18 replies to this topic

[Paul Inouye]

Update for today...

From what it looks like and what QA is currently reporting on internal servers... PTS has a high probability of going up on Monday. The ONLY thing holding it back right now is a double check on the remote servers for the Error 63 issue. IF PTS goes up, it will be late afternoon. I'll update on Monday as QA does their testing.

That being said.. I mentioned a potential inclusion into the FP Update Match Maker yesterday.

That new item is:

Elo will be a unique ZERO SUM Elo used explicitly in FP. PSR/Tiers/SSR etc are no longer going to be part of the FP Match Maker.

The only gamespace that will affect your FP Elo is Faction Play. Quick Play, Solaris, Comp will not affect your FP Elo. There is no special seeding algorithm and everyone is going to start at square one. Your FP Elo will be compared against the average of the opposing team and your movement will be dependent on win or loss as a pure Elo system is supposed to.

Does match score still affect your FP Elo? No it will not.

This FP Elo will not be in PTS however it will be in the upcoming full patch in May.

More details on this will be updated in the next week.

That's it for now. Have a good long weekend (to whom it applies) and will be back on Monday.

-Paul

[justcallme A S H]

Quick Play PSR will not feature or contribute in any way to the FP ELO? And it'll be zero sum?

HALLELUJAH

Now that's an update and a half.

[braveheart95]

Nice!

[VigorousApathy]

Paul Inouye, on 18 April 2019 - 05:18 PM, said:

Elo will be a unique ZERO SUM Elo used explicitly in FP. PSR/Tiers/SSR etc are no longer going to be part of the FP Match Maker.

Bless you paul! Bless you!

[Yondu Udonta]

REJOICE now I can stay at tier 4

[Bishop Six]

"The only gamespace that will affect your FP Elo is Faction Play."

Ha!

Yondu, time is over for Tier 4 trash talk

Good Work ButPaulNotBut, Good Work (so far)!


Edit:
Please, can anyone explain to me what "zero sum" Elo means?

Edited by Bishop Six, 19 April 2019 - 12:04 AM.

[Roland09]

Bishop Six, on 19 April 2019 - 12:01 AM, said:

Edit:
Please, can anyone explain to me what "zero sum" Elo means?

From my limited understanding, the winning party goes up by the amount the losing side goes down. You won't get even a tiny bit up on a loss, no matter your individual score.

[xX PUG Xx]

Great news on both the PTS and the match maker.

I need to stop dropping solo against the murder groups now. Or do I want my ELO to tank? Hmmm... tricky one

[Lotspeech]

So it will be based solely off wins?

[Grus]

Oh thank you baby butpaul

[Khalcruth]

Interesting. Will the new ELO be visible on the leaderboards?

[B0oN]

Colour me impressed .
Good move, SirBut, errr I mean Paul

[xX PUG Xx]

kamichiwa343, on 19 April 2019 - 05:15 AM, said:

So it will be based solely off wins?

Yes, that is my interpretation.

Khalcruth, on 19 April 2019 - 06:18 PM, said:

Interesting. Will the new ELO be visible on the leaderboards?

This is a good question....... nothing to do with epeen shining, oooooh no! Purely for personal interest you understand?

But it is a good question, sarcasm aside.

Edited by xX PUG Xx, 20 April 2019 - 02:06 AM.

[SilentScreamer]

Bishop Six, on 19 April 2019 - 12:01 AM, said:

Edit:
Please, can anyone explain to me what "zero sum" Elo means?

Roland09, on 19 April 2019 - 01:16 AM, said:

From my limited understanding, the winning party goes up by the amount the losing side goes down. You won't get even a tiny bit up on a loss, no matter your individual score.

Zero sum also typically means whatever one player looses in point value the other player gains. So say boths sides have a ELO rank of 100 or 1,000 or 10,000. If the winners get +10, +100 or +1,000 for the win, the other side will get the same amount subtracted from their ELOs.

This is a great system for static populations which engage in regular play. I am not sure it is a good system for a gamemode which is prone to large population fluctuations and irregular experience. For example:
- Team A plays regularly with some success, a 50/50 win ratio against equal ELO opponents should keep your ELO close to starting value in a zero sum system.
- Team B has a bunch of rookies who have never dropped in Faction Play.
The matchmaker considers them equal since their ELOs are similar. Who thinks the match will go well for Team B?

However, while quality of life at the "middle" will not be great, lower-end ELO and top ELO teams should benefit greatly, ASSUMING there is a sufficient population close to their own ELO in queue at any given time.

