Accuracy of Prediction Models

[thebillfella]

I know I'm a bit old school and out of touch here, but I just don't get it guys and girls. Where is the fun in playing this game or sense of achievement if you are getting AI to make all of the decisions for you?!!

[blahblah]

I know I'm a bit old school and out of touch here, but I just don't get it guys and girls. Where is the fun in playing this game or sense of achievement if you are getting AI to make all of the decisions for you?!!

And it's not AI: it's just an old fashioned Computer Program trundling away doing as it's told and doing some Maffs to some inputs inc what is ripped from OPTS, FPL etc

[hancockjr]

AI won’t make any decisions. At best it suggests decisions which one can take or not. But also no-one blindly following a model will do really well, but those that take what the models are good at and supplement that with extra consideration will likely do the best.

For instance is an early BB a good idea? One consideration is whether playing/viable base price players will become apparent by the time you wildcard (if so, later BB will likely pay off). No model will tell you that.

Don’t get me wrong - I far preferred it when info was scarce and you had to rely much more on your own analysis and research. And I was much better at it back then! But the world we are now in is that the key is to understand what info is available, which is quality and which isn’t, and use it to assist decisions.

Anyone is free to ignore it all, but they won’t do well.

I actually found the World Cup comp a bit of a throwback, far fewer models and much less info.

[thebillfella]

So what are the key models/data sources critical these days then? Solio? FFHub? FFScout? NextXI? FFFix? Others? Are people scraping data from all of the above and coming to a consensus or just focusing on one or two as the best sources of truth?

[blahblah]

If you can work out how to use Solio (no idea how much it costs as I only used it when it was a Free Introduction) and don't take the GW guesses at points seriously then it's quite good for seeing how xfers feed through the future GWs.

Very basically it allows branches to be set up so that you can see what should/could happen if you do this or that Xfer or nothing etc for a few GWs and have loads of versions of your team/squad. So you can see all these versions until the end of the season.

Given all the uncertaties this preseason from the new managers/virgin European Clubs and managers to new players to new managers to it being the first season after a WC in a while I'm not sure any maffs based on last season will be of much use until at least the IB.

[Hogmeister]

I far preferred it when info was scarce and you had to rely much more on your own analysis and research. And I was much better at it back then! But the world we are now in is that the key is to understand what info is available, which is quality and which isn’t, and use it to assist decisions.

Anyone is free to ignore it all, but they won’t do well.

Yes exactly. I used to have a huge edge over most people in my minileagues by coming on here and learning from the discussions with other “engaged managers”. These days everyone has near instant access to the same information, and a lot more besides. Sadly those days are gone.

And the Solio model can help with analysing chip strategy by the way, as SP highlighted, it’s not just player points predictions. So on solio you can have, for example, one branch based on BB1 and another on BB2, and it will produce its assessment of each in terms of predicted total points, potential upside/downside (risk curves), likely robustness against unexpected events, etc.

And @Blah… they are still offering access to Solio for free, so you can still have a play without forking out money. Limits you to only a few branches, there are other limits/restrictions, and the planning horizon is capped at 5 weeks, but still (IMO) gives some useful insights.

[blahblah]

Is there an idiots guide to doing the Branches, which was kinda useful to see which players I had vs whoever in a few GWs.

Any idea how to get the Far East betting odds? I thought that was the best bit tbh - I'm not sure other sites used those almost objective stats.

[hancockjr]

Yes Solio can look at chip strategies but won’t help with the specific consideration I raised.

[blahblah]

Yes Solio can look at chip strategies but won’t help with the specific consideration I raised.

I totally agree that it won't. But if you want to see how dumbed down doing well in FPL etc has become just watch Hand Shandy.

There is a good thread on here about an "AI" bod who got his tool to read the transcripts of various Content Creators to decide what to do. He was interesting because he didn't know much about the Prem and people were messaging him about the best Tubers to rip off

The trouble was he ended up copying other aggregators and it got vaguely surreal

[zipnolan]

Is there an idiots guide to doing the Branches, which was kinda useful to see which players I had vs whoever in a few GWs.

There is a series of short YT videos (6 in all) which cover everything: this is the first one


https://www.youtube.com/watch?v=EtL1kwUbq04

This is video 6, on branches:

https://www.youtube.com/watch?v=x7wbyB7Lu7c

[blahblah]

I'll give them a look.

When I saw 6 I was thinking "that's 6 hours"

[Hogmeister]

There is a good thread on here about an "AI" bod who got his tool to read the transcripts of various Content Creators to decide what to do. He was interesting because he didn't know much about the Prem and people were messaging him about the best Tubers to rip off
The trouble was he ended up copying other aggregators and it got vaguely surreal

Yes I remember that thread, and that guy trying to do that. In fact a couple of weeks ago I had a look on youtube, wondering how that AI-generated team had finished up at the end of the season. Looks like the poster got bored with the whole idea and stopped putting videos out halfway through the season.

[blahblah]

Yeah, he stopped for some reason. Maybe litigation from some he was reading?

I was kinda following him on YouTube as he was a good watch compared to the Aggregators (is that the right insult?) that he was Aggregating

[Pirlo's Beard]

I haven't played it, but that seems about right. You need regular and playing players, but ones that rake up almost no points.

I've been unwittingly playing a version of Anti FPL for the last couple of years.

[Smurphy Paw]

f17cba60-cede-4d3a-8fb4-be6dd3321e93.jpeg

Scout are doing OK so far

[Smurphy Paw]

IMG_8005.jpeg

Although for longer term advice Review does best

[Smurphy Paw]

HOWEVER
Apparently the first set of stats has merit and the second one is garbage. Bad maths. So Review don’t like it, never mind those who are behind Review in the pecking order.

Go with the first if you go with any

[Hogmeister]

HOWEVER
Apparently the first set of stats has merit and the second one is garbage. Bad maths. So Review don’t like it, never mind those who are behind Review in the pecking order.

Go with the first if you go with any

If I've understood these twitter/X posts from midnightmule:

The one I posted at the very start of this thread was for the whole of last season - albeit the "bad maths" version. I tend to agree with that bad maths verdict by the way... what matters is the least variance from what actually happens, so the signed variance version ( -1 + 1 = 0) is probably the wrong approach.

While the two posted above are for the first week of the new season? If that's the case, I'd expect that just one week of data (just like picking a player in FPL) is not enough to go on.

[Smurphy Paw]

I thought about posting the explainer - it was hancockjr who summarised it in our 5AS thread and so his to share.
Suffice to say, the guys who run the Scout model think it is garbage and confirmed that the other model operators think the same