|
On November 05 2013 04:13 anessie wrote:Show nested quote +On November 05 2013 03:59 Mendelfist wrote: As far as I know bonus pool doesn't have anything to with ladder placement, except for grandmaster league. What's your source? Just deducting that from this topic and http://www.teamliquid.net/forum/viewmessage.php?topic_id=423477Being inactive = growing bonus pool = mmr decay? Plus personally getting demoted two leagues by having a 1750+ bonus pool at the end of previous season? Bonus pool as far as we know does not affect decay, but the length of the inactivity period does. Please note that you cannot measure time with bonus pool increase as during ladder lock period the bonus pool does not increase.
The minimum time period after the decay kicks in is not known at the moment. It is known that accounts that have been inactive for 3 weeks or more are often affected by the decay. Funny thing regarding this system is that if you don't have time to play for several weeks, it is likely better to lose 1 game e.g. by leaving instantly per week to avoid possible decay after the inactivity period (of course insta-leaving games could be considered 'MMR abuse', but in this case it is the 'lesser evil' and they will not ban for 1 game). Couple of losses drop your MMR less than e.g. 3 weeks worth of decay.
I still haven't had time to look at S14 data more carefully (and check accounts that had less than 3 weeks of inactivity). But some of the tools I developed to speed up the analysis were added to the public release of the MMR tool yesterday.
|
On November 05 2013 04:49 korona wrote: The minimum time period after the decay kicks in is not known at the moment.
Good point, did not think about it like that.
Since I only did my placement match after the season reset and never played for 9 weeks till a few games during ladder lock I probably had a lot of decay.
|
On November 05 2013 04:54 anessie wrote:Show nested quote +On November 05 2013 04:49 korona wrote: The minimum time period after the decay kicks in is not known at the moment.
Good point, did not think about it like that. Since I only did my placement match after the season reset and never played for 9 weeks till a few games during ladder lock I probably had a lot of decay. If you played your placement match when this season began and have not played after that, your account will likely face maximum decay, which is little over MMR range of one league (around 300 points in MMR tool scale) as far as we know at the moment.
|
On November 05 2013 04:13 anessie wrote:Show nested quote +On November 05 2013 03:59 Mendelfist wrote: As far as I know bonus pool doesn't have anything to with ladder placement, except for grandmaster league. What's your source? Just deducting that from this topic and http://www.teamliquid.net/forum/viewmessage.php?topic_id=423477Being inactive = growing bonus pool = mmr decay? Plus personally getting demoted two leagues by having a 1750+ bonus pool at the end of previous season?
The MMR league thresholds are set manually by Blizzard so that each league contains a certain percentage of "active" players. That does NOT mean that the activity metric has any influence on ladder placement. It's just a goal, and they don't necessarily always achieve it.
We don't know exactly how MMR decay works, and I haven't heard of any indications about a connection to bonus pool.
|
On November 05 2013 04:10 Thieving Magpie wrote: David Kim gave out a "number" for how many points it takes to get promoted by the end of a season. He said it early 2010. You misunderstood him. That was just a rule of thumb, and it could only be used if bonus pool was spent. That doesn't mean that spending bonus pool is a requirement or even affects your chances of promotion. It doesn't. We have a pretty good knowledge about how the ladder works now thanks to the guys developing the MMR Tool plugin.
|
United States12181 Posts
The connection between decay and bonus pool is speculative because they have shown in the past to measure inactivity through the bonus pool. Korona's stat breakdown, which hopefully will happen eventually (!) should show us whether there is conclusive evidence to support it.
Now that the leagues are locked I will request another data dump from Shadowed for this season and run through the league apportionment numbers.
|
On November 05 2013 05:27 Mendelfist wrote:Show nested quote +On November 05 2013 04:10 Thieving Magpie wrote: David Kim gave out a "number" for how many points it takes to get promoted by the end of a season. He said it early 2010. You misunderstood him. That was just a rule of thumb, and it could only be used if bonus pool was spent. That doesn't mean that spending bonus pool is a requirement or even affects your chances of promotion. It doesn't. We have a pretty good knowledge about how the ladder works now thanks to the guys developing the MMR Tool plugin.
