Jump to content


Unofficial Fantasy League


  • Please log in to reply
203 replies to this topic

Unknown0ne #161 Posted May 25 2017 - 12:43

    Captain

  • WGLNA Gold League Player
  • 27528 battles
  • 1,104
  • [BULBA] BULBA
  • Member since:
    07-28-2010

View PostTheDukeJ, on May 24 2017 - 17:44, said:

A few quick clarification questions about sat/sun grand finals picking.

Given that were supposed to pick for both days with one roster, this means that 4 teams will play only one set of matches and therefore predictions are incredibly important in picking players correct?
However, what is not too clear are the Sunday matches. Especially regarding a possible 3/4 place match. Do you have any idea if there will be one? (Offline presumably?) This isn't really made clear - but if the prizes break down like previous grand finals ( https://www.esportse...and-finals-2016) it would seem like there has to be? 
Given that this could affect who people pick could this be clarified :D

Thanks! and thanks for doing this unknown!

 

Yes, picking the teams you think will go far in the playoff bracket is an important factor: 4 teams will play 1 match, 2 teams will play 2 matches, and 2 teams will play 3 matches.

 

I'm honestly not sure if there will be a 3/4 place match, but it won't be included for fantasy even if it exists. The matches shown on the fantasy site ( https://fantasytanks...ents/5/schedule ) are the only ones that will be included for fantasy.

 

 

I'm happy to do it; thank you for playing!



TheDukeJ #162 Posted May 25 2017 - 22:35

    Corporal

  • -Players-
  • 9189 battles
  • 24
  • Member since:
    06-07-2014
Gucci

Kebabmunger #163 Posted May 29 2017 - 08:30

    Private

  • Players
  • 17726 battles
  • 8
  • [CAPS] CAPS
  • Member since:
    05-24-2013

when do the prizes get paid out? ty

 



Unknown0ne #164 Posted May 29 2017 - 11:16

    Captain

  • WGLNA Gold League Player
  • 27528 battles
  • 1,104
  • [BULBA] BULBA
  • Member since:
    07-28-2010

View PostKebabmunger, on May 29 2017 - 02:30, said:

when do the prizes get paid out? ty

 

Typically within a week.

BattleshipBob27 #165 Posted Jun 04 2017 - 02:36

    Corporal

  • -Players-
  • 18502 battles
  • 58
  • [SHAM] SHAM
  • Member since:
    09-10-2016
Any word on the prizes.

Unknown0ne #166 Posted Jun 04 2017 - 04:04

    Captain

  • WGLNA Gold League Player
  • 27528 battles
  • 1,104
  • [BULBA] BULBA
  • Member since:
    07-28-2010

View PostBattleshipBob27, on Jun 03 2017 - 20:36, said:

Any word on the prizes.

 

Not currently; I'll look to get an update Monday.

Unknown0ne #167 Posted Jun 05 2017 - 17:12

    Captain

  • WGLNA Gold League Player
  • 27528 battles
  • 1,104
  • [BULBA] BULBA
  • Member since:
    07-28-2010
Fantasy Prizing for the Grand Finals should be paid out within the next few days.

Unknown0ne #168 Posted Aug 15 2017 - 23:23

    Captain

  • WGLNA Gold League Player
  • 27528 battles
  • 1,104
  • [BULBA] BULBA
  • Member since:
    07-28-2010

Improved Fantasy Point Formula

This was the ranked second highest for “What feature would you most like to see added to the Fantasy site?” on the survey held some time ago. Here’s what I’ve come up with.

Preamble

This post covers the very basics about the new formula I’ve created as a candidate for replacing the existing formula. If you want more information, there are a number of more in-depth resources linked at the bottom of this post.

Note: In this post, when I refer to the existing fantasy point formula, I’m referring to a slightly modified version which uses 0.2 (instead of 0.8) for the multipliers for Capture/Defense Points, and only awards points for those metrics to the winning team.

The Existing Formula

First and foremost, the existing formula works pretty well. This is particularly true for larger sets of data (say a season’s worth of data).

