17 replies to this topic

[Cato Phoenix]

Hi guys, do you all have any insights as to the new match score algorithm for QP? Like, what contributes what? Some of the games I've had recently have had me scratching my head, and I don't think its as simple as majorly damage + kills/assists any longer.

[Dr Cara Carcass]

Would like to know that too, strange low matchscores in games with 1k dmg....

[Tarl Cabot]

http://mwomercs.com/...ed-for-science/

Nothing recent but the above thread made my Mcgral18 might be used to get a gist of it (his words ), especially since Private Lobbies no longer require Premium Time. This was back in late 2015 into early 2016.

edit. Since either Sept or Oct patch some have noted some oddities with match scores but nothing had been noted in patches of any changes.

Edited by Tarl Cabot, 25 November 2017 - 02:58 PM.

[Dr Cara Carcass]

Tarl Cabot, on 25 November 2017 - 02:50 PM, said:

http://mwomercs.com/...ed-for-science/

Nothing recent but the above thread made my Mcgral18 might be used to get a gist of it (his words ), especially since Private Lobbies no longer require Premium Time. This was back in late 2015 into early 2016.

edit. Since either Sept or Oct patch some have noted some oddities with match scores but nothing had been noted in patches of any changes.

Yeah i noticed that too as i said. Wouldnt be the first tiem something got changed without beeing mentioned by pgi. Thank for the answer, but i was aware of the thread you linked. I can add that the low match scores arent consistent, i can pull games that are identical withhin 50 damge or so and have +-1 or two difference in areas that are meaningfull for match score. Sometimes a game is vastly different from the others.

[Lukoi Banacek]

Team damage now impacts your match score and heavily. So if you had a 1k damage game but a -100 team damage or a team kill, expect to see a brutal hit to your match score.

[Asym]

Has PGI ever published anything that outlines the math behind all of their calculations?

I wonder if they would so that we'd at least have the algorithm that drives the numbers....

[mistlynx4life]

Asym, on 27 November 2017 - 05:52 PM, said:

Has PGI ever published anything that outlines the math behind all of their calculations?

I wonder if they would so that we'd at least have the algorithm that drives the numbers....

I'm no expert but I've done a little digging here and there and I don't believe they've ever released hard numbers on anything like that. I imagine some folks would rush to build Exploitsville, capital of Puglandia, all in one day if they did.

[arcana75]

Pretty much 1/2 of damage dealt is added to match score. Issue is we don't know the penalty for team damage, but from anecdotal evidence it's higher than 1-to-1. Penalty for a suicide, be it from ejecting or overheat, is pretty big too.

EDIT: ok I used a recent battle result and the data from mcgral18's results and came up with something that seemed to match the results.





Hope the image can be seen!

Most I followed mcgral's data, but some I had to fudge to try to find a sync, eg team damage at -5 weight, KMDD at 4 (mcgral18 said it doesn't affect score), and Scout/Brawl at 1.

Edited by arcana75, 27 November 2017 - 10:06 PM.

[Xavori]

If I get really bored, I'll reverse engineer it. Or if someone is already wanting to know now, the way to do it is to go through everything listed that gets rewarded (damage, kills, sill assist, solo, ams damage, scouting, formation, etc.). Then start saving matches where you know your match score and the list of variables that went into it. You'll then need a bunch of equations so that you can start solving for variables.

For example, let's say that all that mattered was damage, kills, kmdd's, and team damage (it's way more complicated than that, but I'm not going to write out a long enough equation to solve for how complicated I think it is for this example).

Your equation would be:
D(d) + K(k) + KM(km) + TD(td) = match score
The reason for both a big variable and small being multiplied together is that it's almost a certainty that variables have different impact on your match score. And in the case of TD, the td is going to be negative.

You'll then need at least one different match score for each variable so then when you start eliminating or substituting which I why I suspect this would be a bit complicated.

Alternatively, someone could beg Russ on twitter to just go ahead and release the formula.

[Tier5 Kerensky]

It should be too difficult to figure out. Take bunch of results, enter them into spreadsheat, then try different values to get the same results.

