#bot game

[benusaur-gx]

Riot really outdid themselves with these new bots...

[lol-name]

You ok, Ez

[vegeta897]

The April Fools Game

As you may know, I maintain my own bot in a small Discord server with my friends. I enjoy adding functions to it because they’re like mini-projects that I can knock out in a couple hours or less, and I can get instant gratification from seeing people try out the new features.

Over the past few days I was working in secret on a special one-time chat-based game, to be played on April Fools. I called it a riddle, but it was really just a series of Half-Life trivia questions accompanied by images.

The twist was that there were special rule modifiers that randomly activated during play, and every time someone said the wrong thing, the riddle would reset to the beginning.

It was meant to be frustrating, but I knew my audience and they were up to the challenge. It was immensely entertaining for me to watch their struggle, as well as gratifying to have made a game that kept them trying repeatedly to win. After 25 attempts in just as many minutes, they made it through the 16 questions without messing up. I only had to fix a single bug during play, too.

Development of the game was slow-going at first, because I wasn’t sure how to design the game. All I had decided was that there would be a series of questions, but I wanted there to be more to it. I tried coming up with ways of requiring cooperation among players so that one person couldn’t just run through it. I eventually abandoned that pursuit, and it worked out fine in the end simply because I made sure enough people were around to play when the time came.

When a player sent a message, it would be compared against the current stage’s answer with fuzzyset.js. If it was at least 90% similar, it was considered correct. This allowed minor differences in punctuation or spelling.

The random modifiers came in 2 flavors, answer modifications and chat rules. Answer mods transformed the actual text of the questions, and required that the user type their answer in the same style. For example, if the question showed up backwards, you had to answer backwards. Chat rules were arbitrary restrictions on all non-answer messages, like banning the usage of a letter or requiring capitalization.

The definitions for these were set up in a way that would easily let me add a bunch, but unfortunately I couldn’t think of very many. Here’s what the backwards answer modifier looks like:

answerMods.set('reverse', { transform: a => a.split('').reverse().join(''), validGuess: () => true });

The transform function is applied to the question and the answer, and submitted answers are checked against the transformed answer. validGuess is for any special checks that might need to be run on the answer to consider it valid, aside from the fuzzy string matching. This was necessary with certain answer mods like all-caps. The fuzzy matching ignored case, so it considered "THIS ANSWER" to be the same as "this answer". The validGuess function can check that the submitted answer actually was in all-caps.

validGuess: (a, g) => g === g.toUpperCase()

The chat rules were defined in a similar manner, but some required more complexity. The forbidden letter rule worked like this:

chatRules.set('banned-letter', { init: d => d.letter = util.pickInArray([ ...'abcdefghilmnoprstw' ]), message: d => `Don't use the letter ${d.letter.toUpperCase()}`, violates: ({ message: m }, d) => m.toLowerCase().includes(d.letter) });

Since it picked a random letter, there had to be some data persistence so it would know what letter to check for in messages. The d variable holds this data. The rule was initialized with init, which picked a letter and saved it to d.letter. The description of the rule was generated with message which is able to reference this saved data. violates is the function that checks for a violation of the rule.

Here’s a gist of the entire mods file.

Speaking of data persistence, I rewrote my custom JSON storage module during this process. The entire game’s state is kept in a JSON file so that the bot could be rebooted at any time without breaking or restarting the game.

{ "attempts": 5, "stage": 2, "stageBeginTime": 1522735122678, "furthestStage": 7, "lastClueTime": 1522735122678, "clueShownTimes": 1, "nonAnswerMessages": 4, "lastAnswerer": "86928421359198208", "answerMod": "allcaps", "chatRule": false, "chatRuleData": {}, "chatRuleViolations": 0, "usedAnswerMods": [], "usedChatRules": [], "lastChatRuleTime": 1522735015044 }

Some of these properties ended up unused due to iterations on the rule design.

My storage module used to work by just providing the data object and and a save() method. This felt redundant because there would never be a time when you want to change data without saving. So instead of making modifications directly on the object, I added set(key, value) and get(key) methods, like a JS Map. When writing data with set, the module internally calls the save() method.

JSONFile.prototype.set = function(key, value) { this.data.set(key, value); this.save(); };

This made certain kinds of data changes a little more complicated. For example, incrementing a number had to be done like this: data.set('stage', data.get('stage') + 1);. Manipulating arrays or child objects was awkward too. So I created a transaction method that provided the value you want to change, and let you return the changed value to be saved. Now incrementing looks like this: data.trans('stage', s => s++).

