the correct solution for war matching

Discussion in 'Ideas + Feature Requests' started by tmh, Jan 31, 2014.

  1. Here it is.

    There are two separate stat pools that are matched. The first pool is building stats. The second pool is static bonus, BFE plus BFA.

    There are two different matching styles, which clans can select by way of a War Match Mode setting that is a clan-specific setting that is selected by the owner or admins-- it's a switch on the clan page exactly like the Enable/Disable joins button. The two choices are Tight Matching and Any Match.

    Tight Matching has the potential to result in a no-match if there are no other clans warring that are actually a good match for your clan. Any Match means that the system will, after trying to find the best match for you, go with whatever it has come up with. So while you will always get a match (unless an odd number of clans sign up for war with Any Match setting, in which case one clan won't possibly have a partner) with Any Match there is always going to be a chance you get matched against someone weaker or stronger than you, kinda like how the current system works.

    The way the actual matching will occur in either scenario is similar to how it is now, but markedly different. The system will, for each clan, calculate:

    1) the total building stats of all on roster
    2) the total BFA BFE of all on roster
    3) the MEDIAN combined building stats
    4) the MEDIAN combined BFA BFE

    Then for each clan, the system sums up 1 and 2, and sorts the list of clans based on that total, which is the total power of the roster (both dynamic plus static power). Starting from the top of the list, the system takes that clan's calculated values for 1 through 4 above, and compares them against those values for each subsequent clan, calculating the percent difference between the two clans for each of the four values. Some values may be positive while others are negative. Those four values are added together for each of the clans being compared, and whichever clan is closest to 0% percent difference is the best match.

    For clans that have selected Any Match, they will be matched against whatever clan had the value closest to 0% total difference. For clans that have selected Tight Match they will only be paired against a clan if that total % difference is below some set threshold-- I can't say exactly what is correct without running simulations using actual clan signup data, but I would guess that a value of 20% should be pretty reasonable and not result in too severe of mismatches. If that can be lowered to 10% and almost all clans still get a match, great.

    The matching process will start with every clan in it, regardless of whether they have selected Tight or Any. It will go through all clans from strongest first to weakest last, and for each clan that has selected Tight Matching it will find the best match for them (even if the best match is a clan that has selected Any). After all the tight matches have been made, the system will go through the list of clans that have selected Any Match and weren't already matched with a Tight Match clan, and find the best matches there, ignoring the threshold mentioned above (20% total difference, for example).

    One other fix to this matching system is how BFA is considered. BFA is given exactly the same weight as BFE, as it should be-- they are both static bonuses so are equally relevant. At the time a clan signs up for war, at the same time equipment is locked and the clan is locked, the war system will take a snapshot of each player's BFA values for each of the four stats. These are the values that will be used in matching, and these are the values that will apply to the player for the entire duration of the war REGARDLESS IF THEY SELL OR HIRE ANY ALLIES. You can buy and sell allies all you like during war (or have them hired away against your will, just like always) but that will have no bearing on what your BFA power is in the war. It is essentially the exact same thing as the equipment lock, in terms of you losing the ability to cheat the system and sign up for war with $1T out and then hire allies after matching. But it's done in a way that has no effects whatsoever on the ally market.

    Just to point this out for anyone who might not understand the point of giving the medians equal weighting to the total stats... What it does is make it so if you're looking for tight matching, if you are a clan that actually has some range to its builds, actually has some mids in it, you are less likely to be matched up against the clan with a small number of giants and a large number of smalls and no mids. The clans with no mids are given more likelihood of matching other clans with no mids, and the clans with mids are more likely to match other clans with mids. In short, a greater likelihood of fair and fun matchups.

    So there it is, I just fixed the war matching system! Now it's up to the devs to run simulations using actual war rosters and see how it works, and then implement it. It is actually a very simple process, the math and logic couldn't be much more easier. Please devs, take the time to do this and fix the war matching system!

    I'm posting it here in case the devs actually do read these suggestions, but will also post it in a forum that gets put into active topics otherwise nobody will ever see it.
     
  2. I actually like the idea, and it was well thought out and put. That way there won't be anybody complaining about no match due to the fact there's a low risk of having one, and people won't complain about bad matches because they most likely selected any match an that's a risk they were willing to take as a clan. I also like the "snapshot" idea, because there's quite a few clans that unequip their equipment for signup, then equip it after they get a match, and this would make it irrelevant to do that if the system made it so that the have the stats they had when they signed up. Also it would make the annoying mid-war-ally-hires less annoying :p I support, but somebody more learned about the system and loopholes will point out exploits below ;P
     
  3. This isn't a bad idea really.

    Going into Any Match, there's no room to complain.
    More freedom in choosing battles.

    Support!
     
  4. Seems like this is falling on deaf ears
     
  5. Great idea Tmh! Email it to Devs!
     
  6. Trying sumting is mo better than what's we gots
     
  7. kinda like the old system, medium would mean gh/lb would work even better...
     
  8. No it wouldn't deez. The median player in a lb/sh clan would be an SH of around 2m cs. The median player in a normal, real, balanced situation would be maybe 8m cs give or take. So the comparison of them would indicate a 400% ratio of stats. Way way beyond the 20% I proposed. Using matching of the median player exactly solves the problem you say it brings back (which can't be truly brought back as it is the problem right now, it didn't go away)
     
  9. tmh strikes again. Surprised they haven't tried to hire you lol
     
  10. Bump so the devs can find it easier
     
  11. Cool. You were always good at this Bro. :)
     
  12. I've thought of a way to make the matching even tighter, might result in too many No Match if people go with Tight. Not sure. But the idea is, instead of only looking at the clans' total cs and median cs values, also compare them at the 25th and 75th percentiles (where the median refers to the 50th). That adds precision, increases the capability of precision, but has two drawbacks I can identify already. One is it requires faster processing (or increases processing time), and two is with low enrollment in wars I expect it to lead to more No Match for people who are looking for Tight. If there were enough clans warring (say, if roster sizes were 13, you'd have more than double the number of clans warring) then the matching would be great, as long as there is a big enough pool of clans to work with.

    The original proposal is best with low clan enrollment. With high clan enrollment this addition makes it even better.

    So the best system would have both styles coded up, and then at signup time it will look at the number of clans enrolled and determine if it's enough to use the higher accuracy method I just added here.