in the semifinals. Tennis does that, I've always thought that's stupid).
In terms of criteria, they should definitely not be hand-picked by a host or any committee. I have seen too many Melee or Ultimate tournaments where the hosts blatantly seed the tournament to guarantee the matchups they want to see, giving their favourites easier paths or giving their least favourite characters all their worst matchups. To me, that's a complete non-starter. I would also be hesitant on any system that places substantial emphasis on teamtour placements. Ultimately, teamtours are invitational-only, and I just think it's more fair if seeding is determined by performance at open-bracket tournaments.
I agree with this comment from the previous thread. That's why to me, the best way to do it is to follow circuit standings from the previous year. You have the data to seed as many players as you need. You don't require the development of a new formula. You have all of the single-elimination tours seeded in the same way (OU Circuit for OST, the low-tier circuits for Slam Opens, the old-gen circuits for Classic). You're drawing from a sample of several tournaments, which limits volatility. Yes, you're increasing the incentive to participate in subforum tours. Honestly, I was under the impression that that was desirable and that this would kill two birds with one stone.
As an alternative that still doesn't require you to devise an entirely new formula, you can instead seed based on placements in the previous year's iteration. For instance, Fusien would have been the 1st seed at OST this year, Xrn would have been the 2nd seed (since neither oldspicemike nor sunsets signed up), you randomize JspKiC/FFK/Patatexv from 3/4/5, and continue for as many seeds as you want. The incentive, then, is to play OST every year, and each year you can make progress that compounds into a slightly easier path the next year. You can do something similar with the Cups and Opens.
There are other methods, in theory. Developing a new formula is viable if someone wants to try their hand at it and refine it; I think that'd take time, but could work. The thing is, circuit points are already basically a formula that weighs performance at previous tournaments in the tier. It's not a perfect system, but it's clearly good enough for the mains of that tier. Doing so according to ladder ranking by verifying alts sounds like it could be very fun and true to the spirit of laddering being important to improving at current-gen OU. That's quite a bit of work on the host team, though. Doing so by previous-year sheet wins in the tier is probably viable, but I mentioned why I was a little skeptical of that idea. To be clear, I think there is an obvious answer and playing around with these kind of experimental solutions is only worthwhile if someone has major problems with seeding based on either previous year circuit standings or placement in the previous iteration of the relevant tournament.
In terms of criteria, they should definitely not be hand-picked by a host or any committee. I have seen too many Melee or Ultimate tournaments where the hosts blatantly seed the tournament to guarantee the matchups they want to see, giving their favourites easier paths or giving their least favourite characters all their worst matchups. To me, that's a complete non-starter. I would also be hesitant on any system that places substantial emphasis on teamtour placements. Ultimately, teamtours are invitational-only, and I just think it's more fair if seeding is determined by performance at open-bracket tournaments.
I agree with this comment from the previous thread. That's why to me, the best way to do it is to follow circuit standings from the previous year. You have the data to seed as many players as you need. You don't require the development of a new formula. You have all of the single-elimination tours seeded in the same way (OU Circuit for OST, the low-tier circuits for Slam Opens, the old-gen circuits for Classic). You're drawing from a sample of several tournaments, which limits volatility. Yes, you're increasing the incentive to participate in subforum tours. Honestly, I was under the impression that that was desirable and that this would kill two birds with one stone.
As an alternative that still doesn't require you to devise an entirely new formula, you can instead seed based on placements in the previous year's iteration. For instance, Fusien would have been the 1st seed at OST this year, Xrn would have been the 2nd seed (since neither oldspicemike nor sunsets signed up), you randomize JspKiC/FFK/Patatexv from 3/4/5, and continue for as many seeds as you want. The incentive, then, is to play OST every year, and each year you can make progress that compounds into a slightly easier path the next year. You can do something similar with the Cups and Opens.
There are other methods, in theory. Developing a new formula is viable if someone wants to try their hand at it and refine it; I think that'd take time, but could work. The thing is, circuit points are already basically a formula that weighs performance at previous tournaments in the tier. It's not a perfect system, but it's clearly good enough for the mains of that tier. Doing so according to ladder ranking by verifying alts sounds like it could be very fun and true to the spirit of laddering being important to improving at current-gen OU. That's quite a bit of work on the host team, though. Doing so by previous-year sheet wins in the tier is probably viable, but I mentioned why I was a little skeptical of that idea. To be clear, I think there is an obvious answer and playing around with these kind of experimental solutions is only worthwhile if someone has major problems with seeding based on either previous year circuit standings or placement in the previous iteration of the relevant tournament.
I'll say one more thing. The question of whether seeding should be implemented on the basis that it improves these tournaments and the question of how seeding should be determined are different and should be evaluated in turn. If the answer to the former is yes, seeding would improve these tournaments and should be implemented, then I think the goal should be to implement the best version of that. The idea that seeding just shouldn't be implemented if there's no universally agreeable method even though the best option would be an upgrade on the status quo doesn't make sense. It's not nobler to leave tournaments worse than they could or should be through inaction rather than action. I'd love to hear thoughts here; it's been a while since this was discussed formally and I'm sure people will have strong thoughts on the matter.