You misunderstand me.
Being that I said that if you have a high enough winrate over time you can ignore the bonus pool, the statement he made was directed and non-grinders to give them an estimate of how much better they need to be.
Enough games above 60% winrate will promote you, or a shit tonne of games with an average of about 60% (usually eating up your bonus pool)
Hence why I called it a gauge, not a determining factor.
If you play so few games that you can't eat up the bonus pool, you need a high winrate. If you play so few games that you simply stay on par with the bonus pool, you need about 60% winrate If you play so many games that you always have zero bonus pool, then you can't use the bonus pool as a measurement of anything.
Win 1-2 games a day and you'll stay on par with your bonus pool.
|
On November 05 2013 07:02 Thieving Magpie wrote: If you play so few games that you can't eat up the bonus pool, you need a high winrate. If you play so few games that you simply stay on par with the bonus pool, you need about 60% winrate. I'm sorry, but you are just making numbers up. There is no support for anything of that.
|
On November 05 2013 15:23 Mendelfist wrote:Show nested quote +On November 05 2013 07:02 Thieving Magpie wrote: If you play so few games that you can't eat up the bonus pool, you need a high winrate. If you play so few games that you simply stay on par with the bonus pool, you need about 60% winrate. I'm sorry, but you are just making numbers up. There is no support for anything of that.
If you win one game, you have a 100% winrate starting at that game. However, that doesn't mean you will be promoted. If you win 3 games and lose 1, you win 75% of your games. However, that doesn't mean you will be promoted.
How many games do you need to win? Depends on your MMR. How many games can you afford to lose? Depends on your MMR.
If you play often enough that you're able to outpace your bonus pool and end up with about a 60ish% winrate, you will gain the points necessary by the end of a season to be promoted (according to David Kim)
Like I said, its a gauge. Play a bunch of games after you zero out your bonus pool, then the math doesn't work. Have too high a winrate over a large number of games, then you don't have to care about how much bonus pool you have left since you'll be promoted anyway.
|
On November 05 2013 15:32 Thieving Magpie wrote: If you play often enough that you're able to outpace your bonus pool and end up with about a 60ish% winrate, you will gain the points necessary by the end of a season to be promoted (according to David Kim). When did David ever say anything about 60% win rate? That's a new one to me. If you are just speculating, just say so. What you are saying is not established fact, and this: "It was a dishonest number because to get that many points plus having zero bonus pool you needed to have about a 60% winrate otherwise you'd be promoted from winning too many games in a row." Is completely false, especially back then when MMR decay didn't exist. Having enough points just meant you had high MMR, and a high win rate is not a prerequisite for having a high MMR.
|
On November 05 2013 16:50 Mendelfist wrote:Show nested quote +On November 05 2013 15:32 Thieving Magpie wrote: If you play often enough that you're able to outpace your bonus pool and end up with about a 60ish% winrate, you will gain the points necessary by the end of a season to be promoted (according to David Kim). When did David ever say anything about 60% win rate? That's a new one to me. If you are just speculating, just say so. What you are saying is not established fact, and this: "It was a dishonest number because to get that many points plus having zero bonus pool you needed to have about a 60% winrate otherwise you'd be promoted from winning too many games in a row." Is completely false, especially back then when MMR decay didn't exist. Having enough points just meant you had high MMR, and a high win rate is not a prerequisite for having a high MMR.
Win rates affects MMR
If you are at 50% you won't move. If you win 70%-100% of your games you'll be promoted so long as it involves a decent number of games or enough games over a longer period of time.
David Kim gave specific numbers for each league in WoL, but when I did the math of how often you needed to win to get those points, it landed somewhere between 55%-65% depending on league.
My math assumed that you stopped playing that day once you hit zero points and only played when you had a bonus pool available to use. When you did, it wouldatch up with David Kims numbers.
