Of Karma Grinding

by Jochoi

Introduction

The wikidot karma system is of a mechanism unknown to users of wikidot. Rumours are heard that karma4 (very high, abbreviated "k4" below) and karma5 (guru, abbreviated "k5" below) requires 2000 revisions and 8000 revisions respectively.

In order to find out how such mechanism works, a series of data collection is carried out.

---

Concept

There is a speculation about wikidot karma system as follows:

The portion of k5 members out of total number of users is constant. Wikidot promotes the top k4 members to k5 when new users join wikidot and the number of k5 members had been raised according to the portion, and that new k5 slots are available; or when a k4 member has surpassed the relative threshold corresponding to said portion by their activity calculations.

Wikidot source code

The wikidot source code for karma calculation is available here with relevant rules here, which can be read.

First-hand data

A karma grinding script is developed and executed with karma level being monitored.

Second-hand data

The revision count and forum post count for accounts known to have k5 are extracted.

---

Method

Karma grinding script

A karma grinding script is developed in NodeJS, and executed on Karma PoC.

The code is not disclosed to prevent abused usage, but the concepts are as follows:

1. The script logs into the said account.
2. The script sends an edit and save edit request every 3 seconds to a particular page that can be discarded easily.
3. The script sends requests to post a forum post every 60 seconds to per page discussion of the said page.
4. The script retrieves the karma level status of the said account every 60 seconds, and stops steps 2 & 3 from executing after k5 is reached.

Counting guru revisions and forum posts

A revision and forum post counting script is developed in NodeJS.

The code is as follows:

+ Open code

const got = require('got');
const cheerio = require('cheerio');
class WD {
  // Modified from https://github.com/resure/wikidot-ajax/blob/master/index.js
  constructor(baseURL) {
    this.baseURL = `${baseURL}/ajax-module-connector.php`;
  }
  async req(url, params) {
      const wikidotToken7 = Math.random().toString(36).substring(4);
      return await got.post(url, {
        headers: {Cookie: `wikidot_token7=${wikidotToken7}`},
        form: Object.assign({wikidot_token7: wikidotToken7, callbackIndex: 0}, params)
      }).json();
  };
 
  async module(moduleName, params) {
    return await this.req(this.baseURL, Object.assign({moduleName: moduleName},params))
  }
 
  async getUser(username) {
    return await this.module("users/UserSearchModule", {
      query: username
    })
  }
 
  async getrevs(info, id, perpage, page) {
    info = await this.module("userinfo/UserChangesListModule", {
      userId: `${id}`,
      perpage: `${perpage}`,
      page: page
    })
    return {body: info.body, ex: info.body.trim()!==`Sorry, no revisions matching your criteria.`}
  }
 
  async getposts(info, id, perpage, page) {
    info = await this.module("userinfo/UserRecentPostsListModule", {
      userId: `${id}`,
      limit: `${perpage}`,
      page: page
    })
    return cheerio.load(info.body)
  }
 
  async linsearch(username, userId) {
    if (!(userId||username)) return null;
    if (!!userId) {var id = userId};
    if (!userId && !!username) {
      // The following retrieval of user id by username is avoided whenever user id is provided, as the search process is very slow.
      var user = username.toLowerCase()
      var res = await this.getUser(user);
      var names = Object.values(res.userNames).map(x=>x.trim().toLowerCase());
      var id = Object.keys(res.userNames)[names.indexOf(user)];
    }
    // If the user id is already known by other means the id can be directly inputted in the following line.
    //var id = 1404533//zyn////3427263//7happy7////4476849//sekai_s////2941964//m element
    var ret = {revisions: null, posts: null}
 
    // Retrieving number of total revisions
    var info1
    var num1 = 5000, count1 = 0
    while (num1==5000) {
      info1 = await this.getrevs(info1, id, 5000, ++count1)
      var $1 = cheerio.load(info1.body)
      num1 = $1("tr").length
      //console.log(`Got revisions ${$1("span.pager-no").first().text().trim()}`)
    }
    ret.revisions = (count1-1)*5000+num1;
 
    // Retrieving number of total forum posts
    var info2
    var num2 = 3000, count2 = 0
    while (num2==3000) {
      var $2 = await this.getposts(info2, id, 3000, ++count2)
      num2 = $2("div.post").length
      //console.log(`Got forum posts ${$2("span.pager-no").first().text().trim()}`)
    }
    ret.posts =  (count2-1)*3000+num2;
 
    return ret;
  }
 
  async expsearch(username, userId) {
    if (!(userId||username)) return null;
    if (!!userId) {var id = userId};
    if (!userId && !!username) {
      // The following retrieval of user id by username is avoided whenever user id is provided, as the search process is very slow.
      var user = username.toLowerCase()
      var res = await this.getUser(user);
      var names = Object.values(res.userNames).map(x=>x.trim().toLowerCase());
      var id = Object.keys(res.userNames)[names.indexOf(user)];
    }
    // If the user id is already known by other means the id can be directly inputted in the following line.
    //var id = 1404533//zyn////3427263//7happy7////4476849//sekai_s////2941964//m element
    var ret = {revisions: null, posts: null}
 
