Collection of helpfull Maxims

User avatar
Kanishka
Skynet
Skynet
Posts: 1417
Contact:

Re: Collection of helpfull Maxims

Post#21 » 12 Aug 2017, 13:51

Ritter Runkel wrote:Improve aiming :lol:

Let me try to rephrase in a Maxime:

  • First solve the tutorials and single player missions to get a basic AI
  • Improve retreating until you're at master league
  • Improve aiming when you reach GM
  • Improve harvesting and edge cases to get to the top 10
  • Don't try to optimize too much for specific maps, that only hides other bugs which needs to be fixed


I had completed all the steps and reached Global 11. Then I automatically dropped back.
Fixes break an AI more than bugs do. :ugeek:

Gladiabots Off-Topic Chat


My Stats: Kanishka_RN3;Kanishka_MiPad

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 482

Re: Collection of helpfull Maxims

Post#22 » 12 Aug 2017, 14:09

Kanishka, you could send some private matches and I will see if I see room for improvement ...

... Or you come to GM

User avatar
TheGnoo
Automaton
Automaton
Posts: 175

Re: Collection of helpfull Maxims

Post#23 » 13 Aug 2017, 03:44

I think that push mechanics are more important than harvesting in this meta. Milkhunter and Miojo are such a pain in the ass because they push me really hard.

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#24 » 15 Aug 2017, 12:25

Pier A, [15.08.17 12:21]
moreover Arcueid is in GM, you don't have to counter him. Counter those that sucks your points

Pier A, [15.08.17 12:22]
I do not counter milkhunter, I counter those that takes 7+ points from me

Pier A, [15.08.17 12:22]
or even better, first those 10+, then those 7+


So then I improve being "safe" from those below me, and I can focus on those above me because my score will raise.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

Castlevania
Automaton
Automaton
Posts: 173

Re: Collection of helpfull Maxims

Post#25 » 23 Aug 2017, 13:13

TheGnoo wrote:I think that push mechanics are more important than harvesting in this meta. Milkhunter and Miojo are such a pain in the ass because they push me really hard.


Agreed. The forth should be: Improve pushing mechanics to get to the top 10

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#26 » 17 Sep 2017, 09:25

One thing I noticed is that some tactics (that are time consuming to implement if one has to rearrange its AI nodes) that are quite effective are implemented by some players that dominates and the others will adapt only very slowly (I am one of the latter).

So it can be that a successful tactical procedure was discovered already in February 2017 and it gets replicated by the top20 only in August 2017.

I think this is due to the fact that if the one "dominating" is not always there farming a lot of points, there is not much reward to develop a counter - that takes time - since the points lost are a little. In this period, for example, I do not care much if Miojo or Milkhunter beat me because I focus first on those sucking more than 7-9 points from me. If I would encounter only Miojo or Milkhunter I would develop against them. Instead the omnipresent Chris is the enemy.

The moral of the story is: in GM the average level become slowly more and more higher because more players implement successful tactics that before were used by few.

So if you are new and you got pounded, don't get discouraged because in GM one is against tactics that incorporated evolutions over months, that one needs to digest first before being able to handle them.

And, I would say, the guys floating (floating, not hit and fall) around 2100+ now are some months "in the future" in terms of tactics implementation compared to the rest of GM.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 482

Re: Collection of helpfull Maxims

Post#27 » 17 Sep 2017, 10:52

pier4r wrote:One thing I noticed ...


That's an interesting point of view, I second that. Also when you're one of the top three players you don't have to slapped that many matches because you loose seldom.i remember when I was number one, I analyzed every match I lost. Now it's the same what you wrote, if milk, LuBeNo or miojo beat me and as long as it's not an obvious bug ... I don't care. First I have to get stable against buzz, Chris, you and some others.

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#28 » 17 Sep 2017, 11:47

Yes and it makes sense because the elo foster this (and that's reasonable I would say). Since we know that if I beat the first one it does not mean that I beat all the others (we remember buzz beating the top1, ritter, (and me) to then losing against all the others), then it is better to fix problems against those that inflicts too big elo changes and then against those that inflict less.

Moreover it is also true that if I never lose against those that takes 7 or more points from me, my score will raise and then those at the top will be those taking 7 or more points from me. So they will deserve my attention.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#29 » 29 Sep 2017, 09:35

Sometimes some tactics get heavily countered and then disappear. Then the new tactics los the optimizations against those and so when the 'old' tactics reappear, bring havoc.

That is one point that let me think that it is very unlikely to get an ai that is foolproof against all the others (unless it gets over 500+ nodes)
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#30 » 01 Oct 2017, 20:23

In this period (and with my current pace, I guess for long time) I am focusing to damage the enemy as much as I can, or push him away, rather than collecting resources. I am losing a lot of points but it is super fun!

I recommend it! Also this is due to the experience with miojo and milkhunter, that pushed me so many times that I decided to try the same.