Which is why I call it a gauge. The bonus pool does not causate with promotions, but it is possible to use it as a gauge if you only play a few times a day. If you play multiple games past zero points to earn extra points or go in a massive losing/winning streak then the math gets off completely. The gist is that if you win about 2 out of 3 games per day you will matc up with David Kim's numbers with an above 50% win rate and a non-100% win rate preventing you from automatic promotion.
|
trying to get my NA account back into diamond, which has proven harder than I first thought, cause of all the high level players hanging out in gold and plat lol.
|
On November 05 2013 23:21 Thieving Magpie wrote: If you are at 50% you won't move.
Yes you will. Your opponents MMRs are randomly distributed around your MMR, and the sum of a random series with the mean of 0 is not 0. It will diverge. The average will converge, but the sum will not. Your MMR is a sum. But I guess that's beside the point. The point is that here is no specific win ratio that will get you promoted. 50.1% will get you promoted eventually. It will take longer than 50.5%, but you will get there.
On November 05 2013 23:21 Thieving Magpie wrote: My math assumed that you stopped playing that day once you hit zero points and only played when you had a bonus pool available to use.
Isn't that a little bit contrived? In any case, Davids rule of thumb, or the win ratio to be able to use it, is useless and pointless. It applied only to a specific season with a specific length and specific league offsets. You can no longer use the same numbers. Back then we also had division offsets, which made it even more useless.
|
Players like me, high plat, seem to only play ex-diamonds and ex-masters. Dunno why, but honestly, the games seem fair and not one sided.
|
On November 06 2013 01:00 Mendelfist wrote:Show nested quote +On November 05 2013 23:21 Thieving Magpie wrote: If you are at 50% you won't move.
Yes you will. Your opponents MMRs are randomly distributed around your MMR, and the sum of a random series with the mean of 0 is not 0. It will diverge. The average will converge, but the sum will not. Your MMR is a sum. But I guess that's beside the point. The point is that here is no specific win ratio that will get you promoted. 50.1% will get you promoted eventually. It will take longer than 50.5%, but you will get there. Show nested quote +On November 05 2013 23:21 Thieving Magpie wrote: My math assumed that you stopped playing that day once you hit zero points and only played when you had a bonus pool available to use.
Isn't that a little bit contrived? In any case, Davids rule of thumb, or the win ratio to be able to use it, is useless and pointless. It applied only to a specific season with a specific length and specific league offsets. You can no longer use the same numbers. Back then we also had division offsets, which made it even more useless.
My math only applies to season locks because that was what his commentary was responding to at the time.
And yes, over infinite time, 50.1% winrate will get you promoted for the same reason having a 100% win ratio over a short period of time will not get you promoted.
David Kim's comment only matched season lock timeframes wherein he said that if you get X points over 1 season with 0 bonus pool--that it will promote you.
if you don't use up your bonus pool--his numbers don't match. If you do use up your bonus pool--but then play a significant number of games afterwards, the numbers don't match. If your winrate goes to about 70% or higher, you will start getting promoted before the season is over--and hence the numbers don't match. If you win equal to or less than 50%--then the ladder will not move you.
60ish% was the winrate I saw needed to be above the 50% mark and at the same time, have enough games to finish the bonus pool before a season ends. The correlation does not causate--as I have said. It is best used as a gauge and is only relevant to people who don't play very often but are able to play consistently.
The only thing MMR decay adds is that people can no longer play intermittently without drawbacks. Consistency is now more important than streaks.
|
Well this ruins my plan for coming back after inactivity and ista-GMing it... sigh
|
On November 05 2013 23:23 JacobShock wrote: trying to get my NA account back into diamond, which has proven harder than I first thought, cause of all the high level players hanging out in gold and plat lol.
