NRL Fantasy 2023 Part 70 - the Happy Birthday Random thread

Great to hear SJ will be back at Warriors. A chance to build next year with him back and RTS adding some real quality to the backline.

mattnz wrote:Great to hear SJ will be back at Warriors. A chance to build next year with him back and RTS adding some real quality to the backline.

Self imposed is back !

WB

mattnz wrote:Great to hear SJ will be back at Warriors. A chance to build next year with him back and RTS adding some real quality to the backline.

i missed you

Yay Matt’s back

I’ve always maintained fb at bluebet is one of the worst scoring positions to be

That analysis is great, thank you

lukeayee wrote:

Love the enthusiasm and innovation in this analysis.

However, without trying to be too negative, I'm not sure how statistically significant your findings are.

There are a lot of independent variables that effect the players scores. That analysis only looks at performance of X player against Y team. How much of the variation in score could be explained by who's playing, who's injured, weather, momentum of team performance etc etc.

I'm sure there is a correlation of some kind between the teams played, but I really don't think it's as strong as everyone makes out.

Yeah it's really cool but I actually don't really know what to make of it or how to use it.

mattnz wrote:Great to hear SJ will be back at Warriors. A chance to build next year with him back and RTS adding some real quality to the backline.

Good man

Is everyone pivoting away from Fifita now or still bringing him in?

Must win game against Parra for them

mattnz wrote:Great to hear SJ will be back at Warriors. A chance to build next year with him back and RTS adding some real quality to the backline.

Great to see @mattnz back in the ring.
We missed you buddy.
Love to hear your thoughts on Origin III.

ynot wrote:Little bit of analysis for the afternoon:





These will update automatically each week on https://dash.yafss.com.

I've also added some filters for positions, bye weeks and updated the table extension, so it all runs a bit faster.

lukeayee wrote:
Love the enthusiasm and innovation in this analysis.

However, without trying to be too negative, I'm not sure how statistically significant your findings are.

There are a lot of independent variables that effect the players scores. That analysis only looks at performance of X player against Y team. How much of the variation in score could be explained by who's playing, who's injured, weather, momentum of team performance etc etc.

I'm sure there is a correlation of some kind between the teams played, but I really don't think it's as strong as everyone makes out.

I'm a fan too - nice work @ynot !

The way to deal with @lukeayee 's (correct) criticism is to build two models. The model you have here is a model of actual outcomes.

The model you need to complement it and add intelligence (a cute way of saying observed-minus-predicted) is a model of predicted outcomes - some folks here would've seen the xG stuff soccer uses (and is seen in FPL).

This sounds complex, but fortunately you've already darn near everything you need to build the predicted model! Easiest option is to rely on autocorrelation (i.e. tomorrow=yesterday+today+randomness) - a variant of which is how we price players.

A better (still easy) option is throw everything you can think of together in a LASSO or something, and take the top few predictors and back-build the model. Happy to help with R/Python code here if useful.

A full-powered predictive model is a big chunk of work (needs player identification, weather, trends in each, injuries, player-vs-players stats, season point, distance from Origin, etc etc) because it needs separate models for each position&player. But then you'd have a betting forecasting model that'd beat Sportsbet. Again, happy to help but given the quanta of work it'd be less hands-on-code help and more advice.

The Bludger#2 wrote:

Great to see @mattnz back in the ring.
We missed you buddy.
Love to hear your thoughts on Origin III.

Good to be back. Just needed some time out.

Not sure what happened in Origin III, all the real opportunities went to NSW and they made the most of them. The normal desperation from QLD didn't seem to be there, as the series was wrapped up. New recruits Best and Koala played well.

Shady2K wrote:Is everyone pivoting away from Fifita now or still bringing him in?

Must win game against Parra for them

Depends on Carrigan and SJ's availability for me. Most likely one of Cleary or Fifita, but might be able to swing both even with holding Haas. If Cleary doesn't play, then will grab Fifita assuming he plays.

Shady2K wrote:Is everyone pivoting away from Fifita now or still bringing him in?

Must win game against Parra for them

What would be a reason to not bring Fifita in?

He has looked much more involved in attack since the interim coach has taken over, in addition to getting through lots of work. Was my first picked, with the additional trades available.

