|
Online Heroscape Games and Events Play Heroscape online with people across the country and around the globe. |
|
Thread Tools | Search this Thread | Display Modes |
#49
|
||||
|
||||
Re: Online Player Rankings Thread
Quote:
You saw nothing. My Repaints and Mods: Updated 1-13 My Custom Cards: Updated 4-15 My Custom Superheroes: Updated 1-13 My figure images for online games: Alliance figures, Utgar figures My Etsy Store |
#50
|
||||
|
||||
Re: Online Player Rankings Thread
Same.
Last edited by superfrog; October 14th, 2013 at 11:53 PM. Reason: I am allowed to lurk, am I not? |
#51
|
||||
|
||||
Re: Online Player Rankings Thread
It's for your own protection. If you saw my record, we'd have to kill you. The other two must just be glitches.
|
#52
|
||||
|
||||
Re: Online Player Rankings Thread
Huh. It's being caused by a really weird glitch (which I won't be fixing because it's in an outdated spreadsheet anyway). dok will have to switch it over to the S10 sheet, he's the one controlling the Top 25 list.
|
#53
|
||||
|
||||
Re: Online Player Rankings Thread
I'm hoping to get myself into the top 100 here, haha. Gotta start somewhere.
|
#54
|
||||
|
||||
Re: Online Player Rankings Thread
101-148 is all quite low. You could probably get into the top 100 with a single win.
|
#55
|
||||
|
||||
Re: Online Player Rankings Thread
Quote:
Quote:
I suspect that if Robber’s results were normalized to the same amount of games as Megasilver that he would be significantly higher. Robber or Boromir would probably be the best case sample for normalization. If Boromir’s results of 10-3 were projected to a 40-12 record with the exact same opponents played as he did in his first 13, then I suspect he would be top 5 easily. And if that’s the case then it tells me that the number of games is a significant factor. Quote:
1) Does a forfeit loss count in these ratings? 2) Is there a static threshold where number of games played causes the ‘confidence’ value to stabilize, or does that confidence value fluctuate as a percentage of the total number of games played? (or total games played by the player with the highest amount of games played) Example
Spoiler Alert!
3) How is your rating affected by previous opponent’s future results? I have heard that is definitely a factor, but relatively small or relatively large? If I beat someone when they were 0-0 all time and they later are #1 with a record of 19-1, at the point they became 19-1 would I have the same ranking at 1-0 as if I beat them when they were 19-0? And I hope I don’t sound like I’m picking on Megasilver or anything here. His case is the best one to discuss because he is under .500 and on the top 25 list. |
#56
|
|||||||||||
|
|||||||||||
Re: Online Player Rankings Thread
Quote:
That said, because of the way we pair rounds 1 and 2, this naturally leads to them drawing mid-tier opponents for the first two rounds. (This was actually the rationale for getting rid of the artificial boost - we realized that it was redundant given the pairing rules.) Quote:
This is my point, in a roundabout way: very few players actually maintain those sorts of results over an extended period. Quote:
As I've said before, these are conservative estimates of skill. The algorithm is saying "we are very confident that this player is at least this good". In the case of Megasilver, the algorithm is very confident that he's at least fairly good. Quote:
Again, though, it's completely reasonable to be suspicious of a good record built on a small number of games. Heroscape is a game with variance built in. Most players who have started out with 3:1 win ratios (or posted 3:1 win ratios for some short period) have not been able to maintain those win ratios. If we assumed that everyone who played that well for a brief stretch was good enough to do that consistently, we would see more retrospective errors in our rankings than we have. In other words, it's better to underestimate Wanderer for a little than overestimate a lot of other people for similar lengths of time. Quote:
Quote:
Each player's ranking has a mean and a variance. The variance starts high and falls as a player plays more games. Results that are consistent with your previous results will drive your variance down more quickly. Games against players with low variance will drive your variance down more quickly as well (so, beating me will lower your variance more quickly than beating vegie's dad, because the algorithm has more confidence in the significance of a win against me). Quote:
Quote:
Quote:
This is the big difference between Trueskill through Time and regular Trueskill. And I want to emphasize again what an achievement it is that we have TTT in use here. We really are on the cutting edge of ranking algorithms here. Xorlof deserves a huge amount of credit here. Quote:
As always, thanks for the discussion. Last edited by dok; October 17th, 2013 at 04:42 PM. |
#57
|
||||
|
||||
Re: Online Player Rankings Thread
I completely agree with Kevindola here, I think the people with a lot of games are clearly avantaged.
Not so long ago Kinseth was ranked higher than Wanderer whereas there was a quite huge difference between their win ratios. Wanderer is finally placed at a decent place for him but the fact that he was not in the top 5 for so long proves in my opinion that the systems is biased and advantages a lot players with a lot of games played. Anyway the rankings does not matters much to me, we all know dok did this in order to comfort himself. |
#58
|
||||
|
||||
Re: Online Player Rankings Thread
I'm pretty sure that playing more games is a huge advantage too. My dad fell two or three spots when he sat out just one season. Did that many people get better than him or were they just able to get more points in this system from an extra season of games?
I think I'm the perfect example of someone with way too high of a ranking for my record. I do think I'm a good player, but I have some terrible losses and my record of 29-23 really isn't that great. I'm probably one of the most inconsistent players. However, I've played a ton of games, which is what is really keeping me afloat, especially since I played in the early seasons. Playing in the earlier seasons means that I was able to play against the people who are now the Top 25 a ton of times, and those wins are worth a lot now. 10 Years of Gencon/Scapecon Battle Reports - Comic Battle Reports - Probability Calculator App - Reverse the Whip Army Archetypes "It's all about the game." - Sgt. Ernie Calhoun |
#59
|
||||
|
||||
Re: Online Player Rankings Thread
Quote:
Yes, playing games *is* an advantage, assuming you continue to play on the same level, and that level is at least reasonably good. This is because the algorithm gets more confidence in your ranking and the conservative estimate of your skill gets closer to your actual performance. And again, being conservative in our assessment of players skills actually improves the predictive accuracy of the rankings. If we assumed every 5-1 player was super awesome at Heroscape, we would make more mistakes in our predictions of who wins games than if we guessed they were an above average player on a hot streak. Neither of those things. |
#60
|
||||
|
||||
Re: Online Player Rankings Thread
Why would that cause a substantial amount of mistakes?
I think that more often than not, players first seasons are worse than their later seasons, since they learn a lot from online play. How many people have gone 5-1 and then regressed to their mean and had multiple seasons of going 1-3 or 2-2? I'd bet that the number of people who started 5-1 and then continued at a similar winning rate is much greater. I am curious what happened with my dad's ranking then, if not the latter (the question was intended to be rhetorical). 10 Years of Gencon/Scapecon Battle Reports - Comic Battle Reports - Probability Calculator App - Reverse the Whip Army Archetypes "It's all about the game." - Sgt. Ernie Calhoun |
|
Thread Tools | Search this Thread |
Display Modes | |
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Online Heroscape Match Request Thread | Izner | Online Heroscape Games and Events | 2670 | September 15th, 2023 09:37 AM |
Single player versus multiple player activations? | underling | Other Customization & HS Additions | 3 | November 19th, 2012 11:26 PM |
Maryland Multi-player Madness Post tourny thread | clancampbell | Events | 8 | May 26th, 2010 11:41 PM |
Online battle finder thread | pixlepix | Software | 0 | November 15th, 2009 06:37 AM |
Unit Rankings: Master Thread | Elginb | Official Units | 2 | October 12th, 2009 11:46 AM |