会员服务 ·

用 Python 分析了 20 万场吃鸡数据

2018 年 7 月 9 日 Python开发者

（点击上方蓝字，快速关注我们）

作者：张波（简介见末尾）

本文来自作者投稿，首发「Python开发者」微信公号

首先，神枪镇楼

背景

最近老板爱上了吃鸡（手游：全军出击），经常拉着我们开黑，只能放弃午休的时间，陪老板在沙漠里奔波。上周在在微信游戏频道看战绩的时候突发奇想，是不是可以通过这个方式抓取到很多战斗数据，然后分析看看有什么规律。

秀一波战绩，开黑情况下我们团队吃鸡率非常高，近100场吃鸡次数51次

简单评估了一下，觉得可行，咱就开始。

Step 1 分析数据接口

第一步当然是把这些战绩数据采集下来，首先我们需要了解页面背后的故事。去看看页面是如何获取战斗数据的。

使用Charles抓包

抓包实现

在Mac下推荐使用工具Charles来从协议层抓取手机上的流量，原理就是在Mac上开启一个代*理*服务器，然后将手机的网络代*理设置为Mac，这样手机上的所有流量都会经过我们的代*理*服务器了。大致流程如下：

https加密流量的处理

在实际操作的时候发现微信所有的流量都走了HTTPS，导致我们的抓到的都是加密数据，对我们没有任何参考意义。经过研究，可以通过在手机和电脑都安装Charles根证书的方式来实现对Https流量的分析，具体操作可以参考：

charles mac下https抓包和iphone https抓包
解决Charles无法正常抓包iOS 11中的Https请求

安装证书后，我们的流量大致是这样子的

经过上述的配置，我们已经可以读取到https的请求和响应数据了，如下图所示。

windows下用findler可以实现相同的功能

其实这就是一个非常典型的中间人场景

数据接口

接下来就根据这些数据来找出我们需要的接口了，经过分析，主要涉及三个接口

获取用户信息接口
获取用户战绩列表接口
获取用户指定战绩详细信息接口

下面我们一个一个看

1. 获取用户信息接口

request

API	/cgi-bin/gamewap/getpubgmdatacenterindex
方法	GET
参数	openid、pass_ticket
cookie	key pass_ticket、uin、pgv_pvid、sd_cookie_crttime、sd_userid

response

{

    "user_info": {

        "openid": "oODfo0pjBQkcNuR4XLTQ321xFVws",

        "head_img_url": "http://wx.qlogo.cn/mmhead/Q3auHgzwzM5hSWxxxxxUQPwW9ibxxxx9DlxLTsKWk97oWpDI0rg/96",

        "nick_name": "望",

        "role_name": "xxxx",

        "zone_area_id": 0,

        "plat_id": 1

    },

    "battle_info": {

        "total_1": 75,

        "total_10": 336,

        "total_game": 745,

        "total_kill": 1669

    },

    "battle_list": [{

        "map_id": 1,

        "room_id": "6575389198189071197",

        "team_id": 57,

        "dt_event_time": 1530953799,

        "rank_in_ds": 3,

        "times_kill": 1,

        "label": "前五",

        "team_type": 1,

        "award_gold": 677,

        "mode": 0

    }],

    "appitem": {

        "AppID": "wx13051697527efc45",

        "IconURL": "https://mmocgame.qpic.cn/wechatgame/mEMdfrX5RU0dZFfNEdCsMJpfsof1HE0TP3cfZiboX0ZPxqh5aZnHjxPFXUGgsXmibe/0",

        "Name": "绝地求生全军出击",

        "BriefName": "绝地求生全军出击",

        "Desc": "官方正版绝地求生手游",

        "Brief": "枪战 | 808.2M",

        "WebURL": "https://game.weixin.qq.com/cgi-bin/h5/static/detail_v2/index.html?wechat_pkgid=detail_v2&appid=wx13051697527efc45&show_bubble=0",

