Alpha 6.1.3

[pier4r]

All in all, it was a good improvement to the game, but I wonder how new players feel about this. Maybe tagging shouldn't be available from League 1.

Do not agree. It is not (I did not play yet but for what I read) that you are forced to use tags to play. You can play even without. So if a player wants to start with the perfect AI (I know many people with this approach) they will quit nevertheless at the first row of defeats, that will happen in league 1 or 2 or 3, because they feel that they cannot do it.

Instead if they start with the approach "I try and then I improve" they will start without tags and then move slowly with tags. The tag users will, likely, move in upper leagues so the league1 will be fitting for the ones without too advanced AIs.

The problem, if any, is that could be that starting from L2 the competition is way more though than expected (Dunno, did not test) and casual players could be frustrated.

I still think that for L1 a sort of "newbie protection" would be good, so playing 50 games before being promoted unless the score is already very high. Or making a L0 or L_newcomers to collect all those players below 900 elo scores. So actually L0 and L1 have the same rules but L0 collects those that still have to figure out how the game works because dropping from 1000 to 900 in L1 is very fast (3 consecutive losses). Then they get promoted back in L1 once they go over 1000 again.

But this is an extreme because with the current matchmaking and the scores, I strongly believe that in a league with many active players, people are matched with people of similar strength most of the time.

[pier4r]

. Keep in mind GFX47, you are making a GAME, not a programming language.

Ok, rant is over. Merry Christmas!

Well actually we are talking about visual programming, not turing complete, but still programming. SQL is not turning complete too (keeping aside the extensions), but it is a computer language nevertheless.

[HBomb]

<snip>
"It eventually becomes a waste of time"
<snip>

My bots set anywhere between 1 and 4 new tags every single engagement. Do you really notice it that much, as an opponent?

[pier4r]

Just give it some time before saying whether people is dropping the game or not. (That said, I do think tags still need some more, because their form is the the best with these current bots)

On this I'm interested too. As soon as I can I will do with the db dumps of GFX the request of ntk. How much games in a day, how much in the last week given a day, how many players active in the day, how many players active in the last week given the day. So we will start to see how the trend goes.

Because if there are a lot of game but only the same 10 players playing, this is no good.

Said that, there are plenty of successful games (see don't starve, minecraft, etc.) that are enjoyable from the start but then are also quite complicated to get to the last levels (spacechem too!). Also chess (and other board games) are the same. One can enjoy playing chess without much knowledge or can grow studying the patterns of the game.

[mcompany]

<snip>
"It eventually becomes a waste of time"
<snip>

My bots set anywhere between 1 and 4 new tags every single engagement. Do you really notice it that much, as an opponent?

If I'm not watching the battle in fast forwards... definately

[pier4r]

I have to say that the biggest rework from alpha 5.3.1 is to clean up the converted "currently targeted" nodes and readjust the AI without the CT option. Tags so far are used but I would use them way more if actions could do something and leave tags.

[Trigary]

If tagging wouldn't cost a full tick, it would be used a lot more in non-critical situations allowing extremely complex AIs. I like the fact that the current system punishes players who don't use tagging effectively and gives AIs with minimal tagging a (slim, but existing) chance.

[pier4r]

Yes I like it too that one has to pay for tags. But leaving the tags with an action is a sort of payment too (although smaller), but then the fact that every bot would have 4 ticks would not be important.

[pier4r]

I do not know how many will agree with me but since 5.3.1 the matchmaking in League 4 is really pleasant. I rarely run out of pending matches, when other similar strong players are active I match them instead of matching player of whatever score (unless the amount of available games is almost empty due to conditions), it is quite easy, even with less than 100 active players to deploy 20 matches a day in batches of 2 or 4.

for me it is absolutely pleasant compared to alpha 5.0 to 5.2

[Ritter Runkel]

After I played intensive with tagging and subtrees, I want to write down my thoughts about it.

First of all, it changed the whole game. In combination with the one-special-class rule the game is now so more tactic and AI based.