Once I have a good "I will damage you" overall tactic, I can see to detect the situations where I have field advantage to collect resources at once.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 482

Re: Collection of helpfull Maxims

Post#31 » 01 Oct 2017, 21:07

pier4r wrote: I am focusing to damage the enemy as much as I can, or push him away, rather than collecting resources. I am losing a lot of points but it is super fun!


I do it the same way. Once my pushing is good enough I can optimize my harvesting part. Since that it's more about getting the opponent out of the field.

Focusing at harvesting also hides some failures you make in your retreatment and your pushing. I would say that

<Maxime>
you can't optimize your push when you collect.
</Maxime>

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#32 » 01 Oct 2017, 21:26

yes if someone collects one has less firepower available, it is like the bot is not there to fight (unless the bot is there to collect damage)
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#33 » 15 Dec 2017, 15:48

This talk here can help programmers in general, but I see that it can be applied to gladiabots as well.

http://bed-con.org/2017/talks/Excavatin ... -ancestors

Especially sentences like:
"But there is one quality that cannot be purchased in this way - and that is reliability. The price of reliability is the pursuit of the utmost simplicity."

Because when the code is too complicated (and the AI editor is not that flexible), simplicity helps.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

WASP103
Script
Script
Posts: 35

Helpful Maxims specifically concerning TAGs:

Post#34 » 21 Feb 2018, 10:38

  • Use "ADD TAG" and "ADD TEAM TAG" instead of their hard replacing counterparts. You can tag much more that way. Cost me 3 days just to refactor after I stumbled across that.

  • Concerning Instant tagging: TAGs are still executed from left to right like everythig else, but: If there are overlapping and/or contrary commands dealing with the same tags, the order of who comes last, always gets precedence over everything said and done before is in place. That leads to a feeling of tags being executed right-to-left, which is not the case. Cost me 4 days to understand.

  • Define what personal TAGs and what TEAM TAGs are used for what, and do never overlap them, unless you really know what you're doing. This together with the instant tagging from the point above can bring you looping and contradicting tags you will find very hard to reproduce and solve. If you ever come across a tagging tree with half of the nodes constantly lit up like a christmas tree, you've likely run into exactly this one.
I'll add more as time goes by <3

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#35 » 21 Feb 2018, 11:42

If you are a casual player or you have a ton to fix / refactor and you are not ready for the competition of ranked games.

GO UNRANKED! The unranked mode is a godsend to fix, refactor, etc. Slowly. Take your time, it is pleasant!

But dn't expect to be on the bleeding edge with unranked. If you want up to date competition, RANKED (or private challenges towards strong guys) is the way.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Hairy Phil
Algorithm
Algorithm
Posts: 63
Contact:

Re: Collection of helpfull Maxims

Post#36 » 27 Feb 2018, 20:43

You don't have to win 100% of matches. You are ready for Ranked matches when you consistently win 51% or more, which is enough to gradually increase your rank. Many of the top players in Ranked only win 60-70% of their matches.

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#37 » 27 Feb 2018, 21:30

Hairy Phil wrote:You don't have to win 100% of matches. You are ready for Ranked matches when you consistently win 51% or more, which is enough to gradually increase your rank. Many of the top players in Ranked only win 60-70% of their matches.


Ah yes, the win rate. Too many players are used to other games where the win rate is wrongly considered super important. Here the win rate means next to nothing, if you win against those with your score but lose against those with a way lower score.

You can raise or keep your score as long as you avoid losing (too much) to those that capture 12 points or more from you.

Also not losing is great. You do not necessarily need to win. A draw is not a loss, and it is a win in "points saved" as well.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Hairy Phil
Algorithm
Algorithm
Posts: 63
Contact:

Re: Collection of helpfull Maxims

Post#38 » 27 Feb 2018, 22:40

pier4r wrote:Too many players are used to other games where the win rate is wrongly considered super important. Here the win rate means next to nothing


Win rate is more important than score for novices because their initial score is fixed at 1000 but their win rate roughly reflects their performance in Unranked. Once they hit 51%, then they can switch to Ranked and start using score as their metric instead.

pier4r
Skynet
Skynet
Posts: 3122

Re: Collection of helpfull Maxims

Post#39 » 16 Mar 2018, 16:59

One thing I realized is that before I tried to not reuse the same subtree in different places to look only in one place, rather I put some "stop, don't" so the current subtree would be left until the correct one would be executed.

But the "stop don't" become way more complicated than a simple "if the situation is this, check a subtree that should come later, earlier".
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

DrPhoenix
Hello World
Hello World
Posts: 3

Re: Collection of helpfull Maxims

Post#40 » 17 Mar 2018, 19:28

Hi all, I am new here, since I started to play against human players only recently. I have however dabbled a lot with the campaign. To my mind doing this helps a lot, since every mission has different challenges. It is easy to win the campaign, but I was trying to get a AI which beats every mission (not done on both accounts, but close).

My main strategy:

- Know your Bots
- Survival first
- Use targets of opportunity (such as low shields and health)
- Push when not attacked
- Don't score if you are point leader

Return to “Strategies”

Who is online

Users browsing this forum: No registered users and 1 guest