        "DownloadInfo": {

            "DownloadURL": "https://itunes.apple.com/cn/app/id1304987143",

            "DownloadFlag": 5

        },

        "Status": 0,

        "AppInfoFlag": 45,

        "Label": [],

        "AppStorePopUpDialogConfig": {

            "Duration": 1500,

            "Interval": 172800,

            "ServerTimestamp": 1531066098

        },

        "HasEnabledChatGroup": false,

        "AppType": 0,

        "game_tag_list": ["绝地求生", "正版还原", "好友开黑", "百人对战", "超大地图"],

        "recommend_reason": "正版绝地求生，荒野射击",

        "size_desc": "808.2M"

    },

    "is_guest": true,

    "is_blocked": false,

    "errcode": 0,

    "errmsg": "ok"

}

分析

openid是用户的惟一标识。

2. 获取用户战绩列表接口

request

API	/cgi-bin/gamewap/getpubgmbattlelist
方法	GET
参数	openid、pass_ticket、plat_id、after_time、limit
cookie	key pass_ticket、uin、pgv_pvid、sd_cookie_crttime、sd_userid

response

{

"errcode": 0,

"errmsg": "ok",

"next_after_time": 1528120556,

"battle_list": [{

    "map_id": 1,

    "room_id": "6575389198111172597",

    "team_id": 57,

    "dt_event_time": 1530953799,

    "rank_in_ds": 3,

    "times_kill": 1,

    "label": "前五",

    "team_type": 1,

    "award_gold": 677,

    "mode": 0

}, {

    "map_id": 1,

    "room_id": "6575336498940384115",

    "team_id": 11,

    "dt_event_time": 1530941404,

    "rank_in_ds": 5,

    "times_kill": 2,

    "label": "前五",

    "team_type": 1,

    "award_gold": 632,

    "mode": 0

}],

"has_next": true

}

分析
这个接口用after_time来进行分页，遍历获取时可以根据接口响应的has_next和next_after_time来判断是否还有下一页的数据。
列表里面的room_id是每一场battle的惟一标识。

3. 获取用户战绩详情接口

request

API	/cgi-bin/gamewap/getpubgmbattledetail
方法	GET
参数	openid、pass_ticket、room_id
cookie	key pass_ticket、uin、pgv_pvid、sd_cookie_crttime、sd_userid

request

{

"errcode": 0,

"errmsg": "ok",

"base_info": {

    "nick_name": "柚茶",

    "head_img_url": "http://wx.qlogo.cn/mmhead/xxxx/96",

    "dt_event_time": 1528648165,

    "team_type": 4,

    "rank": 1,

    "player_count": 100,

    "role_sex": 1,

    "label": "大吉大利",

    "openid": "oODfo0s1w5lWjmxxxxxgQkcCljXQ"

},

"battle_info": {

    "award_gold": 622,

    "times_kill": 6,

    "times_head_shot": 0,

    "damage": 537,

    "times_assist": 3,

    "survival_duration": 1629,

    "times_save": 0,

    "times_reborn": 0,

    "vehicle_kill": 1,

    "forward_distance": 10140,

    "driving_distance": 5934,

    "dead_poison_circle_no": 6,

    "top_kill_distance": 223,

    "top_kill_distance_weapon_use": 2924130819,

    "be_kill_user": {

        "nick_name": "小旭",

        "head_img_url": "http://wx.qlogo.cn/mmhead/ibLButGMnqJNFsUtStNEV8tzlH1QpwPiaF9kxxxxx66G3ibjic6Ng2Rcg/96",

        "weapon_use": 20101000001,

        "openid": "oODfo0qrPLExxxxc0QKjFPnPxyI"

    },

    "label": "大吉大利"

},

"team_info": {

    "user_list": [{

        "nick_name": "ooo",

        "times_kill": 6,

        "assist_count": 3,

        "survival_duration": 1638,

        "award_gold": 632,

        "head_img_url": "http://wx.qlogo.cn/mmhead/Q3auHgzwzM4k4RXdyxavNxxxxUjcX6Tl47MNNV1dZDliazRKRg",