The weaknesses of the existing formula, from my point of view, is when looking at the results of an individual battle: There are a number of edge cases it handles rather poorly due to the ratios used, and generally speaking doesn’t really account for the differences in difficulty based on context.

Difficulty

By “difficulty” I’m referring to the fact for a given metric--say damage dealt--there is more to it than just the value the player has at the end of the battle.

To start with, there’s a lot of other factors could have made it harder (or easier) for a given player in a given battle to reach that value. For damage dealt, that might be the player’s tank’s starting HP and the total HP of the enemy team.

There is also the nonlinear difficulty increase as any given metric’s value increases. This is to say, for example, getting from 1,000 damage to 2,000 damage is--generally speaking--harder than getting from 0 damage to 1,000 damage, despite in both cases the increase being the same (1,000 damage).

The New Formula Candidate

Addressing these weaknesses is the primary goal of this new formula, a goal I believe the formula has achieved to some degree:

Difficulty is accounted for by considering context based factors like the player’s tank’s characteristics and the enemy team’s HP pool, in addition to using a nonlinear function to help take into account nonlinear difficulty increases.

Likewise, I have not found any edge cases that would award a large number of fantasy points for a arguably poor performance.

Other Major Differences

Lower average fantasy points. With the existing formula, the average fantasy points earned by a player in a battle was 39, versus 19 in the new formula. The highest average for a player (sans Sublimebanana with 1 battle)--in both cases--is Oxmathus, earning an average of 51.6983 fantasy points in the existing formula, versus 19.4974 in the new formula.

There is a higher relative difference in fantasy points earned between the lowest and highest scoring players in the new formula. With the existing formula, the average points earned for the lowest scoring player is 55.04% of the average for the highest scoring player (Oxmathus), versus 47.76% with the new formula.

In terms of what contributed to the total fantasy points earned, damage dealt (+10%), kills (+3%), and penetration ratio (+1%) all saw upticks with the new formula, while spotting assist (-2%) and damage blocked (-1%) saw reductions, and distance traveled (8%) and damaged tanks (3%) were outright removed.

---

So, those are the basics of the new formula. I’m happy to answer any questions, and interested to know what everyone thinks; does this new formula seem better, worse, or simply different than the existing one? Any thoughts on areas of improvement?

For more details:

Design Details: https://docs.google.com/document/d/11Q3l_V_xiNe9yyekn_d5j9LV1l0LeUni1ElOqz0dcX4/edit?usp=sharing

Exact Formula: https://docs.google.com/document/d/1UhA0DAxOWHhNgMbt58jQwTpDIluw85JOF4o1puUFX2c/edit?usp=sharing

Data based comparision between the two formulas: https://docs.google.com/spreadsheets/d/1mAqtssKO67Y6lnohvpYgcGkNnShp1yiJov4vczKpH7Q/edit?usp=sharing

Case study: https://docs.google.com/document/d/1Ohi_KhE0_ulczOVoG8FYAFiS_k5GmYa8UqXHBpiOs9s/edit?usp=sharing

---

Comparsion tool for use with any replay: https://github.com/JoshuaEN/Fantasy-Point-Tester/releases/tag/v1.0

Here’s the replays from the last WGLNA Live Finals for easy downloading: https://drive.google.com/open?id=0B3ix0OKW45noZUZSakpmbi1zR28

The rest of the past season’s replays are available at: http://wotreplays.com/wgleague#?league=na_gold&season=3

You can also try it out on replays from random battles or what not, though it’s not designed around random battles it tends to work alright from what I’ve seen.

 


Jaspo #169 Posted Aug 17 2017 - 04:22

    Staff sergeant

  • -Players-
  • 23860 battles
  • 294
  • [4HIM3] 4HIM3
  • Member since:
    03-12-2015

-spotting assist reduction might unfairly hurt scouts a bit, unless it was simply too high a value before?

-was damaged tanks removed because it was a bit redundant?

-makes sense to remove distance traveled; certain tanks in certain strats will not be moving as much as others and that isn't a bad thing.

