The Future of Matchmaking
02/10/2014
 
Author: Kirill Yudintsev
 
The team and I often get questions like: “What’s with the Battle Ratings thing”? “Any news about them?” Players see an issue in the current balance of Realistic Battles and tie it in with BRs, suggestions are made to discuss and think about variants involving “historical balance” and “balance based on performance”, or to look for other ways to solve the issue.
 
I have been talking about balance based on characteristics and historical balance quite often. At first glance – the ideas seem to be interesting, but after closer inspection it becomes clear that they are not practical systems, but just ideas.
 
Regarding ‘Historical matchmaking’: within the same historical time period there were, for example, Me 262 (which fought from 1944) and La-7\P-51D. Only a small number of Me 262’s were shot down by fighters of that era (and the Me 262 is a formidable adversary with clear advantages even for significantly more powerful late planes, such as F-80A, Yak-15 or MiG-9). Or, for example, the B-17E, that fought in 1942 and B-17G from 1943 – it is a powerful foe for the most fighters that are armed with low-calibre armaments (it wasn’t by accident that there was a project to use the B-17 as a “fighter hunter” for bomber cover). Good, cheap in production, reliable and reputable aircraft were produced and were fighting for a long time, even while being inferior to the recent developments of any country, but being still good at tasks which made this difference not as vital. At the same time, new developments often would be economically impractical for its time: War is firstly economically driven and if speed of production and reliability of one vehicle is inferior to the other, then quite often choices in production would not be set for characteristically better vehicles.
 
It is the same with balance by characteristics. Battles in War Thunder are not a ‘knight fight’ in the sky (as well as real battles) – it is a completion of certain battle tasks, which are usually tied to the elimination of ground vehicles. Victory – is a victory of a team that acts together. It’s a futile task to compare a B-17 and Yak-1 or a P-40 and Il-2 by characteristics. Even a direct comparison between fighters won’t show their real efficiency in the game – some fighters are better against other fighters, others against attackers and bombers.
 
The task of getting the best balance in the game is not solved through the only simple precept of  “better duel balance”. This is a task that should be better solved with complexity.
 
 
 
Detailed and comprehensive analysis shows that partially the issues are linked to non transitivity of the advantages of some aircraft over others in different country pairs. A rough (but real) example: Me 410 overpowers B-17. Ki-61a “is more powerful than” Me 410. But Ki-61a “is weaker” than a B-17.
 
Like this, in some countries, pairs of the same aircraft (for example, it affects the whole Messerschmitt line) can have an efficiency that is different by 2 or more times. So the BR of a Bf 109 G-2 should be a lot different when it fights against the US than when it fights against the USSR (sometimes, the difference is higher than 1 BR).
 
 
 
The other problem is that on 5th rank some planes don’t have any adequate enemies in some countries. The most obvious example is the Me 262, which is more effective than almost all early jets of each country (F-80, MiG-9, Yak-15, etc), in the same way as the best piston fighters, but significantly weaker for all late jets (F-86 and MiG-15). At the same time we in fact have only situations of “Me 262 vs F-80A (Yak-15\17, MiG-9) and top piston planes” or “Me 262 vs F-86\MiG-15” – so situations where the Me 262 is either a significantly advantaged aircraft or one that suffers the most. The Me 262 cannot be at the same battle with both of those because of the BR difference brackets. How often it will get to either of those battles depends only from the amount of certain aircraft in the queue.
 
And finally, a most visible part of the imbalance problem happens because of the fact that at certain hours, matchmaking has no one to gather into a “good” battle. If, for example, there are 8 US aircraft in the queue and 7 of them are B-17’s and one P51D-20, for Germany there are 7 Bf 109 G-2 and 2 Fw 190 A-5, then sooner or latter the balancer will have no choice, but to start the battle with those aircraft in it. If mixed nation battles were allowed, then resulting matches would be almost ideally balanced, but in RB, where vehicles of one country won’t fight the same country – its a dead-end and the fight has to start regardless. This issue is especially visible at higher ranks (4th and 5th) in minimal online hours.
 
 
 
In the end our follow-up program will be:
 
1) We plan to completely remove BR from RB. Instead, we will have tables which will show what vehicles of what country can meet each other in the battle. This table will be available to be viewed by players directly inside the game.
 
We will also have a balance weight for the country pairs “under the hood”, meaning if vehicle CAN meet another vehicle it won’t mean that they will meet often and in equal proportions, so it almost like a battle rating, but only for country pairs. So a hypothetical encounter frequency will be shown for the aircraft (of course, in reality the encounter frequency with different aircraft will depend on who exactly is queued at that moment).
 
2) For top ranks (5th) we going to implement mixed nation battles. In fact there are some kind of Mixed Battles already – 4 of 5 nations have MiG-15s or F-86s, but for matchmaking it is harder to make a good match – 8 F-86s + 2 MiG-15s in Germany and 4 F-86s in USA with 2 MiG-15s in USSR obviously can make a very good match but not now. Another thing to notice – in this mode we have a dire situation, there will be no aircraft equal to the Me 262 for other countries, since the aircraft was unique, without equal. And the queue fullness at top ranks at non peak hours is not giving us any choice (at the moment in nonpeak hours queues are 10 minutes long with majority of the battles starting in 5×5 battles, this increases imbalance issues).
 
These changes are planned to be implemented soon after 1.43
 
 
 
Finally, let’s take a look at Simulator Battles, which also has its own issues with balance, which are looks similar to those in Realistic Battles on first glance.
 
Despite a formal similarity there are substantial differences between them.
 
Firstly, this mode is much harder to learn with a high complexity of control functions. Much more depends on the pilot and teamwork than on the vehicle. This substantially lowers the differences  between different aircraft. In fact, planes that give advantage without dependence on pilot skill are the ones that affect the balance the most – like heavy and armoured bombers.
 
One main change to solve this issue could be in limiting the amount of bombers for the team in a single battle.
 
 
 
Secondly, the majority of the issues with Simulator Battle balance are based on a small amount of the players in queues for the significant part of the day, the exception is peak online hours. Essentially, this problem was evident before the implementation of Simulator Battles in random battles (before the OBT began), but the community asked us to enable them in main Random Battle modes (from the very beginning we planned Simulator Battles only in Events, which wasn’t ready back then, and which gathers the same amount of the players as random battles do at the moment).
 
At any given time there are hundreds of players participating in Simulator battles alone, of course. But since battles in that mode last longer than in any other random battle mode and the game has extremely different planes based on characteristics and historical period, it means that for a good match we need thousands of players, in other words a “peak” online number.
 
Sadly, the worst side of that problem is that at certain hours on some ranks you can wait 10 or more minutes for a battle, where “normal” ones would be ~2 minutes.
 
 
 
Therefore, the main ideas for Simulator could be only ideas which solve queue filling:
 
    Mixed battles (statistically makes situation in 2.5 – 3 times better)
 
    “Axis vs Allies” battles (this makes the situation in 1.5 times better based on statistics, which is noticeably better, but not enough)
 
    New “Battle” mode, where players are joining in one session (currently in development)
 
In the first two variants it doesn’t matter if you count country pairs or not, because of the obvious reasons of missing different country pairs in the queue.
 
In the third suggestion, before-battle balance is only approximate, the main balance should be achieved by in-session mechanics.
 
 
 
Finally, at this moment we usually have only two countries available. That is why there are no big reason to implement a new balance system (based on country pairs) for Simulator. Though it doesn’t mean that it won’t be implemented for unification.