Recruitment Tool Discussion Thread

[Prince1999]

Recruiting is a lengthy chore that prevents me from playing the game itself. As a closed clan that is not full, we need to actively recruit to grow/maintain numbers. I have found that to achieve any sort of viable results I have to recruit by volume. It is not feasible to spend minutes reviewing each player's profile and base. All that time adds up and more often than not wasted.

Players who are actively looking for a clan should also actively have to opt in again after a period of time has passed. This would remove those who have stopped playing the game from the pool or those who have joined new clans and are happy. I do not understand Darian's counter-argument that it should just be on like a spam folder. Our invites are spam?

Players struggle enough as it is keeping their war shield updated red or green - why does Supercell think they will update their labels? I've lost count the number people who have the clan war label on and yet they come into the clan and say they dont like / want to war. Labels are a 'one and done' activity on which we are supposed to based our search controls on...?

TL: DR... Recruiting takes too long to successfully do. Would much rather have better quality of results rather than quantity. The algorithm is learning, but not learning fast enough. How can we help it learn if we don't know what it is taking as 'learnings'.

I agree with your points in general. One concern that has come up in that previous 30 pages is that we can not make the system too complex or onerous for the player to use. It should not be a chore. I feel like too many options to choose and too many refreshes, particularly for lower level players and clans, could be a problem.

Along those lines, do low level players or clans really need to be able to select for minimum trophies, war stars, or CWL stars? It does not seem like they would. But higher level clans and higher TH players do. Is there a way that these functions could be added as they are more appropriate. Could clans be given certain screening functions as they progress in clan levels, as clan perks? Could higher TH players be given these options as they progress? Already things like building the clan castle and joining a clan are gated at TH3 (there are ways that players can buy packs or passes to shorten this.) Troops spells and defenses are all expanded as an account grows. Could the ability to accept invites only from clans in certain CWL Leagues or certain trophy ranges, for example, be something that is earned as the account progresses?

I think that adding more functions as a player and clan progresses could do two important things. First, it would avoid the overload on a new player or clan of having to choose too many options and performing too many recurring tasks. At the same time, it makes those functions available to the higher level and especially end-game level clans that need them. Second, it would ensure that those developing accounts and clans would be available in the general pool to players and clans that they are appropriate for.

[2222]

I agree with your points in general. One concern that has come up in that previous 30 pages is that we can not make the system too complex or onerous for the player to use. It should not be a chore. I feel like too many options to choose and too many refreshes, particularly for lower level players and clans, could be a problem.

Along those lines, do low level players or clans really need to be able to select for minimum trophies, war stars, or CWL stars? It does not seem like they would. But higher level clans and higher TH players do. Is there a way that these functions could be added as they are more appropriate. Could clans be given certain screening functions as they progress in clan levels, as clan perks? Could higher TH players be given these options as they progress? Already things like building the clan castle and joining a clan are gated at TH3 (there are ways that players can buy packs or passes to shorten this.) Troops spells and defenses are all expanded as an account grows. Could the ability to accept invites only from clans in certain CWL Leagues or certain trophy ranges, for example, be something that is earned as the account progresses?

I think that adding more functions as a player and clan progresses could do two important things. First, it would avoid the overload on a new player or clan of having to choose too many options and performing too many recurring tasks. At the same time, it makes those functions available to the higher level and especially end-game level clans that need them. Second, it would ensure that those developing accounts and clans would be available in the general pool to players and clans that they are appropriate for.

Rather than relying on the AI to learn, though, I prefer to let the players learn. Assume the person searching for a new player for their clan has a long list of filters to choose from. Anyone who had done any bit of online shopping is very used to this. They already know the wider they keep their search the more results they get and the more narrow they make it, the fewer results they will get, even sometimes getting no results. Players would quickly learn how wide to cast their net. This is especially true if players being recruited also can set their own filters so they only get invites from clans that are the level they want, the location they want, etc. A player in a level 2 clan searching for max th13s and getting no results (because those players have all selected that they only want invites from level 10 clans) will learn they need to reduce their filter down and start looking for th7s.

[SumoSloth]

Rather than relying on the AI to learn, though, I prefer to let the players learn. Assume the person searching for a new player for their clan has a long list of filters to choose from. Anyone who had done any bit of online shopping is very used to this. They already know the wider they keep their search the more results they get and the more narrow they make it, the fewer results they will get, even sometimes getting no results. Players would quickly learn how wide to cast their net. This is especially true if players being recruited also can set their own filters so they only get invites from clans that are the level they want, the location they want, etc. A player in a level 2 clan searching for max th13s and getting no results (because those players have all selected that they only want invites from level 10 clans) will learn they need to reduce their filter down and start looking for th7s.

I've always agreed with this thought process. What it does differently than the machine learning is a couple of significant things.