My POD pick-up for the week is Fitzgibbon. Started the season really well, back to old form. Then had some injury niggles, so dropped some cash.

Has always been at his best when Ponga is playing great at fullback, Ponga knows how to set him up. Looked great at the weekend and should be a nice depth pick up, that can be looped in when he scores really well.

L-Jimmy wrote:



I'm a fan too - nice work @ynot !

The way to deal with @lukeayee 's (correct) criticism is to build two models. The model you have here is a model of actual outcomes.

The model you need to complement it and add intelligence (a cute way of saying observed-minus-predicted) is a model of predicted outcomes - some folks here would've seen the xG stuff soccer uses (and is seen in FPL).

This sounds complex, but fortunately you've already darn near everything you need to build the predicted model! Easiest option is to rely on autocorrelation (i.e. tomorrow=yesterday+today+randomness) - a variant of which is how we price players.

A better (still easy) option is throw everything you can think of together in a LASSO or something, and take the top few predictors and back-build the model. Happy to help with R/Python code here if useful.

A full-powered predictive model is a big chunk of work (needs player identification, weather, trends in each, injuries, player-vs-players stats, season point, distance from Origin, etc etc) because it needs separate models for each position&player. But then you'd have a betting forecasting model that'd beat Sportsbet. Again, happy to help but given the quanta of work it'd be less hands-on-code help and more advice.

Woweee we have some smart cookies amongst us.

I'm not the only one who has done econometrics?

L-Jimmy wrote:



I'm a fan too - nice work @ynot !

The way to deal with @lukeayee 's (correct) criticism is to build two models. The model you have here is a model of actual outcomes.

The model you need to complement it and add intelligence (a cute way of saying observed-minus-predicted) is a model of predicted outcomes - some folks here would've seen the xG stuff soccer uses (and is seen in FPL).

This sounds complex, but fortunately you've already darn near everything you need to build the predicted model! Easiest option is to rely on autocorrelation (i.e. tomorrow=yesterday+today+randomness) - a variant of which is how we price players.

A better (still easy) option is throw everything you can think of together in a LASSO or something, and take the top few predictors and back-build the model. Happy to help with R/Python code here if useful.

A full-powered predictive model is a big chunk of work (needs player identification, weather, trends in each, injuries, player-vs-players stats, season point, distance from Origin, etc etc) because it needs separate models for each position&player. But then you'd have a betting forecasting model that'd beat Sportsbet. Again, happy to help but given the quanta of work it'd be less hands-on-code help and more advice.

Yeah that's the gap!

To me, knowing Tigers remaining opponents let +1.6 pts per game more than the league average for FBs doesn't really help with a decision to keep Bula or not. It sounds good for Bula! But I'm not really wanting to assess Bula against the league average for FBs.

There are FBs in the red (Drink, Gutho) that could still outperform Bula.

I am not the kind of useful analyst that can write code and shit though.

mattnz wrote:

What would be a reason to not bring Fifita in?

He has looked much more involved in attack since the interim coach has taken over, in addition to getting through lots of work. Was my first picked, with the additional trades available.

My POD pick-up for the week is Fitzgibbon. Started the season really well, back to old form. Then had some injury niggles, so dropped some cash.

Has always been at his best when Ponga is playing great at fullback, Ponga knows how to set him up. Looked great at the weekend and should be a nice depth pick up, that can be looped in when he scores really well.

Played 30 min yesterday and news he has a back niggle or hip

Origin hasn't really changed the sub 300k choices. Shame tigers play first as they provide the best shitto ctr coverage now vtw is out

Khany wrote:Origin hasn't really changed the sub 300k choices. Shame tigers play first as they provide the best shitto ctr coverage now vtw is out

I like Kepaoa more than Talau there. But annoying the cheapest CTRs are both Tigers. Would hate to bring one in and by Sunday arvo VTW is starting again.

mattnz wrote:Great to hear SJ will be back at Warriors. A chance to build next year with him back and RTS adding some real quality to the backline.

Welcome back dude.

lukeayee wrote:

Woweee we have some smart cookies amongst us.

I'm not the only one who has done econometrics?

What about funomics?