-confused by average vs maximum points for the new formula...am I seeing something wrong? Because it looks like average was 19 but the highest player average was only 19.5?



Wildblade #170 Posted Aug 17 2017 - 04:42

    Staff sergeant

  • Players
  • 40881 battles
  • 274
  • [REL2] REL2
  • Member since:
    02-21-2012
I might scale it a little differently.  With only 0.5 of a point between the best and the mean, and only ~10 points between best and worst (I agree with Jaspo this seems wrong, typo?), people are going to have to look at a lot of decimal places to get anything meaningful.  It might be easier for the end user if you scaled everything by a factor of 2 or 5 or something in between.

Unknown0ne #171 Posted Aug 17 2017 - 22:52

    Captain

  • WGLNA Gold League Player
  • 27528 battles
  • 1,104
  • [BULBA] BULBA
  • Member since:
    07-28-2010

View PostJaspo, on Aug 16 2017 - 22:22, said:

-spotting assist reduction might unfairly hurt scouts a bit, unless it was simply too high a value before?

-was damaged tanks removed because it was a bit redundant?

-makes sense to remove distance traveled; certain tanks in certain strats will not be moving as much as others and that isn't a bad thing.

-confused by average vs maximum points for the new formula...am I seeing something wrong? Because it looks like average was 19 but the highest player average was only 19.5?

 

1. The reduction in spotting assist points will probably hurt light tanks the most. I don't have very much data on tier 10 light tanks (and the next WGLNA season is 7/70), but tier 8 light tanks averaged 109 more spotting assist damage than the next closest (tier 10 mediums) tier/type combination. It seems reasonable that would apply to tier 10 lights as well, to some degree. That said, the old formula effectively awarded the same fantasy points for damage dealt as spotting assist damage, which I would argue is too high because dealing damage, particularly at long range, is typically harder (if safer) than spotting the targets being damaged. The new formula awards roughly half the fantasy points for spotting assist damage as damage dealt.

 

It's also worth noting, with regards to damage dealt, light tanks have an advantage in the new formula due to their lower HP pools. On average, light tanks will still likely underperform compared to the medium counterparts in fantasy points. I suspect in no small part because what light tanks do best (spot tanks) has an incomplete set of metrics. There's tanks spotted, but that only counts the first time a tank was spotted; and there's spotting assist damage, which is a good metric, but still limited as a lot of value in a scout can come from gathering intel even if the team is unable to engage the spotted enemies. But, there's no metric for say the total number of seconds a player spotted enemy tanks, so there's no way to award for that.

 

2. Damaged tanks was removed because, in my testing, I wasn't seeing an advantage (more accurate points, to my view) to having it in there. I wouldn't say it was redundant so much as vague. For example, having more kills is always better (other things equal), but having more damaged tanks isn't so clear cut (is having damaged 3 different tanks universally better than having damaged 4, other things equal?). I do think it would be worth further experimentation in the future, perhaps as a penalty metric (awarding less points if say player damaged just 1 tank in a battle).

 

Also, it wasn't a big influence in the existing formula, accounting for just 2.75% of the total points earned.

 

3. This was one of my main lines of thinking for not using it as a primary metric; I don't want to punish players (in terms of fantasy points) for the strat if I can help it, and distance traveled is generally a metric that is dependant on the role the player has in the strat, rather than their own personal skill.

 

For the sake of clarity, I will note distance traveled is used as a modifier for tanks spotted points. Specifically, it reduces the value of each tank spotted by a player if they don't have much distance traveled in the battle.

 

4. Sorry, that's a typo. The average for the new formula is 15 (technically, ~14.56).

 

 

View PostWildblade, on Aug 16 2017 - 22:42, said:

I might scale it a little differently.  With only 0.5 of a point between the best and the mean, and only ~10 points between best and worst (I agree with Jaspo this seems wrong, typo?), people are going to have to look at a lot of decimal places to get anything meaningful.  It might be easier for the end user if you scaled everything by a factor of 2 or 5 or something in between.

 

You're correct, the average was a typo. The average for the new formula is 15 (technically, ~14.56).