        "openid": "oODfo0xxxxf1bRAXE-q-lEezK0k"

    }, {

        "nick_name": "我吃炒肉",

        "times_kill": 2,

        "assist_count": 2,

        "survival_duration": 1502,

        "award_gold": 583,

        "head_img_url": "http://wx.qlogo.cn/mmhead/sTJptKvBQLKd5SAAjOF0VrwiapUxxxxFffxoDUcrVjYbDf9pNENQ",

        "openid": "oODfo0gIyDxxxxZpUrSrpapZSDT0"

    }]

},

"is_guest": true,

"is_blocked": false

}

分析
这个接口响应了战斗的详细信息，包括杀*敌数、爆*头数、救人数、跑动距离等等，足够我们分析了。
这个接口还响应了是被谁杀死的以及组团成员的openid，利用这个特性我们这可无限深度的发散爬取更多用户的数据。

至于cookie中的息pass_ticket等信息肯定是用于权限认证的，在上述的几次请求中这些信息都没有变化，所以我们不需要深研其是怎么算出来的，只需要抓包提取到默认信息后填到代码里面就可以用了。

Step 2 爬取数据

接口已经确定下来了，接下来就是去抓取足够量的数据了。

使用requests请求接口获取数据

url = 'https://game.weixin.qq.com/cgi-bin/gamewap/getpubgmdatacenterindex?openid=%s&plat_id=0&uin=&key=&pass_ticket=%s' % (openid, settings.pass_ticket)

r = requests.get(url=url, cookies=settings.def_cookies, headers=settings.def_headers, timeout=(5.0, 5.0))

tmp = r.json()

    wfile = os.path.join(settings.Res_UserInfo_Dir, '%s.txt' % (rediskeys.user(openid)))

    with codecs.open(wfile, 'w', 'utf-8') as wf:

        wf.write(simplejson.dumps(tmp, indent=2, sort_keys=True, ensure_ascii=False))

参照这种方式我们可以很快把另外两个接口写好。

使用redis来标记已经爬取过的信息

在上述接口中我们可能从用户A的入口进去找到用户B的openid，然后从用户B的入口进去又找到用户A的openid，为了避免重复采集，所以我们需要记录下哪些信息是我们采集过的。核心代码片断：

# rediskeys.user_battle_list 根据openid获取存在redis中的key值

def user_battle_list(openid):

return 'ubl_%s' % (openid)

# 在提取battle list之前，首先判断这用用户的数据是否已经提取过了

if settings.DataRedis.get(rediskeys.user_battle_list(openid)):

return True

# 在提取battle list之后，需要在redis中记录用户信息

settings.DataRedis.set(rediskeys.user_battle_list(openid), 1)

使用celery来管理队列

celery是一个非常好用的分布式队列管理工具，我这次只打算在我自己的电脑上运行，所以并没有用到分布式的功能。我们创建三个task和三个queue

task_queues = (

    Queue('queue_get_battle_info', exchange=Exchange('priority', type='direct'), routing_key='gbi'),

    Queue('queue_get_battle_list', exchange=Exchange('priority', type='direct'), routing_key='gbl'),

    Queue('queue_get_user_info', exchange=Exchange('priority', type='direct'), routing_key='gui'),

)

task_routes = ([

    ('get_battle_info', {'queue': 'queue_get_battle_info'}),

    ('get_battle_list', {'queue': 'queue_get_battle_list'}),

    ('get_user_info', {'queue': 'queue_get_user_info'}),

],)

然后在task中控制API请求和Redis数据实现完整的任务逻辑，如：

@app.task(name='get_battle_list')

def get_battle_list(openid, plat_id=None, after_time=0, update_time=None):

    # 判断是否已经取过用户战绩列表信息

    if settings.DataRedis.get(rediskeys.user_battle_list(openid)):

        return True

    if not plat_id:

        try:

            # 提取用户信息

            us = handles.get_user_info_handles(openid)

            plat_id=us['plat_id']

        except Exception as e:

            print 'can not get user plat_id', openid, traceback.format_exc()

            return False

    # 提取战绩列表

    battle_list = handles.get_battle_list_handle(openid, plat_id, after_time=0, update_time=None)