[arcana75]

Teer Kerensky, on 28 November 2017 - 03:12 AM, said:

It should be too difficult to figure out. Take bunch of results, enter them into spreadsheat, then try different values to get the same results.

I did that. Turns out it seems entirely plausible to take everything shown on the Match Performance screen at face value, except divide damage by 2.

[Xavori]

arcana75, on 28 November 2017 - 04:00 AM, said:

I did that. Turns out it seems entirely plausible to take everything shown on the Match Performance screen at face value, except divide damage by 2.

You realize it's entirely possible to get AMS damage of 1000+, and yet I'm sure that never ever running my Kitfox have I broken 1000 match score

[arcana75]

Xavori, on 28 November 2017 - 04:11 AM, said:

You realize it's entirely possible to get AMS damage of 1000+, and yet I'm sure that never ever running my Kitfox have I broken 1000 match score

I read somewhere that AMS score you divide by 5, seem to recall that in one of the invasion threads. However I'm basing all my calculations now on 14 matches I recorded recently for the Battle for Tharkad event, and none of my piloted mechs had AMS. If you have a data point on AMS, share it.

Clearly the correct way here is to go into a private game with a friend, do exactly 1 thing, then record the results and repeat for each and every act that scores points, as mcgral18 did previously.

For my 14 matches, just taking the Match Performance at face value seems to correspond quite closely to my actual match score.

Edited by arcana75, 28 November 2017 - 04:27 AM.

[Tarl Cabot]

I knew there is a reason I should build a 2nd computer.. with updated items..../// Cyber Monday sales still going on!!

[Asym]

Xavori, on 28 November 2017 - 04:11 AM, said:

You realize it's entirely possible to get AMS damage of 1000+, and yet I'm sure that never ever running my Kitfox have I broken 1000 match score

I think this is where the mech class, function and damage potential needs to be added in to the calculation to give the numbers "relevance"..... My Kit Fox is the same way. One match with 1000+ AMS and an overall damage of <300 may not be a bad game and in many cases, by protecting team assets, crucial..... Or, on the opposite scale, the huge assault with a bazillion weapons that is killed with 10 damage (the other night as a matter of fact - same match I got 600 dmg and 3 kills in my Kit Fox).... Numbers for the sake of numbers have little value....

Relevance. In that place and time, where do I fit in and what is expected?

[arcana75]

Asym, on 29 November 2017 - 02:24 PM, said:

I think this is where the mech class, function and damage potential needs to be added in to the calculation to give the numbers "relevance"..... My Kit Fox is the same way. One match with 1000+ AMS and an overall damage of <300 may not be a bad game and in many cases, by protecting team assets, crucial..... Or, on the opposite scale, the huge assault with a bazillion weapons that is killed with 10 damage (the other night as a matter of fact - same match I got 600 dmg and 3 kills in my Kit Fox).... Numbers for the sake of numbers have little value....

Relevance. In that place and time, where do I fit in and what is expected?

Agree. Playing a specific role and excelling in in, is something that is difficult to capture in the current data available to us.

[Tier5 Kerensky]

Definetly not 1:1 on most. I did some testing with screenshots typed into excel and I got most close results when I took the old 2015 findings as basis, then added all the not mentioned stuff with value of 1, and took team damage as 1:1.


TBH it's a ***** to type all that stuff into sheets. But now that I started, I guess I could continue typing in more results and seeing what common have those that are most off. I also noticed I didn't take account unlisted stuff such as winning or surviving, but those are not the only ones off since their effect would be seen easily.

But still, since everything else expect damage is pretty small, it mostly comes from damage.

[ANOM O MECH]

Cato Phoenix, on 25 November 2017 - 05:45 AM, said:

Hi guys, do you all have any insights as to the new match score algorithm for QP? Like, what contributes what? Some of the games I've had recently have had me scratching my head, and I don't think its as simple as majorly damage + kills/assists any longer.

I don't even know whats going on anymore. Got three kills and 2kmdd's...with 80dmg. 426 match score. I screen shotted and sent to PGI.