 

Likewise on the point of scaling; a multiplier of 6 seems to be the likely candidate, since it gives a much larger range while keeping roster scores for a day well under 10,000 (which is good, I don't want to get into 5 digits on a day to day basis).

 

 

Thank you both for your feedback.

 

---

 

Also, a point of feedback I received via skype was the complexity of the formula is bad because it makes it (much) harder to mentally convert in-game performance to points while one is watching the stream. For example, with the old formula when a player on your roster got a kill, you knew it was worth 2 points. With the new formula (in 7/70, presuming all 10s on the enemy team), it could be anywhere from 1.25 to 1.67 points per a kill, with the exact value depending upon the total number of kills a player has in the battle.

 

Now, this is a valid criticism; but it's a question of is (in theory, I'd have to play around to see the real-world impact) the reduced accuracy worth it to have a simpler formula? Would perhaps listing low-end numbers be a good enough alternative (e.g. saying each kill is worth 1.25 points or more)? Thoughs?



screng #172 Posted Aug 18 2017 - 05:21

    Sergeant

  • Players
  • 15361 battles
  • 164
  • [_SACK] _SACK
  • Member since:
    12-11-2012

Love this discussion! 

 

1) Everything sounds good to me.  You cant really eval the scout changes under a 7/70 format until you have a pile of data.  All of your logic makes sense tho as to why what youve done makes sense until proven otherwise thru actual play. 

2) Id leave the kills value at the new formula (1.25 - 1.67).  I think the number of players this will affect (no disrespect to your skype friend) is very very small, so from a pragmatic stance there is no real benefit to altering it, but secondly i think making tweaks to a complex formula based on what ill call "peer pressure" is a bad precedent. "peer pressure" is way too strong a phrase but i cant think of a weaker idiom at the moment.  Only a small number of people would even notice.  I think youre the Cook in the Kitchen and id stick with what you created (unless 7/70 assisted and scouting real world data suggests tweaks at the end of the season.) 

 

PS It was yoott wasnt it, he needs those kills rounded down to have a chance against me this season!

 



Unknown0ne #173 Posted Aug 28 2017 - 05:25

    Captain

  • WGLNA Gold League Player
  • 27528 battles
  • 1,104
  • [BULBA] BULBA
  • Member since:
    07-28-2010

So, a few things. A lot of new features have been added to the site in preparation for the next season, including roster reserve slots, automatic roster carry over, and user created groups. The details about all of the new features can be found here: https://fantasytanks...elog-2017-08-27

 

The Tournament page for the next WGLNA season is also up, including the regular season prizing details: https://fantasytanks...ments/7/prizing

 

Also, if you want to try out and help test the new roster management features, while also getting a feel for the new fantasy point formula, there is a (no prizing) Fantasy League for the Qualifiers live right now: https://fantasytanks.com/tournaments/8



Vird_Killer #174 Posted Aug 28 2017 - 18:45

    Sergeant

  • Players
  • 26933 battles
  • 149
  • [SG] SG
  • Member since:
    06-10-2011
Amazing stuff as usual UO.  Looking forward to see how you tweaked the points.

KrafterAnonymous #175 Posted Aug 28 2017 - 23:25

    Corporal

  • Players
  • 7952 battles
  • 14
  • [PACNW] PACNW
  • Member since:
    10-12-2013
Can't wait to test it out today! The one suggestion I would make at this point is adding a button to the team window that allows you to clear all players from your team at once. I think this would be beneficial once the daily roster carryover takes effect, and with all the added reserve slots, it seems like it would be easier to just start with a fresh slate when you need to. Thanks for all the great work on the site, and keep it up! <o

screng #176 Posted Aug 28 2017 - 23:47

    Sergeant

  • Players
  • 15361 battles
  • 164
  • [_SACK] _SACK
  • Member since:
    12-11-2012

I also like the nbuke from orbit button idea.

 

my only concern is that i saw 5 reserve slots, expecting one.  Wouldnt it be possible for people (esp if they had multiple accts) to just pick 1-2 top players from each team, fill up their twelve slots, and never change anything? w auto roster carry over, it would always use the best 7 of the 12, which would generate some nice scores (i think?) if you just set it to known performers and then never came back. 