    # 为每一场战斗创建异步获取详情任务

    for room_id in battle_list:

        if not settings.DataRedis.get(rediskeys.user_battle(openid, room_id)):

            get_battle_info.delay(openid, plat_id, room_id)

    return True

开始抓取

因为我们是发散是爬虫，所以需要给代码一个用户的入口，所以需要手动创建一个用户的采集任务

from tasks.all import get_battle_list

my_openid = 'oODfo0oIErZI2xxx9xPlVyQbRPgY'

my_platid = '0'

get_battle_list.delay(my_openid, my_platid, after_time=0, update_time=None)

有入口之后我们就用celery来启动worker去开始爬虫

# 启动获取用户详情worker

celery -A tasks.all worker -c 5 --queue=queue_get_user_info --loglevel=info -n get_user_info@%h

# 启动获取战绩列表worker

celery -A tasks.all worker -c 5 --queue=queue_get_battle_list --loglevel=info -n get_battle_list@%h

# 启动获取战绩详情worker

celery -A tasks.all worker -c 30 --queue=queue_get_battle_info --loglevel=info -n get_battle_info@%h

这样我们的爬虫就可以愉快的跑起来了。再通过celery-flower来查看执行情况。

celery flower -A tasks.all --broker=redis://:$REDIS_PASS@$REDIS_HOST:$REDIS_PORT/10

通过flower，我们可以看到运行的效率还是非常不错的。在执行过程中会发现get_battle_list跑太快，导致get_battle_info即使开了30个并发都还会积压很多，所以需要适时的去停一下这些worker。在我们抓到20万条信息之后就可以停下来了。

Step 3 数据分析

分析方案

20万场战斗的数据已经抓取好了，全部分成json文件存在我本地磁盘上，接下来就做一些简单的分析。 python在数据分析领域也非常强大，有很多非常优秀的库，如pandas和NumPy，可惜我都没有学过，而且对于一个高考数学只考了70几分的人来说，数据分析实在是难，所以就自己写了一个非常简单的程序来做一些浅度分析。需要进行深度分析，又不想自己爬虫的大牛可以联系我打包这些数据。

# coding=utf-8

import os

import json

import datetime

import math

from conf import settings

class UserTeamTypeData:

    def __init__(self, team_type, player_count):

        self.team_type = team_type

        self.player_count = player_count

        self.label = {}

        self.dead_poison_circle_no = {}

        self.count = 0

        self.damage = 0

        self.survival_duration = 0  # 生存时间

        self.driving_distance = 0

        self.forward_distance = 0

        self.times_assist = 0  # 助攻

        self.times_head_shot = 0

        self.times_kill = 0

        self.times_reborn = 0  # 被救次数

        self.times_save = 0  # 救人次数

        self.top_kill_distance = []

        self.top_kill_distance_weapon_use = {}

        self.vehicle_kill = 0  # 车辆杀死

        self.award_gold = 0

        self.times_reborn_by_role_sex = {0: 0, 1: 0}  # 0 男 1 女

        self.times_save_by_role_sex = {0: 0, 1: 0}  # 0 男 1 女

    def update_dead_poison_circle_no(self, dead_poison_circle_no):

        if dead_poison_circle_no in self.dead_poison_circle_no:

            self.dead_poison_circle_no[dead_poison_circle_no] += 1

        else:

            self.dead_poison_circle_no[dead_poison_circle_no] = 1

    def update_times_reborn_and_save_by_role_sex(self, role, times_reborn, times_save):

        if role not in self.times_reborn_by_role_sex:

            return

        self.times_reborn_by_role_sex[role] += times_reborn

        self.times_save_by_role_sex[role] += times_save

    def update_top_kill_distance_weapon_use(self, weaponid):

        if weaponid not in self.top_kill_distance_weapon_use:

            self.top_kill_distance_weapon_use[weaponid] = 1

        else:

            self.top_kill_distance_weapon_use[weaponid] += 1

class UserBattleData:

    def __init__(self, openid):

        self.openid = openid

        self.team_type_res = {}

        self.label = {}

        self.hour_counter = {}

        self.weekday_counter = {}

        self.usetime = 0

        self.day_record = set()

        self.battle_counter = 0

    def get_avg_use_time_per_day(self):

