26 replies to this topic

[Geno1isme]

I don't see how this could even remotely work. The MM already has so (too) many constraints to consider that an additional map priority would either mess with the other parameters or just would not have any noticable effect. A map ban is relatively easy to handle (just skip matches set on the selected maps). Map priority would involve making decisions which players to select for a given match, which is a way more complex issue.

 

In the end it might actually result in the opposite effect than intended: When many players select the same maps as favourites those matches will fill up sooner, and remaining players will see less favored maps more often (assuming matches are in general equally distributed across maps).

[saxsan4]

Geno1isme, on 20 January 2021 - 02:41 PM, said:

I don't see how this could even remotely work. The MM already has so (too) many constraints to consider that an additional map priority would either mess with the other parameters or just would not have any noticable effect. A map ban is relatively easy to handle (just skip matches set on the selected maps). Map priority would involve making decisions which players to select for a given match, which is a way more complex issue.

 

In the end it might actually result in the opposite effect than intended: When many players select the same maps as favourites those matches will fill up sooner, and remaining players will see less favored maps more often (assuming matches are in general equally distributed across maps).

I dont see how that would ever happen?

[Geno1isme]

saxsan4, on 20 January 2021 - 04:44 PM, said:

I dont see how that would ever happen?

It's impossible to say without knowing the detailed inner workings of the MM. But to create a hypothetical example:

Lets say that out of 1000 matches the MM will create 50 on Prokhorovka (to have nice numbers, in reality it will be more like 26). So of 30k players in all matches 1500 will end up on Prokhorovka. Now lets assume 50% of the playerbase would select Prokhorovka as a favored map. That means of the 30k players 15k would like to play on Prokhorovka, but the MM has only spots for 1500, so only ten percent of the people having it selected will actually be able to be picked. Which in turn will result in anyone not having it selected to see it less often than currently, in the most extreme case maybe never.

Also as has been mentioned above maps are generally favored by certain classes. Prokhorovka will generally be picked by light tank players rather than heavy players. One result could be in Prokho games ending up with 5+ lights per team in every match, another that your chance as a light to get put on Prokho becomes even lower as more people will try to use lights when the map selection suggests that they may see Prokho more often.

 

Again, this is a purely hypothetical example making various assumptions about the MM algorithm. I'm just pointing out that ANY additional constraints to the MM can result in very unexpected consequences.

[250swb]

Geno1isme, on 20 January 2021 - 02:41 PM, said:

 

 

In the end it might actually result in the opposite effect than intended: When many players select the same maps as favourites those matches will fill up sooner, and remaining players will see less favored maps more often (assuming matches are in general equally distributed across maps).

 

I think this would be the outcome, maybe not noticeable on a day-to-day basis but over time the stats will start to show the bias. 

[saxsan4]

Geno1isme, on 20 January 2021 - 03:15 PM, said:

It's impossible to say without knowing the detailed inner workings of the MM. But to create a hypothetical example:

Lets say that out of 1000 matches the MM will create 50 on Prokhorovka (to have nice numbers, in reality it will be more like 26). So of 30k players in all matches 1500 will end up on Prokhorovka. Now lets assume 50% of the playerbase would select Prokhorovka as a favored map. That means of the 30k players 15k would like to play on Prokhorovka, but the MM has only spots for 1500, so only ten percent of the people having it selected will actually be able to be picked. Which in turn will result in anyone not having it selected to see it less often than currently, in the most extreme case maybe never.

Also as has been mentioned above maps are generally favored by certain classes. Prokhorovka will generally be picked by light tank players rather than heavy players. One result could be in Prokho games ending up with 5+ lights per team in every match, another that your chance as a light to get put on Prokho becomes even lower as more people will try to use lights when the map selection suggests that they may see Prokho more often.

 

Again, this is a purely hypothetical example making various assumptions about the MM algorithm. I'm just pointing out that ANY additional constraints to the MM can result in very unexpected consequences.

but why would 50% select prok, there are about 50 maps in the game with an even distribution on what maps people like, believe it or not outside the forums prok is  not a well liked map, in general only good players like prok and most of the player base are average or below, also you can make this a premium only feature which means even less likely for the above to happen

 

 

[Geno1isme]

saxsan4, on 20 January 2021 - 05:24 PM, said:

but why would 50% select prok, there are about 50 maps in the game with an even distribution on what maps people like, believe it or not outside the forums prok is  not a well liked map, in general only good players like prok and most of the player base are average or below, also you can make this a premium only feature which means even less likely for the above to happen

*sigh*, why did I state twice that this was a hypothetical example? And even pointed out that the numbers were not realistic?

You cannot demonstrate MM issues with realistic examples, as there are way too many variables involved.

 

Algorithms have to generate usable results for any (legal) input parameters, unrealistic corner cases included. That is why the MM in certain situation is even allowed to create understaffed teams. "less likely" isn't good enough there, esp. as the problem would only present itself in the long run, not within a few days of testing.

 

Once more: I'm not saying that exactly this would necessary be a problem, just that it is a more or less remote possibility that such a change to the MM may have very unexpected side-effects like that. Just like I've always said that the 3-5-7 template system is the most retarded thing ever invented by WG, long before it was introduced (I considered it to be a bad joke when I heard the first rumors). And the general population realized that like 2-3 years later when facing the long-term effects.

[Airidas10]

Geno1isme, on 20 January 2021 - 04:42 PM, said:

*sigh*, why did I state twice that this was a hypothetical example? And even pointed out that the numbers were not realistic?

You cannot demonstrate MM issues with realistic examples, as there are way too many variables involved.

 

Algorithms have to generate usable results for any (legal) input parameters, unrealistic corner cases included. That is why the MM in certain situation is even allowed to create understaffed teams. "less likely" isn't good enough there, esp. as the problem would only present itself in the long run, not within a few days of testing.

 

Once more: I'm not saying that exactly this would necessary be a problem, just that it is a more or less remote possibility that such a change to the MM may have very unexpected side-effects like that. Just like I've always said that the 3-5-7 template system is the most retarded thing ever invented by WG, long before it was introduced (I considered it to be a bad joke when I heard the first rumors). And the general population realized that like 2-3 years later when facing the long-term effects.

 

I guess the way around that could be creating more games with the favorited map. So, if 50% of online players (currently in the queue) have Prokhorovka selected, roughly 50% of games are created on Prokhorovka.

But that introduces/still has new issues which make this idea a no go.

Edited by Airidas10, 20 January 2021 - 07:21 PM.