Most of the game’s code is kind of messy, so I don’t have much else worth showing. Here’s a gist of the whole thing, in case anybody out there is curious. The 2 entry points are the listen and tick functions. The former is a listener that handles messages sent in Discord, and the latter gets called every 1 second to handle time-based logic.

All the testing during development had to be done by me alone, in a hidden channel. A lot more could have gone wrong when it went live, and I’m very lucky it did not, since my complex bot commands usually are full of bugs when I think they’re ready for deploying. Not only were there no game-breaking bugs, but the players really enjoyed it. I couldn’t be more happy and satisfied with the end result.

I’m certainly considering expanding on the format with a new set of questions and some changes on the mechanics based on what worked well and what did not. Just like a real game developer doing a play-test session!

I know it’s trite to thank people for reading, but I am sincerely grateful to anyone who gives my writing any attention. Leave a comment 😊

[nirahsaooc]

Apparently this POTG Against freakin’ Hard Bots annoyed some tryhard. Who proceeded to ask if I was intentionally trying to do ‘pentas and shit, or if I was actually trying to get better at the game.’ 

It’s just kind of amusing that he’s taking a bot match that seriously, especially when Bots are extraordinarily easy to get mass kills on with Ults like D.Va, Junkrat, Hanzo and others that do massive AOE damage.

[king-ddraig]

(via https://www.youtube.com/watch?v=qmrTVYd-v3A)

league of shenanigans bot game rush just to see how fast we could do it haha

[cryomorphsdk]

League of Legends #9 - Feeder Simulation (Training Katarina and reaction time)

Hi there pals! For these who are confused about topic - it's not about ranked game or even normal. I just figured out how to create a small feeder simulation where you actualy can train your skills in reaction time on hyper-carry champions (same as ability usage).

1. Create a Custom normal game. 2. To your team join 2 another bots (beginner) who will go 1 for each lane left. 3. On the enemy team join 5 Intermediate bots who will slain happily your bots. 4. Try to carry the game and win it.

I just played Katarina game against: Top: Master Yi, Darius vs my Xin Zhao Bot: Tristana, Morgana vs my Cassiopeia Mid: me as Katarina vs MalzaharChampion choice was like this on purpose - most of them are OP in actual meta (Morg, Tris) or hidden OP (Darius, Malz). You just tell now: "But they are still bots!". Yes, they are still bots but diffrence betwin beginner and intermediate bots:

Beginner: - Recall in unsafe place or run to base when below 50% hp. - Do not use summoner spells. - Barely use abilities (mostly only autoattack). - Barely trades. - Playing completly unsafe.

Internediate: - Overall AI (Artificial Inteligence) is adjusted based on playerbase skill level counted in first 6 levels. - Bots use summoner spells frequently depends on situations. - Bots buy and put wards in tactical places based on map situation and contest progress. - Bots react on roaming enemy - they roam also (here funny story, got snared by roaming morgana while used Shumpo to harras Malza - becouse i was before in bot to kill Tristana). - Bots coordinate their attacks and do not waste ultimate skills for non-thread champions. - They are scary as fuck when fed - confirmed while leveling another account.

So yeah, top tower down fast, bot tower down fast, it was quite hard work to keep my mid inner tower (first one was down while i roamed to make myself fed). Best part started when they grouped, it was 3v5 basicaly or more like 1(+2 x meatshield) v 4 fed bots + 1 underleveled Malz. Using Zhonya and DFG was like - if i mess this up, i might lose match. After a while (i mean - first Pentakill near my inhib tower) i already started understading Katarina mechanics and that Ultimate is not only way to kill people and using QEW combo / Ward-shumpo (juke skillshots) is very important to do something. Still i died few times from Malz ult + Morgana ult + Darius ult combo (including Tristana dmg output) but at the end i was like 63/8/0, my team bots was around 0/10+/something.

I might make more this type of matches till i will feel confident to jump into rankeds with Katarina. If someone from you reading this play on EUW, poke me here, maybe we could make a training together ^^.

That's all from morning news and as always - HAVE A NICE DAY!

[hel-lo-and-good-bye]

6.19.2014

So a bot got a quadra the other day.