 

I love the automation and idea of making it more friendly to actual players.  i am not so keen to have "bots" just trying to farm prizes. 

 

i.e this is worrying me it will be like fantasy football where you lose to a team that someone forced their significant other to make, they havent edited their roster since draft, have 3 players not playing, but one of their wide receivers has an epic day and you lose to a player who hasnt logged in in a month.

 

 


Edited by screng, Aug 28 2017 - 23:49.


Tanks__ #177 Posted Aug 29 2017 - 00:17

    Corporal

  • Players
  • 27956 battles
  • 60
  • [MAHOU] MAHOU
  • Member since:
    09-12-2011

View Postscreng, on Aug 28 2017 - 14:47, said:

my only concern is that i saw 5 reserve slots, expecting one.  Wouldnt it be possible for people (esp if they had multiple accts) to just pick 1-2 top players from each team, fill up their twelve slots, and never change anything? w auto roster carry over, it would always use the best 7 of the 12, which would generate some nice scores (i think?) if you just set it to known performers and then never came back. 

 

Its not the best 7 of 12 that get used, if the main 7 you chose play all their battles then you get their points, even if someone in the reserves did better. The reserve system is just in case someone didnt play all the matches so you dont get screwed out of points because they didnt play

screng #178 Posted Aug 29 2017 - 00:56

    Sergeant

  • Players
  • 15361 battles
  • 164
  • [_SACK] _SACK
  • Member since:
    12-11-2012

View PostTankDestroyer590, on Aug 28 2017 - 17:17, said:

 

Its not the best 7 of 12 that get used, if the main 7 you chose play all their battles then you get their points, even if someone in the reserves did better. The reserve system is just in case someone didnt play all the matches so you dont get screwed out of points because they didnt play

 

Ahh yes i knew that, my reply was unclear sorry.  Let me try again.  

 

With 12 slots, *we now have more slots than teams*. I pick Oxmathus, or the Oxmathus equiv, from each team, at the start of the season, and never change my roster.  I am guaranteed to always have the #1 player from each team in my roster even if i have no idea which teams are playing that night.

  

 

 

 



Unknown0ne #179 Posted Aug 29 2017 - 01:31

    Captain

  • WGLNA Gold League Player
  • 27528 battles
  • 1,104
  • [BULBA] BULBA
  • Member since:
    07-28-2010

View PostKrafterAnonymous, on Aug 28 2017 - 17:25, said:

Can't wait to test it out today! The one suggestion I would make at this point is adding a button to the team window that allows you to clear all players from your team at once. I think this would be beneficial once the daily roster carryover takes effect, and with all the added reserve slots, it seems like it would be easier to just start with a fresh slate when you need to. Thanks for all the great work on the site, and keep it up! <o

 

Automatic roster carry will happen after rosters for a day have already locked, specifically during scoring. As such, your roster for a newly unlocked match day won't be prefilled by automatic roster carry over (though there will be a link you can click to copy over your previous roster, if you want).

 

Regardless, it wouldn't hurt to have a clear roster button.

 

View Postscreng, on Aug 28 2017 - 17:47, said:

I also like the nbuke from orbit button idea.

 

my only concern is that i saw 5 reserve slots, expecting one.  Wouldnt it be possible for people (esp if they had multiple accts) to just pick 1-2 top players from each team, fill up their twelve slots, and never change anything? w auto roster carry over, it would always use the best 7 of the 12, which would generate some nice scores (i think?) if you just set it to known performers and then never came back. 

 

I love the automation and idea of making it more friendly to actual players.  i am not so keen to have "bots" just trying to farm prizes. 

 

i.e this is worrying me it will be like fantasy football where you lose to a team that someone forced their significant other to make, they havent edited their roster since draft, have 3 players not playing, but one of their wide receivers has an epic day and you lose to a player who hasnt logged in in a month.

 

The 5 reserve slots is just for the Qualifiers to try and test the full set of functionality the reserve system can support (also because there's basic zero info on which players for each team will be playing). For the actual season, rosters will have just 1 reserve spot.

