Jump to content

g00gly eyes

Member
  • Content Count

    18
  • Joined

  • Last visited

  1. Vegas is included in Elo ratings. You are correct in that I initially had planned to leave open events out of it because of the several problems it creates, which I explained in detail. However, whether to include it or not is really just a choice about what we want Elo to capture and how. I initially got no feedback, but then an influential member complained carefully articulated his concerns about not including OC, so I revised the methodology and began including all official open events. I can explain in more detail later, but the way open events work pretty much is that only matches among pro teams are counted (you don't get points for crushing amateur teams).
  2. No worries, I'm happy to explain: First: why don't all teams start Season 2 at the same Elo rating (which would be the average, 1500)? That's because the Elo ratings actually go back to NA Regionals and HWC 2016, which I used to initialize Season 1 Elo ratings. I had to start somewhere, so I started everyone at 1500 for NA Regionals and progressed from there. Second, you may ask: why do all teams' Elo ratings "move back" towards the middle at the beginning of Season 2? Shouldn't they start where they left off at Season 1 finals? This is called "regression toward the mean," and it's standard methodology in Elo ratings. The reasoning behind this is that all teams "reset" a little between seasons (i.e. bad teams catch up, good teams lose some advantages, new maps, new gametypes, etc). I revert all teams by one third to start each season. How did Optic reach a higher Elo rating in Season 2 than they did in Season 1 when they dominated Season 1 much more? Two reasons: Optic started at a significantly higher Elo rating in Season 2 than they did in Season 1. This of course is precisely because of their dominance in Season 1. In this regard, you can see that Elo ratings carry over a "legacy" or "dominance" component, which again is a feature. For example, an NFL team might reach its highest Elo rating after dominating for some consecutive years. That is, their highest Elo rating would measure the apex of their historical run. Season 2 had open events, meaning it had more matches than Season 1. This makes it a little harder to directly compare the two seasons since it means more potential opportunities to rack up points. For example, if Season 1 had more matches, Optic might have won those anyway, meaning they would have accumulated a higher Elo rating. This is why it says "Not seasonally adjusted." I will probably seasonally adjusted it later, but I didn't want to do it mid-season and create more confusion. Finally, if I understand correctly, why did Envy's Elo rating drop a second time? From what I can see, Envy's Elo rating dropped three key times: Week one when you lost to TL, OC when you lost to OG, and at finals in the winner's bracket when you lost to OG the first and only time in that event. So the last dip you see is that last loss, after which you came back to beat OG twice, causing the large swing in Envy's favor. And in case you're wondering, Finals matches have a higher "weight" than regular season matches, specifically 30 and 20 respectively. These numbers won't mean anything on their own, but you can see their relative weight. This type of weight adjustment is also standard in Elo methodologies. Of course, I've had to adapt Elo in several ways to develop a methodology that works for esports, but I believe I've done it well. Happy New Year!
  3. OG reached their highest Elo rating right after beating NV in the winner's bracket of the finals, but then NV came back to win the next two series in the grand finals to take the tournament. At the very end, NV barely had a higher Elo than OG in part because of those two final head-to-head victories. With teams this evenly matched, head-to-head matches are key because they produce a swing on both ends (i.e. NV goes up while OG goes down).
  4. Hi Everyone! I finally got some free time to put together the final fall season Elo ratings: Envy barely edged out Optic at the end, but Optic did reach the highest Elo rating of the season (and beat their own unadjusted Season 1 record). It was a great season and I'm really looking forward to HWC 2017!
  5. I just hope this doesn't compromise my position at Snipedown Industries...
  6. Since @LethuL asked so nicely, here are the end-of-season Elo ratings including OC and LV. Notes and observations below for those interested:
  7. Lol. No, I have not, at least for now. These open tournaments create some methodological problems, as I mentioned in my previous post -- which received a lot of feedback, btw. Thanks, guys. If I include OC, I have to include LV. That could actually favor Envy...
  8. End of regular season Elo ratings: Interesting to note: Envy could potentially surpass Optic in Elo rating after finals, or Optic could win out and break their own record from last season. Also, right before Week 7, Allegiance broke last season's low, held by the former Optic team. Finally, Str8 went on two impressive runs to make it into finals.