    // Retrieving number of total revisions
 
    var info1={body:undefined, ex:true}, $1, page, hiin1=8, index1=8, count1=256, num1=20;
    while (num1==20) {
      info1 = await this.getrevs(info1, id, 20, count1);
      $1 = cheerio.load(info1.body);
      num1 = $1("tr").length;
      page = $1("span.pager-no").first().text().trim()
      if (page.split(" ").length!=4) {
        if (index1<0) {index1=0};
        if (info1.ex) {
          if (index1>=hiin1) { count1*=2; hiin1++; index1++; }
          else { count1+=2**index1; index1--; }
        } else {
          count1-=2**index1; index1--; num1=20;
        }
      } else {
        info1 = await this.getrevs(info1, id, 20, page.split(" ").pop());
        $1 = cheerio.load(info1.body);
        num1 = $1("tr").length;
      }
    }
    ret.revisions=(count1-1)*20+num1
 
    // Retrieving number of total forum posts
    var info2, $2, page, hiin2=7, index2=7, count2=128, num2=20;
    while (num2==20||num2==0) {
      $2 = await this.getposts(info2, id, 20, count2);
      num2 = $2("div.post").length
      page = $2("span.pager-no").first().text().trim()
      if (index2<0) {index2=0};
      if (num2) {
        if (index2>=hiin2) { count2*=2; hiin2++; index2++; }
        else { count2+=2**index2; index2--; }
      } else {
        count2-=2**index2; index2--;
      }
    }
    ret.posts=(count2-2)*20+num2;
 
    return ret;
  }
}
 
var a = new WD("http://zok6hoi2.wikidot.com")
 
// If the user id is already known by other means the id can be directly inputted in the following line.
var id = 1404533//zyn////3427263//7happy7////4476849//sekai_s////2941964//m element
 
// Use linsearch() or expsearch() to your needs. Or do Promise.race() with both of them for the best of two worlds.
a.linsearch(null, id).then(res=>console.log(res))

The code is also usable as a service at https://zh.xjo.ch/wd/userinfo , with query string parameters userId or user_id for wikidot user id, or username for wikidot username if user id is not provided.

Try using user id and avoid using username if revision amount is over 20000 or forum posts amount is over 10000 because converting username to user id takes an additional 10 seconds or so.

If revisions or forum posts are too many then the tool may error out with gateway timeout on server end.

Many thanks to Unihedron for advice on using exponential search algorithm.

+ Revisions and Forum Posts Totals Tool

---

Results

Wikidot source code

As the wikidot source code is over 10 years old and may not reflect the current process of karma calculation accurately, it serves for reference purposes only.

By reading the wikidot source code, it can be observed that there are certain minimum activity scores that has to be reached in order to qualify for different karma levels, however the division of user karma levels are done according to proportions of the overall wikidot userbase. This confirms the speculation put forth in the concept section, at least for wikidot 10 years ago.

Karma grinding

Karma PoC reached k4 at around 2000 revisions and 0 posts.
Said account reached k5 at 11631 revisions and 101 posts.

A possible error: wikidot may be slow to update karma level, such that the script stops well after the threshold had been surpassed.

Statistics of current gurus

Sorted in alphabetical order
All data retrieved are of 03:30 10th May, 2020, UTC+8
Retrieved data may be inaccurate as deleted revisions and forum posts are not counted

User	No. of revisions	No. of forum posts
7happy7	29917	653
djkaktus	6871	2463
feitag	9346	3244
Helmut_pdorf	19940	8590
Ihp does not match any existing user name	8472	2716
Karma PoC	11631	101
Lt Flops	10702	700
M Element	15108	2699
NatVoltaic	11197	1404
notgull	5629	1752
pokm	5685	2285
Rounderhouse	11272	713
Sekai_s	5489	629
Woedenaz	8355	353
Zyn	8410	48175

---

Conclusion

First, the 8000 revisions rumour is considered false, as it does not take into account the other factors that affect karma levels. Examples include pokm and Sekai_s at under 6000 revisions. As for first-hand example, k5 is only reached at about 11000 revisions and about 100 posts.

Second, regarding how exactly the karma system works, it can be said that the "proportional karma" speculation was true for the open-sourced wikidot source code from over 10 years ago, however there are no conclusions that can be drawn as of current, as there are multiple possible errors.