Yea same here...
|
i just started playing again a couple days ago after ~ 2 years of only dota. i was high platinum last time i played regularly, now im bronze xD. truth is i suck really hard but im still winning quite a lot just cause 70% of my opponents fail their cheese. i agree with the mmr decay, it makes sense.
|
United States12181 Posts
On November 06 2013 01:13 Thieving Magpie wrote:Show nested quote +On November 06 2013 01:00 Mendelfist wrote:On November 05 2013 23:21 Thieving Magpie wrote: If you are at 50% you won't move.
Yes you will. Your opponents MMRs are randomly distributed around your MMR, and the sum of a random series with the mean of 0 is not 0. It will diverge. The average will converge, but the sum will not. Your MMR is a sum. But I guess that's beside the point. The point is that here is no specific win ratio that will get you promoted. 50.1% will get you promoted eventually. It will take longer than 50.5%, but you will get there. On November 05 2013 23:21 Thieving Magpie wrote: My math assumed that you stopped playing that day once you hit zero points and only played when you had a bonus pool available to use.
Isn't that a little bit contrived? In any case, Davids rule of thumb, or the win ratio to be able to use it, is useless and pointless. It applied only to a specific season with a specific length and specific league offsets. You can no longer use the same numbers. Back then we also had division offsets, which made it even more useless. My math only applies to season locks because that was what his commentary was responding to at the time. And yes, over infinite time, 50.1% winrate will get you promoted for the same reason having a 100% win ratio over a short period of time will not get you promoted. David Kim's comment only matched season lock timeframes wherein he said that if you get X points over 1 season with 0 bonus pool--that it will promote you. if you don't use up your bonus pool--his numbers don't match. If you do use up your bonus pool--but then play a significant number of games afterwards, the numbers don't match. If your winrate goes to about 70% or higher, you will start getting promoted before the season is over--and hence the numbers don't match. If you win equal to or less than 50%--then the ladder will not move you. 60ish% was the winrate I saw needed to be above the 50% mark and at the same time, have enough games to finish the bonus pool before a season ends. The correlation does not causate--as I have said. It is best used as a gauge and is only relevant to people who don't play very often but are able to play consistently. The only thing MMR decay adds is that people can no longer play intermittently without drawbacks. Consistency is now more important than streaks.
Do you have a link to these Dayvie posts? As I recall he's only ever gone over the ladder system very briefly, with most of the posts being written by the (now former) systems designer. Not to suggest he's out of touch with how the ladder operates, but he wouldn't be the main point of contact for the explanations behind it because his position is balance designer.
Win rate only affects MMR in that when you win, your MMR increases, and when you lose, it decreases. There's no forced 50% win rate, there's no win rate requirement for promotion, there's no mystique behind it. MMR is just a number and it goes up and down based on the gap between your number and your opponent's number. You have an equal chance at facing someone 1% higher than you as 1% lower than you, an equal chance at 5% higher versus 5% lower, and an equal chance at 8.235% higher vs 8.235% lower. Let's say that by luck (maybe it's late at night on the server and not many people are queued), you faced the same guy three times in a row and his MMR is lower than yours with the gap predicting that you have a 75% chance to win, and the outcome was 2-1 in your favor. Your MMR would be net negative because you might earn +8 +8 -24 MMR, leaving you at -8 from where you started even though your win rate in that series was 66%.
The promotion boundaries are also just numbers which Blizzard changes themselves based on their internal activity metric, which we know is based on remaining bonus pool. If they say "let's change the league values based on everyone who has less than 4 weeks' worth of bonus pool" then they would run their algorithm to find out what number fits where 2% of the population from last season lands, and where 20% lands, and where 40% lands, and 72%, and 92%. There is no specific targeting of players who don't meet the activity requirement, although we would commonly see posts from players saying "well the leagues are based on active players and I wasn't active so that's why I didn't get promoted." The MMR decay function operates independently though, so while that could be a contributing factor in HotS toward a missed promotion opportunity, it's not why that happened in the past.
|
It would be nice if there were an official Blizzard post about this. As it is, there is a lot of speculation.
|
|
|
|