3000+ Skill Rating Data and Analysis (now including DCs)

someone did the math on here and there are about 460 ways to cycle 11 different players to make a match for you. So once you’ve played about 700 games, your SR can be relatively certain to be relatively accurate :slight_smile:

Kaawumba, do you think, there is any point in analyzing the pattern in team SR difference? I can clearly see it, judging by my data and your data from top guys’ matches. Top guys certainly have some insane team SR differences, but in lower ranks these 30 average SR play a huge role, with PBSR in effect.
Maybe there is some math in play?

Whats the use for this exactly?

The team SRs are averages of the public SRs, so they include decay. As such, they are not as useful in diamond+ as we might hope. I still record them anyways, because I can. I look at and analyze them from time to time, but haven’t found anything interesting enough to post.

Your data I am interested in, but the error bars are still too big to demonstrate the effect that I’m looking for, unfortunately. I’m continuing to follow and watching as the error bars shrink with more data.

20 more characters…

Ok. I’m already in placements on new account. I wonder how will they turn out :slight_smile:

Here’s what I was talking about. If everyone is forced to use this system, the worst that can happen is that it will all equal out in the end and they wasted a bunch of money and queue times are longer. but aside from that, it’s a win win :slight_smile:

My proposal:

If I win a game of Comp, give all five of my teammates 1 point.
If I loose a game of Comp, give all five of my teammates negative 1 point.
When queueing prefer people with the most points.
After in queue for 5 minutes, disregard all points and find me a match ASAP.

This could all work in the background, or we could be shown what our teammates’ current point tally is.

You can opt out of giving points, but can’t opt out of having points given to you.
You can toggle the queue time thingy too. Like make it 5,10 or 15 minutes, or forever.

On my alt account,

I started to gain 50 sr per win for the last 3 games, I am on a 10 game winning streak but I don’t remember matchmaking giving me this much sr even on a streak since season 3, is the modifier for winning streaks back? Any news on this? I thought winning streak modifiers were negligible, since season 3. Keep in mind this is diamond with no performance based sr.

This is probably why it won’t happen. High cost, small benefit. Note how they moved from two to three “avoid this player” and acted like it is a big deal. They are very worried about queue times going up. Whether or not this fear is justified is something I can’t see with out Blizzard’s internal data.

1 Like

How long has it been since you’ve played on that account? Inactive accounts get their uncertainty modifier bumped up so you gain/lose more per game. And if you keep winning, I expect that uncertainty modifier remains high for a while.

It’s not streaks. See Streamer Data - Google Sheets → Dmum 11, which was a new account, and has 18 wins in a row, with steadily decreasing gains as his account loses the new account uncertainty. If you give me your SR after each game this season, I may be able to say more.

All that hard work means nothing, because everyone’s still hardstuck in plat :smile:

LULxD Your hard work and knowledge means nothing because it’s 2018, OP!

It has 32 hours of comp history but with really good stats and I skipped two or 3 seasons.

My friends started getting 50 sr per win too, perhaps my sr is trying to catch up to my mmr since my stats are so high?

I averaged 37 sr per win yesterday. Super strange in diamond.

This is why. Reread my first reply to you, and let me know if and how it is not clear. As you play more and your rating uncertainty goes down, your gains will fall down to the normal +/- 24.

No. As you’ve said, performance doesn’t matter in diamond+.

I’d ask your friends the same questions I asked you.

I don’t think its new enough I played last season 10 plus hours or so, before I skipped 2 seasons. You’re probably right, every time I place its a career high and I did gain 30 sr per win during placements. However if uncertainty is the case I will lose a lot of sr next time I lose, if I don’t I think it’s a simple case of mmr being substantially higher than sr, and I’m getting additional sr per win. Correct or no?

Yes, your losses should be larger for a time as well.

K I will see when I finally lose xD, otherwise something might be up.

Thanks to Pandaman, I have data on disconnects, and it looks the same as decay. Details above.