        # print "get_avg_use_time_per_day:", self.openid, self.usetime, len(self.day_record), self.usetime / len(self.day_record)

        return self.usetime / len(self.day_record)

    def update_label(self, lable):

        if lable in self.label:

            self.label[lable] += 1

        else:

            self.label[lable] = 1

    def get_team_type_data(self, team_type, player_count):

        player_count = int(math.ceil(float(player_count) / 10))

        team_type_key = '%d_%d' % (team_type, player_count)

        if team_type_key not in self.team_type_res:

            userteamtypedata = UserTeamTypeData(team_type, player_count)

            self.team_type_res[team_type_key] = userteamtypedata

        else:

            userteamtypedata = self.team_type_res[team_type_key]

        return userteamtypedata

    def update_user_time_property(self, dt_event_time):

        dt_event_time = datetime.datetime.fromtimestamp(dt_event_time)

        hour = dt_event_time.hour

        if hour in self.hour_counter:

            self.hour_counter[hour] += 1

        else:

            self.hour_counter[hour] = 1

        weekday = dt_event_time.weekday()

        if weekday in self.weekday_counter:

            self.weekday_counter[weekday] += 1

        else:

            self.weekday_counter[weekday] = 1

        self.day_record.add(dt_event_time.date())

    def update_battle_info_by_room(self, roomid):

        # print '  load ', self.openid, roomid

        file = os.path.join(settings.Res_UserBattleInfo_Dir, self.openid, '%s.txt' % roomid)

        with open(file, 'r') as rf:

            battledata = json.load(rf)

        self.battle_counter += 1

        base_info = battledata['base_info']

        self.update_user_time_property(base_info['dt_event_time'])

        battle_info = battledata['battle_info']

        userteamtypedata = self.get_team_type_data(base_info['team_type'], base_info['player_count'])

        userteamtypedata.count += 1

        userteamtypedata.award_gold += battle_info['award_gold']

        userteamtypedata.damage += battle_info['damage']

        userteamtypedata.update_dead_poison_circle_no(battle_info['dead_poison_circle_no'])

        userteamtypedata.driving_distance += battle_info['driving_distance']

        userteamtypedata.forward_distance += battle_info['forward_distance']

        self.update_label(battle_info['label'])

        userteamtypedata.survival_duration += battle_info['survival_duration']

        self.usetime += battle_info['survival_duration']/60

        userteamtypedata.times_assist += battle_info['times_assist']

        userteamtypedata.times_head_shot += battle_info['times_head_shot']

        userteamtypedata.times_kill += battle_info['times_kill']

        userteamtypedata.times_reborn += battle_info['times_reborn']

        userteamtypedata.times_save += battle_info['times_save']

        userteamtypedata.damage += battle_info['damage']

        userteamtypedata.top_kill_distance.append(battle_info['top_kill_distance'])

        userteamtypedata.update_times_reborn_and_save_by_role_sex(base_info['role_sex'], battle_info['times_reborn'],

                                                                  battle_info['times_save'])

    def get_user_battleinfo_rooms(self):

        user_dir = os.path.join(settings.Res_UserBattleInfo_Dir, self.openid)

        r = [room for room in os.listdir(user_dir)]

        r = [rr.replace('.txt', '') for rr in r]

        return r

class AllUserCounter:

    def __init__(self):

        self.hour_counter = {0: 0, 1: 0, 2: 0, 3: 0, 4: 0, 5: 0, 6: 0, 7: 0, 8: 0, 9: 0, 10: 0, 11: 0, 12: 0, 13: 0, 14: 0, 15: 0, 16: 0, 17: 0, 18: 0, 19: 0, 20: 0, 21: 0, 22: 0, 23: 0}