#1 - It gives players the sense of control, which the machine learning does (even if ultimately it ends up doing the same thing)

#2 - It allows players to decide how to expand their search parameters, rather than a machine arbitrarily deciding which parameters to expand often then including results that are unwanted and become a waste of time to wade through.

#3 - It allows a clan to be fluid, changing and flexible. My assumption is (of course it can only be an assumption because we aren't privy to any behind the scenes understanding of how it actually works) machine learning is based on past clan interactions, historical data, which is often irrelevant if a clan decided they want to go a different direction, or new leadership takes over or even if the clan suddenly is one TH 10 short for war at a given time. There is no way the machine learning can take that into account, where as we can if we were able to select our own parameters.


2222 is 100% right, it's much better for us to do the learning, if we do, it at least puts the onus on us to decide how to go with it. As someone who is shopping for a new car right now, I can tell you that when I go to a dealership website and they don't have the exact model I'm looking for, the first thing I do is check the other dealerships in my area. The second thing I do is expand my search criteria (maybe I include all trims in that make, maybe I open it to all SUVs in that dealership). If the list is too big for me than I tighten the search criteria to more specific things (Maybe I narrow it to one specific colour, etc.). Now no matter how big or small that list is, I know that it at least is filled with things I want. Which ultimately gives me the sense of control I think most players are looking for in this process.

[Piper139]

I've always agreed with this thought process. What it does differently than the machine learning is a couple of significant things.

#1 - It gives players the sense of control, which the machine learning does (even if ultimately it ends up doing the same thing)

#2 - It allows players to decide how to expand their search parameters, rather than a machine arbitrarily deciding which parameters to expand often then including results that are unwanted and become a waste of time to wade through.

#3 - It allows a clan to be fluid, changing and flexible. My assumption is (of course it can only be an assumption because we aren't privy to any behind the scenes understanding of how it actually works) machine learning is based on past clan interactions, historical data, which is often irrelevant if a clan decided they want to go a different direction, or new leadership takes over or even if the clan suddenly is one TH 10 short for war at a given time. There is no way the machine learning can take that into account, where as we can if we were able to select our own parameters.


2222 is 100% right, it's much better for us to do the learning, if we do, it at least puts the onus on us to decide how to go with it. As someone who is shopping for a new car right now, I can tell you that when I go to a dealership website and they don't have the exact model I'm looking for, the first thing I do is check the other dealerships in my area. The second thing I do is expand my search criteria (maybe I include all trims in that make, maybe I open it to all SUVs in that dealership). If the list is too big for me than I tighten the search criteria to more specific things (Maybe I narrow it to one specific colour, etc.). Now no matter how big or small that list is, I know that it at least is filled with things I want. Which ultimately gives me the sense of control I think most players are looking for in this process.

I agree with the idea of filters. Especially on something other than badges. The devil is in the details however. The data table is going to be huge. How often do you refresh? You sure as heck don't want thousands of clans running queries against the live game database. I would imagine it doesn't store the data optimally for searches anyway. The more filters you have, the more complex the query. Indexing is going to be key. I would start with language / location. That will be the one that eliminates the most matches in my opinion. After that, it gets tricky. Maybe town hall or war stars. Not sure. Getting that right will make or break the search engine. I hope they think it is worth the effort.

[ClasherAK47]

But consider why I spent 5 minutes doing the test above... I was confident I was going to get nonsense!

This sentence sums up whole recruitment process,
When you are this confident that you won't get anything meaningful then you got to admit that the AI not learning anything "meaningful"

On paper it might look that results are 100 % what player is looking for but in reality it's exactly what OnyxDS said "nonsense"

First stab in the back of clans sustainability was when they changes how clean search was changed, hundreds of thousands clan went on life support after that, and when global was switched off and we were given this silly system that pulled the plug of the life support

More controls on how we filter will give us some satisfaction that what we are looking for is not available out there and like someone said earlier you'd lower your expectations and try again rather than AI bringing up nonsense

I've lowered my expectations and invited 10's of th11 with 100ish war stars in a hope it maybe a second account, but when they attempted fc half way through the fc my foot started itching and kicked him out

For 14 months we kept hearing we've made changes that you can't see but system works as intended but our results are just rubbish.

For me they are rubbish because I got no interest in even looking at Arabic, Chinese, Korean or any non English character name profiles because we only speak "English" in clan chat nothing else. Try solving why AI thinks these players are suitable for us, to me just don't show them to me and waste my time. More you try using it more it starts feeling like a chore

I even tried recruiting from discord but found myself using discord more than I was actually playing the game, 7 th13 accounts with over 12,000 war stars and losing the will to pick up my phone/tablet and rally my clan with dwindling numbers

[2222]

I agree with the idea of filters. Especially on something other than badges. The devil is in the details however. The data table is going to be huge. How often do you refresh? You sure as heck don't want thousands of clans running queries against the live game database. I would imagine it doesn't store the data optimally for searches anyway. The more filters you have, the more complex the query. Indexing is going to be key. I would start with language / location. That will be the one that eliminates the most matches in my opinion. After that, it gets tricky. Maybe town hall or war stars. Not sure. Getting that right will make or break the search engine. I hope they think it is worth the effort.