It would be nice if there was also an "experience" factor to prevent the mismatch case above. Teams with less than 10 faction play matches per player would be matched seperately from those with more than 10 average matches per player. But that is asking a lot of the matchmaker.....

Edited by SilentScreamer, 20 April 2019 - 07:28 AM.

[Grus]

xX PUG Xx, on 20 April 2019 - 02:06 AM, said:

Yes, that is my interpretation.



This is a good question....... nothing to do with epeen shining, oooooh no! Purely for personal interest you understand?

But it is a good question, sarcasm aside.

It would be nice to know so you can see how bad the MM boned you over when you lose.

[xX PUG Xx]

Grus, on 20 April 2019 - 07:47 AM, said:

It would be nice to know so you can see how bad the MM boned you over when you lose.

Or hard you carried ..... ep33n man, ep33en!!

[SmokeGuar]

Depending how FP Elo forms... playing mode, ie match score does not affect it?

Only either winning or damage + kills?

[Khalcruth]

Roland09, on 19 April 2019 - 01:16 AM, said:

From my limited understanding, the winning party goes up by the amount the losing side goes down. You won't get even a tiny bit up on a loss, no matter your individual score.

I would also note that, normally, the amount you go up / down is dependent on the difference in ELO of you and your opponent. For example

Two people of similar ELO play against each other. The winner goes up a moderate amount, and the loser goes down the same amount

You have a person with a very high ELO play against a very low ELO person, and the person with the high ELO wins. The high ELO will go up only a very small amount, and the person with the low ELO will go down only a small amount. This is because this result was largely expected.

You have a person with a very high ELO play against a very low ELO person, and the person with the low ELO wins. The low ELO person will go up a very large amount, and the person with the high ELO will go down a very large amount. This is because this result was very unexpected.

I will note that it is somewhat possible to game this system to get an extremely high ELO - as long as you win at least 50.00000000000001% of your games, on average your ELO will increase. By playing an infinite number of games, you can theoretically get an infinitely high ELO.

I am also curious as to how having 12 different players per side will work. All the games I have ever used this system for were one on one games. So it was quite easy to figure who had the better ELO. With 12 people per side, I'm not sure what you use - is it a straight average, or perhaps some sort of weighted average? Does the type of game mode count? For example, just because I can use a stealth locust to smoke dive in scouting and get a lot of wins, that doesn't mean I will be able to fight my way out of a paper bag in a straight up 12 vs 12 fight.

Also, will everyone on the winning / losing team have the same ELO change? For example, when comparing ELOs, it may have been a gigantic upset for me to win a game, but for my teammate it may have been completely expected. Do I go up a lot, and he goes up a little? Or do we both go up a moderate amount?

Edited by Khalcruth, 20 April 2019 - 02:06 PM.

[RJF Volkodav]

Paul Inouye, on 18 April 2019 - 05:18 PM, said:

Your FP Elo will be compared against the average of the opposing team and your movement will be dependent on win or loss as a pure Elo system is supposed to.

That sounds good. I hope you consider spreading ELO between players by their performance?

I.e. (bad example) if match result causes 1200 ELO to be added to winners and 1200 ELO to be sub from losers then would it just add +100 to each winner and -100 to each losser? In that case that wont be an adequate system. Example - player 1 did 15 kmdd/kills and carryed a team, other one did 0 kmdd/kills and still gets the same amount. After like 10 games together potatoe guy will get lots of ELO and when players who carried leave he will have problems when matched against higher ELO players. So in this case system carries bad players by a good ones. Mybe after some months of playing that will converge, but seems like it allows tonns of ELO fluctuations for lower skill players.

If system can take into account match outcome for a players then the sum of ELO for a match could be divided between players accourding their influence. Matchscore seems to be a bit bugged and can be farmed by not lossing a mechs so i think measure like KMDD should be used to avoid system exploits. I.e. if match result causes 1200 ELO to be added to winners and 1200 ELO to be sub from losers then:
1) Part of ELO could be randomly distributed between players. I.e. each player from a winning team gets 200/12 ELO boost for a win.
2) Other part of ELO left could be distributed according match results - if player 1 of winning team have 12 KMDD of 48 then he gets additional 1000/4 = 250ELO, players 2 having 0KMDD gets no additional ELO as for low match outcome influence.

The same could be applied for a loosing team subtracting more for a loss without influence and less if you got some good results.