 

 

As to your second reserve related point, the reserve isn't just additional rosters spots, with the top 7 fantasy point earners (of the entire roster) being counted for scoring.  Rather, the reserve only comes into play if 1 (or more) players on the primary roster doesn't play in all of the battles their team did for the day. Even then, reserve spots are checked in order; if, for example, just 1 player on your primary roster hasn't played in all of their battles, and the 1st player on your reserve (the player closest to the "Reserve" divider) did play all of their battles, the system will completely ignore the rest of the players on your reserve. Thus, you can't just cheese the system by picking someone you know won't play in your primary roster to automatically get the best scoring fantasy player in your reserve.

 

Basically, the entire reserve system is intended to help alleviate one very specific problem: Having someone on your roster who doesn't play in all battles for their team on a match day, which in turn really hurts your roster for the night. I have done my best to close off any loopholes that would allow players to use the reserve system, to their benefit, for something other than that. I may not have succeeded, only time will tell, but doing so was one of my primary design goals for the reserve system.

 

 

As to automatic roster carry over, given the above, I don't think that is going to be a major issue: A fire and forget roster is, on average, going to underperform compared to the rosters from a player who is creating a roster specifically for each match day. That said, players who set their roster once and don't return are probably going to win some prizing over the course of the season, especially with the depth of the daily prizing; but, someone who takes the time to make a roster each match day should win more prizing more consistently.

 

"With 12 slots, *we now have more slots than teams*. I pick Oxmathus, or the Oxmathus equiv, from each team, at the start of the season, and never change my roster.  I am guaranteed to always have the #1 player from each team in my roster even if i have no idea which teams are playing that night."

 

First, as a note, in the upcoming season every team is playing every match day.

 

I don't think picking the #1 player from each team would be a winning strategy, versus rosters which were picked with other factors in mind as well, like the likely score line (shorter battles equate to less points) and which team will ultimately win (the players on the losing team tend to score fewer points). To that point, last season even Oxmathus wasn't in the top 7 every night he played.

 

 

Mind, this is all conjecture on my part at the moment. Once the season is has been going for a bit, the data should help clarify what impact automatic roster carry over is having.



Unknown0ne #180 Posted Aug 29 2017 - 09:55

    Captain

  • WGLNA Gold League Player
  • 27528 battles
  • 1,104
  • [BULBA] BULBA
  • Member since:
    07-28-2010

Monday's rosters have been scored for the quals league. Thank you everyone who made a roster. 

 

Everything appears to have worked as intended The rosters for Tuesday have been unlocked.

 

 

Having a day scored now, there's a few things of note. Firstly, pretty much everyone who made a roster for Monday with now have a "Roster has Issues" notice for their Fantasy roster, since automatic roster carry over is enabled and the system detects your Monday roster is suboptimal for Tuesday (because half the teams were eliminated, so if you have anyone from one of those eliminated teams, that's a wasted roster spot).

 

Second, you can see the fantasy points per a player per a battle for the two scored matches: https://fantasytanks...s/8/matches/231 and https://fantasytanks...s/8/matches/232

The site doesn't have a per-a-metric breakdown of fantasy points at the moment, but if you have any questions about the breakdown of any of the fantasy point scores, just let me know as I can get that information. Related to this, I plan on doing a full writeup on the new fantasy point formula, what's changed from the previous draft, and data from the quals replays so far in the morning.

 

That said, I am interested in what everyone's "gut feeling", if you will, is in regards to the point values players received.

 

 

Third, there's now some nice examples of how the reserve system works: https://fantasytanks...ters/1000327843 which shows reserve players not being swapped in due to player per a team limits, https://fantasytanks...ters/1006322191 which shows reserves being completely ignored if a player has a valid top 7, and https://fantasytanks...ters/1000040901 which shows the order reserve list only being searched until a player with 100% battles played (and would keep the roster valid if swapped in) is found.






3 user(s) are reading this topic

1 members, 2 guests, 0 anonymous users


    vetro_xD