EPT[Fantasy] Shinhan PL Round 2 - Page 17
| Forum Index > BW General |
Hot_Bid
Braavos36362 Posts
| ||
OneOther
United States10774 Posts
| ||
|
SingletonWilliam
United States664 Posts
On November 30 2008 16:07 OneOther wrote: if bisu beat calm and calm was on a 3 game winning streak, does it count as 1 point? No, all streaks started when the deadline ended. | ||
semioldguy
United States7488 Posts
On November 30 2008 08:53 Hot_Bid wrote: while we could've pro-rated the current streaks players were on into the player costs, i don't think semi accounted for already existing streak value, so i would assume all streak statistics are now reset, but wait for semi to make the official call All streak values are reset. The reason is to put all players on equal footing at the start of the game. | ||
Aesop
Hungary11233 Posts
What is the logic behind giving points for appearance in the lineup? The point is this. If a player is fielded and loses, he loses his LU point again and gives you 0 points in total. If a player wins, he gains 3 points, 2 for win, 1 for lineup. So why even value lineup and not give "3 points per win" and "0 if they are not on the lineup or lose their game"? Normally I would think you want to reward players even making the lineup, but currently it only happens if they win, else they are +-0. | ||
|
Hot_Bid
Braavos36362 Posts
Your alternate classification is just a different way to show the scoring, which really is +3 per win and 0 for a loss. You can do this for a lot o fthings, for example Ace appearance can be reclassified as +1 appearance and +1 ace bonus, which is the same as just +2 ace appearance. Also, they don't have to win to get rewarded, if your player is the 4th game and the match doesn't progress that far (3-0 win or 0-3 loss) he still gets +1 for being scheduled to play. | ||
|
Aesop
Hungary11233 Posts
On December 01 2008 07:31 Hot_Bid wrote: I think that's exactly the logic, you get rewarded for making the lineup, it means your player is good. Your alternate classification is just a different way to show the scoring, which really is +3 per win and 0 for a loss. You can do this for a lot o fthings, for example Ace appearance can be reclassified as +1 appearance and +1 ace bonus, which is the same as just +2 ace appearance. Also, they don't have to win to get rewarded, if your player is the 4th game and the match doesn't progress that far (3-0 win or 0-3 loss) he still gets +1 for being scheduled to play. The case of "being scheduled game 4 but not getting to play" is interesting, but right now it's the only (and comparably rare) case where making the lineup is rewarded. Apart from that, players who are not played at all and players who make the lineup and lose are not treated differently. Of course it's too late to change the rules now (people picked based on rules), but removing the -1 for losing appears like the most logical to me to introduce proper differentiation. | ||
|
Hot_Bid
Braavos36362 Posts
| ||
|
Aesop
Hungary11233 Posts
On December 01 2008 08:02 Hot_Bid wrote: why should the -1 be removed? i don't understand Right now: No differentiation between {making the lineup and losing} and {not making the lineup}. a) 0 if not in lineup b) 0 if in lineup and lose c) 3 if in lineup and win As you stated, you want to reward players if they are being put into the lineup, compared to not being considered at all. So the logical step would be to remove the -1, so players are worth in points: a) 0 if not in lineup b) 1 if in lineup and lose c) 3 if in lineup and win Thus creating a difference between the cases a) and b). | ||
|
Hot_Bid
Braavos36362 Posts
On December 01 2008 08:12 Aesop wrote: Right now: No differentiation between {making the lineup and losing} and {not making the lineup}. a) 0 if not in lineup b) 0 if in lineup and lose c) 3 if in lineup and win As you stated, you want to reward players if they are being put into the lineup, compared to not being considered at all. So the logical step would be to remove the -1, so players are worth in points: a) 0 if not in lineup b) 1 if in lineup and lose c) 3 if in lineup and win Thus creating a difference between the cases a) and b). a) and b) are not the same, in a) you have no chance for win points and b) you have an opportunity to get a win. that's real value even if it doesn't necessarily show up on the score sheet, and a significant difference than not being in the lineup at all, which is guaranteed no points (save for team wins). | ||
|
SpiralArchitect
United States2116 Posts
On December 01 2008 04:23 semioldguy wrote: All streak values are reset. The reason is to put all players on equal footing at the start of the game. KK thank you. | ||
|
LonelyMargarita
1845 Posts
On November 30 2008 06:01 traced wrote: Okay so you did some math. You overlooked important things, though. 1- Projecting the high end. We both had Flash at greater than 5 points per dollar. The thing is, Flash, Jaedong, Stork, and Best are more likely to keep a > 5ppd, whereas guys like Calm and Pusan are not. Flash, Jaedong, and Best, though I didn't do the math for the previous season, have similar stats last season. They're consistently elite, and their point output is likely to be similar. Calm and Pusan may perform just as well, but they are less likely to. 2- Predicted points per dollar is greatest at the highest end and lowest end. $1 players may not necessarily be worth it. I had Anytime anywhere from a 3-9 point value. $0 players are. Because... 3- $0 players give you 5-10+ points for free. Each one essentially make your other players worth more points per dollar. The important factor here is team wins. SkyHigh and Han benefit from 1 lineup appearance already, which projects for 1-2 points, in addition to their 5 points from team wins. Also, simply appearing in the lineup once makes it more likely for them to appear in the lineup again. Shuttle is also a decent pick because STX will likely win 6 or more games, and he is a candidate to make an appearance or two. I'm sure there are others. 