2019 FF Original discussion thread


#422

Probably Jones’ best score for the year (that’s not saying much though), but Sinclair got slaughtered by Gawn. Was hoping/needing a better opening.


#423

Doesn’t matter because he was a late out, but Lockhart doesn’t seem to be on my list.


#424

Ross and Bolton to make their Saracens debut. The new era is (kinda, splutteringly ) underway. A late GWS out with Shipley coming in would be outstandingly handy.


#425

I don’t really have a lot to say about my four Swans boys last night.
In a word: adequate.


#426

I’m still a little confused about it’s purpose.
Dont we already have live scoring through UF?

If we’re looking at getting projected scores, then seems a bit of extra work. We should just each pay an extra $1 entry fee to cover the $20 UF Insider Pass which gives a live projected score.

eg from League 2
Capture


#427

Unless they’ve drastically improved their algorithm since the last time I had access to their predicted scores, I wouldn’t be in favour of wasting people’s money to get an inferior product to Neela’s (assuming he has the time to make it work).


#428

Agree with Dunlop. Their predictions are useless.
I trust Neela’s ones a lot more.

I can help download the data through the weekend if @Neela let’s me know how he does it.
I’m also hoping to use parts of what he’s done on my sheet. Mines been mainly manual entry which I can do, but would really love to be able to smarten my own xls sheet up to match.


#429

Just thought I would poke my head in here and I have just had a look at the UF predictions for our comp.

  • For the first 3 rounds of the season the UF predictions for BBFFL#2 were within 100 points of the actual score on 2/3rds (36 out of 54) of the time.

  • 6 times the score was within 10 of the prediction - including one for @saladin which was bang on.

  • So far for the season only 2 scores have been out by over 200 points from the prediction.

  • Precentage wise 43 out of 54 scores were within 10% of the UF prediction.

  • There is an even spread with 26 predictions higher than the final score and 27 lower.

Whilst not perfect I think the UF scores are a reasonable guide.


#430

Getting my 2 * UF and 2* SC(draft,classic) teams set properly will take precedence, but I’ve got it bookmarked and will try to fill it out as soon as i am able.

Prediction has me getting flogged by AllBlack as expected!


#431

Gosh, I’ve made some baaaaad decisions with my team over the last 9 months.
A lot of work to do to get back where I was.


#432

I’m not off to a great start.


#433

Lmao. Schache played an uninjured four quarters for a total of 9 points. And I reckon he had most of them by 1/4 time.

Outstanding drafting by me.


#434

Yup, we can do that. This was going to have other data such as best +/- differential for the round (actual - average), it would tie into Dunlop’s draft spreadsheet. It’s there but not quite, it would be cool to automate any updates from Ultimate Footy to it.


#435

I’m one short, Weller was on the field but has a hammy issue now.


#436

Sounds great then mate. You’ve done some excellent work with it.


#437

Thanks for the feedback. On those stats, they’ve evidently considerably improved the algorithm then, as when Blummers bought the stats package for the league a few years back it was roughly 60% reasonable predictions, 5% zeros for guys named and 35% random outliers (guys predicted to smash their PBs by considerable margins or regular 100 scorers predicted to score sub 50).


#438

248 from my three Cats in the early game, please.
Keep up that 90 average.


#439

Not a great start by the donkeys. Mayne’s 10 is going to sting. Down 60-odd with +2 players. Will take some effort to crawl this one back.


#440

They have been okay this year. But that is pretty common early in the year.

My xls sheet does it’s own predictions based on an average of the following:

  • what the player scored last week
  • last 4 weeks average
  • highest score for the season
  • lowest score for the season
  • average for the season

From round 5 onwards, it works a treat due to having more data available, however early in the season, it struggles due to lack of data so sometimes I add last years average in lieu of last four weeks or a expected average for the season.

I generally rely on UF to predict the opposition’s scores and I’ve found them to be flawed in the past. I’ve had a look and their predictions this year seem okay. They used to struggle on the players who haven’t played games or who have played very few games. For instance someone like Sam Walsh would be predicted to average around 55 in his first game but I’d say a 70-75 would be a better prediction for a number 1 draft pick midfielder.

I still pay for the UF predictor but mainly so that I can read their draft guide as Inside Football no longer publish.
What I did notice was that their expected player averages during the draft were still flawed. For memory, they predicted James Bell from Sydney to average 80. There were other ‘obscure’ players who would be unlikely to get a game be expected to average 70 to 75, yet someone like Bailey Smith had an average of 55. I realise he’s currently sitting on a 55 average, but I’d bet that it becomes 70 to 75 by the end of the season if not more.

Other things of note…
For memory UF had Josh Smith of West Coast averaging around 70ish for the season during the draft, but was predicted to score 55 in the game he played. Looking at his UF profile, he’s only due to score above 60 twice this year and the rest are below 60 with four in the 40s. I would have thought off scoring 61 in his first game, his yearly average would drop to around 60 to 65ish (from the 70ish pre draft), but they’ve downgraded him to 55ish.

They have Callan Ward projected to score 99 this week, 112 next week, then 92, then 106, then 110.
Now, I’ve owned Callan Ward since he first played a game of footy for the Bulldogs. The last time he went 90+ in 5 consecutive weeks was in 2012. This is coming off a limited pre-season and a rough injury as well. I would be pleased with 95 every week, let alone an average of 104.

They bump up established player projections and reduce younger players projections and this helps them get to their predicted score that they had prior to the game beginning. But once the game starts, due to these innaccuracies, their predictions through the weekend are useless and are dependant on who you still have to play.

I just don’t know how they come about with their predictions.


#441

No late changes cats/GWS , so there goes my last hope to field 18 players via Shipley.