Swiss breakdown - S9 Adv vs Challenger

#1597
Topic created · 17 Posts · 175 Views
  • alt text

    Rather than write another big essay on this, I figured I’d just present my findings when I investigated how Swiss System worked in Advanced this season. Teams with Earned Points in green made playoffs, red teams did not make playoffs.

    G-League is excluded, as they died and earned 0 points, and wins against them are essentially considered equivalent to wins against bye week, though obviously bye week and g-league are not the same caliber of opposition 😉

    https://docs.google.com/spreadsheets/d/1tOXEc1-hsISrVVjBg1ZTw_MnoMzYtsyg51_zv2HPq78/edit?usp=sharing

    Discuss.

    Edit: changed permissions on doc so you guys can see it

  • alt text

    To have some sort of control group here, I added a tab to the google doc where you can see a similar breakdown for Challenger this season. Note that round robin will always have the top scoring teams with the “easiest” opponents because they didn’t have to play against themselves. Also note that the Willmatic team forfeited two games, so their opponents’ wins against them are essentially counted as against bye week, aka a 0 point team.

    I’m sure there’s a cleaner statistical way to break this down and feel free to do so, I just don’t think I want to spend all day researching this lol

  • @tua tkk beastmode

  • Can anyone explain the significance of these graphs to a layperson

  • rr good swiss bad

    personally i think this graph becomes more meaningful if you order the teams in seeding order so that you can see where the trends naturally fall with placement rather than needing to mentally reconstruct everything, like the top teams having the least mp against becomes naturally apparent that way and you can actually see outliers

    really cool to see this though, thanks!

  • @Alto Ask and you shall recieve. via windows paint.
    Swiss vs RR

    and yes, the TL;DR: Swiss is bad.

    the two teams who had the hardest schedule were both in the bottom 3.

  • @Mothership said in Swiss breakdown - S9 Adv vs Challenger:

    the two teams who had the hardest schedule were both in the bottom 3.

    Isn’t that still going to be the case in RR, as an inverse of the “top teams will have the easiest schedule because they can’t play themselves” thing that Tua pointed out?

  • @Mothership said in Swiss breakdown - S9 Adv vs Challenger:

    @Alto Ask and you shall recieve. via windows paint.

    excellent, thank you.

    @scaredy-bat said in Swiss breakdown - S9 Adv vs Challenger:

    Isn’t that still going to be the case in RR, as an inverse of the “top teams will have the easiest schedule because they can’t play themselves” thing that Tua pointed out?

    yes, it’ll be the case in RR but it shouldn’t be the case in swiss, as bad teams in swiss will presumably play other bad teams leading to lower opponent MP (OMP), whereas good teams play other good teams and thus accumulate more OMP. there is still some basis for it (if you say there are 6 top teams and 6 bottom teams, a top team has a 6/11 chance to play a bottom team vs a 5/11 top team and vice versa), but it shouldn’t be as pronounced. it’s very difficult to draw any hard data from this because ultimately teams don’t control their schedules and good teams will win regardless of who they play and bad teams will lose regardless of who they play, but you can likely use it as a basis to investigate other trends. I was going to say something about asian being an outlier here, but apparently their match against :3 wasn’t counted and they should have 71.67.

    you have to be careful with this stuff though because nothing here is conclusive and it’s very easy to use the data to support an argument that may not be right. that being said, you could choose to interpret the data here (combined with the fact that everyone in adv other than :3 was within a game of each other) to come to the conclusion that since making playoffs or missing it were a matter of one game, the fact that there’s a slightly higher point total to the bottom implies that teams who were “unlucky” (played 4 top teams and 3 bottom teams leading to a higher OMP) missed out on playoffs over the team who were “lucky” (played 3 top teams and 4 bottoms for a lower OMP). that’s a sound argument but not necessarily a correct one; you’d have to go and define top and bottom teams and then actually compare individual team schedules.

    and even if you can confirm that, it doesn’t necessarily mean anything. the best team in the div could get “lucky” and play 2 top teams/5 bad teams because they happened to catch bad teams when they were relatively over-ranked by swiss, and still end up rolling the div because they were the best team. stuff like this is (really) fun to look and conjecture over but not nearly as indicative of trends as it may seem, with no offense meant to tua for making it.

  • @Alto Thanks for walking through it! I reckon it’s the middling ones that it affects most as you say, rather than tippety top and bottomest bottom.