        self.weekday_counter = {0: 0, 1: 0, 2: 0, 3: 0, 4: 0, 5: 0, 6: 0}

        self.times_reborn_by_role_sex = {0: 0, 1: 0}  # 0 男 1 女

        self.times_save_by_role_sex = {0: 0, 1: 0}  # 0 男 1 女

        self.user_count = 0

        self.battle_count = 0

        self.every_user_use_time_per_day = []

        self.top_kill_distance = 0

    def avg_use_time(self):

        return sum(self.every_user_use_time_per_day) / len(self.every_user_use_time_per_day)

    def add_user_data(self, userbattledata):

        self.every_user_use_time_per_day.append(userbattledata.get_avg_use_time_per_day())

        self.battle_count += userbattledata.battle_counter

        self.user_count += 1

        for k in userbattledata.hour_counter:

            if k in self.hour_counter:

                self.hour_counter[k] += userbattledata.hour_counter[k]

            else:

                self.hour_counter[k] = userbattledata.hour_counter[k]

        for weekday in userbattledata.weekday_counter:

            if weekday in self.weekday_counter:

                self.weekday_counter[weekday] += userbattledata.weekday_counter[weekday]

            else:

                self.weekday_counter[weekday] = userbattledata.weekday_counter[weekday]

        for userteamtype in userbattledata.team_type_res:

            userteamtypedata = userbattledata.team_type_res[userteamtype]

            for k in userteamtypedata.times_reborn_by_role_sex:

                self.times_reborn_by_role_sex[k] += userteamtypedata.times_reborn_by_role_sex[k]

            for k in userteamtypedata.times_save_by_role_sex:

                self.times_save_by_role_sex[k] += userteamtypedata.times_save_by_role_sex[k]

            if userteamtypedata.top_kill_distance > self.top_kill_distance:

                self.top_kill_distance = userteamtypedata.top_kill_distance

    def __str__(self):

        res = []

        res.append('总用户数\t%d' % self.user_count)

        res.append('总战斗数\t%d' % self.battle_count)

        res.append('平均日耗时\t%d' % self.avg_use_time())

        res.append('最远击杀\t%d' % max(self.top_kill_distance))

        res.append('男性角色\t被救%d次\t救人%d次' % (self.times_reborn_by_role_sex[0], self.times_save_by_role_sex[0]))

        res.append('女性角色\t被救%d次\t救人%d次' % (self.times_reborn_by_role_sex[1], self.times_save_by_role_sex[1]))

        res.append('小时分布')

        for hour in range(0, 24):

            # res.append('\t%d: %d' % (hour, self.hour_counter[hour]))

            res.append('\t%d: %d %.2f%%' % (hour, self.hour_counter[hour], self.hour_counter[hour]/float(self.battle_count)*100))

        res.append('星期分布')

        # res.append(self.weekday_counter.__str__())

        for weekday in range(0, 7):

            res.append('\t%d: %d %.2f%%' % (weekday+1, self.weekday_counter[weekday], (self.weekday_counter[weekday]/float(self.battle_count)*100)))

        return '\n'.join(res)

def get_user_battleinfo_rooms(openid):

    user_dir = os.path.join(settings.Res_UserBattleInfo_Dir, openid)

    # files = os.listdir(user_dir)

    r = [room for room in os.listdir(user_dir)]

    r = [rr.replace('.txt', '') for rr in r]

    return r

if __name__ == '__main__':

    alluserconter = AllUserCounter()



    folders = os.listdir(settings.Res_UserBattleInfo_Dir)

    i = 0

    for folder in folders:

        i+=1

        print i, '/' , len(folders), folder

        userbattledata = UserBattleData(folder)

        for room in userbattledata.get_user_battleinfo_rooms():

            userbattledata.update_battle_info_by_room(room)

        alluserconter.add_user_data(userbattledata)

    print "\n" * 3

    print "---------------------------------------"

    print alluserconter