  9. Everyone -- this is your chance to have input into how the Halo Elo ratings work! But first: Now the fun part: how should open (or other) tournaments be handled in regards to the Halo Elo ratings? Should they be included? What about official/sanctioned versus third-party tournaments? And if included, how much "weight" should these tournament matches get compared to pro league matches (for simplicity: more, less, same)? As you can see, I have not included the Orange County tournament for now. These type of tournaments pose some problems: are all pro teams competing? How do you handle amateur teams (who don't have Elo ratings)? What about mixed/one-off teams (e.g. Truce)? I have solutions in mind for these problems, and some I've already had to address when initializing the Halo Elo ratings. But in the end it really comes down to preference and (e)sport-specific knowledge, so first I want to get your thoughts, ! Note: the season finals will of course be included -- as they were last season -- and they received a higher weight. The question is in regards to tournaments outside the scope of the Pro League itself. P.S. Again, I pull scrim results from halodatahive, so thanks to @Bad Dub for the great site.
  10. Thanks! I completely agree. An important way of bridging the gap with traditional sports is in analytics and statistics. And you're right, esports actually has the potential to surpass traditional sports in this regard. Good luck with your project. Let me know how it's coming along!
  11. Good question! For this season, they start at the rating they had at the end of last season, but reverted to the mean by one third (this is standard practice among Elo methodology in traditional sports). However, when individuals form new teams, they carry forward one fourth of their previous team's rating. But you're right that at some point I had to make all the teams equal, and that is exactly what I did for the start of the NA Regionals. From there the ratings carried over to the HWC Finals, and then to the Pro League Season 1.
  12. Hi guys! I was in Europe the two weeks before the league started and have been really busy since then, but I finally made the time to start up the Season 2 Elo Ratings: Interesting to note, NV is the closest a team has gotten to OG (formerly CLG) since the preseason of the first Pro League! The Elo rating composition of the league is also interesting: two teams at the top, two at the bottom, and four clustered in the middle. Something similar happened around this time last season before teams separated themselves, but with the parity we've seen so far I wonder if that will happen again this time. And that makes me wonder if we'll ever see a team reach the level CLG did last season. P.S. I've been trying to catch up on the forums since I got back. Just gotta commit to mad reading one of these days! P.P.S. Shoutout to @Bad Dub for the excellent source of scrim data.
  13. Hi guys! Here are the complete HCS Season 1 Elo ratings: Calculating Elo ratings for this season was a fun exercise, in part because the nature of esports poses interesting methodological challenges compared to traditional sports. However, as I mentioned earlier, I think bridging the gap with traditional sports is very important, and analytics is just one of the many areas in which we could do so. While I'm on the topic, but without going into great detail, I do want to say that everything revolves around the players, and they are key to the success of Halo as an esport. This is the case in all traditional sports. Competition does not happen in a vacuum. People care about people. They care about their personalities, their performances and their storylines in the context of the sport. Competitive Halo needs to transition from a sequence of tournaments to a cohesive product like modern sports leagues. I think this has been happening, but not quickly enough. In this regard I think 343 and the current league/esport organizers could do a lot more. As the owners of the intellectual property and esport, it really is up to them (and in their best interest) to develop the right type of institutions that would foster growth around Halo as an esport. (Great examples of player development are the Naded X Games video and more recently the Ninja Bud Light All Star video.)
  14. Hi guys. Here are the HCS Elo ratings at the end of the regular season. Will CLG continue to grow the gap into the finals? I thought it'd be interesting to look at the top 5 upsets, as determined by Elo rating difference. Obviously, E6's upset over CLG is the largest. Unfortunately for Liquid, their low Elo rating throughout the season meant that any time they beat a top team it was a large upset. However, they did manage to beat top teams, even if only at scrims. It's a shame they weren't able to translate that to the pro league. Also, thanks to the guys from HaloDataHive and Halo5Arena for compiling scrim results (among the other awesome stats). Fantastic work!
×

Important Information

By using this site, you agree to our Terms of Use & Privacy Policy.