Thread has been deleted
Last comment
Neural network predictions
 | 
Russia _g00s3_ 
What do you think about doing predictions on csgo matches with trained neural network? There are good training dataset - over 70k matches for past 7 years I already tried doing very simple neural network and trained it with 5k tier1-2 matches, prediction here: hltv.org/matches/2344719/big-vs-vitality.. I want to continue testing different models and datasets in future but im lazy.
2020-10-22 10:52
Topics are hidden when running Sport mode.
#1
 | 
United States E_40
so you think i should bet on Over2.5 ?
2020-10-22 10:56
3 replies
#5
 | 
Russia _g00s3_
I can't say it good prediction or not. 1/0.65 = 1.54 so if you find some where odds more than @1.54 it might be profitable bet theoretically. But if you want to make a bet on it don't bet much because i did just a simple example of neural network..
2020-10-22 11:42
2 replies
#16
 | 
United States E_40
please dont do predictions anymore thanks for loose
2020-10-22 15:42
1 reply
#17
 | 
Russia _g00s3_
Hey? 1. I said several times it is just simple network for fun. 2. I didn't mean that this prediction will 100% win (network said it has 65% if winning) 3. Have you ever heard "Conditional expectation"? For example if there are 50% chance of winning and @2.2 odds it it *PROFITABLE* bet, because 2.2 > 2.0 If you have 50% chance and @1.8 odds you will lose money if you do alot of bets. If you lost one single bet it dont mean that you will lose all other bets. 4. Who the fuck told you to bet on it without any tests before?
2020-10-22 15:57
#2
zorte | 
Finland ye6ok
a lot of people have models already, even for handicaps, but if you made this one public then I guess it would be nice for some
2020-10-22 10:57
7 replies
#6
 | 
Russia _g00s3_
I will continue trying to find good model for csgo and then if it will show good results i will share all predictions with people
2020-10-22 11:43
6 replies
#7
zorte | 
Finland ye6ok
good luck good luck :D I dont really know what your formula has but "current form ranking" stat helped me a lot in balancing model, if that makes sense
2020-10-22 11:45
5 replies
#8
 | 
Russia _g00s3_
I will try to add as much input data as it possible with hltv info
2020-10-22 11:49
4 replies
#9
zorte | 
Finland ye6ok
glgl, just try to not get banned. A buddy of mine got hltv ip ban for scraping data
2020-10-22 11:50
3 replies
The world would be so much better if hltv released an official api
2020-10-22 12:03
2 replies
#14
zorte | 
Finland ye6ok
hltv still stuck in 2000s xd
2020-10-22 12:05
#15
 | 
Russia _g00s3_
It would be very nice... Parsing fucking html pages for every single match isnt good
2020-10-22 12:10
nothing would make cs less gambling
2020-10-22 10:58
1 reply
but hey your idea is interesting, keep it up, maybe you'll create something interesting of it
2020-10-22 10:59
It's fun if you're a programmer, but of course nothing can predict stuff, not even statistics.
2020-10-22 11:50
1 reply
#12
 | 
Russia _g00s3_
Yes i just doing it for fun, i don't except really good results
2020-10-22 11:54
Very nice!
2020-10-22 11:51
I tried it for league of legends cause they have several good apis and got an accuracy of 60%. My model didnt have some sort of recurrency tho, thats why it liked to predict teams that have been good in the past for a long time, rather recent good teams. Later I added just the last 5 games (0 or 1 for winning/loosing and enemy team name). But I have no idea how to tell the nn that the teamnames in different inputs are meant to be the same. Didnt really test this method on new data, but on old data I got 60% accuracy, too
2020-10-22 19:45
1 reply
#19
 | 
Russia _g00s3_
I don't think you actually need to tell nn that X1 and X2 teams are same team. Just put more input data like each player stats, head to head stats, and as much as you can other stuff 🤷‍♂️
2020-10-23 00:50
can you tell how you became a data scientist? like what should I study is chronological order?
2020-11-08 12:56
1 reply
#21
 | 
Russia _g00s3_
im not data scientist at all. It is second time in my life i doing something like this. Im just interested in machine learning
2020-11-08 12:58
Nexus
1.74
Apeks
2.01
OG
3.60
MIBR
1.29
Chaos
3.88
Vitality
1.23
Bet value
Amount of money to be placed
Winning
Odds total ratio
-
Login or register to add your comment to the discussion.