  • @Alto
    No offense taken, it seemed pretty clear to me that the results were not really super damning (or un-damning) for swiss, hence why I published it without any sort of real narrative or argument; i wanted people to interpret it freely and get a chance to look at what the numbers were before voting on the survey this season

    I still believe RR is the best way to go and I’d much rather increase the amount of divs, more of which are running RR, than decrease the amount of divs and lump everyone together, that way every team is judged against much more similar criteria for making playoffs. We could even run something like what ESEA does with Premier now, two RR groups where top 4 from each make playoffs https://www.hltv.org/events/5885/esea-premier-season-37-europe

    An alternate route would be just to play more matches, extending the season to 8 or 12 weeks so you have more datapoints to work with; the more matches played, the lower the chance of a meh team fluking their way to playoffs by getting disproportionate amounts of easy teams to play against

  • can we please extend the seasons to be 99999 weeks so that i never ever have to stop playing highlanfer

  • @tua said in Swiss breakdown - S9 Adv vs Challenger:

    I still believe RR is the best way to go and I’d much rather increase the amount of divs, more of which are running RR, than decrease the amount of divs and lump everyone together, that way every team is judged against much more similar criteria for making playoffs. We could even run something like what ESEA does with Premier now, two RR groups where top 4 from each make playoffs

    i think rr offers diminishing returns after inv and ch. i remember looking at the prolander league standings this last season and wondering how the line between div 13 and 14 is set, especially when admins are in charge of multiple divisions anyway

    multiple groups for a div (conferences?) is a solid idea for handling larger divs like main and advanced, but you run into the issue of having not enough admins if you decide to have one admin per conference, and the issue of having “easy” vs “hard” conferences if one is stacked and one isn’t

    i don’t think it should be considered for inv or ch unless the number of solid teams increases. right now the gap in those divs is large enough

  • Given the teamdeath issue in RR I understand not extending it across the board, but I reckon most folks thought RR Adv, at least, was nice in S8.

    If we do extend the season I think nine would be nice, at least until we get more league-ready Steel-likes – an odd number, anyway, to preserve balance between KOTH and Payload. Bonuses: same number of weeks as sixes has (less confusion when people cross over?), same number as the number of players (okay, that one is just for fun).

    Anyway, I would be super interested to see this kind of breakdown across more seasons and/or divs, if any stats nerds out there are feeling bored some afternoon 🙂

  • i think personally that there are multiple viable solutions, but i think that in smaller divs, like imo 16 or smaller given 8*2, swiss isn’t really viable imo. i’d much rather have more RR OR fewer, but bigger, swiss divs. like possibly an esea type div structure for the lower divs couldnt hurt imo.

    nc, amateur, im, invite? where inv is RR, nc is newcomer, am is am-low main, im is mid main to high adv/chal

    just brainstorming rly

    edit: just scrolled up and tua has good points on why the former is preferable

  • ETF2L does multiple smaller groups with the top of each making playoffs with Division 2 A/B. It really depends on how the numbers work out.

    Swiss accuracy will always be lackluster when you have only 7 matches, and in a div like advanced this season which wasn’t far off of being able to run RR, there’s an argument both ways, it really just depends on how structure matches team count ultimately.

  • @dlphn yeah I feel you, I think that at lower levels esp. when people don’t have much experience to base placements off of and it’s harder for admins to gauge the skill levels of players, swiss is best. I’d personally draw the line at either adv or mid-high main, because I think as long as we can reliably place teams into small divs by skill (which mind you, is pretty hard), we should run RR. Plus, it encourages more qualifiers, which I think are healthier for the competitive scene than having the only placement method being admins trying to determine which teams are better

    @scaredy-bat I actually think team death in RR is actually a little better than in Swiss, though you could debate it either way. Chances are, it’ll be a bad team in the div that dies, usually like 3-4 weeks in. So half of the remaining teams will have played the bad team and likely won (considering the unlikelihood of a good team dying), the other half of teams will get a bye week, so essentially the majority of teams, if not all, will get a bunch of points from the dead team. Everyone in the div gets a chance to farm points from the shit team.

    Meanwhile, in Swiss, if the dead team dies, they might be replaced by a bye week or they might not, depending on the number of teams in the div. For divs that are like 12 teams for example, if a 0-3 team dies, maybe 4 teams get a bye week, and if the div is competitive, then it could be a team that would normally finish the season 3-4 or 2-5 that gets the bye week and is able to finish the season 4-3 and sneak into playoffs. At least in 20 team swiss, there will be more potential 0-7, 1-6, etc. teams to get the ffw instead of a borderline playoff contender team in a div tight enough where they might be in the bottom few teams, even though they’re one win away from playoffs. Sound too hypothetical?

    https://rgl.gg/Public/LeagueTable.aspx?g=503&s=98&r=24 in Season 7, Crit Sandvich made playoffs over the Flow (arguably a better team) by less than one point because Crit Sandwich got a bye week and Flow didn’t.

  • This post is deleted!
Log in to reply