I have spoken to a few people with models for European soccer and asked them how much soccer they watch and they barely watch any.
I asked them about players and they didn't know any.
I asked them what stats they looked at for individual players and they didn't take individual player stats.
I asked them what use stats were useful from last season for a team that has at least 5 new players in their starting XI and they said it didn't matter.
I asked them if they were profitable and they said they were.
They have a 'machine learning' model which 'learns over time'. I don't understand what that means but clearly the output is only as good as the input and they are using obsolete data in my opinion.
Still don't know how all that can match up.
Before the current stream of advanced data-driven models, the sharpest guys had models that simply used data that we now view as very basic, like shots, passes, time-of-posession, corners, cards, etc etc. Even today, some very simple models can be fed with data found on whoscored.com and I'd argue that they wouldn't be too far off the advanced xG models, which currently are in fashion.
Today, modelers that can pay for OPTA data and develop their own player-driven models, have a much higher chance of of originating +EV bets in soccer.
No. Probably depends on the market. I'm saying if you are using a model that's using past stats for players are no longer on the team, you better be making adjustments for that.
Or team level models driven by players who are not playing in that next game (which can also be equivalent to 'no longer on the team' as Durito said)
I dunno they quoted me something like 1/4 million $ for the data I wanted. Seemed a little excessive, so I just stole it.
That's impressive. Why don't you just return the data after you're through with it. Then contact them and tell them you just "borrowed" it for a short time. :-)
Comments
Before the current stream of advanced data-driven models, the sharpest guys had models that simply used data that we now view as very basic, like shots, passes, time-of-posession, corners, cards, etc etc. Even today, some very simple models can be fed with data found on whoscored.com and I'd argue that they wouldn't be too far off the advanced xG models, which currently are in fashion.
Today, modelers that can pay for OPTA data and develop their own player-driven models, have a much higher chance of of originating +EV bets in soccer.
Bad experience/price?
Or you don't think their data is relevant?
Well-known issue.
Or team level models driven by players who are not playing in that next game (which can also be equivalent to 'no longer on the team' as Durito said)
That's impressive. Why don't you just return the data after you're through with it. Then contact them and tell them you just "borrowed" it for a short time. :-)