4- $3-4 players all have lower ppd than the highest and lowest end. I had them consistently anywhere from 3-4.5 ppd. Then they share a synthesis flaw: 5- Though a few (Bisu) $5 players are worth more ppd than the highest end, all of them force you to take a $3 or $4 player. There are even more good $6 values, but they force you to take a $2 or $3 player. Flash or Jaedong plus Han or SkyHigh is simply more valuable than any combo you can create. Remember, you're buying consistency - Flash and Jaedong are very likely to repeat performances - some $3 or $4 WILL outvalue them this season, but at this point you cannot project any individual to, if you understand what I mean. 6- Most people overlooked this one. Teams just score less points per dollar than players. It is actually more efficient to choose a $3 team that wins 4 games over a $4 that wins 6 games, because the extra dollar in players makes up the difference and then some. That SKT will probably go at least .500 this season makes them the obvious choice. My lineup is Jaedong, Flash, Stork, Thezerg, Han, and Skyhigh. I'm projecting my players for a combined >5 ppd (without factoring in SKT). 1) Flash was exactly 5 pt/dollar. He also had the most wins of any player, the most ace appearances, and the longest streak. He was the absolute best performing player in round 1, and among the best for any player of any season since proleague has been around. And you're trying to argue he's not only going to improve, but by enough to end up at a greater points/cost than 5 point players that improve. Combine the fact that 8-9 pointers have the greatest potential to underperform last season with the fact that they have the least room for improvement, and I don't see anyone winning with them. It's much more likely a 6-7 pointer will perform as well as last season's 9 pointers, and that strategy could win. Stork and Jangbi I think are the best bets, but I didn't like that strategy quite as much, so I didn't spend much time on it. 2-3) Anytime is a MUCH better choice than any zero point player. Why? a) He's a better, more experienced player overall and more importantly (pending something really weird), b) He will play in AT LEAST 9 games. I'm guessing he'll play in 11 because of an ace match or two. He's ace's best player of any race, multiple maps already favor his race, and he's the new star of the team (fandom wise). He will not get as many team points as any zero pointer on any team besides ace or estro. But why are zero pointers zero points? Because they played zero games last round. Coach wouldn't even try them on TV once. The best performing zero pointer will likely win 3 games in 5 appearances, and that's one guy out of the whole pool of 0s. sKyhigh is the best bet, especially if Iris slumps. But if he loses his 1st game, there's a decent chance he won't get more than 1 or 2 more tries. CJ will likely play 2 zerg in 9+ matches, and Iris will get most the terran slots, meaning a likely 5 appearance best possible situation for sKyHigh (bear in mind CJ just moved azang onto the roster as well). I'll take Anytime's 11 over sKyhigh's possible 5 anytime for 1 more point. 4) I don't have any 3-4 pointers, nor do I have any 2 pointers or 6+ pointers. I have anytime and five 5 pointers. This was the basis of my strategy, as I briefly outlined. And I think your math may be off: I got that 3 pointers were about 4.33 pt/dollar, and 4 pointers were about 4.56 pt/dollar (beating or tying 5-, 6-, and 7-pointers). Besides Anytime, Midas, Thezerg, by.hero, Firefist, Movie, and Much, I feel all 1-3 pointers were bad. There were a few possibly good 4 pointers, including GGPlay, Oversky, and Mind. I thought anytime was a better pick than midas, didn't have room for a 2 pointer in the team I chose, and found there was always a 5 pointer better than a 4 pointer of the same race. As for much and movie, it's too much of a risk that they'll split the games (especially with jjonga joining the roster); compound it with the fact that toss is CJ's weakest race, and that's likely a maximum 4-6 appearances each. 5) Like I said, I have Anytime and then all 5-pointers. No 3s, no 2s, not even any 4s, 6s, 7s, 8s, or 9s. I don't know why you'd mathematically think 5 pointers then require 2-3 pointers, unless you assumed everyone chose STX for their team. STX was a very bad team choice. CJ, Stars, and SKT were the good choices. And your next point is the exact opposite of how fantasy leagues work. You are absolutely not buying consistency if you want even a remote chance at winning. Buying consistency will get you an average-below average team in the final standings (since the average person's team will probably beat the average points/team using last rounds stats just based on making decent picks). If you want to win, you are betting on not only improvement, but noted improvement. You want a team where EVERY SINGLE PLAYER can get 3 more wins than last season. As soon as you buy a 8-9 point player, it's absolutely impossible that will happen. No one's going to win 13 games in round 2. It would be the greatest proleague performance by any player in any proleague ever. The winning team will probably have an average improvement of more than 2.5 wins/player. Although there's a small chance someone could win with Flash or Jaedong on their roster, it will likely be because of a dark horse 0 pointer getting 5+ wins. Not sKyHigh, han, or young, but someone on only a couple people's teams. And not projecting a low-value player to improve? This is a gambling game. You have to gamble to win. Like I said, mathematically you have no room to improve with high-pointers. If every player can move up or down a maximum of 4 points from last season with a maximum of 10 (the best season any player has ever had), the most potential comes from a team with all players of 6 or fewer points. They also have the most potential to get last (since 6 5 point players could all be 1 point players next season), but you need the potential to win or you'll just finish average. 