If they only use the data tracked on our profile already, such as war stars, xp, hall level, friend in need, wins this season, trophies, etc. does that make it better? I’d be very happy to filter my search by what I can already see on every profile.

[Piper139]

If they only use the data tracked on our profile already, such as war stars, xp, hall level, friend in need, wins this season, trophies, etc. does that make it better? I’d be very happy to filter my search by what I can already see on every profile.

Maybe. Maybe not. It depends on how it is organized. It also does not include what (in my opinion) would be the most effective index which is language or location. To make queries efficient, your first data element should eliminate the largest number of records to review. So, if I select English or USA, it doesn't look at the other criteria you selected if the first column isn't English. Say you use townhall next. That could eliminate a lot too. Maybe we are down to English speaking th11s and up. Next criteria would be up for debate. Now that I think about it though, the absolute first criteria should be looking for a clan assuming its something that needs to be updated regularly. Anyway, my point is it sounds simple. It is in a way. In others its very not simple. It could be very worthwhile though. Even just matching on actively looking for a clan, speaks my language, and appropriate town hall level would be better than what we have.

[PaluEF]

Just a thought (to all that propose using existing data like achievements, etc...): why would SC have created the tags when implemnting the (new) recruiting tool?

There may be an issue with using the data already existing... why that is maybe the key to using it in future. Which means also that it is not a solution as easy as some think. Which could be why Darian keeps speaking of *all* countries, ie using a *new* country list, not the ones already in place (either language list, clan country list or flag list).

Which means existing data is not usable for recruiting, for any reason.

Is the fact that the team developing the recruitment tool is separate from the development team at issue?
Should solutions rely (almost) only on new attributes (not currently available in game)?
An obvious I forgot to mention: intellectual property...

I think somehow a new line of thoughts should be started, to avoid running back into the same issues (if the above is confirmed)

[2222]

Maybe. Maybe not. It depends on how it is organized. It also does not include what (in my opinion) would be the most effective index which is language or location. To make queries efficient, your first data element should eliminate the largest number of records to review. So, if I select English or USA, it doesn't look at the other criteria you selected if the first column isn't English. Say you use townhall next. That could eliminate a lot too. Maybe we are down to English speaking th11s and up. Next criteria would be up for debate. Now that I think about it though, the absolute first criteria should be looking for a clan assuming its something that needs to be updated regularly. Anyway, my point is it sounds simple. It is in a way. In others its very not simple. It could be very worthwhile though. Even just matching on actively looking for a clan, speaks my language, and appropriate town hall level would be better than what we have.

Location is recorded already too, though not seen on the profile. It would be an important filter, I agree.

[Cersei80]

This sentence sums up whole recruitment process,
When you are this confident that you won't get anything meaningful then you got to admit that the AI not learning anything "meaningful"

On paper it might look that results are 100 % what player is looking for but in reality it's exactly what OnyxDS said "nonsense"

First stab in the back of clans sustainability was when they changes how clean search was changed, hundreds of thousands clan went on life support after that, and when global was switched off and we were given this silly system that pulled the plug of the life support

More controls on how we filter will give us some satisfaction that what we are looking for is not available out there and like someone said earlier you'd lower your expectations and try again rather than AI bringing up nonsense

I've lowered my expectations and invited 10's of th11 with 100ish war stars in a hope it maybe a second account, but when they attempted fc half way through the fc my foot started itching and kicked him out

For 14 months we kept hearing we've made changes that you can't see but system works as intended but our results are just rubbish.

For me they are rubbish because I got no interest in even looking at Arabic, Chinese, Korean or any non English character name profiles because we only speak "English" in clan chat nothing else. Try solving why AI thinks these players are suitable for us, to me just don't show them to me and waste my time. More you try using it more it starts feeling like a chore

I even tried recruiting from discord but found myself using discord more than I was actually playing the game, 7 th13 accounts with over 12,000 war stars and losing the will to pick up my phone/tablet and rally my clan with dwindling numbers

Unfortunately this is true,nonsense! For example when I perform a search it throws up everything except what I'm looking for. Tobe fair I don't use the tool that much as I find i spend too much time looking and less time playing,and the issue for me at least is it's not in real time,while I wasn't a fan of global its absence is deeply felt when it comes to recruitment!

As for recruitment outside of the the game well what can I say,it should NEVER have to be done! I'm sorry but I don't want my 14 yr old child (if I allowed them to play clash) to have to resort to chatting to strangers on discord or whatever!

Clash is about forming a team of players that work well together to achieve a collective goal wether it CW,CWL,CGs or simply just farming,not having real time recruitment hinders this whatever way you want to look at it,and no amount of so called machine learning is ever going to change that IMHO of course.