Subtrees: I use subtrees for two reasons

• I use them for subroutines which I use in other AIs too. That works well and makes my AIs much more maintainable. Nice feature
• I use subtrees for adjusting my tactics during the fight. That is a real game changer. I push to hard and lose a bot ... OK ... Use AI for pushing less hard. I see many resources and no enemy, let's change the strategy, stop pushing, start collecting. And so on and so on. These tactics are strongly map related but it's not choreography, it's all about when to behave how. Today I had a match with nullpointers at meeting point. We both changed tactics for some bots several time during the game. It was one of the most interesting matches I ever had

Tagging:

• I tag enemys, works fine, even when I miss currently targeted. Nothing special about it
• I tag to save states. Are we on pushing mode or do we collect? Are we on an resource rush? Is the enemy rushing? Whenever I sense something like that, I tag my own base to save the state and to change the bot behavior through using different subtrees.
• I count with tags to grab different resources at the same time
• I tag to mark which enemy not to shoot (dancers, tanks, ...) And which resource is not worth to collect.
• I tag with bottags to run some actions only once on the match (IF NOT selftag AND blabla ... DO blablabla and selftag. Awesome feature
• I tag groups to priorize directions on big maps

So tagging in combination with subtrees and well ballanced classes (imho they are well ballanced now) makes a really big difference. It's really hard now to understand for some matches why I lost them. But I concentrate only on my bots, what should they have done. When I can answer that question, I code the answer in the AI tree.

Also when beginning with a new level or an enhancement of my AIs. I start with a paper, I draw what they should do, I think about if some bots need a bodyguard, where to rush to, what to do when I reached the rushable point and so on.

I think the game is much more concentrated on designing the AI. It's also about maintainance and debugging but not to much. Even when I'm not any longer in the top 5 (where I was at 5.x) it's much more fun to play now. Even when losing I'm often appreciating to see the opponents behavior in different situations.

Very rewarding.

[NullPointer]

The game is really great and 6.1.3 is the best version by far. I like that I have a lot of modules/subtrees that I can easily plug into any AIs and I can improve them independently on the map.

On the other, although the tagging system made possible a lot of different strategies that we couldn't even consider before, it added a lot of complexity to the game. I'm feeling the complexity crossed a line that may be too much for a game. The same thing happened with me in 6.1.3 that is happening in every release: creating 1 AI per bot is more efficient than creating a few generic AIs. I have something like 20 AIs not including subtrees because programming for the details of each map gives me a considerable edge against any generic AI. I do have generic AIs, but I only use in Split Team and as well as a starting point for my specific AIs.

I don't know what is the best solution for this problem, but the amount of time needed to maintain AIs for 8 maps is just too much. Probably less maps or only random maps will remove this tiring part of the game. Or maybe having only 2 or 3 unique maps per week.

With the upcoming Strategic Points and similar features, the complexity will just increase, encouraging even more specific AIs.

Overall, the game is awesome, none of that means that game is not good, it just means I would like less complexity.

[pier4r]

Gfx already said something about this, rotating maps, but this is even better! Thanks nullpointer! What about choosing a pool of accepted maps for Multiplayer (so for example all the ones that we play in 6.1.3) and then drawing from this pool only 3 per week (or every two, but then it starts to be boring) so people can focus on them and the work does not seem too much? In the Multiplayer screen there would be a tab telling which maps are in the current rotation. So the game can include more and more maps, old maps get never inactive and players handle few maps at once focusing on ais and not every possible generic situation.

[NullPointer]

2 or 3 per week from a pool of maps sounds good. Or instead of a pool, just create 2 or 3 brand new maps every week, so no one have to keep old AIs. But any approach that limits the amount of maps will be good.

We could also have 2 modes of ranked matches:
- maps selected from a pool
- randomly generated maps that you can't see before deploying

First is for those who like to create specific AIs and the second for generic AIs.

[Ritter Runkel]

...it added a lot of complexity to the game. I'm feeling the complexity crossed a line that may be too much for a game

I think the account of complexity now is still on the edge. But I also would strongly recommend not to add more complexity (with more filters or marking points).
With the objects and filters we have, we can do almost anything. The fun thing is also to think about how to do it. For example it took me some time to find a good way to mark the two meeting point groups of the opponent. But it is possible and finding the solutions is part of the fun.

randomly generated maps that you can't see before deploying

When I first read that I thought ... Noooo ... How? ... Why? ...
But after thinking about that more on deep, this could be a very good approach. You are forced to make good generic AIs. You also can adjust them to the map by making a quick new AI with some conditions and AIs as subtrees. So every match needs one to five minutes to be setted up, what's very nice in my opinion. You can setup a game while waiting for the metro. You can setup a game when waiting at the supermarket or at a break. You can setup a match but you need some brainpower for that, it's not only clicking ten times on the same button (like it is now after you adjusted your AIs).

Kudos for that I like this idea.

[pier4r]

2 or 3 per week from a pool of maps sounds good. Or instead of a pool, just create 2 or 3 brand new maps every week, so no one have to keep old AIs. But any approach that limits the amount of maps will be good.

We could also have 2 modes of ranked matches:
- maps selected from a pool
- randomly generated maps that you can't see before deploying

First is for those who like to create specific AIs and the second for generic AIs.

+1 for the two modes.