6) On average, yes: Average team is 4.21 pt/dollar and average player listed was about 4.41 pt/dollar. Average also includes Ace though, and going by the average is pointless, since you can pick out the winners and losers. CJ Entus outscored any player last round with 5.25 pt/dollar, and with their lineup as the most well-rounded team they could easily be the best team in round 2. If their 0-3 loss had gone 1-3, they'd have actually scored the most before factoring in the +4 for overall round winner. SKT has the most potential for improvement, but are so lopsided with 2 powerhouse Protoss players and not much else. Since 3-0s and 3-1s are worth so much, they won't be the best team next round, but could be a decent value. Woongjin could also be great if a Terran steps up on the team and balances them out. *** Basically, your strategy could be described by words like "average," "consistency," and "risk-free." And that's just what I expect will happen. Your team will do about as well as last season, but gain a few points over the mathematical average because of the 0 pointers. The only chance you have at winning is a 10 win season by stork, and a huge season by thezerg, along with flash and jaedong continuing their unlikely 10 win seasons. You can't gain on skyhigh and han because too many people have them on their roster. You'll probably finish between 50 and 100 out of the 200+ entries. The guy who wins will likely have a similar strategy to quite a few people near the bottom, because he'll have lots of room for improvement on all players, but also lots of room for them to do worse. You can't win without risk. My strategy combined maximizing potential gain stat-wise with maximizing appearances. All my players will play in at least 7 games, and they should all play in 9+ games. I don't see how anyone can argue that this is a bad strategy, unless you think the object of the game is to do average. There's nothing lost for coming in last place, but there is reward for coming in 1st, so all that matters is winning. You don't get points for being above average. My team should have a decent chance at finishing in the top 20, but also a decent chance at being in the bottom 50. The risk balances out, and most people following my strat will finish around average, but I think more than likely the winner will have followed it. We'll see in about 6 weeks. | ||
|
semioldguy
United States7488 Posts
On December 01 2008 08:12 Aesop wrote: Right now: No differentiation between {making the lineup and losing} and {not making the lineup}. a) 0 if not in lineup b) 0 if in lineup and lose c) 3 if in lineup and win As you stated, you want to reward players if they are being put into the lineup, compared to not being considered at all. So the logical step would be to remove the -1, so players are worth in points: a) 0 if not in lineup b) 1 if in lineup and lose c) 3 if in lineup and win Thus creating a difference between the cases a) and b). As Hot_Bid mentioned... the difference is the player got the opportunity to get more points. It then is the player's responsibility to win to earn points. The point for appearance is there not to benefit players that lose, but so that a player who didn't get a chance to play and win could still be worth something. | ||
|
traced
1739 Posts
On December 01 2008 12:53 LonelyMargarita wrote:--- First issue - you're confusing risk with variance. You described your strategy as "gambling," which is appropriate; your build buys variance at a loss of value, ie. typical gambling. You can still have a variant team without sacrificing value. You say you're more likely to finish in the top 20 or bottom 50 than me, most likely to finish 50-100. It's more realistic to project my likely range anywhere from 20-100 with a 60 mean, while yours 20-150 with an 85 mean. Though honestly both of us are just guessing. Second issue - you based a lot of your entire argument on my use of the word "consistency," which was my mistake in phrasing. Regression-resistant is a more appropriate term. My point was, basically, Jaedong and Flash are both consistently the highest scoring players in proleague per dollar, and least likely to regress back to the mean. Third issue - mainly because I used "consistency" to describe Flash and Jaedong, you conclude my entire team is "consistent" and "average" and "risk-free." SkyHigh, Han, and TheZerg are all players with a high degree of positive variance. Stork is relatively variant, especially considering he's a high value player. And we agree SKT is probably the most variant team, other than ACE. It doesn't matter that quite a few people have the $high/$high/$high/$1$2/$0/$0 build. Player choice and lineup synthesis will be more than enough to distinguish teams and create a significant spread. Fourth issue - You say "We'll see in 6 weeks," as if the final standing will prove everything... Come on. We're arguing over statistical theory, we both know result-oriented analysis is useless. The hypothetical 14 point team in the results thread is a perfect example. The final standings will prove very little - only the art of our debate can determine a clear winner  | ||
|
semioldguy
United States7488 Posts
On November 30 2008 06:01 traced wrote: 6- Most people overlooked this one. Teams just score less points per dollar than players. It is actually more efficient to choose a $3 team that wins 4 games over a $4 that wins 6 games, because the extra dollar in players makes up the difference and then some. That SKT will probably go at least .500 this season makes them the obvious choice. Actually teams were priced based on the same point output ratio as players. So any player worth 5 points scored more total points in round one than any team worth 4 or less points | ||
| ||
