【大小球让球实战】——深度剖析足球大小球数据分析软件，找到一个临界点，把准确率提高到70%？

最近在跟一些圈内朋友聊天，发现每个人都有一套自己的玩球逻辑，比如有些人看亚盘，有些看欧盘，有些玩绝杀，有些看盘口，是否可以自定义的去玩球？结合多年的经验，把分析的过程大概写了一下。

数据爬取

采集汇聚全球各大足球赛事的即时比赛数据。从进球瞬间到红黄牌警告，每一刻赛场动态都捕捉
采用先进的数据清洗与校验技术，确保每一条即时比赛数据都精准无误。
无论是球队控球率、射门次数还是传球成功率，力求做到分秒不差。

获取实时滚球数据

这里简单用python爬取某站的滚球数据

	t = int(time.time() * 1000)
    liveUrl = "https://xxx.com/"
    # 全天比赛信息
    match_url = liveUrl + "/xxxx" + str(t)

    response = requests.get(match_url, headers=headers)
    match_content = ''
    # print('---正在获取比赛信息---')
    if response.status_code == 200:
        match_content = response.content.decode('utf-8')
        print(response.text)
        print(match_content)
    if response.status_code != 200:
        print('---赔率获取结果异常---', response.status_code)

数据清洗

这里简单的计算球队的elo，将多余数据简化

"""
    计算elo
    :param match:
    :param elo_cache:
    :param info:
    :return:
    """
    home_id = match['home_id']
    away_id = match['away_id']
    match_res = match['sfp_res']
    if home_id in elo_cache.keys():
        home_info = elo_cache[home_id]
        home_po = home_info['new_elo']
    else:
        home_po = 2000
        home_info = {
            "old_elo": 0,
            "new_elo": home_po,
        }
    if away_id in elo_cache.keys():
        away_info = elo_cache[away_id]
        away_po = away_info['new_elo']
    else:
        away_po = 2000
        away_info = {
            "old_elo": 0,
            "new_elo": away_po,
        }

    home_pn = get_pn(home_po, away_po, HOME_MATCH_RES_ELO[match_res])
    away_pn = get_pn(away_po, home_po, AWAY_MATCH_RES_ELO[match_res])
    home_info['old_elo'] = home_po
    home_info['new_elo'] = home_pn
    away_info['old_elo'] = away_pn
    away_info['new_elo'] = away_po

    elo_cache[home_id] = home_info
    elo_cache[away_id] = away_info
    home_team_info = {f"home_{key}": val for key, val in home_info.items()}
    info.update(home_team_info)
    home_team_info = {f"away_{key}": val for key, val in away_info.items()}
    info.update(home_team_info)

清洗之后的数据显示

自定义条件设置

比如说：如果即时盘中未进球（70分钟前），降盘1个球以上，选大球，赔率大于1.9
想要拆选这种数据，那么先写个SQL验证一下

select *From fb_roll where mins<=70
and daGA1>=0.9
and overallGoals=0
and (daDbCA2-daDbGA2)>=1

得到结果

分析结果

需要对这一结果进行历史验证
看了一下，57%的胜率，勉强能盈利，完结撒花！！！

select *From fb_roll_his where mins<=70
and daGA1>=0.9
and overallGoals=0
and (daDbCA2-daDbGA2)>=1

写到最后，推荐一款基于大模型+条件预测的平台，有兴趣可以了解一下：http://lcsjfx.com/web/sjfxpt.html